When it comes to heart health and how to keep it, lab results and risk scores often contradict each other. That happens more frequently once past age 60.

Why it matters

When your risk estimate hovers around a treatment recommendation threshold, it pays to do a “due diligence” of your lab reports. After all, at that age, we don’t have much margin for error anymore.

Your takeaway

The 5 most common contradictions on your cardiovascular lab reports and how to bring them up to your doctor for better-informed shared decision-making.
50% to 100% discount on the Tiebreaker do-it-yourself course and self-assessment toolkit for heart-related blood tests (for annual and founder members respectively).

Table of Content
· The following scenario could be yours (trust me, I have seen it before)
· Why your numbers disagree in the first place
· Why this happens more often after 60
· 1. Two LDL values for the same blood sample
· 2. LDL says one thing, non-HDL another
· 3. Your ApoB doesn’t match your LDL
· 4. Two risk scores, two different answers
· 5. The risk score says one thing, the calcium scan another
· The one rule to take away
· Want to run these checks on your own numbers?
· Cited References

The following scenario could be yours (trust me, I have seen it before)

You’re 64, you feel fine, and the results of your latest blood test have just arrived.

At 98 mg/dL (2.5 mmol/L), your LDL cholesterol sits within the reference range. But further down is a marker your doctor had ordered for the first time,

• ApoB: At 118 mg/dL, the report flags it as high.

• An online calculator puts your 10-year heart risk at 7%.

• The printout your doctor hands you says 12%.

• Your last year’s coronary artery calcium score, the one that came back with an enviable “zero”, doesn’t seem to have made it into any of the risk calculations.

You start to wonder whether your cholesterol is really fine and whether your risk is high (>10%) or maybe not.

The numbers disagree because they measure different things, in different ways, against different reference groups. And in a 15-minute appointment, you need to know which questions to ask when your doctor suggests you take a statin for the rest of your life.

That’s what this post will help you with. Most contradictions on a lipid report, or between risk scores, can be resolved. Below are the five you’re most likely to meet, what the new 2026 ACC/AHA cholesterol guideline says about each, and the questions to bring to your doctor.

Why your numbers disagree in the first place

Three reasons explain almost every case:

• Some numbers are measured, others calculated. Most labs don’t measure LDL at all. They calculate it from total cholesterol, HDL, and triglycerides, and the formula matters.

• Some measure mass, others count particles. LDL cholesterol measures the weight of cholesterol the particles carry. ApoB counts the particles. Lp(a) shows weight when measured in mg/dL but particle numbers when measured in nmol/L. You can’t convert one into the other because particles vary in size, and plaque risk aligns more with particles, less with mass.

• Risk scores answer different questions. Each was built in different people, to predict different events. So the same person gets different answers.

Why this happens more often after 60

These contradictions happen across the adult age spectrum, increasingly though after 60. At that age they carry more weight, for the simple reason that we don’t have much time left to correct their consequences:

• The numbers for cholesterol mass and particles drift apart. In the large ApoB study I’ll come to below, higher-than-expected ApoB was more common in older adults and in people taking statins. Statins lower LDL cholesterol more than they lower the particle count, so on treatment the two numbers often part ways.

• Triglycerides tend to creep up, especially in women after menopause. The old LDL formula (Friedewald) tends to underestimate LDL when triglycerides move up. But Friedewald still reigns supreme, especially outside the US, so check your lab report’s small print.

• Age powers the needle on the risk dial. Age is the weightiest factor in every risk equation. Using the latest PREVENT equation, 3 of 4 people below 60 sit in the low-risk category [1]. At age 70 plus, that category is empty, and 8 of 10 people find themselves in the intermediate and high risk category. These risk algorithms grew out of population studies that only considered readily available biomarker data. So for many people over 60, the choice of equation, or one extra test such as ApoB, Lp(a), or a calcium scan, can flip the risk category, and with it the treatment recommendation.

• The equations’ hand-over. PREVENT stops at 79, MESA at 84, and in Europe SCORE2 hands over to SCORE2-OP at 70. Around these ages, the same person can get different numbers not only from different models but also from the same model just by crossing an age threshold.

• Calcium becomes common. By your 60s, a calcium score above 0 is more common, so how your score compares with others of your age and sex matters.

1. Two LDL values for the same blood sample

In our opening example, your lab report (using the Friedewald equation) puts your LDL at 98. The Friedewald equation was developed in the early 70s. Today, the AHA/ACC guidelines recommend the 2013 Martin-Hopkins equation [2], because it more accurately reflects LDL when triglycerides are elevated [3]. The image below illustrates the discrepancy:

I have created that calculator on my webpage for you to use and play around with

Depending on your TG level, Friedewald underestimates LDL a lot, as you’ll see when you play around with the numbers. A gap under 10 mg/dL doesn’t matter. A bigger gap usually means the Martin-Hopkins value is the closer one.

Ask: “Which LDL value should we use for decisions, and, if the lab value is based on Friedewald, will the switch to Martin-Hopkins make a difference?”

2. LDL says one thing, non-HDL another

Non-HDL cholesterol is simply total cholesterol minus HDL. You’ll probably find it on your lab report and might be tempted to skip it. Here is why you shouldn’t:

The guideline pairs every LDL goal with a non-HDL goal 30 mg/dL higher: LDL below 100 goes with non-HDL below 130, LDL below 70 with non-HDL below 100. When your non-HDL lands in a worse band than your LDL, the extra cholesterol is traveling either in VLDL remnants or IDL. Non-HDL can raise a first yellow flag that LDL alone doesn’t. But there is another more pointed question you can raise.

Ask: “My non-HDL is in a higher band than my LDL. Would measuring my ApoB particle count provide a clearer picture of my atherogenic particle burden than relying on non-HDL-C alone?”

3. Your ApoB doesn’t match your LDL

In our opening example, it was 118 mg/dL. It is the number that mattered most in that case.

Every LDL particle carries exactly one ApoB molecule, so ApoB counts particles. Two people with the same LDL can carry very different numbers of them. I discussed this issue in detail in an earlier post.

In a study of nearly 12,700 US adults not taking statins, 95% of those with an LDL of 100 mg/dL had ApoB values between 66 and 99 mg/dL [4]. An ApoB of 118 with an LDL of 98 is far outside that range: many more particles than the LDL suggests.

The image is a screenshot taken of the self-assessment tool I created, “The Tiebreaker, When Your Lab Values and Risk Scores Disagree” (Link is at the end of the post)

When LDL and ApoB are discordant, risk tends to follow ApoB. Positive discordance (not positive in the colloquial sense, as it translates into “more ApoB than expected for a given LDL) is more common in men, older adults, and people with obesity, diabetes, or high triglycerides, but the spread within each group is wider than the differences between them. Only an (inexpensive) lab test for ApoB will tell you precisely where you stand.

That recommendation made it also into the 2026 guideline, which now calls ApoB testing reasonable for people on cholesterol-lowering treatment, to decide whether to go further once LDL and non-HDL goals are met, and possibly useful before treatment to sharpen the risk estimate.

Ask: “My ApoB is higher than expected for my LDL. Should my treatment goal be based on ApoB rather than LDL?”

4. Two risk scores, two different answers

In our opening example, the 7% and the 12% risk estimates came from the same PREVENT risk prediction algorithm. 7.2%risk of suffering a fatal or non-fatal heart attack or stroke within the next 10 years. It’s 12.2% if you also count heart failure, which PREVENT’s total-CVD estimate includes (HF being driven mainly by adding BMI into the equation).

The image is a screenshot taken of the self-assessment tool I created, “The Tiebreaker, When Your Lab Values and Risk Scores Disagree” (Link is at the end of the post)

The 2026 guideline switched to the new US risk equation, PREVENT, which is built on more than six million recent patients [5], gives dramatically lower numbers than the older Pooled Cohort Equations (PCE), but hasn’t made it into every doctor’s risk estimation system. Those lower estimates also drove the guideline’s risk thresholds down: borderline now starts at 3%, intermediate at 5%, high at 10%.

SCORE2, the European score, counts heart attacks, strokes, and cardiovascular deaths (similar to PREVENT’s ASCVD risk plus deaths from other cardiovascular causes), and is calibrated to each region of Europe. Which is why, in our example, using a moderate-risk region, the estimate is a bit higher than PREVENT.

The MESA calculator sees your risk at 8.7% (for our opening example) when ignoring your CAC score, but dramatically reducing it to 3% when recognizing your last year’s zero score. That’s the subject of the next section.

Ask: “Which risk equation is this number from, and which risk category does it put me in under the 2026 guideline?”

5. The risk score says one thing, the calcium scan another

PREVENT and SCORE2 estimate your risk from commonly available risk factors: age, cholesterol, blood pressure, smoking, diabetes. They ignore coronary artery calcium (CAC). Only the MESA calculator bridges this gap. I discussed MESA and the role of CAC in an earlier post about a reader’s actual case. When a risk score and the scan disagree, the scan usually wins.

• Borderline or intermediate risk, but CAC of 0: CAC “zero” dramatically reduces your risk such that even the guidelines recommend as reasonable to defer cholesterol-lowering medication and repeat the scan in 3 to 7 years. That doesn’t apply if you have diabetes, smoke, have an LDL of 190 or more, or a strong family history of early heart disease.

• Low risk score, CAC above 0: plaque is already there. The guideline recommends treatment, with tighter LDL goals the higher the score. And once your CAC reaches 300 or more, the risk equations no longer decide: the 2026 guideline considers your risk comparable to someone already treated for heart disease, and recommends lowering LDL by at least half, to below 70 mg/dL.

There is a catch (as if anything was ever without a catch): a score of 0 has an expiry date (depending on sex, ancestry, and risk factors) [6], and in younger adults it can miss early plaque that hasn’t calcified yet. You’ll find all that in my earlier post. In our example, last year’s “zero” doesn’t cancel out this year’s ApoB of 118.

Ask: “My calcium score is [X]. What LDL and ApoB goal applies to me, and when should we rescan?”

The one rule to take away

When two of your numbers disagree, always ask which number is closer to what actually causes the disease.

Particles beat cholesterol content. A picture of your arteries beats an estimate.

That’s why ApoB outranks LDL when they clash, why a calcium score outranks a risk equation, and why an appropriately calculated LDL outranks an outdated one. Doing this due diligence of your lab reports will arm you with better questions for your doctor and, ultimately, promote better-informed shared decision-making.

Subscribe now

Want to run these checks on your own numbers?

The Tiebreaker is a do-it-yourself course and self-assessment toolkit for heart-related blood tests:

• The toolkit (Excel): enter the results once, and it runs every cross-check.

• The course: a 9-chapter guide to every number on the report.

• The source library: an interactive notebook that answers in plain English.

The toolkit:

• recalculates LDL with Martin-Hopkins

• tests LDL, non-HDL and ApoB against the 12,700-person study above

• calculates 10-year risk three ways (PREVENT, MESA with and without a calcium score, SCORE2) and explains why they differ

• runs “What If” scenarios (e.g., “what if I reduce my LDL by [X]?”)

• separates real change from normal fluctuation over time

• turns the results into questions for the doctor

Every equation comes from its original publication and was checked against the official calculators.

The Tiebreaker; When your lab results and heart-risk scores disagree →

Laboratory members: Founding members get it free; annual members get 50% off.

Your codes are behind the paywall.

If your results conflict in ways no workbook can settle, or you face a treatment decision, start with the free fit check on lutzkraushaar.com.

Cited References

[1] T. S. Anderson, L. M. Wilson, and J. B. Sussman, “Atherosclerotic Cardiovascular Disease Risk Estimates Using the Predicting Risk of Cardiovascular Disease Events Equations,” JAMA Intern. Med., vol. 184, no. 8, pp. 963–970, Aug. 2024, doi: 10.1001/jamainternmed.2024.1302.

[2] R. S. Blumenthal et al., “2026 ACC/AHA/AACVPR/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA Guideline on the Management of Dyslipidemia,” JACC, vol. 87, no. 19, pp. 2624–2757, May 2026, doi: 10.1016/j.jacc.2025.11.016.

[3] S. S. Martin et al., “Comparison of a novel method vs the Friedewald equation for estimating low-density lipoprotein cholesterol levels from the standard lipid profile.,” JAMA, vol. 310, no. 19, pp. 2061–2068, Nov. 2013, doi: 10.1001/jama.2013.280532.

[4] A. Sayed, E. D. Peterson, S. S. Virani, A. D. Sniderman, and A. M. Navar, “Individual Variation in the Distribution of Apolipoprotein B Levels Across the Spectrum of LDL-C or Non-HDL-C Levels,” JAMA Cardiol., vol. 9, no. 8, pp. 741–747, 2024, doi: 10.1001/jamacardio.2024.1310.

[5] S. S. Khan et al., “Development and Validation of the American Heart Association’s PREVENT Equations,” Circulation, vol. 149, no. 6, pp. 430–449, 2024, doi: 10.1161/CIRCULATIONAHA.123.067626.

[6] O. Dzaye et al., “Warranty Period of a Calcium Score of Zero: Comprehensive Analysis From MESA,” JACC Cardiovasc. Imaging, vol. 14, no. 5, pp. 990–1002, 2021, doi: 10.1016/j.jcmg.2020.06.048.

Disclaimer: Similarities to the feeding practices in your household are neither intended nor coincidental…but inevitable.

Just kidding.

The Epidemiologist Had a Theory. His Wife Had the Receipts (to Disprove).

Enter: my wife.

Age 70, going on 40 (in energy, fitness, and mental acuity), she had developed a ‘circumferential’ problem.

Until a year ago, she had weighed in at 44 kg (97 pounds), which, at her height of 150 cm (4 feet 11), translated into an enviable BMI of 19.5. Since then she had added 4 pounds (The Horror!) At the resulting BMI of 20-point-something, the adjectives “fat” and “obese” had wormed themselves into her descriptions of her waist.

And since having a husband must be good for something, she tasked hers, the epidemiologist, to sleuth for the cause and the solution! “What are you waiting for?! Now!!”

I explained to her that the human metabolism inevitably slows down over the aging process, as has been shown in several papers, such as…. Well, she wasn’t out for an excuse. She wanted a solution.

Given her utterly useless husband, she took the sleuthing into her own hands, asking the simplest question the epidemiologist hadn’t thought of: did we, or she, change our food intake over the past year? Thanks to my wife’s detailed grocery records, we know, for example, that in 2025 we consumed

• 152 kg of bananas,

• 58 kg of blueberries,

• 471 avocados,

• 82 kg chicken,

• 46 kg of pork

• … you get the picture.

Altogether 717 kg of fruit, 143 kg of meat and… 7.7 kg of roasted, salted peanuts.

There was the key: While nothing much changed for all the other foods, the peanut record had already climbed to 14 kg! And 2026 was only 10 months old.

She had eaten most of the extra peanuts because, unbeknownst to me, she had acquired a daily habit of snacking on a handful of them. After all, peanuts are healthy, aren’t they? Especially to older people. There is even a study to prove that, she said [1]. She remembered having read a press release about that paper in late 2025.

Since we all know that press releases by study funders with a vested interest in exploitable study outcomes (in this case, the Peanut Institute Foundation, Albany, Georgia) are reliably unbiased, what can go wrong with adding a handful of peanuts to one’s diet every day?

Since the brain-benefit message is worth repeating, the researchers (and the Peanut Institute) justifiably flogged their study for a follow-up paper exactly one year later [2]. That’s when it pinged on my daily HypeWatch radar, raising the suspicion that my wife might have been sacrificing her waistline for the potentially empty promise of a better-functioning brain.

So I decided to check the evidence.

What Did the Peanuts Actually Do to the Brain?

The study, published on September 30, 2026, in Food & Function, was a randomized crossover trial that enrolled 36 adults aged 60 to 75, of whom 31 completed the experiment.

Participants consumed 60 grams of unsalted, skin-roasted peanuts daily during a 16-week period, and avoided peanuts altogether for another 16-week period. An eight-week washout separated the two periods, with their sequence randomly assigned.

Such a crossover design is a prudent choice for a small nutrition experiment. It ensures that each participant serves as their own control, which reduces the variability of outcomes that otherwise affects the comparison between different people. In other words, I have no beef with the study design.

More interesting is the question the researchers wanted to answer: Does daily consumption of unsalted, skin-roasted peanuts affect brain insulin responsiveness in healthy older adults?

You’ll need a bit of background knowledge to appreciate the question and its results. Insulin’s job in the body is not limited to regulating blood sugar. It also functions as a signaling molecule in the brain where it modulates neuronal activity. The hippocampus is (a) essential to memory consolidation and (b) the part of the brain that is most densely populated with insulin receptors. Long story short, when hippocampal insulin receptors respond to the molecule, the net effect is a calming down of overactive nerve cells and enhancement of neural efficiency.

That’s of particular interest to older adults because, with aging, but independently also with metabolic or neurodegenerative disorders, the hippocampus and other brain regions can become less sensitive to the insulin signal. Since the brain’s perfusion with blood is tightly linked to neuronal activity, you can imagine that in an insulin-resistant state, a brain shows higher perfusion than the same brain in an insulin-sensitive state.

In other words: A temporary drop in blood flow after an insulin dose suggests that the observed brain region responds to the signal. If that particular brain region were “insulin resistant,” the cells would ignore the insulin, and blood flow wouldn’t change.

That’s why blood perfusion in response to an (intranasal) insulin challenge can serve as a surrogate measure for a brain’s insulin sensitivity (which we can’t measure directly), and hence as an indicator of peanuts’ effects on insulin sensitivity.

That’s exactly what the researchers found: Compared to the no-peanuts period, after the 60-gram-daily period the participants’ brains responded with greater sensitivity to the same intranasal insulin dose, i.e., with measurably reduced blood flow in some brain regions.

BUT…

…there are at least three caveats to a prematurely enthusiastic interpretation of the results.

First, we don’t exactly know which brain regions we are talking about.

Brain imaging doesn’t deliver a map of the brain with cartographically precise boundaries between its individual regions, like between California and Oregon, or France and Germany. All we have is the Harvard–Oxford brain atlas, which assigns probabilities that particular imaging locations belong to specific brain structures. For our purpose, the highest probabilities (20% and 14%) were assigned to the lingual gyrus and hippocampus, respectively; the rest to memory-relevant structures.

Second, depending on the brain region and physiological circumstances, (a) insulin can increase or decrease cerebral blood flow, and (b) the relationship between these blood-flow responses and cognitive performance comes with lots of question marks [3]. In any case, neither reduced nor increased blood flow is automatically better.

Third, the statistically significant decrease in blood flow (though at borderline p=0.044) showed no correlation with a measurable improvement in cognitive function. The researchers’ 2025 paper reported a statistically significant 5.8% improvement in verbal memory during one of the memory tests, the Verbal Recognition Memory (VRM). But this 5.8% translates into just 1.4 more words correctly identified on recall.

The peanuts didn’t improve free recall, immediate recall, or any aspect of executive function and psychomotor speed tasks. You should keep that in mind before gorging yourself on peanuts in the hope of remembering where you left the car keys, whether you locked the door, or why you left the house in the first place.

So here is what we have: The researchers found a small patch on a brain scan where blood flow responded differently to insulin depending on short-term peanut exposure. Based on a probabilistic brain atlas, that patch overlapped some anatomical brain areas associated with memory.

But to translate changed blood flow into an upgrade of the brain’s memory circuitry is a stretch. However, that’s exactly what happened in the media.

From Brain Scans to Brain Boosters

On October 5, The Peanut Institute, which funded the research, issued a press release highlighting the potentially beneficial effects of peanuts on brain insulin responsiveness. It also introduced a rather more consequential possibility: insulin’s role in brain function “may help protect against Alzheimer’s disease.”

To transition a localized cerebral blood-flow response to a protection against losing one’s marbles is, shall we say, hype?

Not to be outdone, some media outlets ran with that.

On October 12, Griffon News declared that eating peanuts daily “could improve memory and protect against Alzheimer’s disease.”

To be fair, other outlets showed more restraint.

Yahoo Health, on October 9, explained that the imaging result did not demonstrate protection against Alzheimer’s disease or dementia. So did South Korea’s Health Chosun, also on October 9.

That’s actually good news about the news: There are still some journalists out there, capable of separating an intriguing physiological observation from the clinical benefits readers might mistakenly infer from it.

Now here was the bigger question for me: How to tell my wife?

The 340-Calorie Catch

The study participants consumed 60 grams of peanuts every day. The image below illustrates their calorie and macronutrient content. 340 calories, of which 78% came from fat, don’t exactly qualify as negligible.

Now, peanut fat isn’t inherently bad. It’s mostly unsaturated fats, along with protein and fiber. But even healthy calories are calories. The laws of thermodynamics don’t give a damn about “healthy” calories, and deposit them wherever you dislike them most: on your waist and hips.

The researchers addressed this issue, too. According to their dietary assessments, the participants consumed approximately 215 additional calories per day during the peanut period. That level of accuracy is, quite frankly, BS. Half of the participants in a recent validation trial had to be excluded for physiologically implausible underreporting of calorie intake [4].

That observation actually makes the peanut study’s results even more puzzling. 215 excess calories daily translate into 2.6 kg of extra fat over 16 weeks, which should have shown on the scale, the waist, or on body composition. But according to the researchers, there were no changes.

215 excess calories don’t just disappear into thin air. So, if body weight and fatness didn’t change, dietary intake must have compensated for the additional calories. And that begs the question: was the observed brain signal specific to peanuts or reflected a broader dietary change that accompanied their consumption? After all, participants were explicitly instructed to follow the Dutch food-based dietary guidelines.

Anyway, nothing of that sort interests my wife. Her waistline is far closer to her than the nitty-gritty questions of an anally retentive epidemiologist husband.

Putting an abrupt stop to her peanut consumption, she dropped her weight by a kilo within 4 weeks. That doesn’t mean it’s all due to the peanuts. I also restructured her daily 60-minute cardio exercise routine to optimize fat burn. But that 1-kilo drop in weight put me in her good books again. Which is all that counts.

Is that proof that peanuts caused her weight gain? Nope. This wasn’t a controlled experiment. I also didn’t notice any drop in cognitive or memory performance. She remembers perfectly well how many glasses of Spanish Red I had the previous day (particularly when I can’t remember it).

Like your health science without the hype?

As an Epidemiologist, DrPH, MSc. I help people over 60, and women through menopause, see through health hype and understand what their lab reports and medical studies mean for them. That’s what Double Check Doc is about.

Subscribe now

Cited References

[1] L. Kerkhof, R. P. Mensink, J. Plat, K. M. R. Nijssen, and P. J. Joris, “Longer-term skin-roasted peanut consumption improves brain vascular function and memory: A randomized, single-blind, controlled crossover trial in healthy older adults,” Clin. Nutr., vol. 55, pp. 170–179, Dec. 2025, doi: 10.1016/j.clnu.2025.10.020.

[2] L. Kerkhof, R. P. Mensink, J. Plat, K. M. R. Nijssen, and P. J. Joris, “Daily consumption of unsalted, skin-roasted peanuts beneficially affects regional brain insulin responsiveness in older adults: results of a longer-term randomized controlled trial,” Food Funct., vol. 17, no. 20, pp. 9354–9364, Oct. 2026, doi: 10.1039/d6fo03208g.

[3] K. M. R. Nijssen, R. P. Mensink, and P. J. Joris, “Effects of Intranasal Insulin Administration on Cerebral Blood Flow and Cognitive Performance in Adults: A Systematic Review of Randomized, Placebo-Controlled Intervention Studies,” Neuroendocrinology, vol. 113, no. 1, pp. 1–13, Jan. 2023, doi: 10.1159/000526717.

[4] P. Putz, B. Kogler, and I. Bersenkowitsch, “Reliability and validity of assessing energy and nutrient intake with the Vienna food record: a cross-over randomised study,” Nutr. J., vol. 18, no. 1, p. 7, Jan. 2019, doi: 10.1186/s12937-019-0431-9.

One reader had his coronary artery calcium (CAC) scan done just before my previous post came out. His score came back at around 700. Before his doctor’s appointment, he asked me for an independent review of his results. That is the work I do for clients, and this post walks you through what we worked out together. What he did next might serve you as an inspiration for your own case.

What is Vital?

When a CAC score comes back not only high but with most of the calcium sitting in the left anterior descending artery (LAD), aka the widow-maker, most patients’ knee-jerk reaction is one or more of the following: that artery must be blocked, I need some more scans, and maybe I should go for having it stented. The first and the third are mostly wrong; the second depends on a lot more than just the score.

Why it Matters to You, if You are 60+

Your calcium score answers the question of how much calcified plaque your heart carries, but not how much that plaque narrows it. Your cardiologist is in the best position to determine the optimal follow-up. But to have a better-informed shared decision-making, you ought to understand which procedures and which decisions really help prevent what you fear most.

Your Takeaway

A seven-step walk-through of what a high score does and doesn’t tell you, and the questions to take to your doctor.

Table of Contents

• A Reader’s Email

• Step 1: Separate the Two Questions

• Step 2: Know Which Question Drives Heart-Attack Risk

• Step 3: Why a “Preventive” Stent Isn’t Prevention

• Step 4: What Your Score Can and Can’t Tell You

• Step 5: Turn Your Number Into People

• Step 6: Look Past Your LDL

• Step 7: The Questions to Take to Your Doctor

• The Lucky One

• Got a Number You Don’t Know What to Make Of?

• Cited References

A Reader’s Email

He, let’s call him Peter, is around 70, active, doesn’t smoke, has normal blood pressure, and has type 2 diabetes that is well controlled on medication. His cholesterol has always been high, though still below the threshold that would trigger a familial hypercholesterolemia (FH) suspicion. Statins (he tried several ones) caused intolerable side effects, which is why he is done with that drug.

Peter read my statin articles, brought the subject up with his doctor, who agreed it was time for a calcium scan. He also gave me permission to share his story (name changed and exact numbers rounded).

His scan came out with a total Agatston score of about 700, which puts him between the 75th and 90th percentile for men his age (i.e., 75-90% of men his age have lower scores). Calcium showed up in all four main coronary arteries, and about two-thirds of it sat in the LAD.

The LAD is called the widow-maker for a reason: it’s the largest of the coronary arteries, supplying roughly half of the left ventricle’s muscle (the one that pushes oxygenated blood into the arterial circulation). Historically, blockages of the LAD had middle-aged men keel over with no warning, so the name widow-maker stuck.

Following Peter’s scan, his first three thoughts arrived in my email:

1. “I’m concerned about the LAD portion.” Translation: that artery must be badly blocked.

2. “I’m guessing I should get a CT angiogram next to evaluate openings and soft plaque.”

3. “If I end up needing a preemptive stent at some point, that would not be the worst thing.” Two of his friends survived widow-maker heart attacks and are doing great years later.

The second thought is defensible. The first and the third deserve a closer look, because they rest on a popular misunderstanding that a calcium score tells you how blocked your arteries are.

It doesn’t.

Step 1: Separate the Two Questions

When your CAC score comes back >0, it answers one question and leaves another one open.

The answered question: How much plaque is there overall?

Think of it as: how much limescale sits in the heart’s plumbing? The Agatston score adds up every calcified spot above a density threshold and weights it by its density, as I explained in my post on deciphering your CAC score.

The open question: Is an artery significantly narrowed? Think of it as: how much does the limescale in the pipe narrow its lumen? A CAC scan can’t see lumen. Two different tests can deliver that answer. One is a stress test (does any narrowing limit blood flow when the heart works hard, e.g., while stressed with exercise?), the other is a CT angiogram (how narrow is the lumen, and what does the plaque look like?). Alternatively, a catheter can also deliver the answer, but it’s an invasive procedure.

So, Peter now knows that he has quite the calcium sitting in his LAD, but not whether that calcified plaque narrows his artery by, say, 20% or by 70%. That matters because cardiologists stratify blockages into different severity classes, as you can see from the table below [1].

To complicate matters a little further, even the degree of blockage does not equate with what you really want to know: what is my heart attack risk?

Step 2: Know Which Question Drives Heart-Attack Risk

A plumber would say: the narrower the pipe (because of a limescale deposit), the greater the chance that it clogs. Three studies illustrate that this is where the plumbing analogy falls apart.

In 2020, Danish researchers followed 23,759 people who had a CT angiogram for chest symptoms [2]. Within the 4-year follow-up, the degree of blockage barely mattered, as you can see in panel B of the illustration below. In Peter’s band (a score of 400 to 1,000), the event rate was 30.2 per 1,000 person-years without, and 28.6 with significant narrowing.

Grandhi et al. evaluated 5,192 acute emergency department chest pain patients to demonstrate that a calcium score of zero reliably excludes obstructive disease and the need for revascularization in over 99% of acute cases [3]. And 4 of 5 patients with CAC > 0 have < 50% stenosis. CAC = 0 is not Peter, but the < 50% may still apply to him.

A third study explored the same question from a different angle [4]. The ICONIC researchers took people who suffered an acute coronary event and had a CT angiogram taken around 4 years earlier. In three out of four cases, the culprit plaques that triggered the event had narrowed the affected artery by less than half.

So, greater narrowing does not necessarily translate into greater-risk plaque. Another study that used deep-learning methods to evaluate CT angiography found that narrowing still had a small but significant predictive value [5]. That keeps the degree of narrowing relevant; it’s just not the main driver of what you fear most: a heart attack.

So, let’s condense these 4 studies into three takeaway points:

1. Plaque Burden Drives Risk: It still drives the risk of future heart attacks, major cardiovascular events, and overall mortality

2. Stenosis Is an Imprecise Predictor of Events: Within identical plaque burden categories, the presence of a tight stenosis adds little predictive value.

3. Only A CAC score of zero is a reassuring reading: It excludes obstructive disease and the need for revascularization in over 99% of acute symptom cases, and it is far from being a rare finding, even at Peter’s age, as I explained in an earlier post. High overall plaque burden should trigger you to have a hard look at your options for preventive medical therapy.

Which brings us to Peter’s third thought.

Step 3: Why a “Preventive” Stent Isn’t Prevention

Sounds like common sense, but trial evidence doesn’t support it.

The COURAGE trial randomised 2,287 people with stable coronary disease to stenting plus medication, or medication alone [6]. After a median of 4.6 years, the stent group had no fewer deaths or heart attacks. That was 20 years ago, and the trial’s inclusion criteria don’t necessarily match Peter’s or your profile (the trial only enrolled people with known coronary stenosis ≥70% stenosis in at least one artery, and 1 in 3 participants had a history of heart attacks).

The 2020 ISCHEMIA trial looked at 5,179 people with stable disease and moderate to severe reduced blood flow on a stress test [7]. Half of the participants were randomized to routine cardiac catheterization followed by revascularization (stent/PCI or bypass surgery/CABG) when deemed feasible, plus medical therapy.

The other half received Conservative Strategy consisting of medical therapy alone (catheterization/revascularization was reserved for those in whom medical therapy failed).

Over five years, the main outcome (cardiovascular death, heart attack, and a few related events) occurred in 16.4% versus 18.2%, with the difference of 1.8% not being statistically significant. Death rates, for example, didn’t differ at all.

In stable patients, stents may relieve chest pain, but their track record for preventing heart attacks and deaths is virtually nil.

That shouldn’t come as a surprise after what you know now from the preceding section: If we can’t divine which plaque causes an event, why would we expect that gifting any specific plaque with a stent will prevent an event? After all, most heart attacks arise from a mildly obstructing plaque that nobody would stent.

One caveat I need to mention. ISCHEMIA excluded people with a significant narrowing of the left main artery, the short trunk that feeds both the LAD and the circumflex. For that artery, other evidence and guidelines apply. But Peter had only a negligible amount of calcium in his left main.

Peter’s observation that two friends survived widow-maker heart attacks deserves another caveat that you should apply to yourself when (no, actually “before”) making similar comments and drawing conclusions therefrom. The stories we hear, whether it’s the two-friends story or the my-mother-smoked-a-pack-a-day-and-survived-to-age-95, come mostly from those who survived. In statistics, we call this survivorship bias. And it makes any danger look smaller than it is.

Step 4: What Your Score Can and Can’t Tell You

In my earlier CAC post, I described the Jekyll-and-Hyde quality of coronary calcium. Higher calcium volume equals higher risk; higher calcium density equals less risk. That’s because dense calcium reflects the body’s track record of stabilising a plaque [8]. And the Agatston score, perversely, rewards density with a higher number.

But…

1. The density bonus fades at high scores. A 2023 analysis of the MESA cohort, with 17 years of follow-up, found the protective signal of dense calcium only at low calcium volumes [9]. Above a volume of about 130 mm³, it was no longer statistically significant. A CAC score of 700 sits far above that range. So in Peter’s case, “maybe it’s the stable kind” is not a reassurance the data support.

2. On the other hand, a rising CAC score can be good news. Intensive cholesterol lowering by statins has shown to shrink plaques and make them more calcified [10]. So, if you start treatment and your next scan shows a higher score, that’s not automatically a reason to despair; maybe on the contrary.

3. Spread matters. At the same total score, more affected vessels mean more risk [11], which, again, shouldn’t come as a surprise. If each plaque represents a ticket in the heart attack lottery, more tickets mean better chances at ‘winning’ the lottery. Oh, and calcium in the left main is kind of a specially (un-)lucky ticket [12] .

You know already that the soft, non-calcified plaque is invisible to the CAC scan. From the SCOT-HEART trial, we know that the amount of low-density soft plaque burden on a CT angiogram predicted heart attacks independently of both the calcium score and vessel narrowing [13].

“Low-attenuation plaque burden was the strongest predictor of myocardial infarction…, irrespective of cardiovascular risk score, coronary artery calcium score, or coronary artery area stenosis.”

That is why Peter’s second thought, a CT angiogram, is a reasonable question to put to his cardiologist.

Step 5: Turn Your Number Into People “like you”

When you walk out of your doctor’s office with a calcium score of 700, or whatever yours might be, you probably can’t grasp what that means in terms of risk. I like to translate such numbers into a how-many-of-a-hundred-people-like-you-will-suffer-an-event image. Creating such an image is embarrassingly simple. The only thing you need is one of the official, association-sanctioned, web-hosted risk algorithms (PREVENT for the US and SCORE2 for Europe).

For Peter, I could have used the PREVENT algorithm. The only problem with PREVENT is that it explicitly rules out people with CAC scores >300. That’s where the MESA risk score comes in. It was developed from the Multi-Ethnic Study of Atherosclerosis, and it combines your traditional risk factors with your calcium score and estimates your chance of a coronary event (mainly heart attacks) over the next ten years [14]. It is valid from age 45 to 84, and here is its online version.

Read the strip from left to right to see how the CAC score affects his risk, and how lowering his cholesterol reduces it:

• Risk factors only: about 14 in 100 men like him would have a coronary event within ten years. About 86 would not.

• With his calcium score: about 23 in 100. The scan added roughly 9 people to the count. That is what a high score does: it moves you from a statistical average to your own arteries.

• With LDL cholesterol 25% lower: about 20 in 100. Three of the 23 are spared.

• With LDL cholesterol 50% lower: about 17 in 100. Six or seven are spared.

The higher your starting risk, the more people are spared by a given change in cholesterol. That is why the calcium score matters so much for this conversation.

A word of caution before you take these numbers to the bank. The last two panels compare men who are identical except for their cholesterol. They do not tell you what a particular drug or diet would achieve for you. That question needs trial data, which is a different (and longer) story.

MESA offers one more way of reading the result, which I like because it connects to a theme I have addressed in several earlier posts, for example here and here.

Based on his risk factors alone, the MESA calculator estimates Peter’s “coronary age” at 78. That is, a 9-year-older man would typically carry his estimated risk. Add his calcium score, and that hypothetical age jumps to 90.

That sounds dramatic, but remember that this is what happens when you translate a probability into an age-equivalent. It’s not a measure of how old his heart is physically. And unlike calendar age, you can modify it.

With LDL 50% lower, the same calculator puts him at 82. And if he had a zero CAC score, his coronary age would have dropped to 51 (not shown). Which is why a zero score calls for a little celebration, including a cholesterol-rich steak and a couple of beers to wash it down.

Just kidding. Or maybe not.

Any other tests that would make sense for Peter? You bet.

Step 6: Look Past Your LDL

In earlier posts about statins, I explained why the amount of LDL cholesterol is an imprecise measure of plaque risk. Far more revealing is the number of particles that ferry the cholesterol load through your arteries. Apolipoprotein B (ApoB) is a molecule that associates with the LDL carriers: exactly one ApoB molecule per LDL vehicle. So, getting your doctor to order an ApoB test (inexpensive) can put your LDL number into perspective. That ‘perspective’ can cut both ways because people’s ApoB/LDL ratio shows a wide spread [15].

At Peter’s LDL-C of about 175 mg/dL, the middle half of people have an ApoB between roughly 123 and 139 mg/dL. His was about 145. That puts him into “positive discordance” territory, meaning that he carries more plaque-building particles than his LDL-C suggests.

Men, older adults, and people with diabetes (Peter’s profile) tend to sit above the median line. It matters, because when the two measures disagree, positively or negatively, risk tends to follow the particle count. If your lab report shows only LDL-C (many standard U.S. panels don’t include ApoB), it is worth asking for an ApoB measurement at least once.

What target should he aim for? For people with a calcium score of 300 or higher, the 2026 ACC/AHA dyslipidemia guideline sets the tightest LDL-C goal: below 55 mg/dL [16]. Peter’s LDL-C is more than three times that. With statins off the table for him, how to close that gap is what he needs to explore with his cardiologist. I discussed the 7-step Lipid Lowering Ladder of therapies in an earlier post.

Step 7: The Questions to Take to Your Doctor

Here are the 6 questions I gave Peer to take to his upcoming PCP appointment. Adapt the numbers to yours.

• 1: “My calcium score is [X], and my LDL is [Y]. What LDL and ApoB targets apply to me, and if statins aren’t an option, which other medicines could get me there?”

The Lucky One

In his email, Peter wrote: “I am focusing on how lucky I am that my doctor and I decided on a calcium scan, and that I have it done now. Just wish I’d done it sooner.”

He is right; the scan answered the question of whether a more aggressive prevention is worth the trouble. With a score of 700, the answer is “probably yes”, and the first consequence is that his PCP referred him to a cardiologist for stress testing. That test is only a few weeks off, and I’m eager to see the results.

The not-so-lucky course of events would have been him walking out with a scan that only produced fear with a blocked-artery story and possibly a stent that, in all likelihood, wouldn’t fix the problem. Turning that CAC score, or any other lab number, into the right questions for his doctor is what made the difference.

Got a Number You Don’t Know What to Make Of?

Most people get a lab report or a calcium score with a one-line comment and a follow-up appointment weeks away. That is where Peter stood. The detailed review I worked out for him before his PCP appointment is what this post is built on.

If you would like the same for your own results, have a look at the three types of reports on my website, lutzkraushaar.com. Each comes with a full sample report:

The first step is a free fit check, so you know before you commit which report is an appropriate match for your question.

BTW, Founding members of my Double Check Doc Substack receive one Cardiovascular Lab Audit per year as part of their membership, plus 20% off one further review. If you’re already a founding member, simply mention it in your fit check.

For Paid Subscribers: Ask the Sources About Your Own Numbers

I’ve put the full text of all 16 papers cited in this post into an interactive source library. You can ask it questions in plain English, including about your own numbers. For example:

• “My calcium score is 450, mostly in the RCA. What did the studies find for people in that range?”

• “My LDL is 140 and my ApoB is 120. Is that discordant, according to Sayed?”

• “I have no symptoms. Did ISCHEMIA include people like me?”

Every answer comes from the papers themselves, with the source cited. It tells you what the evidence says about numbers like yours. It won’t tell you what to do; that conversation belongs with your doctor, or in a personal review.

Privacy tip: share numbers, not names. Your questions are processed by Google’s Gemini.

As always: this post is an independent evidence review for educational purposes. It is not medical advice and does not replace your doctor’s judgment.

Find the link to the Interactive Source Library behind the paywall:

Subscribe now

Cited References

[1] R. C. Cury et al., “CAD-RADS^TM 2.0 – 2022 Coronary Artery Disease-Reporting and Data System,” JACC Cardiovasc. Imaging, vol. 15, no. 11, pp. 1974–2001, Nov. 2022, doi: 10.1016/j.jcmg.2022.07.002.

[2] M. B. Mortensen et al., “Impact of Plaque Burden Versus Stenosis on Ischemic Events in Patients With Coronary Atherosclerosis,” JACC, vol. 76, no. 24, pp. 2803–2813, Dec. 2020, doi: 10.1016/j.jacc.2020.10.021.

[3] G. R. Grandhi et al., “Coronary Calcium to Rule Out Obstructive Coronary Artery Disease in Patients With Acute Chest Pain,” JACC Cardiovasc. Imaging, vol. 15, no. 2, pp. 271–280, Feb. 2022, doi: 10.1016/j.jcmg.2021.06.027.

[4] H.-J. Chang et al., “Coronary Atherosclerotic Precursors of Acute Coronary Syndromes,” JACC, vol. 71, no. 22, pp. 2511–2522, Jun. 2018, doi: 10.1016/j.jacc.2018.02.079.

[5] A. Lin et al., “Deep learning-enabled coronary CT angiography for plaque and stenosis quantification and cardiac risk prediction: an international multicentre study,” Lancet Digit. Health, vol. 4, no. 4, pp. e256–e265, Apr. 2022, doi: 10.1016/S2589-7500(22)00022-X.

[6] W. E. Boden et al., “Optimal Medical Therapy with or without PCI for Stable Coronary Disease,” N. Engl. J. Med., vol. 356, no. 15, pp. 1503–1516, Apr. 2007, doi: 10.1056/NEJMoa070829.

[7] D. J. Maron et al., “Initial Invasive or Conservative Strategy for Stable Coronary Disease,” N. Engl. J. Med., vol. 382, no. 15, pp. 1395–1407, Apr. 2020, doi: 10.1056/NEJMoa1915922.

[8] M. H. Criqui et al., “Coronary Artery Calcium Volume and Density: Potential Interactions and Overall Predictive Value: The Multi-Ethnic Study of Atherosclerosis,” JACC Cardiovasc. Imaging, vol. 10, no. 8, pp. 845–854, 2017, doi: 10.1016/j.jcmg.2017.04.018.

[9] H. S. Bhatia, R. L. McClelland, J. Denenberg, M. J. Budoff, M. A. Allison, and M. H. Criqui, “Coronary Artery Calcium Density and Cardiovascular Events by Volume Level: The MESA,” Circ. Cardiovasc. Imaging, vol. 16, no. 2, p. e014788, Feb. 2023, doi: 10.1161/CIRCIMAGING.122.014788.

[10] R. Puri et al., “Impact of Statins on Serial Coronary Calcification During Atheroma Progression and Regression,” JACC, vol. 65, no. 13, pp. 1273–1282, Apr. 2015, doi: 10.1016/j.jacc.2015.01.036.

[11] M. J. Blaha et al., “Improving the CAC Score by Addition of Regional Measures of Calcium Distribution: Multi-Ethnic Study of Atherosclerosis,” JACC Cardiovasc. Imaging, vol. 9, no. 12, pp. 1407–1416, Dec. 2016, doi: 10.1016/j.jcmg.2016.03.001.

[12] S. J. Lahti et al., “The association between left main coronary artery calcium and cardiovascular-specific and total mortality: The Coronary Artery Calcium Consortium,” Atherosclerosis, vol. 286, pp. 172–178, Jul. 2019, doi: 10.1016/j.atherosclerosis.2019.03.015.

[13] M. C. Williams et al., “Low-Attenuation Noncalcified Plaque on Coronary Computed Tomography Angiography Predicts Myocardial Infarction,” Circulation, vol. 141, no. 18, pp. 1452–1462, May 2020, doi: 10.1161/CIRCULATIONAHA.119.044720.

[14] R. L. McClelland et al., “10-Year Coronary Heart Disease Risk Prediction Using Coronary Artery Calcium and Traditional Risk Factors Derivation in the MESA (Multi-Ethnic Study of Atherosclerosis) with Validation in the HNR (Heinz Nixdorf Recall) Study and the DHS (Dallas Heart Stud,” J. Am. Coll. Cardiol., vol. 66, no. 15, pp. 1643–1653, 2015, doi: 10.1016/j.jacc.2015.08.035.

[15] A. Sayed, E. D. Peterson, S. S. Virani, A. D. Sniderman, and A. M. Navar, “Individual Variation in the Distribution of Apolipoprotein B Levels Across the Spectrum of LDL-C or Non-HDL-C Levels,” JAMA Cardiol., vol. 9, no. 8, pp. 741–747, 2024, doi: 10.1001/jamacardio.2024.1310.

[16] C. P. Guidelines, “2026 ACC / AHA / AACVPR / ABC / ACPM / ADA / AGS / APhA / ASPC / NLA / PCNA Guideline on the Management of Dyslipidemia: A Report of the American College of Cardiology / American Heart Association Joint Committee on Clinical Practice Guidelines,” Circulation, 2026, doi: 10.1161/CIR.0000000000001423.

[17] F. Kronenberg et al., “Lipoprotein(a) in atherosclerotic cardiovascular disease and aortic stenosis: A European Atherosclerosis Society consensus statement,” Eur. Heart J., vol. 43, no. 39, pp. 3925–3946, 2022, doi: 10.1093/eurheartj/ehac361.

This month’s research finding that almost everyone above 60 has atherosclerotic disease is good news and bad news, just not in the way the authors intended. If everyone has a “disease”, the “disease” isn’t a disease anymore; it’s a part of life.

Why it Matters to You, if You are 60+ (or Intend to Get There)

The paper omits to mention that 9 out of 10 of its “diseased” people won’t have a heart attack or stroke over the next 10 years. Since medical practitioners have a documented track record of “inumeracy” when it comes to the statistics of risk, and the screening recommendations they derive therefrom, you should be prepared for when your doctor brings up that study.

Your Takeaway

An epidemiologist’s (surprisingly?) simple walk-through of medicine’s most common misconception that screening can do no harm. And a guide to the two questions you should ask your doctor if he or she brings up this NEJM paper that is creating waves in the media right now.

Table of Contents

· The 98% Club
· REACT: Why Calculate When You Can Take Pictures?!
· What React Actually Found (and What It Didn’t)
· From “Detectable” to “Cure”: The Claim Ladder
· DANCAVAS: Rung 4 Has Already Been Tested. In Denmark.
· The Party-Poopers’ Rain on REACT’s Parade
· A Screening Quiz Most Doctors Fail
· Two Illusions That Make Every Screening Test Look Good
· The Trial REACT Should Run, and the One It Plans
· My HypeWatch Verdict and the Two Questions for Your Doctor
· Cited References

Subscribe now

The 98% Club

(Not so) great news for everyone either hitting their 60th birthday soon or having that event already in the rear-view mirror. If you are a man and get scanned, you have a 98% chance (it’s 92% if you are a woman) of being labeled a heart patient during your next doctor’s visit, even if you considered yourself heart healthy until yesterday. That is, if the data just published in a New England Journal of Medicine study reaches your doctor before you do [1].

That study (REACT) found atherosclerotic plaque in about everybody: from 1 in 3 men and 1 in 5 women in their thirties to 98.1% of men and 91.9% of women in their 60s.

Henning Bundgaard, the lead author, says that finding and treating plaque early “is essential.” Borja Ibáñez, a co-author, wants portable ultrasound to become “the standard tool for atherosclerosis screening from the early stages of adult life.” Valentin Fuster, another co-author, even suggests that REACT will answer whether early intervention can “cure the disease”.

Once your doctor falls for that disease label, guess what you’ll be in for: more diagnostics and more treatment (from age 20 onward, mind you). That’s great news for the medical industry, which, with its admirably altruistic ethics, funded the study. It’s not so great news for you, because, as you’ll see, 9 of every 10 “diseased” men wouldn’t need that treatment in the first place.

Why did I say that your doctor might “fall for a disease label”? Do I, an epidemiologist, really claim that medically trained experts won’t see through a misinterpretation of basic statistics?

Oh yeah, big time.

And unfortunately the IP rights to that claim aren’t mine. That credit goes to Gerd Gigerenzer, a statistician, psychologist, and director of the Harding Center for Risk Literacy at the University of Potsdam, Germany. His pet subject is what he has coined

He found that inability widespread not only among the media scribes and journalists (where it may be forgivable), but also among medical professionals (where it isn’t as forgivable).

I’m a fan of Gigerenzer. And also of Ingrid Mühlhauser, a physician and emeritus professor of Health Sciences and Nursing Sciences at the University of Hamburg, Germany. Her pet subject: dismantling statistical illiteracy and transforming patient info from promotional marketing materials into clear, unbiased, and honest numbers. Gigerenzer’s and Mühlhauser’s work is right up my alley.

We’ll use their insights to dissect the NEJM paper and the claims that have started to sprout from it.

So, if you feel like having a better-informed discussion with your doctor about your plaque “disease”, let’s wade through the noise of that study.

REACT: Why Calculate When You Can Take Pictures?!

The European and American risk calculators (SCORE2 and PREVENT) estimate a person’s 10-year risk of a cardiovascular disease event (e.g., heart attack, stroke) based on easily available markers such as age, blood pressure, cholesterol levels, smoking history, and diabetes.

Of these markers, “age is the strongest predictor of incident CVD” [2], such that if you are:

• below age 40, your absolute risk rarely exceeds 5% or 7.5%, even when five of your risk factors are seriously elevated.

• age 70 and above, your risk is always in the “high” category (>10%) no matter how squeaky-clean all your other risk factors are.

The REACT researchers wanted to see whether it makes sense to move from these risk algorithms’ probability estimates to actually visualizing the presence or absence of atherosclerotic plaque since the latter is the major prerequisite for those CVD events.

So, REACT enrolled 16,808 adults from Denmark and Spain aged 18 to 70 without prior history of heart disease, and then used vascular ultrasound and CT angiography to scan their arteries in three locations: neck (carotid), leg (femoral), and heart (coronary).

The two institutions that administered the study were the Spanish CNIC (Centro Nacional de Investigaciones Cardiovasculares) led by Dr. Borja Ibáñez and Dr. Valentín Fuster, and Denmark’s Rigshospitalet led by Prof. Henning Bundgaard. CNIC recruited the Spanish participants via open advertisements; in Denmark recruitment was based on random draws from the population register.

Novo Nordisk Foundation funded the study (174 million DKK / ~€23 million). Philips Healthcare, GE Healthcare, and Zeiss provided equipment and in-kind imaging support.

What React Actually Found (and What It Didn’t)

The results:

The image shows the presence of silent plaques for both genders combined across each age group.

The real prevalence numbers might even be higher than those in the image. The reason is a potential ‘healthy volunteer’ bias, or ‘self-selection’ bias. Typically, participants in such studies are more concerned about their health than the general population is on average. Hence, they may also have healthier lifestyle habits which, in turn, lead to lower atherosclerotic loads.

Even the authors concede that this bias may have pushed their prevalence figures down. But let’s ignore that; the numbers speak for themselves. They aren’t actually as surprising as you might think.

Early autopsy studies in young soldiers killed in Korea in the 1950s already alerted us to the presence of plaques in young adults [3]. And the Bogalusa study, which the authors themselves cite, found plaques post-mortem in children and adults aged <30 [4].

What we need to keep in mind here is that REACT is a snapshot. It:

• followed up nobody

• counted no events (heart attacks, strokes or deaths), and

• treated nobody.

That is important to keep in mind when looking at the claims that have started to sprout from REACT’s cross-sectional observation.

From “Detectable” to “Cure”: The Claim Ladder

If you claim that a screening study might help prevent or even “cure” the very disease that your screening is aimed at, you jump 4 rungs of a claim ladder, each demanding its own evidence:

1. Detect: the abnormality.

2. Predict: finding the abnormality tells you who will get sick.

3. Change treatment: knowing the abnormality leads to changing preventive care.

4. Improve outcomes: changing preventive care means fewer heart attacks, strokes, and deaths (without harms that cancel the gain).

REACT sits on rung 1. Here is where the messaging around it stands:

The above table is a prime example of how to prematurely enthuse an observational snapshot into a population-wide screening standard. The abstract is still sober. The discussion adds a careful “may”, which the press releases drop before hallucinating the first step toward a cure.

And you thought only AI hallucinates.

The authors themselves apparently get carried away. In the American College of Cardiology’s write-up, Bundgaard says early identification and treatment “could help reduce” the disease burden. In the CNIC release, distributed through EurekAlert and ScienceDaily, the same quote morphs into “is essential to reducing” the disease. The first was a hypothesis; the second accepts it as a proven necessity.

Contrary to the authors’ exuberance, Fox News’ coverage quoted the more sober Sirisha Vadali, a cardiologist not involved in REACT, as saying that (a) finding plaque “doesn’t mean someone is about to have a heart attack”, and (b) the study alone does not prove that screening everyone will prevent heart attacks or strokes.

Right.

Fortunately, we already have evidence for rung 4 of the claim ladder.

DANCAVAS: Rung 4 Has Already Been Tested. In Denmark.

In 2022, the NEJM published the results of a trial (again done in Denmark) aimed at rung 4 of our ladder and enrolling the age group where screening is expected to have its best shot.

The DANCAVAS trial randomized 46,526 Danish men aged 65 to 74 [5]. About a third were invited to a comprehensive cardiovascular screening (including a CT scan for coronary calcium, CAC); the rest received usual care. About 63% of those invited turned up.

With respect to mortality (the primary outcome), the difference between the groups was not significant. After 5.6 years, 12.6% of the screening group had died, compared with 13.1% of the usual-care group (p = 0.06).

That’s an absolute risk reduction of 0.5% (if we generously overlook the non-significance of the result).

The hazard ratio (HR) for stroke (secondary outcome) was barely significant at 0.93 (95% CI, 0.86 to 0.99) with an absolute risk reduction of 0.5% (7.5 men per 100 non-screened participants experienced a stroke versus 7 men per 100 screened participants).

The numbers for heart attacks were not significant at all: 2.6% in the screened group versus 2.8% in the usual-care group (HR 0.91; 95% CI, 0.81 to 1.03).

So, in the age group that contributes most CVD events to our population statistics, a screening program that included plaque imaging bought, at best, half a percentage point less death and stroke over 5.6 years, and no measurable drop in heart attacks. These results shouldn’t appear in the same sentence with “essential” or “cure”.

How much smaller would the gain be in 20- to 40-year-olds, whose event rates are a fraction of those of 70-year-olds? Your guess is as good as mine, because nobody has run that experiment yet. For a good reason: Because of the low incidence rates in this age group and the decades of follow-up required, such a trial is prohibitively expensive and therefore unlikely to happen soon.

Why is the benefit so small, even where it should be largest? Part of the answer is simple arithmetic.

The Party-Poopers’ Rain on REACT’s Parade

An image is worth a thousand words.

The 60 to 70-year-old men in the REACT cohort had a median 10-year risk of a fatal or non-fatal heart attack or stroke of 7% (calculated by SCORE2). That is, over the prediction period, approximately 7 of 100 men will suffer such an event. But 98 will test positive for plaque. You do the math.

Of course, you can only do the math if you read the paper AND its supplementary data appendix, specifically Table S4, which is the only source for the predicted clinical event rates.

I’m not the only party pooper in the theatre of screening programs. Here is Ingrid Mühlhauser quoting Sir Muir Gray [6]:

Once you are a screen-detected plaque-loaded “diseased” person, life isn’t going to be the same anymore, whether you ultimately suffer the prognosticated event or not.

The inevitable anxiety is just one part of the harm Mühlhauser and Gray are talking about. Misdiagnoses, unnecessary treatments, and their side effects are the other. Our simple calculation above shows that most of the 98% screen-detected “diseased” men are exposed to that harm.

So, before a screening program is justified, the following criteria ought to be met:

1. The disease progression is known: we can tell a plaque that will trigger an event from one that won’t.

2. Treating the condition before symptoms beats treating it when symptoms appear. Showing that a treatment works in sick people is not enough.

3. Randomized trials show that the screening programme reduces illness or death, not just that it changes a marker.

4. Randomized trials show that the benefits demonstrably outweigh all harms: false alarms, further testing, overtreatment, anxiety and cost.

REACT doesn’t deliver on any of these 4 points.

To be fair, we have some evidence for point #2 (though certainly not from REACT): Statin trials in asymptomatic people show that treating risk before symptoms can reduce events (I discussed the limitations of these trials in earlier posts). What no trial shows is that finding plaque improves on that.

When doctors treat sick patients, they do so based on evidence, typically available from randomized controlled trials (RCTs).

No treatment is supposed to enter, or remain in, the doctors’ toolboxes if side effects outweigh benefits.

But when doctors, RFK jr. or any other health policy promoting agent solicit healthy, asymptomatic people with screening or preventive promises, the burden of proof should be equally high, if not higher: their must exist solid RCT evidence to prove that what they promote causes more good than harm. Keep that in mind for your next doctor appointment. I have discussed this issue in the context of mammography screening in an earlier post.

So, if the rules are so relatively simple, why do smart people, like doctors, fall for statistical illusions?

A Screening Quiz Most Doctors Fail

Which of the following cancer screening programs would you recommend to your aging dad, brother, or significant other?

• Test A improved five-year survival from 68% to 99% and found more cancers at an early stage.

• Test B reduced cancer deaths from 2.0 to 1.6 per 1,000 people and increased the number of cancers diagnosed.

That’s the question Odette Wegwarth, Gerd Gigerenzer and colleagues asked 412 US primary-care physicians in a study [7]. They built both scenarios from real prostate cancer screening data (though, to avoid potential bias, they did not disclose the type of cancer).

Test A showed how PSA (prostate-specific antigen) screening looks through observational survival statistics, while Test B showed how the very same test performs in a randomized trial measuring actual mortality.

Presented with the same screening test for the same disease, 69% of doctors recommended it when screening performance was expressed in survival rates, whereas only 23% recommended it based on the actual mortality rates.

Oh, and when asked specifically, 76% said that better survival proves screening saves lives, and almost half (47%) said that finding more cancers in screened people proves it too.

How can the same test look so deceivingly different depending on which glasses you are wearing?

Two Illusions That Make Every Screening Test Look Good

Imagine identical twins whose arteries develop the same plaque from age 35. Anna is scanned at 35 and told she has atherosclerosis. Her sister Berta learns it at 62, after experiencing chest pain. Both have a heart attack at 70.

Measured from diagnosis, Anna had 35 event-free years and her sister only 8. On paper, screening bought Anna 27 more healthy years compared to her sister. In reality, it bought her nothing, except 27 extra years of living as a patient.

Survival measured from diagnosis always improves when diagnosis moves earlier.

That’s the trap behind Test A’s 68% to 99%. It’s also called “lead-time bias”.

A one-off scan is a snapshot of who has plaque at the time of screening. People with plaque that sits quietly for decades are more likely to populate screening results than those whose plaque rather rapidly progresses to a heart attack.

So when a study compares people who were diagnosed with the disease (based on their screening result) with people diagnosed only after symptom onset, the study compares apples with oranges. The screened ones will always look healthier, which exposes the funny fact that the plaques screening finds best are the ones that matter least.

That is why only one specific RCT design can tell us whether screening works: It needs to

• randomize people to screening or usual care,

• then count heart attacks, strokes, and deaths in everyone, screened or not.

Those fast-growing plaques that the screening missed will show up where they belong: in the screened group. DANCAVAS did that, and the results were more “duuh” than dazzling.

But REACT isn’t the study team’s last word.

The Trial REACT Should Run, and the One It Plans

The REACT team has a sequel on their drawing board. The CNIC release says that, if funded, a second phase called REACT-PROTECT will run from 2027 to 2032. It will be a large randomized trial testing whether imaging-guided prevention “can halt or reduce the systemic progression of atherosclerosis”.

But progression of atherosclerosis is only a surrogate for the only tangible results that really interest you: Fewer heart attacks and strokes. In the press release they appear only as the “long-term view”.

A surrogate is not an outcome. A trial of plaque progression cannot answer Fuster’s question of whether early intervention can cure the disease. And it cannot tell us whether screened people live longer or better lives.

Here is the trial I would want to see before anyone buys a handheld scanner for a GP’s office:

• Randomize people to imaging-guided prevention or current care, in the age groups where screening is actually proposed.

• Make heart attack, stroke, cardiovascular death, and all-cause death the primary endpoints.

• Count every harm: incidental findings, further tests, drug side effects, and the effect of being labelled.

• Follow people long enough for the outcomes to happen.

Until a trial like that has been conducted, here is where REACT leaves you.

My HypeWatch Verdict and the Two Questions for Your Doctor

• The results of REACT: Solid. REACT is the best map we have of how common silent plaque is across adult life.

• The question it answers: How many people have plaque? Answer: most, and almost everyone by their sixties.

• The question it doesn’t answer: Does looking for plaque help anyone? REACT can’t answer that, and the planned follow-up trial probably won’t either.

• The spin: A gradual climb from “detectable” to “may warrant” to “essential” to “cure”, with the evidence staying on the bottom rung throughout.

If you are in your sixties and someone offers you a plaque scan, you now know the most likely result. Ask two questions before you agree:

• What will you do differently if the test shows plaque?

• And is there trial evidence that doing so helps someone like me live longer or avoid a heart attack or stroke?

The dangerous statistic is not the 90%. The dangerous inference is the “therefore screen.”

Don’t take my word for it. Interrogate the sources.

For paid subscribers, I’ve set up an interrogation room for this post: a Gemini notebook loaded with the REACT paper, the press releases, the DANCAVAS trial, the Wegwarth physician study and Mühlhauser’s papers on screening. Ask it anything. Did I quote the authors fairly? What exactly did DANCAVAS find for heart attacks? How do lead-time and length-time bias apply to a plaque scan? Every answer points back to the passage it came from.

And when your question is no longer about REACT but about your own scan report, calcium score or lab results, that's what my written evidence reviews are for. They come in three levels, from a single lab report to a full decision review around one prevention question, and a short fit check tells you which one, if any, suits your situation before you pay anything. lutzkraushaar.com

Subscribe now

Cited References

[1] H. Bundgaard et al., “Prevalence of Silent Atherosclerosis across Adult Life,” N. Engl. J. Med., vol. 0, no. 0, doi: 10.1056/NEJMoa2609059.

[2] S. S. Khan et al., “Novel Prediction Equations for Absolute Risk Assessment of Total Cardiovascular Disease Incorporating Cardiovascular-Kidney-Metabolic Health: A Scientific Statement From the American Heart Association,” Circulation, pp. 1982–2004, 2023, doi: 10.1161/cir.0000000000001191.

[3] W. F. Enos, “CORONARY DISEASE AMONG UNITED STATES SOLDIERS KILLED IN ACTION IN KOREA”.

[4] G. S. Berenson et al., “Atherosclerosis of the aorta and coronary arteries and cardiovascular risk factors in persons aged 6 to 30 years and studied at necropsy (the Bogalusa Heart Study),” Am. J. Cardiol., vol. 70, no. 9, pp. 851–858, Oct. 1992, doi: 10.1016/0002-9149(92)90726-F.

[5] J. S. Lindholt et al., “Five-Year Outcomes of the Danish Cardiovascular Screening (DANCAVAS) Trial,” N. Engl. J. Med., vol. 387, no. 15, pp. 1385–1394, Oct. 2022, doi: 10.1056/NEJMoa2208681.

[6] I. Mühlhauser, “Zur Überschätzung des Nutzens von Prävention,” Z. Evidenz Fortbild. Qual. Im Gesundheitswesen, vol. 108, no. 4, pp. 208–218, 2014, doi: 10.1016/j.zefq.2013.11.006.

[7] O. Wegwarth, L. M. Schwartz, S. Woloshin, W. Gaissmaier, and G. Gigerenzer, “Do Physicians Understand Cancer Screening Statistics? A National Survey of Primary Care Physicians in the United States,” Ann. Intern. Med., vol. 156, no. 5, pp. 340–349, Mar. 2012, doi: 10.7326/0003-4819-156-5-201203060-00005.

The flop of Novartis’ Lp(a) drug pelacarsen gave the media two clickbaity stories in one: a billion-dollar pharma failure, and a scare about cholesterol’s nastiest cousin that the drug was supposed to tame.

Why it matters for you (in two bullet points)

• When a drug reduces a risk factor (Lp(a)) but not the actual risk (heart attacks), you may need to rethink how to read, and act on, your lab reports.

• When the media translates a pharma company’s zero-data two-paragraph press release into medical certainty, you need to question the media’s message.

Your takeaway

Your Lp(a) result is just one piece of cardiovascular evidence. Placed within your overall risk landscape, you’ll have neither a reason to panic, nor to ignore it.

Disclaimer: This post is for informational and educational purposes only. It is not intended to be a substitute medical advice, diagnosis, or treatment. As an epidemiologist and health scientist, I do not recommend for or against treatments or drugs.

Table of Contents
· The Drug Lowered Lp(a) and Still Missed the Point
· Three Questions Camouflaged as One
· How “High Lp(a)” Translates into Absolute Risk
· What Did HORIZON Actually Test?
· Four Things I Wanna See Before Calling HORIZON a Failure
· Your High Lp(a): What to Do With It Now?
· Your Access to the Interactive Evidence Interrogation Room
· Cited References

The Drug Lowered Lp(a) and Still Missed the Point

In medicine’s pantheon of middle-finger-flashing trial results, Novartis’ Lp(a)HORIZON just qualified for an induction. Lipoprotein(a) (Lp(a)), probably the most unfavourable cholesterol carrier your genes can imagine (and build), has evolved for years as one of cardiovascular medicine’s most promising targets. Its association with heart attacks and strokes…

• is causal (as far as genetic studies go) and

• fails to yield to dietary and exercise interventions

So, developing a drug that lowers Lp(a) sounded like a good idea. At least until September 4 this year, when Novartis announced that, in the phase III Lp(a)HORIZON trial, its drug pelacarsen

HORIZON had randomized more than 8,000 people with established cardiovascular disease and elevated Lp(a) to receive pelacarsen or placebo. The drug apparently did what it was designed to do: it lowered Lp(a). But it didn’t do squat to what really counts: reducing event risk.

I said “apparently” because all we have right now is Novartis’ two-paragraph press release, promising to present the detailed trial data during one of the upcoming congresses (and hopefully in a paper we can analyze).

Nevertheless, the media were quick to pick up the story in their characteristically low-key voice:

Reuters, WSJ, Bloomberg, CNBC, and Medical Daily offered their versions of why the drug may have failed. None of them wondered whether

• pelacarsen had really no effect,

• there was a modest effect that failed to clear the statistical threshold, or

• different cardiovascular outcomes moved in different directions.

One thing is for sure: HORIZON exposed the fun fact that even a causal exposure may not automatically be a valid therapeutic target, at least not for every exposed individual.

And that’s where you come in: What does a high Lp(a) result mean for you?

Three Questions Camouflaged as One

When your doctor tells you that Lp(a) is a cardiovascular risk factor, you might read this as a “yes” to three different questions:

First: Do people with higher Lp(a) have more cardiovascular events?

Definitely yes. Large population studies show that as Lp(a) rises, so does the risk for CVD events (specifically heart attacks and stroke). There is no magic threshold at, say, 50 mg/dL or any other commonly used cut-off. Risk increases along a gradient [1].

Second: Does Lp(a) actually cause all or some of that additional risk?

Probably yes. Mendelian randomization (MR) studies show that variants in the LPA gene that produce lifelong differences in Lp(a) also associate with differences in coronary disease and myocardial infarction [2].

Deep-Dive Box: Think of Mendelian randomization as Nature’s equivalent of clinical randomization. The purpose of the latter is to make a trial’s two randomized groups as similar as possible with respect to their biomarker profile. If randomization does its job, the two groups will not differ significantly in any of the measured biomarkers (age, sex, whatever) and, by extension, also no differences in the unmeasured biomarkers (e.g., dietary slip-ups, exercise, whatever). That’s the whole purpose of randomization: the elimination of unmeasured confounding.

Since we inherit our genes randomly before we were born, the group of people with genes that produce lifelong “low Lp(a)” and the group of people at the opposite end of the Lp(a) spectrum will be the same in every other biological or behavioral marker (unless, of course, that LPA gene also drives some of those other biomarkers, too, which is a phenomenon to be kept in mind, but which does not appear to be an issue in our scenario). That’s Nature’s way of randomization, which allows us to read the genetic association as causation. If that’s the case, the third question should have a positive answer, too.

Third: Does lowering Lp(a) with a drug actually reduce the risk for CV events?

That’s the question HORIZON actually tested (check its details in the Clinical Trials Registry).

Its answer: “Apparently not.” After all, pelacarsen was designed to dial down Lp(a).

BUT:

Knowing that lifelong exposure to something contributes to disease does not automatically mean that removing part of that exposure:

• decades later,

• in people who already have developed Lp(a)’s disease consequences,

• for only a few years of observation time

will prevent or slow down disease progression quickly enough to show up during a clinical trial.

Long story short: To the three questions about association, causation, and reversibility, HORIZON has delivered a disappointing answer only to the third one. And even that answer comes with a caveat: we haven’t seen the numbers yet.

How “High Lp(a)” Translates into Absolute Risk

When a lab report flags your Lp(a) value as high, it’s time for a little reality check. In my earlier post on Lp(a), I illustrated (among other things):

1. how Lp(a) amplifies LDL’s atherosclerotic potential,

2. what information your lab result’s choice of metric obscures, and

3. why Lp(a)’s dramatic-sounding relative risk escalation fizzles into far less sinister absolute risk elevations

That latter point just got an update, published in Circulation this September 2026 by Jeevanathan and colleagues [3].

The researchers followed (for a median of 12.5 years) 103,341 participants (of the Copenhagen General Population Study) aged 20 to 100 years who had no known ASCVD at baseline. Every 50 mg/dL (roughly 105 nmol/L) increase in Lp(a) came with a 12-18% relative increase in disease hazard. That relative increase held steady across all age groups. Of course, the older you get, the larger the absolute risk of suffering an event because advancing age inherently amplifies ASCVD vulnerability.

To get an idea about what those numbers mean in absolute terms, let’s look at the age group of 60-69-year-olds.

Relative to the group with the lowest Lp(a) level (0.1–29 mg/dL; 0-59 nmol/L), the absolute risk increased by just 3% until 90 mg/dL (190 nmol/L). So, instead of 11 of 100 people suffering an event, 14 will do so.

What these numbers tell you: Whatever drives your risk for ASCVD events, your membership in any of the Lp(a) categories above the lowest amps up your baseline probability for an ASCVD event.

What these numbers don’t tell you: Whether lowering your Lp(a) would also lower that amped-up risk. You can’t look to HORIZON for an answer, though.

That trial’s population was nowhere near similar to the Copenhagen guys. The latter were a primary prevention cohort with no known ASCVD at enrollment. In contrast, HORIZON only recruited secondary-prevention prospects who had already suffered an event.

So, just don’t take your Lp(a) number as an immovable verdict.

What Did HORIZON Actually Test?

Keep in Mind:

Once HORIZON’s numbers are out, we will probably be able to tell which option applies.

Think about it this way: the genetic evidence that propelled Lp(a) into the therapeutic limelight reflects lifelong exposure. That’s what Mendelian-randomization studies are about. The person entering the HORIZON trial represents a very narrow sub-set of that lifelong exposure: the subset of people with elevated Lp(a) who have already suffered an event. If you don’t meet these criteria, HORIZON won’t tell you much, because it only began lowering Lp(a) after decades of high-level exposure and then looked for effects for only about 4 years.

So, HORIZON wasn’t a general test of:

“Is high Lp(a) causal? And does lowering it reduce the risk of ASCVD?”

It asked a much narrower question :

“If we substantially lower circulating Lp(a), …

• starting late in life,

• in people who already have established cardiovascular disease, and

• who already are on secondary prevention treatments,

• …will that protect a statistically significant number of people from suffering recurrent events?”

That is a much tougher proposition because by the time someone has already had a myocardial infarction, stroke, or peripheral artery disease event, Lp(a) is only one actor among many. ApoB, blood pressure, diabetes, obesity, and smoking history (to name just the major culprits) have driven a decades-long process of plaque formation. And whatever treatment the patient is already receiving also contributes to what happens next.

We need to see the HORIZON study’s numbers to tell whether its “failure” represents an absence of evidence for benefit—or evidence for the absence of a clinically meaningful benefit.

Four Things I Wanna See Before Calling HORIZON a Failure

First, the hazard ratio (HR) with its 95% confidence interval (95% CI) is a dead give-away. If it comes out close to 1.00 with a narrow 95% CI, then, yes, that’s evidence against a clinically meaningful effect (but, mind you, only in that specific population).

If, on the other hand, the HR sits further away from 1.00, say, at 0.90 with a 95% CI of 0.80–1.01, the trial would still have “failed” its significance test (the p-value would be >0.05), but the data would hint at a clinically meaningful benefit.

Second, the component outcomes might tell another interesting story. The researchers lumped cardiovascular death, non-fatal myocardial infarction, non-fatal stroke, and urgent coronary revascularization into one primary endpoint. That may hide different component outcomes. For example, if one or two of those components showed a beneficial result, but the rest did just the opposite, the resulting neutral composite result may hide some uncomfortable truth: That the Lp(a)-lowering strategy may actually do more damage than good. That wouldn’t be a first for the pharma industry.

Pfizer’s hyped torcetrapib flamed out in a phase-3 trial when it increased HDL cholesterol (the drug’s targeted effect) but paradoxically also increased major cardiovascular events and mortality, but not non-fatal heart attacks [4].

Third, I want to know how much the drug lowered Lp(a) (in percent and in absolute mg/dL or nmol/L terms). The genetic studies that suggested Lp(a) to be a causal driver of disease events drew their conclusions from absolute lifetime exposure differences.

Fourth, there are other factors: treatment adherence, safety signals, and whether the event curves ever began to separate.

When the full trial data appear, I’ll subject it to one of my next HypeWatch autopsies.

Your High Lp(a): What to Do With It Now?

Don’t ignore an elevated Lp(a) just because its effect on absolute risk isn’t as dramatic as the media make it sound. But also don’t panic about it. Your lab result identifies your exposure to a genetically determined, and probably causal, contributor to elevated ASCVD risk. In itself it is neither a diagnosis nor the “writing on the wall”.

You and your doctor need to read it in the context of your entire cardiovascular profile:

Coronary artery calcium (CAC), for example, tells us something about how much calcified plaque has already accumulated. On the other hand, a CAC score of zero may substantially reduce your risk estimate (I presented the details and how you can calculate your risk adjustment in an earlier post).

Assessing LDL-C together with ApoB gives us a much clearer picture about the atherogenic particle burden. Blood pressure, diabetes, smoking, obesity, kidney and endothelial function, and established cardiovascular disease change the risk profile further. And most of them are yours to modify.

So if your Lp(a) has come back high, the key question is:

“How much does this result change the interpretation of the rest of my cardiovascular evidence?”

That is the kind of question my cardiovascular evidence reviews are designed to help answer: putting isolated lab values back into the context of the rest of your evidence, so you have something more useful to discuss with your physician than a single red flag on a lab report. That is particularly relevant when your results point in different directions.

Your Access to the Interactive Evidence Interrogation Room

To help you interrogate the scientific evidence for Lp(a) with your personal questions, I have assembled a Gemini Notebook. It is freely accessible to Double Check Doc subscribers (yes, that includes free subscriptions).

Subscribe to DCD (free or paid) to access the bonus content

And if you want to get up-to-speed on Lp(a):

Cited References

[1] H. S. Bhatia et al., “Independence of Lipoprotein(a) and Low-Density Lipoprotein Cholesterol–Mediated Cardiovascular Risk: A Participant-Level Meta-Analysis,” Circulation, vol. 151, no. 4, pp. 312–321, Jan. 2025, doi: 10.1161/CIRCULATIONAHA.124.069556.

[2] S. Burgess et al., “Association of LPA Variants With Risk of Coronary Disease and the Implications for Lipoprotein(a)-Lowering Therapies: A Mendelian Randomization Analysis,” JAMA Cardiol., vol. 3, no. 7, pp. 619–627, Jul. 2018, doi: 10.1001/jamacardio.2018.1470.

[3] J. Jeevanathan, P. E. Thomas, P. R. Kamstrup, and B. G. Nordestgaard, “Lipoprotein(a) and Atherosclerotic Cardiovascular Disease Among Adults Aged 20 to 100 Years,” Circulation, p. CIRCULATIONAHA.126.079412, Aug. 2026, doi: 10.1161/CIRCULATIONAHA.126.079412.

[4] K. Honey, “Drug designed to raise HDL levels falls down,” J. Clin. Invest., vol. 117, no. 2, pp. 282–282, Feb. 2007, doi: 10.1172/JCI31253.

• Reader 1 asks: Who exactly was studied?

• Reader 2 asks: Who, within that selected population, actually benefits most?

Reader Hans Martens’ Questions:

Below is a screenshot of his comment on Medium

Hans asked the question that I hoped to find in the comments to my post.

Before I answer it, Hans’ underlying assumption of “clean” study populations needs a small correction (unless I misunderstand his caveat):

A double-blind trial does not have to use a clinically “clean” population (if by “clean” you mean free from disease). Randomization and controlling (in a randomized controlled trial, or RCT) are the strategies to minimize the potential for confounding and other “noise” in the data.

This combo of strategies works equally well in study populations that are not clinically “clean” in the sense of the diseases they carry. STAREE’s investigators deliberately selected relatively healthy, independently living people because they wanted to answer the question:

So let me paraphrase your first question and then answer it with data for the Australian and US populations (since most readers of the post reside in North America):

Q1. How many people over 70 actually resemble the STAREE participants?

Quick and dirty answer:

• ≈40% of Australian 70-plus-year-olds, and

• ≈30% of the same age group in the US.

That sounds pretty close to the ≈45% (9,971 randomized participants of 22,305 screened individuals). But these percentages need some clarification. First, let’s look at what STAREE’s screening funnel tells us:

These numbers are optimistically biased: GPs had already pre-vetted the 22,305 individuals as likely eligible. 70% of those were found to be eligible based on their health status:

• No previous cardiovascular disease, diabetes, or dementia.

• Independent community living. No substantial physical disability.

• No moderate-to-severe kidney or liver disease.

• No illness expected to kill you within five years.

That’s a lot to ask of a 70-plus-year-old. The gap of more than 5,000 individuals between the 15,722 disease-free eligible ones and the 9,971 randomized participants is largely due to drop-out (declining to participate or failure to pass the pre-randomization pill adherence test).

Now let’s look at the prevalence of STAREE-eligibility in the populations of the US and Australia.

These are back-of-the envelope numbers extracted from various statistics published in the US and Australia (exact percentage are difficult to derive from these population statistics because the exclusion conditions overlap):

• Australian Bureau of Statistics (ABS) “Long-Term Health Conditions”

• US National Health Interview Survey

• Diabetes in America [1]

• Trends in Statin Use in Australia [2]

So, my answer to Hans’ question:

STAREE participants probably represent the healthiest half to two-thirds of Australians aged 70+, and a somewhat smaller share of their US peers (simply because chronic disease prevalence is higher in the US). If you additionally consider current statin use, the proportion of fully eligible STAREE participants shrinks to roughly one-third. Also, you can’t extrapolate the trial results to the oldest-old (those aged 85+) because they were severely underrepresented in STAREE.

But STAREE participants were by no means hallmarks of health. At baseline:

• ≈20% took 5+ medications

• ≈46% were on antihypertensives

• mean LDL was 126.6 ± 28 mg/dL (3.27 ± 0.72 mmol/L)

So “healthy” needs a qualifier: free of some exclusionary conditions, but not free of risk, but still functioning independently.

That qualifier evaporated on the way from the methods section to the newspaper headlines.

Q2. Why would somebody with no cardiovascular disease take atorvastatin in the first place?

Well, Hans, that’s the definition of primary prevention. It is all about NOT WAITING until you keel over from a heart attack. It is about trying to prevent that in the first place. Statins are all about prevention.

Also, “No known cardiovascular disease” doesn’t equate to “no cardiovascular risk.” Age itself is a risk factor, so is hypertension, LDL/ApoB discordance and exposure, smoking, kidney function, family history, Lp(a), and coronary plaque, for example.

So STAREE asked a perfectly legitimate primary prevention question:

That question had not been answered before, because that age group was underrepresented in previous trials. STAREE delivered an answer that, as I discussed in my post, may not appear to you as a resounding “Yes”.

Once you look behind the 30% relative risk reduction, the absolute benefit may not convince you to opt for the drug. And that’s not something I or anyone else can or should decide for you. You have to decide for yourself, together with your doctor. My purpose is to help you have a better-informed conversation with your physician.

Reader Creig S.’ Question:

Does the Same 30% Benefit Apply to Everyone?

Let me paraphrase Creig’s question:

After all, STAREE participants probably had a wide range of LDL levels, blood pressures, degrees of underlying atherosclerosis, and different starting cardiovascular risks.

The straightforward answer is ‘Yes’.

That is why I often emphasize that trial averages hardly ever translate into identical results in individuals. In fact, the “average” participant is a statistical ghost. And, correspondingly, so is the average patient.

I illustrated this phenomenon in an earlier post.

But first, I need to correct one misconception in Creig’s comment.

Statins are not prescribed simply because somebody’s LDL number has crossed some specific threshold.

BTW: None of those thresholds, be it for blood pressure, blood sugar, or blood lipids, are red lines dictated by biology. Rather, they represent a consensus among the relevant medical associations’ alpha animals about the values above which the risk for an event starts to increase sufficiently fast to warrant a preventive intervention.

Indisputably, LDL cholesterol—or more precisely the burden of ApoB-containing lipoproteins– causally contributes to atherosclerotic cardiovascular disease and its events (e.g., heart attacks, stroke). But LDL is far from being the major trigger. These events are the result of a perfect storm scenario, as I explained in an earlier post.

In a typical type-1 heart attack this perfect storm requires [3]:

1. the presence of atherosclerotic plaques,

2. a plaque to break open, resulting in the formation of a blood clot,

3. the blood clot to be large enough to block blood flow to downstream tissue, and subsequently…

4. failure to clear that blood clot in time.

Only if all these 4 steps happen together in an artery of the heart (coronary artery), the muscle cells that depend on the artery’s blood and oxygen supply will die, which is the definition of a heart attack.

Some fun facts to illustrate the dominant role of chance [4]:

Approximately 80% of adults have some degree of coronary atherosclerosis; probably one in three has a history of plaque ruptures and clot formation, but the yearly incidence rate of acute coronary events (ACEs) in people older than 40 is somewhere between 0.2% and 1%. Moreover, plaque ruptures happen every year in millions of people without them even noticing.

Knocking out any one of the four conditions will prevent a heart attack from happening. Hence, cholesterol, being a contributor to the first of those 4 conditions,  will not have the predictably same effect in any two individuals with differing vulnerabilities to any or all of the remaining 3 conditions. So, Creig’s suspicion is correct:

Whether lowering LDL is worth doing in a particular person depends heavily on the probability that this person would suffer an event in the first place.

Did STAREE account for this level of differentiation?

Yes, to a degree.

The investigators ran prespecified subgroup analyses, investigating whether stratification for age, sex, baseline LDL-C tertiles, blood pressure, BMI, kidney function, and cognitive score revealed any significant differences in outcomes. Nada.

But “Neither subgroup interaction was statistically significant” does not mean “everyone benefited equally.”

Subgroup analyses are notoriously underpowered for detecting treatment-effect differences that may be clinically meaningful but too small to discover given the sub-sample sizes. So, the answer to Creig’s question is:

What about LDL specifically?

STAREE inclusion criteria kept the participants’ LDL levels at a relatively moderate mean of 127 mg/dL, or 3.27 mmol/L.

People with very high total cholesterol—above 7.5 mmol/L—were excluded.

The trial’s question was whether lowering LDL in healthy-ish statin-naive older people could prevent enough cardiovascular events to justify a treatment recommendation.

In summary, Creig’s point is spot-on:

There probably are STAREE-like individuals for whom the trial-averaged 2.3-percentage-point absolute benefit substantially understates the achievable personal benefit.

Likewise, there will be others for whom those 2.3 percentage points are an overstatement.

The challenge is to identify those individuals in clinical practice. The headlines’ exuberant statement that

is simultaneously true and also useless for an individual. The real questions are:

• Thirty percent of what?

• For whom?

• Over what period?

• Against what starting risk?

• And with what competing risks?

And then came the anecdotal evidence (an N-of-1 case, actually)

While I was writing this response, another reader, Don Salmon, supplied the perfect reminder of why (a) trial averages are so misleading, and (b) drugs aren’t necessarily your cardiovascular risk lever.

Don is 74, physically exceptionally active and fit, with a strong paternal family history of heart attacks and strokes. His doctor had recommended a statin for years. He finally tried atorvastatin this spring. Soon afterwards, he developed severe fatigue and unbearable sensory symptoms that took several weeks to resolve after stopping the drug.

You can read his remarkable account in his own words here → [original Medium comment].

So, Don took matters into his own hands.

He switched his diet to a calorie-controlled, Portfolio-style diet and sharpened an already serious exercise routine. That lifestyle change dropped his weight from 179 to 167 pounds, his LDL from 179 to 116 mg/dL, and his blood pressure from a 3-meds-controlled 120/80 mmHg to 115/75 mmHg. That is more than impressive. It’s motivating, which is why I mention Don’s case here.

But these are also one person’s numbers.

We cannot tell from his experience how much of the LDL change came from food composition, calorie restriction, weight loss, exercise, or measurement variation. We also cannot tell whether another 74-year-old will achieve similar numbers when they adopt the same routine as Don.

I, for one, would be unable to sustain the severely meat- and dairy-restricted Portfolio diet. I can forego meat for a day, maybe two, but by day three at the latest, I become irritable, and nothing seems to satisfy my appetite (nothing plant-based anyway). I actually live on meat, full-fat milk, and lots of fruits (not so much vegetables).

15 years ago I did what Don did this year. I deliberately designed my diet for heart and arterial health. I made breakfast the anchor meal of the day, composing a smoothie that contains (almost) everything aimed at keeping my cardiovascular system in mint condition.

I called it an anchor meal because it relieves me from having to overthink all the foods I eat during the day (but processed meats, like sausages, and fast carbs, including bread and pasta, are only occasional treats). You can read about it in detail in two earlier posts, one about my anchor meal, and one about how to construct yours specifically to lower cholesterol.

The image below is, again, only anecdotal evidence.

My LDL is almost exactly half of Don’s. My HDL is substantially higher than LDL.

Again, which part of my diet, whether my weight (BMI 23), and how much my exercise routine (60’ cardio & 30’-45’ strength training Mo-Fri) contributes to those values, and whether you’d see the same results if you copied my lifestyle routine remains unknown (or unknowable)

But it illustrates what disappears when the media is lording messages like “statins reduce cardiovascular risk by 30%.” over their readers: The potential benefits and challenges of individualized preventive medicine (including lifestyle medicine) get lost.

Synthesis: The Two Questions STAREE Cannot Answer for You

STAREE tells you whether atorvastatin generates some benefits in the study population. It does not tell you whether and how much of this benefit would accrue to you. The answer to that question hinges on two criteria:

The study’s external validity: Do you resemble the people who were studied?

Absolute treatment benefit: Given your baseline risk, how much do you personally stand to gain?

To help you have a better-informed discussion with your doctor, I created the downloadable two-page “Five Questions Before You Decide on a Statin After 70” guide, with the absolute-risk numbers, PREVENT and SCORE2-OP risk calculator links, and a checklist for your next doctor conversation.

It’s free for Double Check Doc subscribers.

Subscribe to DCD (free or paid) to download this bonus content

Subscribe now

Cited References

[1] L. M. Mitchell, M. Huisingh-Scheetz, and E. S. Huang, “Diabetes in Older Adults,” in Diabetes in America, J. M. Lawrence, S. S. Casagrande, W. H. Herman, D. J. Wexler, and W. T. Cefalu, Eds., Bethesda (MD): National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), 2023. Accessed: Sep. 21, 2026. [Online]. Available: http://www.ncbi.nlm.nih.gov/books/NBK622602/

[2] R. Ofori-Asenso, J. Ilomäki, E. Zomer, A. J. Curtis, S. Zoungas, and D. Liew, “A 10-Year Trend in Statin Use Among Older Adults in Australia: an Analysis Using National Pharmacy Claims Data,” Cardiovasc. Drugs Ther., vol. 32, no. 3, pp. 265–272, Jun. 2018, doi: 10.1007/s10557-018-6794-x.

[3] A. Arbab-Zadeh and V. Fuster, “The Myth of the ‘Vulnerable Plaque’ Transitioning From a Focus on Individual Lesions to Atherosclerotic Disease Burden for Coronary Artery Disease Risk Assessment,” J. Am. Coll. Cardiol., 2015, doi: 10.1016/j.jacc.2014.11.041.

[4] A. Arbab-Zadeh, M. Nakano, R. Virmani, and V. Fuster, “Acute coronary events,” Circulation, vol. 125, no. 9, pp. 1147–1156, 2012, doi: 10.1161/CIRCULATIONAHA.111.047431.

Major press releases spin-doctored the latest statin trial’s spectacularly anemic results into thrilling news for older people who are still sitting on the statin fence: a 30% risk reduction for heart attacks and strokes.

The Big Thing About STAREE

The trial was the largest and latest intervention study designed to deliver a true cause-and-effect estimate for statins. It:

• randomized almost 10,000 participants to atorvastatin or placebo

• focused on the hitherto neglected healthy 70+ year olds

• steered clear of pharma funding

That matters to…

you, the baby-boomer, because, upon closer inspection,

• the 30% shrinks to 2%;

• the pill doesn’t buy more years lived,

• not even more disability-free years,

• and it has no demonstrable effect on dreaded dementia.

But if your profile is close to that of the participants, your doctor will almost certainly bring it up. Read on, if you want that discussion to be a better-informed one.

Your Practical Takeaway

• A 5-question cheat sheet helping you discuss these results with your doctor (who, I bet, will definitely bring it up if you don’t).

Disclaimer: This post is for informational and educational purposes only. It is not intended to be a substitute medical advice, diagnosis, or treatment. As an epidemiologist and health scientist, I do not recommend for or against taking a statin. That’s your and your doctor’s decision. This post is my attempt at enabling you to have a better-informed shared-decision-making conversation with your doctor.

Table of Contents
· Four Questions. One Made the Headlines.
· First: 30% of What?
· The Other Primary Endpoint Most Headlines Buried
· And Dementia? We’re Still Waiting
· Almost Half Stopped the Assigned Pills — Not the Trial
· What Happened on the Way to the Headline?
· What STAREE Means for Your Statin Decision After 70.
· Your Takeaway
· Cited Reference

Four Questions. One Made the Headlines.

The STAREE trial delivered answers to the four questions that have kept many aging baby boomers sitting on the statin fence [1]:

1. Do statins reduce major cardiovascular events? Yep, by 30%

2. Will statins make me live longer? Nope!

3. Ok, do statins at least let me enjoy more disability-free years? Fat hope, but nope.

4. How about dementia? No effect either.

By the time the press-release and media squads had “Peskov-ed” the results, the first answer had become the story, while the other three got buried as footnotes.

That matters because STAREE stands out from the rest of statin trials. Not only because its investigators reinvigorated the long-lost cultivation of drug trials unwarped by pharma funding. They also focused on the 70+ age group. Previous trials largely ignored them. From among those, STAREE randomized 9,971 relatively healthy, primary-prevention participants to atorvastatin or placebo and followed them for a median of 5.9 years.

That’s exactly the group of people who wonder whether statins protect, harm, or don’t do squat to the aging brain. I know that, because I’m uncomfortably close to that age range. Close enough to have dedicated an entire article to that question, which I had closed by telling you that STAREE will answer it.

It didn’t. At least not with the confidence that I had hoped for.

With a hazard ratio (HR) of 1.03, the dementia result landed almost exactly on “zero effect”. And with its 95% confidence interval ranging from 0.87 to 1.21, it’s compatible with anything from a modest protective effect to a clinically meaningful increase in dementia risk.

And that was only the first surprise.

The headlined 30% cardiovascular risk reduction didn’t apply to stroke (there was no statistically significant reduction in stroke risk). Also, halfway through the trial, the researchers had to broaden the primary endpoint of major adverse cardiovascular event (MACE) because those events happened at a much slower rate than what the researchers had expected when they designed the trial. The unexpectedly low pooled event rate forced the trialists to move from a 3-point composite consisting of cardiovascular death, nonfatal myocardial infarction (MI), and nonfatal stroke to add coronary revascularization as a fourth component.

That’s all legitimate if done correctly, which was the case. But it probably makes you smell the rat that I have addressed in several of my previous articles: the uncomfortable gap between relative risk reduction (the 30%) vs. absolute risk reduction (ARR).

Another curious point is that by the end of the study, half of all participants had dropped out. Ironically, more in the placebo group (56.8%) than in the atorvastatin group (47.6%).

Taken together, that’s what I meant with “spectacularly anemic”.

And that is where this ‘HypeWatch’ begins.

First: 30% of What?

The “30% reduction” sounds like something you’d be stupid not to care about. The image below may challenge that notion.

The image shows the absolute risk reduction (ARR) of the two key components of the adjusted 4-component primary outcome (i.e., heart attack and bypass/stents). The difference of 43% vs. the headlined 30% MACE reduction is due to the fact that the other two MACE components (stroke and cardiovascular death were not statistically significant between treatment groups.

Here is how you read these numbers:

In the untreated (placebo) group, roughly 3 (2.7) of 100 people experienced a heart attack during the 6-year follow-up. In the atorvastatin group, roughly 2 (1.6) people suffered the same fate. That’s an absolute reduction of 1.1% (1.16% to be exact).

That difference of 1.1%, the so-called absolute risk reduction (ARR), means that 86 people need to take a statin to prevent one single heart attack.

That “86” is the number needed to treat (NNT), which I explained in detail in an earlier post (“To Statin Or Not To Statin: What Neither Side Is Telling You”).

In other words, for 85 of those 86, the pill will not prevent a heart attack because they wouldn’t suffer one in the first place (at least not over the next 6 years). If that doesn’t pull you off the statin fence, I can empathize.

Just to keep the record straight: the original three-component endpoint still favored atorvastatin: HR 0.73, 95% CI 0.62–0.87. The ARR was 1.6%, which translates into an NNT of 63.

As I explained above, the switch from a 3-component to a 4-component primary endpoint wasn’t an illicit moving of the goalposts halfway through the trial. But it affects the qualitative interpretation of the headline: A heart attack happens to you, and it is a fairly straightforward diagnosis. A coronary revascularization is a clinical procedure, triggered by a certain dose of physician judgment and local practice patterns. It also occurred about 40% more often than MIs.

I hope the numbers above have helped clarify why I called the headline “statins cut heart attacks and strokes by 30%” a Peskov-ed translation of what the trial actually found.

The Other Primary Endpoint Most Headlines Buried

That’s the question on many baby-boomers’ minds.

“Nope” was STAREE’s straight answer.

Disability-free survival occurred at roughly the same rate in both groups: HR 0.94, 95% CI 0.84–1.05, P=0.25. So did overall survival, which means that the statin users’ fewer cardiovascular events failed to translate into more years lived free of death, dementia, or other disability.

For an aging baby boomer, that is NOT a footnote.

And Dementia? We’re Still Waiting

This was the part I had personally been waiting for.

In my previous article on statins and dementia, I illustrated the paradoxical observation that statin use associates with fewer Alzheimer diagnoses, yet no obvious reduction in Alzheimer pathology at autopsy, and no reduction in vascular dementia. This evidence comes exclusively from observational studies. They only allow us to find associations, not cause-and-effect relations. So, I honestly expected STAREE, an interventional trial, to deliver a clear verdict.

And it did, in the oracle-of-Delphi sort of way.

Dementia was diagnosed in 290 atorvastatin participants and 280 placebo participants: HR 1.03, 95% CI 0.87–1.21.

The point estimate sits almost exactly on “no effect.” Unfortunately, the confidence interval is wide enough to include anything from a 13% lower to a 21% higher dementia hazard.

So the jury is still out about statins’ effects on dementia: could they have a preventive effect, or might they actually drive dementia risk up?

But let’s look at the bright side: if after almost 60,000 person-years of follow-up, the possible effect of statins on dementia hasn’t shown up, that effect, if any, is probably vanishingly small (though it could go in either direction).

Almost Half Stopped the Assigned Pills — Not the Trial

The drop-out number is equally deserving of our attention.

By year 5, 47.6% of statin users, and 56.8% of the placebo group had discontinued the drug (with one in five participants already off their pills by the end of year 1). While that number is humongous, it does not reflect trial dropout. The researchers ascertained the vital status of 99.9% of participants. It’s just that half of them didn’t complete it with the pill they had been randomized to initially.

Here comes the fascinating part:

7.2% of the statin users dropped out because of side effect complaints. Their doctors, all of them blinded to whether the pills they administered contained the statin or a placebo, recorded their patients’ complaints as “possibly, probably, or definitely” related to the trial regimen.

Now, before you let those 7.2% feed your statin side-effect suspicion: 6.1% of the placebo group dropped out for the same reason.

That’s a 1.1% difference. Its confidence interval from +0.04% to 2.0% makes it

• (a) statistically significant enough to not rule out side effects, and

• (b) suggestive that those side effects may occur in less than half a percent to 2% of statin users.

Now, what happens in a trial like STAREE when a patient’s doctor, blinded to what’s in the pill they administer, feels that their patient definitely needs a statin prescription?

That’s when open-label prescribing kicks in. The doctor, remaining blinded, will switch the patient to the “open-label” prescription and discontinue the trial pill. That is, if the patient is on a placebo, they will henceforth receive the statin; if they had been on the statin, the newly prescribed statin simply replaces the trial statin. By the end of the trial, this type of open-label switch had happened to one in five placebo participants.

That potentially impacts the results.

In an intention-to-treat analysis (ITT), researchers strictly compare the results of the groups to which the participants had originally been randomized. Any placebo participant who open-label-switched by, say, year 2 to receiving the statin will have his/her outcome recorded in the placebo group. That narrows the effect difference (if any) between the groups.

Unfortunately, while the trial design explicitly allowed for an additional per-treatment analysis, the published paper doesn’t provide that calculation. So, I can’t comment on it. But I suspect that if it had produced a markedly more favourable result than the advertised 30%, we would have been informed enthusiastically.

So the real message about drop-out is far less dramatic but highlights a different type of problem:

long-term pill persistence is generally poor (not only in trials, but also in real life).

What Happened on the Way to the Headline?

STAREE’s home is Monash University, in Melbourne, Australia. I lived in Melbourne for about two years, where I, the probably typical stone-faced, humor-deficient Kraut, fell in love with Aussies’ far sunnier disposition. That disposition shines through Monash’s own press release, its authors apparently unrestrained by stricter reporting guardrails that (hopefully) apply to their academic colleagues.

The headline dressed up a measly 2% absolute risk reduction into a factually correct, but potentially misleading, euphoric 30% message.

The article focused entirely on the positive cardiovascular result, failing to disclose that the trial found no improvement in disability-free survival or overall lifespan.

The author(s) even added a promotional dose of longevity exuberance, quoting lead researchers as saying that the trial

“…answers fundamental questions about healthy longevity” and helps older adults “remain fit and healthy”.

That’s a stretch.

Equally unqualified is the characterization of atorvastatin as a “safe and effective measure” without mentioning that the statin-attributable side effects (1.1%) were almost exactly as “large” as the reduction of heart attack risk (1.16%).

To be fair, comparing the two only by their frequency is like comparing apples with oranges. A heart attack is a far more serious event than a muscle ache or an elevation of liver enzymes. The latter two are typically reversible; the former isn’t.

Now before the North American reader breaks out into a smug smile about the Aussie’s less than factually faithful reporting, the ACC’s (American College of Cardiology) press release wasn’t an example of exactitude either. While its headline carried the more sober question “Can Statin Therapy Help Prevent CV Events in Healthy, Community-Dwelling, Older Adults?”, its answer contained factually incorrect statements:

• “The trial enrolled nearly 5,000 adults over the age of 70 years…” No, it was close to 10,000 adults.

• “Results showed a primary cardiovascular endpoint event… occurred in 297 participants receiving atorvastatin after six months compared with 412 participants receiving placebo.” No, it was after 6 years, not 6 months.

That’s a level of sloppiness unbecoming of the ACC.

In its defense, the ACC did quote the trial’s lead researcher acknowledging that:

“…the lower incidence of major, nonfatal cardiovascular events with atorvastatin than with placebo did not result in a corresponding improvement in disability-free survival.”

It also mentioned that 80% of the deaths (actually 82%; 680 of 828) were of non-cardiovascular causes.

That observation might give you food for thought: Once you cross the 70-year threshold with an apparently intact cardiovascular system, the Bogeyman of death and disability is far more likely to catch you with another of his tools.

And that begs the question:

What STAREE Means for Your Statin Decision After 70.

If your profile matches that of the STAREE participants — being relatively healthy, living independently, having no known cardiovascular disease, diabetes, or dementia — atorvastatin can reduce your risk for major cardiovascular events over about six years.

Over the trial population that benefit was real. But the absolute benefit, which is what interests you, the individual, far more, is modest. And it is concentrated in a lower risk of heart attacks and revascularization procedures, not of stroke, death, dementia, or disability-free survival.

So, your real question is not whether statins work in the population.

It is: “How much can I personally expect to gain, and is that gain worth six years of daily treatment?”

The answer depends very much on your starting risk. A 30% relative reduction means a lot more to someone with a baseline risk of, say, 20% than to someone whose low cardiovascular burden puts him or her at 5%. That is where your coexistent risk markers — blood pressure, smoking, ApoB, Lp(a), CAC, or kidney function matter far more than your birthday alone.

You also have to ask yourself what outcome you care about most. Avoiding a heart attack or a stent may be worth a daily pill even if it doesn’t buy you more years, not even more disability-free years. But you may also come to the opposite conclusion.

Anyway, Monash’s press release concludes by quoting the trial’s lead researchers Zoungas and Nelson as saying:

“In light of this new evidence, older patients should have a conversation with their GP about whether statins are an appropriate treatment for their cardiovascular health to help them remain fit and healthy.”

&

“Answering fundamental questions about healthy longevity has become a universal imperative”

Your Takeaway

To prepare you (or your loved one) for that conversation, I have assembled a two-page “Five Questions Before You Decide on a Statin After 70” guide, with the absolute-risk numbers, PREVENT and SCORE2-OP risk calculator links, and a checklist for your next doctor conversation.

It’s free for Double Check Doc subscribers

Subscribe to DCD (free or paid) to download this bonus content

Subscribe now

PS. I owe Mr Peskov an apology, of course. None of the anonymous media authors I criticized here possesses anywhere near his skills of pretzel-bending facts and truths into their exact opposite. In that respect, we have to concede the Gremlin’s (sorry, Kremlin’s) unassailable supremacy.

Cited Reference

[1] S. Zoungas et al., “Atorvastatin, Cardiovascular Events, and Disability-free Survival in Older Adults,” N. Engl. J. Med., p. NEJMoa2607314, Aug. 2026, doi: 10.1056/NEJMoa2607314.

The New AI Purity Badge

I never [had sex, talked business, used AI] with [that woman, with anybody, in my writing].

The problem with categorical denials is that they leave no wiggle room once the beans have been spilled. One would more leniently define “sex” and “business” if one knew beforehand how much of the denied story will eventually leak out.

Now you can (know beforehand, that is). At least if you are a writer with an illicit AI affair.

On Substack: In July this year, Substack introduced a clever way of getting your denials just right. Once you complete your post, you can let Pangram, an AI detector, inspect it. Then, with Pangram’s tally comes the option to craft a “How I make this” statement.

Of course, you could also massage your post until the machine’s manipulated estimate of provenance matches your favourite creator eulogy. Or you could just disable Pangram for your readers altogether.

On Medium: Many major publications are now guardrailing their submission guidelines against AI contamination. One editor even designed a no-AI logo.

Does knowing whether a machine helped produce the sentences really tell us whether those sentences are worth reading? Almost everybody and their uncle seems to subscribe to that premise. I think that idea is addle-brained. For two reasons:

• Bias: Judge the work, not the tool.

• Credibility: A badge cannot certify honesty.

My rationale covers non-fiction writing only. As a scientist, that’s where I’m at home.

Judge the Work, Not the Tool.

• Suppose I develop the argument, gather the evidence, interrogate it to write the entire draft, and then ask Gemini to tighten three clumsy paragraphs. Has the article now ceased to be mine?

• Suppose I develop the argument like in the first example, submit the rough draft to Claude, and let it turn the draft into coherent prose. The ideas, evidence, sequence, and conclusions are mine, but many of the final sentences are not. Is that human writing or AI writing?

• Suppose I develop the argument, but use ChatGPT to search the literature, summarize papers, identify weaknesses in my argument, and find counterevidence I had overlooked. I still write the entire article, yet AI may have impacted its intellectual content far more than in the previous example. Has the article ceased to be mine?

• Suppose I develop an argument, ask an AI to challenge it, and it proposes an objection that had not been on my radar. Upon literature review, I discover that it has merits, is central to revising my argument, and eventually becomes the core of the final article. At what point did that idea become mine?

In the latter scenario, replace “AI” with “thesis supervisor” or “colleague,” and what you get is a description of the process behind most PhD theses and excellent scientific papers worth reading. Is that article unworthy of your eyeballs because silicon chips did what had been a human brain’s job until LLM’s entered the scene?

Now suppose I engage an AI in a brainstorming session. It proposes a genuinely new hypothesis. I initially doubt it. But when I test it against the evidence, it holds water; I refine it, and then publish it. Is the idea mine or the machine’s?

This is not a hypothetical scenario. AI systems have already generated new biomedical hypotheses and drug targets that not only had not previously been conceived by humans but subsequently also survived laboratory testing. That includes the spread of genetic elements between bacterial species, new treatments for acute myeloid leukaemia (AML), liver fibrosis, and dry age-related macular degeneration (AMD) [1] [2].

But don’t rush to your AI buddy to brainstorm a novel cure for politicians’ pseudologia fantastica, or mythomania (I swear, I’m not making this up), otherwise known as ‘compulsive lying’. The novel medical insights I mentioned didn’t emerge from cleverly prompted LLMs. Complex multi-agent systems are behind these discoveries. The point I want to make:

Come to think of it: Let’s hope that AI never catches our “bias” virus. It might develop a racist streak against us, the human slop. Since we all know that life imitates art, you might want to watch Terminator again. Skynet, anyone?

A Badge Can’t Certify Honesty

Given the absolute honesty and truthfulness that invariably infects people once they enter the social media stage, we can put total faith into their increasingly elaborate AI declarations.

Just kidding.

But whatever their credibility, I put it to you that these declarations actually answer the wrong question: The question about provenance, or the production method. And that’s entirely beside the point.

The originality of the idea, the soundness of the argument, the depth of the evidence, and whether the author’s qualification deserves your attention in the first place, is what matters. Not which intelligence — biological or artificial — helped produce the article in front of you.

Content creators (Gosh, how I hate that term) already produced competent sounding, but empty, unverified chloroform-in-print when Gemini still was a birth sign, Claude went with Monet, and telling people you asked ChatGPT would have made you the hallucination suspect.

The only difference now is that with AI everybody can produce that slop on an industrial scale. At zero cost, and in near-zero time. And that’s ironically where those AI statements may help you uncover the true credibility of their creators. Just ask two questions.

Two Red Flags That Matter More Than an AI Badge

Q1: Does the author’s claimed production process plausibly match the volume, breadth, and depth of what they publish

Here is a confessedly subjective baseline for comparison purposes. I need about one working day for every 4-minute read. That’s what has crystallized over the more than 100 deep-dive articles that I published on Medium and Substack over the past three years.

OK, maybe I’m particularly slow. But developing an argument, assembling the source corpus, cross-checking references (which means actually reading their methods and results, not just the abstract and conclusions), drafting the whole thing, struggling over how to make the material accessible to lay-readers, redrafting the text, and creating intuitively comprehensible illustrations, takes tiiiiiime. At least for my brain.

So, when your favorite health (or other science) author ceaselessly appears in your inbox with a new deep-dive post every two days or so, you might rightfully suspect that he/she is either spectacularly productive (which actually could be the case), or possibly employs AI.

But if his/her AI statement contradicts how Pangram, Copyleaks, and other detection tools consistently grade his/her texts, that should at least raise a red flag. Not that these tools are perfect. Far from it. But a writer who is willingly inaccurate about AI use may also willingly accept inaccuracies in the subjects they cover.

As a caveat: this is not proof of undisclosed AI use. Old manuscripts, assistants, repackaging, or ghostwriting may be legitimate explanations. But once the AI-free claim becomes incompatible with the output, you are entitled to downgrade your trust.

Q2: Can you independently trace and audit the evidence?

“Independent” means: does the author cite original science papers or only what the press releases, media articles, and other sources (like his/her own books) say about them? And can you access those papers? A mismatch between the message in the data and the message that your favorite author delivers tells you a lot more than a “100% human” claim.

In short: Authors whose work you can check deserve a lot more trust than authors who merely certify how they produced it.

Here is a case in point: One of the most prolific science writers on Medium (active on Substack too), who also happens to be the chief editor of one of Medium’s largest publications, not only “guardrailed” that publication’s submission criteria against involving AI, but also posted about how he competes as a human writer against AI slop.

Of course, we all can choose to believe that that person’s 70+ books published over the past 7 years are totally AI-free. We can also choose to believe that the self-proclaimed “AI-free” badge applies to the torrent of deep-dive articles that person releases weekly – for example: one AI-free deep-dive post every 42 hours on average over the past 30 days (17 posts from 6 August to 5 September).

Or we can believe Pangram, which flagged 9 of the 17 posts as 100% AI-written, and the remaining 8 posts as 82% to 97% AI-written.

Whatever your belief, the AI-statement has become…

A Provenance Ritual Masquerading as a Quality Standard

The morality play around AI writing treats machine involvement as contamination.

• A little spell-checking? Acceptable.

• Grammar correction? Forgiven.

• Rephrasing? May be tolerated.

• Research assistance? Already iffy.

• Paragraph generation? Beginning to look suspicious.

• Generating an argument? Definitely intellectual fraud.

As I argued in the previous section, we should abandon that rationale.

Substack’s solution isn’t exactly going that way. Whether Medium will do better (or not) is something I’m curious to find out during the upcoming Medium Day. Its special track about “how top experts are navigating the AI shift to write more while amplifying their work and protecting their unique voice” is reserved for us Friends of Medium. Why reserved for us? Your guess is as good as mine.

Actually, Medium itself recently made part of my argument:

That’s fine, but we should not stop there. Even perfect knowledge of provenance, as I have laid out in the hypothetical scenarios of this post’s opening, will never tell you whether an idea is original, whether the argument is sound, or whether the piece is actually worth reading.

And this is, ironically, where the much-maligned use of AI can actually pivot from contaminant to cure. That’s not just my belief. I’m practicing to prove it.

Don’t Trust My Process. Check My Work.

AI actually hasn’t shortened my production time (remember the one working day for every 4-minute read). Quite the contrary. I use AI to catch and rein-in the threat of confirmation bias. That often leads me to conclusions that deviate from what I hypothesized initially.

And as proof to readers, I now open every new post’s full source corpus in a shared Gemini Notebook. It is for you to challenge my arguments, interrogate the sources with personal questions that my posts may not have answered, and, just for kicks, try to prove me wrong. If an argument of mine doesn’t survive that test, I want to know.

My message:

In short: Forget the provenance ritual, adopt the quality check. That’s what AI now makes dramatically easier to do.

Now go kick the tires of one of my posts’ interrogation rooms:

Cited References

[1] A. E. Ghareeb et al., “A multi-agent system for automating scientific discovery,” Nature, vol. 655, no. 8122, pp. 497–505, Jul. 2026, doi: 10.1038/s41586–026–10652-y.

[2] J. Gottweis et al., “Accelerating scientific discovery with Co-Scientist,” Nature, vol. 655, no. 8122, pp. 487–496, Jul. 2026, doi: 10.1038/s41586–026–10644-y.

What is new

An ESC conference poster about Xylitol’s association with heart disease mutates into a media frenzy about a mortal sugar substitute.

Why it matters

On the way from a study’s data to what you read in the news, the claims often get inflated at every handoff: from data to researcher to press release officer to journalist. That game of telephone rejigged a simple xylitol-heart-disease association into a definitive cause-and-effect story. Ironically, in this case, an observational study actually strengthened the causal case.

Your takeaway

• A blow-by-blow illustration of how the media’s game of telephone played out in this case (an almost universal template).

• A scientist’s sober-minded suggestion on how to handle that alt-sugar. And what he himself does about it.

The xylitol warning is back—and getting amplified.

At the recent ESC Congress, researchers reported a 47% higher adjusted hazard of major cardiovascular events (MACEs) among people with the highest blood xylitol levels. The ESC’s press release boosted that figure to 57%. The media outlets, including Fox, The Guardian, Newsweek and Medical Daily, swallowed it hook, line and sinker. Others, like Discover Magazine, even elevated the adjusted “association” to an “independent effect.”

So, what is it with Xylitol that enthused the journalists into adding another existential threat to the year-round news Halloween we have been enduring for a while? Simple: it’s almost impossible for you to avoid the stuff. Unless, of course, your diet steers completely clear of ice cream, cookies, candy, chewing gum, and anything processed and labeled “sugar-free”. The 200,000 metric tons produced every year need to go somewhere, don’t they [1]?

Here is what you probably didn’t expect from me:

I don’t think the study should be dismissed based solely on the mantra you often encounter in my posts—association can never prove causation. That still holds true. But this study actually strengthens a causal case that the same research group began building two years ago [2].

In a string of studies, the researchers…:

• linked higher circulating xylitol to cardiovascular events (observational)

• proved that xylitol promotes blood clotting (test-tube experiment)

• demonstrated that human consumption of xylitol drives up blood xylitol and platelet reactivity

The study presented at the ESC congress is a logical next step. Compared to the original observational study, the researchers increased the cohort size fivefold, and the observation time window from 3 years to 6 years and 30 years. (You can download the abstract here, since it has not yet entered the National Library of Medicine)

The (big) fly in the ointment of the new study: there is no full paper to inspect. I contacted both Marco Witkowski, who presented the work in Munich, and first author TM Zheng and asked for the manuscript. I wanted to review it for this article. Neither responded. So, for now, we don’t have the methodological detail required for a proper audit. The evidence ends at the abstract, which is where the media’s storytelling begins.

What the new study actually found.

The researchers analyzed the association between circulating xylitol and MACEs in 17,710 participants of two large general population cohorts: the Canadian CLSA (n=7,651, 6 years follow-up) and the British EPIC-Norfolk (n=10,059, 30 years follow-up). Compared to those in the lowest quartile of blood xylitol, those in the highest quartile had :

• a 47% higher risk in CLSA (hazard ratio 1.47)

• an 18 % higher risk in EPIC (hazard ratio 1.18)

“Adjusting for traditional risk factors” apparently didn’t move the needle on those associations (at least not to a statistically significant extent).

But “traditional risk factors” is about as precise as a politician’s “slightly higher gas prices”: What is “slightly higher” (particularly when the politician waffling about it is a billionaire, and has probably never seen a gas station from the inside for the past 20 years)?

Detailed subgroup or middle-quartile data are also absent. So, we can’t verify what the ESC’s press release scribes seem to know, but the abstract keeps mum about: that those middle quartiles revealed a dose-dependent pattern—the higher the xylitol level, the higher the event rate. Absent the data, we just have to believe that.

Despite these limitations, it is reassuring to see that this latest observational exercise replicates the findings of the earlier 2024 study that was done in a population of high-risk patients. It found a 57% increase in MACE risk. That actually makes me wonder whether the ESC reporters lifted the 57% from there and transplanted it into their latest press release (by accident, I hope).

Replicating study findings is one (very important) thing, but before you throw out all the foodstuff with the X-name on its label, or fall into deep depression over your diet’s decades-long zero-sugar doctrine, we need to answer two more questions:

• Where does the xylitol in your blood actually come from, and

• why should it lead to a heart attack or stroke?

Blood xylitol is not the same as dietary xylitol.

Those 10 volunteers that I mentioned (in the 2024 study) took 30 grams of xylitol after an overnight fast. That’s a dose comparable to what you’d get from a pint of xylitol-sweetened ice cream, a xylitol-sweetened bakery good, or several pieces of xylitol-sweetened candy.

The image below shows how quickly (within 30 minutes) and how high (1000-fold) xylitol levels rise, but also how fast they return to baseline levels (within 6 hours). The estimated elimination half-life is just 15 minutes.

At that time, the authors concluded that the fasting xylitol levels associated with cardiovascular events probably didn’t represent the xylitol people had eaten, but endogenous xylitol production.

None of the observational cohorts delivered any information on dietary xylitol intake. The same goes for the new abstract. Unless and until we can (causally) link circulating xylitol to xylitol consumption, a news article’s claim “Popular artificial sweetener tied to higher heart risk, study finds” is more fiction than fact.

So when a headline tells you that a “popular artificial sweetener” is tied to higher heart risk, it quietly inserts a link the study never measured: that the xylitol people ate produced the blood xylitol associated with their cardiovascular events. Until that link is demonstrated, the headline tells a more causal dietary story than the data actually support

But this caveat doesn’t make the findings disappear.

Fortunately, the answer to the second question has more meat to it.

Why xylitol could plausibly increase cardiovascular risk

It’s all about platelets. Their primary job is to clump together to form clots. They rescue you from bleeding out when your kitchen knife has proved a little too sharp for your cutting skills.

But once your platelets become hyperreactive, they don’t wait for an injury to clot together. They might do it just for kicks. The image below illustrates how xylitol can trigger three key steps that precipitate that hyperactivity disorder.

Hyperreactive platelets can more easily trigger a blood clot, particularly in plaque-lined arteries when atherosclerotic plaque ruptures or erodes.

In a string of studies published in 2024, the research group behind our ESC poster first linked higher circulating xylitol to cardiovascular events (again, an observational study). Then:

• In an ex vivo ‘test-tube experiment,’ they showed that xylitol exposure renders human blood platelets more reactive (i.e., more prone to clotting).

• Then they confirmed that this increase in reactiveness escalates actual thrombotic clot formation in a follow-up experiment with mice.

• And finally, an experiment with 10 human volunteers showed that consumption of typical dietary doses of xylitol can increase its blood concentration a thousand-fold while also dialing up platelet responsiveness

Btw, I dissected the biology, methods, and results in two earlier posts (here and here). Blood xylitol and platelet responsiveness increased not just on average over the group, but in every one of the ten participants.

Those who reached the highest post-meal xylitol concentrations also showed the largest increases in platelet aggregation. The correlations between blood xylitol and the two aggregation responses were 0.71 and 0.74.

Put the experiments together, and a coherent biological chain emerges:

That does not yet complete the overall arc:

But it provides something that is conspicuously absent from many nutritional associations: a plausible mechanism demonstrated at several biological levels, from isolated human platelets to whole blood, a living thrombosis model, and finally humans actually swallowing the substance.

And that is why simply saying, “Well, this is all observational, so association does not prove causation,” does not dispose of the xylitol story.

The more interesting question is how close this collection of evidence actually gets us to causation.

So has dietary xylitol now been shown to cause heart attacks?

Definitely not. But the reflex dismissal that “association does not prove causation” is equally misguided.

Of course we won’t run the risk of encountering this objection in the news. It’s just not click-baity enough.

And don’t expect proof of a causal link between xylitol consumption and MACEs to appear any time soon. We can’t run RCTs that randomize people to xylitol consumption or ‘placebo’ and wait until enough of them have keeled over with a heart attack or stroke to establish a statistically significant difference between the groups. Only the Putin Institute of Ethics would nod our application through.

Our only alternative is to observe free-living individuals in a prospective observational study that allows us to measure long-term dietary xylitol intake. That’s virtually impossible:

We can’t reliably quantify xylitol intake over years when people consume constantly changing mixtures of foods that contain the stuff, often without knowing which sweeteners they contain, let alone how many grams they consumed.

So, no: the evidentiary chain “eating xylitol → heart attack or stroke” hasn’t been proven. But we could also easily reject a potential xylitol industry dismissal that “xylitol just happens to correlate with cardiovascular disease”.

Here is how:

Higher circulating xylitol repeatedly predicts cardiovascular events, while experimentally increasing xylitol exposure makes human platelets more thrombosis-prone (the major cause of heart attacks and strokes).

That, of course, sounds a lot less sexy than “eating xylitol causes heart attacks.”

How that distinction started disappearing on the way from the congress presentation to the media articles is illustrative of how you can end up manipulated or misinformed.

The telephone game: How ‘associated with’ became ‘effect’

The scientific abstract says that circulating xylitol was “independently associated with incident MACE risk” after it was “adjusted for traditional cardiovascular risk factors”.

That is standard epidemiological lingo. “Independent” here means independent of all the measured factors (the so-called covariates, in this case the “traditional risk factors”, whatever that means). But you can’t account for what you didn’t measure. It does not mean that xylitol has been experimentally shown to have an independent causal effect on cardiovascular events.

So, the authors’ conclusion that

is causal-leaning language that the new cohorts alone cannot establish. Because “promoting” something means causing it, and “causal” is not what the observational cohorts really demonstrate.

Then the study entered the ESC publicity machinery.

The Canadian hazard ratio of 1.47 suddenly became a 57% higher risk in the official press release.

As we saw earlier, 1.57 happens to be the exact adjusted hazard ratio reported in the 2024 validation cohort. Maybe that’s an honest error. Without the full paper or presentation slides, I cannot tell you which explanation is correct.

The ESC press release adds the claim that, taken together with the middle quartiles, the two cohorts showed a “dose-dependent relationship”: the higher the xylitol level, the higher the cardiovascular event rate. If the complete quartile analyses show a clean gradient, it would strengthen the case considerably.

But the abstract doesn’t reveal that.

My point: the press release starts giving us evidence that the abstract itself does not allow us to inspect.

Then comes the expert commentary. The ESC press release quotes Professor Dan Atar, speaking for the ESC Communication Committee, as saying that “it appears that xylitol has a negative long-term effect on cardiovascular events in the general population.”

The abstract says “associated with”; Atar says “effect”. A linguistic upgrade. To Atar’s credit, he immediately adds that, as with all observational studies, causality is “difficult to infer”. But by then “causality” has already entered the narrative.

And that’s exactly what the press feeds on. Discover Magazine repeats the 57% figure and explains that the multivariable adjustment.

“… suggests that xylitol has an independent effect on the increased risk, …”

No, it doesn’t. Multivariable adjustment does not establish that, as I explained above.

And this is how hype often works in science communication. Ignore for a moment the extreme cases of journalists fabricating a ridiculous headline or scientists presenting fraudulent studies. More often, every handoff moves the narrative just enough so that, as in our example:

• “Associated with” becomes “effect”.

• HR 1.47 becomes 57% higher risk.

• A dose-response pattern, not mentioned in the abstract, worms itself into the press release. Not because its scribe picked it up during the presentation. I read the abstract and the press release the day before the actual presentation.

Each individual step in that telephone game looks small enough to shrug off. But put them together, and the message arriving in your news feed is more dramatic than the scientific abstract from which it started.

Now don’t read this as “the xylitol warning is nonsense”.

What I would do with xylitol now.

• A 30-g xylitol challenge can raise circulating concentrations roughly 1,000-fold above fasting levels.

• Observational evidence reproducibly links circulating xylitol with cardiovascular disease.

• Experimental studies show a plausible prothrombotic mechanism.

Taken together, avoiding xylitol, or at least choosing not to consume large quantities of it, sounds to me like a reasonable precaution. At least while the clinical question remains unresolved. You may think differently, of course.

If you want the detailed walkthrough of the remarkable 2024 experiments, I dissected the original paper here:

I subsequently examined the scientific dispute around that paper—specifically Peter Attia’s argument that the findings could be safely ignored:

Share

Subscribe now

Cited References

[1] Y. Delgado Arcaño, O. D. Valmaña García, D. Mandelli, W. A. Carvalho, and L. A. Magalhães Pontes, “Xylitol: A review on the progress and challenges of its production by chemical route,” Catal. Today, vol. 344, pp. 2–14, 2020, doi: 10.1016/j.cattod.2018.07.060.

[2] M. Witkowski et al., “Xylitol is prothrombotic and associated with cardiovascular risk.,” Eur. Heart J., vol. 00, no. 0, pp. 1–14, 2024, doi: 10.1093/eurheartj/ehae244.

What is New?

Observational studies report fewer Alzheimer diagnoses among statin users. Yet autopsy studies find no less Alzheimer pathology in their brains—and, surprisingly, all studies find little evidence of less vascular dementia.

How can all three be true?

That contradiction is the mystery we are going to solve.

Why it Matters

The media tends to turn these findings into two popular shortcuts: “statins starve the brain of cholesterol” vs. “statins prevent Alzheimer’s”. Accept either at face value, and you may end up making a long-term treatment decision based on a half-truth.

Your Takeaway

A forensic autopsy of the statin–dementia evidence—and why the numbers may be telling us the wrong story about Alzheimer’s and vascular dementia.

Plus: A downloadable seven-question primer that can turn your next “To Statin or Not to Statin?” appointment into a better-informed shared-decision-making with your doctor.

Table of Content

• Two Half-Truths, One Bad Question

• Half-Truth #1: “The Brain Needs Cholesterol, So Statins Must Starve It”

• Half-Truth #2: “The Blood-Brain Barrier Keeps Statins’ Effects Out”

• The Statin–Dementia Paradox

• Three Ways the Evidence Gets Scrambled

• The Verdict: A Vascular-Threshold Shield Is Plausible, Not Proven

• Seven Questions for Your Next Statin Conversation

• Paid-subscriber bonus: Take the investigation one step further

• Cited References

Two Half-Truths, One Bad Question

The Indictment:

Your brain, with its 2% of your body mass, stores 20% of your body’s cholesterol. Neurons need the stuff for their myelin sheaths that wrap around them. Therefore, lowering cholesterol with a statin means denuding your cognitive wiring of an essential commodity.

The Rebuttal:

Statins can’t harm the brain because the blood-brain barrier (BBB) completely separates the systemic from the cerebral cholesterol pool.

Both arguments begin with a simple truth. Both end with a conclusion that is as persuasive as it is wrong. Because both arguments violate Einstein’s simplicity tenet:

It’s the “but not simpler” part which escapes each argument’s proponents (which also propels them into the good company of politicians, lawyers, and blithering social media influencers).

Both arguments peddle an answer to the question that might have begun bothering you (particularly if you are of an age where you can remember rotary phones, bell-bottom jeans, and beehive haircuts):

Can statins promote dementia or protect from it?

In the previous post we saw that trial results span the entire spectrum from “nasty” to “nice”:

• Nasty: Case studies report reversible cognitive symptoms, but leave their true frequency unknown.

• Neutral: RCTs fail to show substantial cognitive harm on average, but can’t rule out the existence of vulnerable subgroups.

• Nice: Observational studies show a protective association, but cannot prove that statins actually prevent Alzheimer’s disease.

Buried in that binary choice of protection vs. harm is a scenario nobody seems to acknowledge:

Statins may fail to alter Alzheimer pathology and still help preserve cognition—by preventing the vascular brain injury that pushes a vulnerable brain across the threshold of clinically manifest dementia.

Whether that could be the case was the concluding question of my preceding post. In this post, we will try to answer that question. We will take a look “under the hood” to:

• expose the flaws of the simplified arguments

• empower you for better-informed shared-decision-making with your doctor

Before we go there…

A disclaimer: This post is for informational and educational purposes only. It is not intended to be a substitute for medical advice, diagnosis, or treatment. As an epidemiologist and health scientist, I do not recommend for or against taking a statin. That’s your and your doctor’s decision. This post is my attempt at enabling you to have a better-informed shared-decision-making conversation with your doctor.

Half-Truth #1: “The Brain Needs Cholesterol, So Statins Must Starve It”

The human brain is the most cholesterol-rich organ [1]. It packs 20% to 25% of the body’s total cholesterol pool into a mere 2% of body mass (the remaining 75% to 80% of cholesterol make up barely 1% of the body mass that contains it).

70% to 80% of brain cholesterol is wrapped around its cognitive wiring (its nerve cells’ axons), much like a power cable’s insulator sheaths wrap around its wires. The name of the insulators’ biological equivalent is “myelin”. The better the myelin insulation, the faster the nerve signals travel along the axons, the more efficient the communication between cells, and the quicker your repartee to a smart-ass’s wise-crack.

OK, I made up the latter, but I suspect that women may be better equipped in the myelin department than men. My wife, for example, shoots down any smart-ass’s taunt within milliseconds, whereas my brain needs a day or two to throw out a snarky comeback (by which time the offending smart-ass has disappeared).

Because myelin and the nerve cells’ synaptic membranes depend on cholesterol to maintain their structural integrity and fluidity, the simple idea that cholesterol-lowering drugs could “starve” the brain of this essential commodity “has garnered significant attention in recent research” [2]. That’s where the “but not simpler” part enters the stage.

One Body, Two Cholesterol Pools

The flaw of the “brain starvation” premise is the assumption that cholesterol in the systemic circulation can freely migrate into the cerebral circulation, and vice versa. But Mother Nature has installed a barrier between those two: the blood-brain barrier (BBB). It protects the brain against toxins and pathogens that might create havoc in our heads. Under normal conditions, lipoproteins circulating in the bloodstream (such as LDL, VLDL, and chylomicrons) are simply too large to pass through the BBB’s wall of endothelial cells.

Fortunately for the wine and beer aficionados among us (I cop to my membership in that group), the same cannot be said about alcohol. Ethanol is small enough to get a free rapid-access backstage pass to the brain. Nicotine and caffeine are equally privileged.

But the brain’s cholesterol pool is structurally and functionally independent of the cholesterol circulating in your bloodstream. Even aggressively lowering blood lipids using statins and PCSK9 inhibitor drugs (I described them in an earlier post about the lipid-lowering ladder of drugs) has no corresponding direct effect on the brain’s cholesterol pool [3]. It certainly doesn’t “drain” or “denude” the brain’s lipid supply.

Noticed that I used the word “direct”? An indirect effect comes into play, and that’s exactly what trips up the “impenetrable BBB” objection.

Half-Truth #2: “The Blood-Brain Barrier Keeps Statins’ Effects Out”

The claim in brief: The BBB prevents cholesterol from entering the brain → the brain relies on its own cholesterol production → Statin-based cholesterol reduction can’t starve the brain of the stuff.

That argument ignores that (a) statins do enter the brain (to various degrees, depending on the type of statin), and (b) while the BBB effectively separates the two cholesterol “hoods”, it does not make them metabolically oblivious to one another. To understand their crosstalk, we first need to explore their metabolisms’ remarkable differences.

The Brain’s Slow-Motion Cholesterol Economy

The image below saves you a thousand-word reading.

In terms of turnover and half-life, the two compartments reside at opposite ends of the spectrum. The stable myelin sheaths contain 70%-80% of the brain’s cholesterol, where its half-life is estimated at 5 years [4]. The remaining 20%-30% still have a half-life of several months. That makes the brain’s cholesterol molecules virtual Methuselahs compared to their peers in the systemic circulation whose half-life is measured in days [5].

Nevertheless, cholesterol molecules can get compromised, and since the brain can’t replenish them from circulating cholesterol carriers (e.g., LDL, vLDL, chylomicrons), it needs to be self-reliant. Here is how it works:

The brain’s astrocytes continuously synthesize cholesterol, ship it in ApoE-containing lipoprotein particles to neurons and other (glial) cells, where it feeds into a continuous turnover process [6]. No cholesterol uptake from blood-cruising cholesterol-carrying LDL required.

Deep-dive Box: Neurons are the information-processing and transmission cells; astrocytes are their metabolic and structural support system; and other glial cells (e.g., oligodendrocytes and microglia) take care of maintenance, cleanup, and immune surveillance. All depend on cholesterol as a critical raw material, though they use it in different ways.

Now don’t think of the “continuous turnover” as a neat “just-in-time”-type cholesterol production to simply replace the molecules that have become compromised. Our biology rarely works like that. Mother Nature adopted a far less error-prone policy that you may know from your local bakery:

Rather than trying to exactly replace the buns that have been sold (or got somehow damaged), the baker will continuously produce new batches to keep the shelves sufficiently stocked to meet fluctuating demands. So, whatever happens, no customer leaves the shop unsatisfied, and at the end of the day, the baker can always donate the leftover buns to the neighborhood nursing home.

The difference in the brain is that the “leftover buns” need to be converted to a molecule that can pass through the BBB. The name of that molecule is 24S-hydroxycholesterol (in case you wanna look up the process in relevant study papers). For the rest of this post, let’s stick to its other, better-fitting name: cerebrosterol.

Deep-dive Box: Here is a number that tells you how efficient the brain’s recycling process really is:

While the average human brain holds roughly 35 grams of cholesterol [7], its daily clearance is about 3-7 milligrams only (i.e., roughly 0.02%) [8]. Compare that to systemic cholesterol, which is cleared at 20-30 times that rate (0.7% - 1%).

Cerebrosterol (the ‘leftover buns’) can be measured in blood samples and therefore tell us…

What happens in the brain when we feed the body a statin?

All statins can breach the BBB. How well they get through depends on the class and even the type of statin. Pharmacologically, we divide them into two classes depending on their solubility: Hydrophilic (water-soluble) vs. lipophilic (lipid-soluble). Since cell membranes (including those that make up the BBB) consist mostly of lipids, lipophilic statins have it much easier at getting past the brain’s BBB bouncer than hydrophilic statins. I discussed this issue in the context of statins’ muscle-pain side effects in a previous post.

• Lipophilic Statins (e.g., Simvastatin, Lovastatin, Atorvastatin): Simvastatin has one of the highest BBB penetration capacities, followed by atorvastatin [4]. They all can passively diffuse through the BBB.

• Hydrophilic Statins (Pravastatin, Rosuvastatin): They depend on active carrier-mediated transport, which makes these statins far more “hepatoselective” (liver-targeted) with minimal brain exposure.

Same Drug, Different Feedback Loop

While triggering the same biochemical pathways, statins’ net effects on cholesterol metabolism depend entirely on the “hood” in which they do their statin thing. The image below illustrates it.

In both the liver and brain, statins dial down de novo synthesis of cholesterol, trigger the cellular low-cholesterol alarm (via up-regulating SREBP-2), and then continue to have vastly different effects on organ cholesterol.

• The liver responds by sucking up more cholesterol-carrying LDL vehicles from the systemic circulation. The net (and desired) effect is a reduction in circulating LDL.

• The brain responds by ramping up the cholesterol synthesis machinery and optimizing cholesterol recycling and uptake (by the cells that need it). The net effect is less cholesterol being expelled via the cerebrosterol pathway, and consequently maintenance of brain cholesterol [5] [4]

So, how does the latter show up in the systemic circulation? As a reduction in cerebrosterol.

Cerebrosterol: A Window, Not a Fuel Gauge

In some, but not all studies, aside from a reduction in cerebrosterol, the cerebrosterol/cholesterol ratio is also reduced. However, both are too crude a measure to judge whether the brain might actually suffer damages from “cholesterol starvation”. If there was a physiologically relevant damage (relevant to cognitive performance), we would expect to see it in the RCTs that I mentioned before. Only we don’t, as I explained in the previous post.

A longer-term experiment by Evans et. al offers a hypothesis that unifies the conflicting evidence from case reports, RCTs, and observational studies [9]:

Over a period of 2 years, the researchers monitored the cerebrosterol response of statin-naive adults to simvastatin and atorvastatin. That study is remarkable for several reasons. First, it measured its markers not in peripheral blood but directly in the cerebrospinal fluid (which offers a much more direct view of the brain circulation). Second, it also monitored lathosterol, a direct marker of cholesterol synthesis. Third, no other study has replicated this design (as of 22nd August 2026), which severely limits its generalizability.

The results: cerebrosterol and lathosterol reached a nadir around 6–7 months. Both subsequently recovered, reaching baseline at about 15 months, overshot it at around the two-year mark, and then tended back toward baseline. That pattern looks less like progressive depletion and more like a perturbed homeostatic system adjusting to the intervention.

An initial drop might well play out in some individuals as the cognitive impairments they experience initially. The return to baseline after first under- then overshooting it could unify the cognition-protective results of observational studies with this post’s central question.

The Statin–Dementia Paradox

You are probably used to whiplash-inducing contradictory media messages about practically any health issue. The statin-dementia subject is no exception.

No Less Alzheimer Pathology

If statins had any effect on Alzheimer’s, we would expect the brains of statin-using AD patients to have fewer amyloid plaques, fewer neurofibrillary tangles (tau pathology), or at least a lower Braak stage (a neuropathological scoring system of tau/tangle pathology) than the brains of statin non-users.

That’s what Jana Crum and her lab hypothesized in 2018. They then analyzed autopsy data of 1,816 deceased people who had both pre-mortem clinically diagnosed and postmortem autopsy-confirmed Alzheimer’s disease [10]:

• Statin non-users: 1,558 patients

• Statin users: 258 patients

The differences between these two groups? Nada. Not on a single marker did these brains show any significant difference:

• Not on Plaque density (Amyloid CERAD score)

• Not on Tangle Severity (Braak stage)

• Not on Tau and Amyloid totals

• Not even on APOE4 genetic influence: Statins showed no effect on pathology in both APOE4 carriers and non-carriers alike.

Contrast that with the more cheerful results of another study published in the same year:

Yet Less Diagnosed Alzheimer’s?

Chu and colleagues meta-analyzed 25 prospective cohort studies of roughly 3 million patients who were cognitively healthy at baseline, following them for 7 years on average [11]:

• Alzheimer’s Disease: statin use associated with a 28.1% risk reduction.

• Vascular Dementia (VaD): no statistically significant association between statin use and the risk of developing vascular dementia

That presents a paradox: since statins supposedly reduce atherosclerosis, improve endothelial function, and preserve blood flow, we would expect VaD (not AD) to be the most likely beneficiary of statin therapy, wouldn’t we?

Just in case you are unfamiliar with the differences between AD and VaD:

What I failed to mention about the meta-analysis: it contained only 3 VaD studies with barely 6,000 subjects. It may well have lacked power to discover a statistically significant effect. But since then

• (in 2020) a larger meta-analysis (9 million subjects) [12]

• (in 2025) a systematic review of close to 7 million people [13].

found no statistically significant effect of statins on VaD either.

HypeWatch box: In case you are prone to put my statements under the microscope, and actually read the papers that I reference, you’ll be tempted to throw the following quote of the 2025 review’s abstract at me:

“Statin use linked to lower risks of all-dementia, AD, and VaD.”

Even researchers aren’t entirely immune to the seduction of knitting a little hype into their abstracts (often the only thing a lay reader bothers to consume). When you analyze the methods and results section, you’ll notice that the authors looked at two statistical metrics: the odds ratio (OR) and the hazard ratio (HR). The former is a snapshot metric; the latter is a time-to-event metric that we epidemiologists prefer as the more robust one for longer-term cohort data.

The HR turned out to be non-significant. The OR was significant, suggesting a 29% risk reduction, but it came out of only 2 studies and didn’t survive sensitivity analysis. I’ll get back to that point in a minute.

Long story short: How do we reconcile

• the lack of a statin effect on the physical manifestation of AD,

• an association of statins with reduced functional consequences (cognitive performance leading to AD diagnosis), and

• a lack of effect on VaD?

After all, the established clinical benefit of statins is vascular. They lower ApoB-containing lipoproteins (e.g., LDL & Co.), reduce atherosclerotic cardiovascular events, and reduce ischemic stroke risk in certain patient subgroups.

So why would their apparent dementia benefit show up more clearly under the label “Alzheimer’s disease” than under “vascular dementia”?

For one reason: Over-simplification. It mucks up our data all the way from pathology to diagnostics to the statistics applied to them. How much? You’ll be surprised to learn that the truth may be just the opposite of what the studies suggest. And no, that’s not just me saying this.

Three Ways the Evidence Gets Scrambled

Contrary to what the media and some health gurus might have impressed on you, AD’s physical hallmarks (amyloid plaques and tau tangles) are not a sentence of inevitable cognitive decline.

Again, a post-mortem study sets the record straight.

Pathology: Plaques Are Not Dementia

Perez-Nievas and colleagues compared three groups of deceased older adults [14]: cognitively healthy controls, cognitively healthy “mismatches” (who had significant AD pathology but no clinical dementia), and individuals with pre-mortem diagnosed AD dementia.

They found that amyloid-β plaque burden and the number of neurofibrillary tangles did not significantly differ between the cognitively normal “mismatches” and the demented AD patients.

In other words, looking at an individual brain, you won’t be able to tell just from the “gum” in there whether its owner is cognitively normal or can’t remember what they had for breakfast. Reassuringly, the “mismatches” aren’t a rare exception. Roughly one-in-three older adults without cognitive impairment until death met the pathological criteria for AD on autopsy [15].

A 2022 study by Wagner et al. confirms these findings [16]. More importantly, it adds an enlightening wrinkle to the statin-dementia story that helps us unify our (well, the media’s) seemingly contradictory messages into a neat theory.

As we age, many of us accumulate silent microvascular injuries in our brains (even if we do not suffer a visible/noticeable stroke):

• micro-strokes (not noticeable)

• arteriolosclerosis (hardening of the tiny blood vessels in the brain)

• atherosclerosis (fatty plaques in the larger brain vessels)

• inflammation (in brain blood vessels)

These damages accumulate independent of any AD amyloid and tau tangles.

So, even if you possess the “cellular software” capacity to tolerate Alzheimer’s pathology (and you very likely have quite some capacity as the autopsy studies revealed), experiencing a microscopic stroke or progressive arteriolosclerosis (both affecting one-in-three older adults) may just push you over the tipping point of your cognitive reserve, and start driving you into dementia territory.

Now look again at the above list of brain vascular injuries. Notice how some of these are precisely the events and conditions statins are supposed to dial down? That’s why the idea that statins, which have no effect on AD build-up, might just keep you at a safe distance from that tipping point. That’s the vascular-threshold shield hypothesis.

Diagnosis: Mixed Brains, Simple Labels

The studies have been based on a diagnostic foundation that crumbles on close scrutiny.

Most dementia patients suffer from a mixed pathology of Alzheimer’s and VaD. The resulting symptom overlap complicates diagnosis. Up to 87% of patients who had been diagnosed with vascular dementia alone showed concomitant AD on autopsy [17]. “Pure” vascular dementia without any co-existing Alzheimer’s pathology is actually rare.

That degree of diagnostic inaccuracy can have a major effect on study results [18]:

Studies that rely on billing codes (ICD codes) systematically show a highly significant, protective effect for statins.

Studies with large numbers of participants typically rely on health records and billing codes because personal differential assessment of dementia types is time-consuming and requires special training.

That’s why, in studies that apply in-person clinician evaluations, the protective effect of statins either vanishes or pulls back close to the null (no effect). These studies get much closer to the pathological reality of their study participants, but, unfortunately, the impressive large-number analyses with their mucked-up diagnostics dominate the “market”.

In these studies, systematic misclassification artificially empties the VaD category of the very patients in whom statins could demonstrate their greatest preventative benefit:

If you siphon off the slow, lipid-sensitive (hence also statin-sensitive), and difficult-to-diagnose cases of atherosclerotic SVD into the Alzheimer’s pool, you are left with a highly selected, underpowered VaD pool.

And since the AD pool is “enriched” with patients whose slow-onset atherosclerotic SVD remains undetected:

How far may these misclassifications swing the study results away from the truth?

Statistics: The Vascular-Dementia Bucket Is Too Small—and Too Mixed

I already mentioned the severely underpowered Chu et al study. With just 422 VaD cases spread over 3 studies, the result shouldn’t come as a surprise. It says, in essence:

Statins could either reduce the risk for VaD by 38%, or actually increase it by 65%.

That’s how you paraphrase the 95% confidence interval (95% CI) from 0.62 to 1.652 (a 95% CI that includes 1.0 always indicates a statistically non-significant result).

Mind you, this result is NOT evidence for statins having no effect on VaD. It’s evidence for an imprecise estimate that sits on the “null effect”, because there were too few data points (participants).

Throwing several of these underpowered studies into one meta-analytic blender doesn’t resolve the issue. But it gives researchers the opportunity to answer another question: “Are the studies at least showing similar effects and effect sizes?” Statisticians call this, fittingly, a test for heterogeneity.

I won’t bore you with the details, so here is the summary: Where researchers had at least enough VaD studies to run these tests, the results were pretty clear: No effect of statins on VaD [13].

Here is why that shouldn’t come as a surprise.

Not All Vascular Brain Injury Is Statin-Sensitive

The comparatively few VaD patients in our analyses hide one more “over-simplification” problem. VaD serves as a clinical bucket that fails to differentiate between arteriolosclerotic and atherosclerotic disease [17]:

Arteriolosclerotic small vessel disease (SVD)

Pressure-driven hardening of the small intracerebral vessels is “the most common vascular contributor to dementia”, providing the “most robust substrates of cognitive impairment.” Statins do not act on these. Statins act on …

Atherosclerotic SVD and large-vessel disease

Atherosclerotic plaques and cardioembolic disease are the primary causes of large, visible strokes (macroinfarctions). These events are driven by circulating lipids like LDL-cholesterol, on which statins do have an impact.

So, within the small numbers of VaD patients, only the smaller subgroup of those with lipid-sensitive atherosclerotic disease may benefit from statins. That downsizes the already insufficient pool of potentially reactive VaD patients even further.

Now, let’s put it all together.

The Verdict: A Vascular-Threshold Shield Is Plausible, Not Proven

Given the well-documented flaws in our chain of data, from pathology to diagnostics to statistics, all those meta-analyses that impressed you with their sheer numbers of millions of included patients may actually present the exact opposite of reality: Not Alzheimer’s benefits from statins, but VaD. Or maybe, statins do diddly-squat to either dementia career.

How will we know which is which?

What STAREE and PREVENTABLE Can Still Tell Us

Two currently running trials (STAREE and PREVENTABLE) have been specifically designed to eliminate the diagnostic flaws and answer that question (I introduced them in my previous post). And reassuringly, both study groups steered clear of pharma funding. STAREE results may come out any time now; PREVENTABLE results probably in 2027.

But whatever the outcome, whether statins alter the progression of AD or VaD, the vascular-threshold shield hypothesis survives. Even if it turns out that statins have no effect on either type of dementia. After all, it’s your lifestyle choices that may either “put the pedal to the metal” for the progression of atherosclerotic and hypertensive vascular disease, or hit the brakes on both. I discussed these effects in an earlier post.

Seven Questions for Your Next Statin Conversation

What does all that mean for your next “to statin or not to statin” conversation with your doctor?

Don’t ask your doctor whether statins are “good or bad for the brain.” I turned the evidence into seven questions that make that conversation more useful. Download it in the bonus content section for paid subscribers.

There you’ll also find the link to the 70-sources interrogation room.

By now, I hope, you are with me on the dangers of over-simplification.

Ironically, even Einstein’s quote is an over-simplification of what he really said [20]:

“Surrendering the adequate representation” of many data is exactly what ails the indictment and rebuttal pitches in the statins vs. dementia debate.

Paid-subscriber bonus: Take the investigation one step further

Reading the evidence is one thing. Using it to make a better decision is another.

Download the 7-Question Statin Shared-Decision Primer
Seven questions—and the reason for asking each—to take into your next “To Statin or Not to Statin?” conversation with your doctor.

Enter the DoubleCheckDoc Evidence Interrogation Room
Cross-examine the curated studies behind this investigation yourself. Ask about statins, cognition, brain cholesterol, Alzheimer’s disease, vascular dementia, study methods, conflicting findings, or any claim made in this post. The notebook is source-locked: it answers only from the evidence included in the room.

Cited References

[1] I. Björkhem and S. Meaney, “Brain cholesterol: long secret life behind a barrier,” Arterioscler. Thromb. Vasc. Biol., vol. 24, no. 5, pp. 806–815, May 2004, doi: 10.1161/01.ATV.0000120374.59826.1b.

[2] K. Li, Y. Li, X. Jiang, and Y. Zhu, “The effects of cholesterol-lowering drugs on neurocognitive function: systematic review and meta analysis,” Front. Neurol., vol. 17, Feb. 2026, doi: 10.3389/fneur.2026.1696228.

[3] R. Kazibwe, R. Rikhi, S. Mirzai, N. P. Ashburn, C. L. Schaich, and M. Shapiro, “Do Statins Affect Cognitive Health? A Narrative Review and Critical Analysis of the Evidence,” Curr. Atheroscler. Rep., vol. 27, no. 1, p. 2, 2025, doi: 10.1007/s11883-024-01255-x.

[4] I. Savulescu-Fiedler et al., “The Cross-Talk Between the Peripheral and Brain Cholesterol Metabolisms,” Curr. Issues Mol. Biol., vol. 47, no. 2, p. 115, Feb. 2025, doi: 10.3390/cimb47020115.

[5] J. Zhang and Q. Liu, “Cholesterol metabolism and homeostasis in the brain,” Protein Cell, vol. 6, no. 4, pp. 254–264, Apr. 2015, doi: 10.1007/s13238-014-0131-3.

[6] L. Delac and S. Maioli, “Cholesterol metabolism and oxysterols in neurodegenerative disorders: Spotlight on Alzheimer’s disease,” Curr. Opin. Endocr. Metab. Res., vol. 41, p. 100590, Dec. 2025, doi: 10.1016/j.coemr.2025.100590.

[7] M. Orth and S. Bellosta, “Cholesterol: Its Regulation and Role in Central Nervous System Disorders,” Cholesterol, vol. 2012, p. 292598, 2012, doi: 10.1155/2012/292598.

[8] M. Wu et al., “Connecting the Dots Between Hypercholesterolemia and Alzheimer’s Disease: A Potential Mechanism Based on 27-Hydroxycholesterol,” Front. Neurosci., vol. 16, Apr. 2022, doi: 10.3389/fnins.2022.842814.

[9] B. A. Evans et al., “Long-Term Statin Therapy and CSF Cholesterol Levels: Implications for Alzheimer’s Disease,” Dement. Geriatr. Cogn. Disord., vol. 27, no. 6, pp. 519–524, May 2009, doi: 10.1159/000221835.

[10] J. Crum, J. Wilson, and M. Sabbagh, “Does taking statins affect the pathological burden in autopsy-confirmed Alzheimer’s dementia?,” Alzheimers Res. Ther., vol. 10, p. 104, Oct. 2018, doi: 10.1186/s13195-018-0430-7.

[11] C.-S. Chu et al., “Use of statins and the risk of dementia and mild cognitive impairment: A systematic review and meta-analysis,” Sci. Rep., vol. 8, no. 1, p. 5804, Apr. 2018, doi: 10.1038/s41598-018-24248-8.

[12] T. N. Poly et al., “Association between Use of Statin and Risk of Dementia: A Meta-Analysis of Observational Studies,” Neuroepidemiology, vol. 54, no. 3, pp. 214–226, May 2020, doi: 10.1159/000503105.

[13] F. L. Westphal Filho et al., “Statin use and dementia risk: A systematic review and updated meta-analysis,” Alzheimers Dement. Transl. Res. Clin. Interv., vol. 11, no. 1, p. e70039, 2025, doi: 10.1002/trc2.70039.

[14] B. G. Perez-Nievas et al., “Dissecting phenotypic traits linked to human resilience to Alzheimer’s pathology,” Brain, vol. 136, no. 8, pp. 2510–2526, Aug. 2013, doi: 10.1093/brain/awt171.

[15] J. A. Schneider, N. T. Aggarwal, L. Barnes, P. Boyle, and D. A. Bennett, “The Neuropathology of Older Persons with and Without Dementia from Community versus Clinic Cohorts,” J. Alzheimer’s Dis., vol. 18, no. 3, pp. 691–701, Oct. 2009, doi: 10.3233/JAD-2009-1227.

[16] M. Wagner et al., “Quantifying longitudinal cognitive resilience to Alzheimer’s disease and other neuropathologies,” Alzheimers Dement. J. Alzheimers Assoc., vol. 18, no. 11, pp. 2252–2261, Nov. 2022, doi: 10.1002/alz.12576.

[17] R. N. Kalaria, “Neuropathological diagnosis of vascular cognitive impairment and vascular dementia with implications for Alzheimer’s disease,” Acta Neuropathol. (Berl.), vol. 131, pp. 659–685, 2016, doi: 10.1007/s00401-016-1571-z.

[18] W. B. Wong, V. W. Lin, D. Boudreau, and E. B. Devine, “Statins in the prevention of dementia and Alzheimer’s disease: A meta‐analysis of observational studies and an assessment of confounding,” 2013, doi: 10.1002/pds.3381.

[19] B. G. Schultz, D. K. Patten, and D. J. Berlau, “The role of statins in both cognitive impairment and protection against dementia: a tale of two mechanisms,” Transl. Neurodegener., vol. 7, p. 5, Feb. 2018, doi: 10.1186/s40035-018-0110-3.

[20] A. Robinson, “Did Einstein really say that?,” Nature, vol. 557, no. 7703, pp. 30–30, 2018.

You normally give me 15–20 minutes of your time to read something I have written. But sometimes I can’t shake the suspicion that I’m living entirely in my own head.

That’s why this time, I would like to reverse the direction.

Give me about five minutes — not to read what I think matters, but to tell me where health decisions are difficult for you and what you would really want me to help you solve. In return I’ll give you something that, ironically, might reduce your need for my help.

I am developing the next stage of Double Check Doc around one recurring problem (at least I think it is): what happens when laboratory results, imaging, risk estimates, physician recommendations and your own priorities do not all point in the same direction.

But before I build anything, I want to know whether I am solving the problems you actually have rather than the ones I assume you have.

Take the 5-Minute Survey

As I said, I want to give you something useful in return.

Everyone who completes the survey gets immediate access to my new Double Check Doc Interrogation Guide: Seven Prompts for Testing What the 2026 Cholesterol Evidence Actually Says About Your Situation, together with access to the source-grounded Statin Evidence Notebook.

The guide is not another list of questions such as “Should I take a statin?” Its prompt templates force the Notebook to examine the relevant population, outcome, evidence type, comparisons and uncertainty before drawing a conclusion. It also shows you how to make the AI challenge its own first answer.

In other words, you give me five minutes of information about what you need. I give you a practical tool for interrogating medical evidence more rigorously yourself.

Oh, and you’ll get three prompts that force the Notebook to Double-Check its own answers (of course, you can use them in your favourite AI too).

Once enough of you have responded, I will report back on what you collectively told me about difficult healthcare decisions and where the current system is failing you.

Take the 5-Minute Survey

As promised, here is your complimentary

Double Check Doc Interrogation Guide: Seven Prompts for Testing What the 2026 Cholesterol Evidence Actually Says About Your Situation.

Download the Interrogation Guide

Open the 2026 Statin Evidence Notebook

The guide shows you how to interrogate the source-grounded notebook more rigorously in the context of your personal profile:

• how to separate what the evidence directly demonstrates from inference,

• compare conflicting studies,

• examine absolute rather than merely relative effects,

• and identify what the available sources cannot answer.

Use any or all of the seven prompt templates to interrogate the evidence:

These templates force the Notebook to do more than produce a plausible answer: they specify the population, outcome, evidence type, comparison, and uncertainty that must be examined before any conclusion is drawn.

That makes the response far more useful for personal decision-making than a generic question such as “Should I take a statin?”, because it turns the AI from an answer generator into a structured evidence interrogator.

Then make it Double-Check its own answers:

I’ll report back to you

Once enough readers have responded to the survey, I will also report back on what you collectively told me about difficult healthcare decisions and where the current system is failing you.

Thanks again for contributing.

Lutz

Subscribe now

Statins either protect your brain from Alzheimer’s, propel it toward the disease, or have no effect. Either message may show up in your media feed, depending on which echo chamber you are residing in.

Why It Matters

Each side in the statin vs. dementia debate presents seemingly persuasive evidence for their argument. But relying exclusively either on individual case reports of brain fog, on chronicles of protective effects in large-scale observational studies, or on null-results from randomized controlled trials may turn a headline into a bad decision.

Your Takeaway:

In this episode, you will learn what the FDA’s drug label really says, why randomized trials come with reassurance but also with a blind spot, and how observational studies involving millions of patients confuse association with causation. You’ll walk away with the ability to

• spot the media’s flawed arguments

• have a better-informed shared-decision-making conversation with your doctor

• run your own questions in a dedicated DoubleCheck Interrogation Room (a Gemini Notebook containing the vetted papers behind this series), rather than relying on generic web summaries or unreferenced AI answers.

Disclaimer: This post is for informational and educational purposes only. As an epidemiologist and health scientist, I do not recommend for or against taking a statin. That’s your and your doctor’s decision. This post is my attempt at enabling you to have a better-informed shared-decision-making conversation with your doctor.

Table of Contents

• “Statins Cause/Prevent Dementia!”: Before You Pick a Side, Define the Outcome

• Can Statins Cause Short-Term Cognitive Symptoms?

• Randomized Trials: Reassuring on Average, Incomplete for Individuals

• When the Brain Changes but Performance Does Not

• The Funding Footnote You Should Never Skip

• Do Statin Users Develop More or Less Dementia Over the Long Term?

• The Bias Minefield Between Association and Causation

• The Studies You Never Get to Read

• The Trials Built to Settle What Observational Studies Cannot.

• The Verdict So Far—and the Question All the Media Miss

• Cited References

“Statins Cause/Prevent Dementia!”: Before You Pick a Side, Define the Outcome

Dementia isn’t the monolithic malaise on which the media articles’ magnetism so much depends. Episodes of forgetfulness after starting a drug aren’t the same as measurable changes on a cognitive test, or a diagnosis of dementia ten years later. And even dementia, the demon dreaded by many of us, is not one disease. Alzheimer’s, cognitive impairment, and vascular dementia are not interchangeable diagnoses [1]. At least not to the diagnosing physician.

You, the patient, might not give a damn about the fine art of nomenclature when the diagnoses’ common denominator is the tottering dotard to which each of them can reduce us eventually.

But the distinction matters for your today’s prevention decisions. Not because we have defined interventions that prevent one but not the other scenario. We still know far too little about the causes, and hence the treatment, of dementia. That’s unfortunately true also for its most infamous incarnation: Alzheimer’s.

The distinction matters because, given this uncertainty, you need to know what information to dismiss as useless, in favor of what to accept as (f)actually helpful for minimizing, or ideally preventing, dementia in your particular case. That information comes from four different types of studies which investigate four different questions:

1. Can statins trigger short-term, reversible cognitive symptoms in some people?

2. Do statins produce measurable cognitive impairment, on average, in randomized trials?

3. Do statin users develop more or less dementia over the long term?

4. Could statins influence cognition indirectly by preventing ischemic strokes and other vascular brain injury, even without preventing Alzheimer pathology?

Case reports, randomized controlled trials (RCTs), observational studies, and mechanistic (in-vitro) studies of possible pathways can deliver answers to each of these questions [2]. The media’s temptation is to present any one of those studies as the latest verdict for one side or the other in the court of “Statins vs. Dementia”. That’s why we are going to separate them now.

Subscribe now

Can Statins Cause Short-Term Cognitive Symptoms?

Remember George Burns? The comedian lived and performed until his death at age 100. All the while smoking 12-20 cigars per day. Is he evidence for the longevity benefits of smoking? I don’t think so.

How about the newly prescribed statin user who complains of sudden brain fog? Are they evidence for the drug’s cognitive side effects?

That’s the type of observation researchers elevate to ‘anecdotal evidence’ when they write about and publish their ‘case studies’. It’s a legitimate effort. The FDA mandates post-marketing pharmacovigilance, and case studies are one of the tools.

But anecdotal evidence is what the extremists on both sides of the statin aisle often misinterpret. The statin villifiers tend to inflate these case reports into evidence that the drugs cause Alzheimer’s. The statinistas, on the other hand, claim that no randomized trial ever demonstrated an average cognitive decline in statin users vs. non-users. But leaping from there to dismissing an individual’s complaint as imagined and not real is as flawed as scaremongering using anecdotal evidence.

So, isn’t the FDA-mandated drug warning about statins’ “association with memory loss, forgetfulness, amnesia, memory impairment, confusion” evidence for cognitive side effects [3]?

Nope!

First, association never proves causation. You already know this from several of my earlier posts.

Second, that warning was triggered by (among others) a series of 60 case reports recorded in the FDA’s register for such reports “MedWatch: The FDA Safety Information and Adverse Event Reporting Program”. Now look at the image below, which illustrates Wagstaff and colleagues’ evaluation of the 60 cases [4],

For only 14 cases we know that their symptoms actually improved after drug withdrawal. That doesn’t mean that the others didn’t improve. It simply means that the evidence is full of holes.

The outcome fields in the MedWatch filings had often been left blank.

Only four of the patients were exposed to a drug rechallenge. And for them the symptoms reappeared.

That sequence is clinically relevant. When a symptom appears after a drug is started, resolves after the drug is withdrawn, and returns after it is restarted, the probability that the drug is the culprit increases. Not to the level of certainty, though.

Ideally, the patient should be blinded to the intervention: they shouldn’t know whether the rechallenge phase presents them with the active drug or a placebo (easy in theory, not so easy in day-to-day clinical practice). That matters a lot here, because the reports contained only subjective symptom reports, not objective cognitive-test results [4]. Add to that the self-selection bias that inevitably affects studies relying on self-selected participants, and you can draw virtually any conclusion you like.

A later survey collected reports from 171 people who also believed that statins had affected their cognition (again, at the risk of self-selection bias). This time, the design minimized loss to follow-up [5]. The image below illustrates the results.

Of the 143 respondents who stopped treatment because of cognitive issues, 128 (90%) reported improvement. Of the 29 rechallenged patients, 19 (66%) reported a recurrence of symptoms. Now think about those numbers:

If one of three re-challenged patients didn’t re-experience the memory problems, we need to give some credit to the idea that, maybe, what they experienced in the first round of drug prescription wasn’t related to the drug.

Some additional case studies also report statin-associated cognitive decline that resolved within weeks of discontinuation but recurred upon drug rechallenge [6], [7], [8].

On first blush, that sounds like “yeah, I definitely have to worry about losing my marbles when the doc brings up the statins”. But look closer at those case reports, and judge for yourself whether the patients’ profiles match yours:

1. a 64-year-old with a history of bipolar disorder, depression, suicide attempts, COPD, and diabetes whose daily drugs included three antipsychotics and antidepressants.

2. a group of 18 Alzheimer’s patients, mean age 76, who saw a 1.9 point improvement of their (30-point) Mini-Mental State Examination results upon statin discontinuation and a statistically significant decline upon rechallenge.

3. A 51-year-old man, with hepatitis, extensive coronary artery disease having required several surgical procedures, who developed subjective memory loss after having been on simvastatin for 12 months, was switched to pravastatin, whereupon his cognitive issues resolved within 1 month.

In case you are wondering now about potential drug interaction (Case 1), the clinical relevance (hardly any) of a 1.9-point MMSE change in Alzheimer’s patients (Case 2), or the confounding role of vascular cognitive impairment that often follows extensive cardiovascular disease and post-surgery (CABG) status (Case 3), congratulations, you are paying attention.

Yet, despite all their shortcomings, these studies deliver an answer to Question 1: Short-term, reversible cognitive symptoms following statin exposure have been observed in some individuals.

In how many individuals (of all those taking statins)? Case reports can’t answer that. They can only sound an alarm that something may happen, not how often it happens. To move from detecting a signal to estimating its frequency (Question 2), we need a different type of study: The randomized controlled trial (RCT).

Randomized Trials: Reassuring on Average, Incomplete for Individuals

If statins consistently impaired cognition, randomized trials should reveal at least some deterioration in average cognitive performance. Spoiler alert: Overall, they don’t, but…

Before we get to the “but” part, let’s look at what the statin evangelists want you to know.

The “PROSPER” trial randomized 5,804 adults aged 70–82 to pravastatin or placebo and repeatedly tested the participants across three cognitive domains (attention/executive functioning, processing speed, and memory) for, on average, 3.5 years [9]. As the researchers expected at trial outset, both treatment groups declined in function over time. But there was no significant between-group difference.

You need to think of this result in this way: At the 70+ age range, cognitive performance deteriorates over time. The tests were sensitive enough to pick up that deterioration. Had they shown no deterioration, we would have reason to question the tests’ sensitivity to detect meaningful differences between the treatment groups.

The HOPE-3 trial randomized 2,361 adults aged 70+ to rosuvastatin or placebo and tested the participants at study start and end for psychomotor speed/attention, memory, and executive function [10]. The observation over 5.7 years mirrored that of the PROSPER trial: a decline in function over the trial period that was not different between placebo and drug groups.

A more recent systematic review of 42 RCTs (including PROSPER and HOPE-3) confirmed the observations of zero-difference between intervention and placebo groups across more than 150,000 participants [11].

So much about measurable cognitive performance. But how about statins’ effects “under the hood”, i.e., on the neurological wiring itself?

When the Brain Changes but Performance Does Not

Taylor and colleagues designed an experiment to investigate that question [12]. The researchers randomized 150 young-ish (mean age 48) hypercholesterolemic adults to either 6 months high-dose atorvastatin (80 mg) or placebo.

The participants completed neuropsychological tests and fMRI imaging during the study and two months after drug cessation. Again, neither participant group showed any adverse cognitive effects. Functional brain imaging, though, did reveal some drug-associated changes in neural activation during some tasks. These differences resolved post-treatment and, more importantly, they were not accompanied by worse cognitive performance.

That study illustrates a distinction the statin villifiers tend to ignore in their arguments: a measurable change in brain activity is not automatically evidence of brain damage or cognitive impairment. Quite on the contrary, as you might be surprised to learn.

The human brain is an astronomically complex system. As I have outlined in an earlier post (and in a paper in the journal Heart, Lung and Circulation [13]), one defining hallmark of complex systems is their ability to generate the same functional output using different internal pathways. The more complex a system, the more optional pathways it has, and the more robust it becomes against external and internal disturbance and damage.

When imaging findings suggest that a drug affects neural processing, a claim of clinical relevance requires evidence that people actually remember, or function worse under the drug. If they don’t, their brain does its ‘complexity thing’ rather well.

Can we then close the book on statins’ cognitive side effects? I’d rather not. Here is why:

The Funding Footnote You Should Never Skip

In medical research, the hand that signs the check also guides the pen.

That’s the result of a Cochrane Database Systematic Review [14].

The PROSPER trial was funded by an investigator-initiated grant from Bristol-Myers Squibb. AstraZeneca co-funded HOPE-3. The authors of the fMRI study reported research grants, consulting fees, and advisory roles across the who’s who of statin drug manufacturers. And the systematic review acknowledges that the pharma industry funded two-thirds of their reviewed RCTs.

Now, how should you read all that:

The randomized evidence fails to support the claim that statins produce a substantial, common cognitive impairment (with the caveat of potential bias).

BUT…

The transient shifts in regional brain activation seen on fMRI hint at some underlying biological mechanism that may well be the neurological signature of the “memory lapse” and “brain fog” that some people experience, and that show up in the pharmacovigilance databases and case studies.

That’s where the statinistas’ carte-blanche reassurance needs a statistical qualifier:

Failure to detect an average effect does not prove that no individual can be affected.

Averaging the results of thousands of participants can conceal a small group moving in one direction while everyone else remains unchanged.

Moreover, the RCTs have focused on short time windows of a few years only. By design, they are incapable of answering the long-term dementia question. The absence of a measurable average cognitive decline over the few years does not prove that statins are incapable of causing dementia over the following decades.

That requires a different type of study, which comes with a different set of caveats.

Subscribe now

Share

Do Statin Users Develop More or Less Dementia Over the Long Term?

If you believe most observational studies, statin users develop less dementia, not more.

The most recent, currently largest, meta-analysis pooled 55 observational studies involving more than seven million people. Statin use was associated with a 14% lower risk of all-cause dementia and an 18% lower risk of Alzheimer’s disease [15].

Now, who can argue with 7 million people?!

You can (and should). And not only because association DOES NOT equate causation.

Here is why you should probe (and how):

Let’s just recall what questions the case studies and the RCTs answer.

A case study illustrates what happened to one individual. It can’t tell us whether the same event happens to one out of 100 statin users (common) or one out of 100,000 (very rare).

BTW, the Council for International Organizations of Medical Sciences (CIOMS) standardized the verbal descriptors that you find on your drug labels (see image below) [16]:

RCTs typically detect the “very common” and “common” side effects. They happen frequently enough to register within the pre-approval trials and their relatively small patient groups.

The rare side effects, however, only show up during post-marketing surveillance once a drug has been released “into the wild”. Large-scale observational studies can follow hundreds of thousands or millions of people for long enough to count, for example, dementia diagnoses. But observational studies carry massive bias risks that can tilt the outcome and interpretation in either direction: negative or protective.

Here is an impressive example:

Rajan and colleagues followed 4,807 older adults (mean age 72) for 10 years [17]. At baseline, none of the participants used statins, and all were free of Alzheimer’s dementia. Those who subsequently started statin treatment had a 19% lower rate of Alzheimer’s disease. BUT, that benefit materialized only in the carriers of the APOE ε4 gene variant (this allele is recognized as a strong predictor of Alzheimer’s). Among non-carriers, statin use didn’t make a difference.

Now here comes the opposite direction:

Ye and colleagues analysed 371,019 UK Biobank participants, with a follow-up of 9 years (pretty close to the Rajan study) [18]. Their statin users had a 19% higher rate of Alzheimer’s disease during nine years of follow-up. Yet, similarly to the Rajan observation, participants carrying two APOE ε4 alleles appeared to benefit from the use of statins.

If you are a media scribe, that is pure gold. Depending on which result is your flavour of the day, you can write it up as statins…:

• reduce Alzheimer’s risk;

• increase Alzheimer’s risk;

• protect mainly APOE ε4 carriers;

• or do nothing measurable in people without that genotype.

Now, how the hell are you supposed to find out whether any of the conclusions applies to you, and if so, which?

It’s in the details.

The Bias Minefield Between Association and Causation

First, let’s get the main difference between RCTs and observational studies out of the way:

Randomization: When you randomize, say, 1000 participants into two (more or less) equally sized groups, you can be pretty sure that these groups are statistically indistinguishable in their measured and unmeasured biometrics and other study-relevant profile data. That equality is what authors typically illustrate in Table 1 of their papers.

Observational studies can’t do randomization. In our two examples, the researchers observed who took the drugs and who subsequently received a dementia diagnosis. They had to acknowledge that these two groups were significantly different in biometrics that have a bearing on their carriers’ dementia risk.

Healthy Users Can Make Statins Look Healthier

In the Rajan study, participants who ended up taking the statins had entered the study with significantly and substantially higher scores on all cognitive tests than their never-statin peers (global cognition, episodic memory, executive function, and MMSE score). Their greater cognitive reserve, rather than the drug, may well have been the driver of their lower Alzheimer’s risk.

Even if you statistically correct for that, you still have a problem. People who voluntarily initiate preventive medications in old age often practice better health habits and healthcare-seeking behaviors, which independently protect against cognitive decline. That’s a prime example of the “Healthy User” bias, also known as “Healthy Initiator” bias.

Sick Users Can Make Statins Look More Dangerous

The Ye study shows the reverse pattern, the “Sick User Bias” (also known as “Confounding by Indication”). While statin users had a 19% higher risk of developing Alzheimer’s, that risk was entirely carried by those younger than 60 (mean age 56). Statin users of this younger population entered the study with significantly lower global cognitive scores. They also carried a higher burden of cardiovascular disease.

While we have methods to statistically adjust for known differences, we can’t adjust for unknown or unmeasured differences (which is what randomization does exceedingly well).

Long story short, evaluating observational studies means walking through a minefield of biases and confounders that most media articles remain blissfully unaware of. The image below summarizes the key issues [19].

The Studies You Never Get to Read

Combining dozens of observational studies into meta-analyses with millions of participants doesn’t move the results closer to the truth. For a simple reason. Its official name is “Publication Bias”, colloquially known as the File Drawer Effect”. The authors of that hitherto largest meta-analysis openly acknowledge it [15]:

As a researcher, if your trial’s result fails to support your favorite hypothesis, you are tempted to lock it away rather than admitting that your hypothesis turned out to be a dud. After all, the discovery of a novel protective side effect will juice up your career far more than a paper that simply says “Duh”.

As a side note: I’m very grateful to my doctoral thesis supervisor, Prof. Alexander Krämer, who had hammered it into his students’ conscience never to ignore or hide a negative result. In medical science, discovering what doesn’t work is as informative as discovering what does.

The Trials Built to Settle What Observational Studies Cannot.

Fortunately, all is not lost. Two large RCTs are currently on their way to get to the truth. Well, kind of; at least in older adults.

The STAREE and PREVENTABLE trials enrolled 10.000 and 20.000 participants, respectively, to explore the question of whether and how putting cognitively and physically healthy older adults (70+) on statins will impact their cognitive performance, aside from preventing cardiovascular disease (CVD) events [20], [21].

Why should we pay more attention to these two trials’ outcomes than to any other before them?

Because both trials were specifically designed to avoid the biases we just discussed, and both trials report no industry funding.

Unfortunately, both trials started only recently, so we’ll have to wait a few years until the results start to trickle in.

That begs the question: What should you do in the meantime?

The Verdict So Far—and the Question All the Media Miss

Episode 1 in three sentences

Possible individual reaction: Reversible cognitive symptoms have been reported, but their true frequency is unknown.

Average trial effect: Randomized trials do not show substantial average cognitive harm.

Long-term dementia: Observational studies usually look protective, but they cannot prove that statins prevent Alzheimer’s disease.

Buried in that contradicting data, though, is a scenario nobody seems to acknowledge:

A statin may fail to alter dementia pathology and still help preserve cognition—by preventing the vascular brain injury that pushes a vulnerable brain across the dementia or Alzheimer threshold.

That is the question I will address in the second and final episode. We will look “under the hood” and examine why the following two popular arguments are both severely flawed:

• The brain, accounting for just 2% of your bodyweight, contains 20% of your body’s cholesterol, which it needs to function. Dialing down cholesterol with statins means dialing down your brain’s function. Verdict: Severely Flawed

• Statins have no effect on the brain because the blood-brain-barrier (BBB) is impenetrable to the drugs. Verdict: Flawed

Your practical takeaways in the next and final episode:

• If you are a patient: a downloadable seven-question primer to help you prepare for a more informed conversation with your doctor about memory symptoms, dementia risk, and the benefits and uncertainties of statin treatment.

• If you are a doctor or healthcare provider: a concise, ready-to-share patient-education handout that explains what is known—and still uncertain—about statins, short-term cognitive symptoms, and long-term dementia risk, so patients can receive a balanced explanation without requiring an extended consultation.

In the meantime:

Cross-examine the evidence yourself: enter the DoubleCheck Evidence Interrogation Room and run your personal questions against the >50 vetted studies behind this series.

Share

Leave a comment

Cited References

[1] R. N. Kalaria, “Neuropathological diagnosis of vascular cognitive impairment and vascular dementia with implications for Alzheimer’s disease,” Acta Neuropathol. (Berl.), vol. 131, pp. 659–685, 2016, doi: 10.1007/s00401-016-1571-z.

[2] M. C. Power, J. Weuve, A. R. Sharrett, D. Blacker, and R. F. Gottesman, “Statins, cognition, and dementia—systematic review and methodological commentary,” Nat. Rev. Neurol., vol. 11, no. 4, pp. 220–229, Apr. 2015, doi: 10.1038/nrneurol.2015.35.

[3] “FDA Changes Label on Statin Drugs | National Lipid Association Online,” archive.ph. Accessed: Jul. 28, 2026. [Online]. Available: https://archive.ph/thBJB

[4] L. R. Wagstaff, M. W. Mitton, B. M. Arvik, and P. M. Doraiswamy, “Statin‐Associated Memory Loss: Analysis of 60 Case Reports and Review of the Literature,” Pharmacotherapy, vol. 23(7):871–880, 2003, doi: 10.1592/phco.23.7.871.32720.

[5] M. A. Evans and B. A. Golomb, “Statin-Associated Adverse Cognitive Effects: Survey Results from 171 Patients,” Pharmacother. J. Hum. Pharmacol. Drug Ther., vol. 29, no. 7, pp. 800–811, 2009, doi: 10.1592/phco.29.7.800.

[6] K. P. Padala, P. R. Padala, and J. F. Potter, “Simvastatin-Induced Decline in Cognition,” Ann. Pharmacother., vol. 40, no. 10, pp. 1880–1883, Oct. 2006, doi: 10.1345/aph.1H014.

[7] A. Orsi, O. Sherman, and Z. Woldeselassie, “Simvastatin-Associated Memory Loss,” Pharmacother. J. Hum. Pharmacol. Drug Ther., vol. 21, no. 6, pp. 767–769, 2001, doi: 10.1592/phco.21.7.767.34577.

[8] K. P. Padala, P. R. Padala, D. P. McNeilly, J. A. Geske, D. H. Sullivan, and J. F. Potter, “The Effect of HMG-CoA Reductase Inhibitors on Cognition in Patients With Alzheimer’s Dementia: A Prospective Withdrawal and Rechallenge Pilot Study,” Am. J. Geriatr. Pharmacother., vol. 10, no. 5, pp. 296–302, Oct. 2012, doi: 10.1016/j.amjopharm.2012.08.002.

[9] S. Trompet et al., “Pravastatin and cognitive function in the elderly. Results of the PROSPER study,” J. Neurol., vol. 257, no. 1, pp. 85–90, Jan. 2010, doi: 10.1007/s00415-009-5271-7.

[10] J. Bosch et al., “Effects of blood pressure and lipid lowering on cognition,” Neurology, vol. 92, no. 13, pp. e1435–e1446, Mar. 2019, doi: 10.1212/WNL.0000000000007174.

[11] K. Li, Y. Li, X. Jiang, and Y. Zhu, “The effects of cholesterol-lowering drugs on neurocognitive function: systematic review and meta analysis,” Front. Neurol., vol. 17, Feb. 2026, doi: 10.3389/fneur.2026.1696228.

[12] B. A. Taylor et al., “The effect of high-dose atorvastatin on neural activity and cognitive function,” Am. Heart J., vol. 197, pp. 166–174, Mar. 2018, doi: 10.1016/j.ahj.2017.10.027.

[13] L. E. Kraushaar and P. Bauer, “Dismantling Anti-Ageing Medicine: Why Age-Relatedness of Cardiovascular Disease is Proof of Robustness Rather Than of Ageing-Associated Vulnerability,” Heart Lung Circ., vol. 30, no. 11, pp. 1702–1709, 2021, doi: 10.1016/j.hlc.2021.05.105.

[14] A. Lundh, J. Lexchin, B. Mintzes, J. B. Schroll, and L. Bero, “Industry sponsorship and research outcome,” Cochrane Database Syst. Rev., vol. 2017, no. 2, p. MR000033, Feb. 2017, doi: 10.1002/14651858.MR000033.pub3.

[15] F. L. Westphal Filho et al., “Statin use and dementia risk: A systematic review and updated meta-analysis,” Alzheimers Dement. Transl. Res. Clin. Interv., vol. 11, no. 1, p. e70039, 2025, doi: 10.1002/trc2.70039.

[16] Guidelines for preparing core clinical-safety information on drugs: report of CIOMS Working Group III. Geneva: CIOMS, 1995.

[17] K. B. Rajan, E. A. Mcaninch, R. S. Wilson, A. Dhana, S. Evans-Lacko, and D. A. Evans, “Statin Initiation and Risk of Incident Alzheimer Disease and Cognitive Decline in Genetically Susceptible Older Adults,” Neurology, vol. 102, no. 7, p. e209168, Apr. 2024, doi: 10.1212/WNL.0000000000209168.

[18] Z. Ye et al., “Association of statins use and genetic susceptibility with incidence of Alzheimer’s disease,” J. Prev. Alzheimers Dis., vol. 12, no. 2, p. 100025, Feb. 2025, doi: 10.1016/j.tjpad.2024.100025.

[19] R. Kazibwe, R. Rikhi, S. Mirzai, N. P. Ashburn, C. L. Schaich, and M. Shapiro, “Do Statins Affect Cognitive Health? A Narrative Review and Critical Analysis of the Evidence,” Curr. Atheroscler. Rep., vol. 27, no. 1, p. 2, 2025, doi: 10.1007/s11883-024-01255-x.

[20] S. Zoungas et al., “Statins for extension of disability-free survival and primary prevention of cardiovascular events among older people: protocol for a randomised controlled trial in primary care (STAREE trial),” BMJ Open, vol. 13, no. 4, p. e069915, Apr. 2023, doi: 10.1136/bmjopen-2022-069915.

[21] J. Joseph et al., “PRagmatic Evaluation of Events and Benefits of Lipid Lowering in Older Adults (PREVENTABLE): Trial Design and Rationale,” J. Am. Geriatr. Soc., vol. 71, no. 6, pp. 1701–1713, Jun. 2023, doi: 10.1111/jgs.18312.

What is New?

Vinay Prasad — hematologist-oncologist and, until this April, director of the FDA’s Center for Biologics Evaluation and Research (CBER) — argues that some health decisions belong to a data expert, not a disease specialist. A borderline heart-risk score, he says, is one of them. That will irritate the medical establishment. Good. Some irritation is overdue.

Why it Matters

Prasad’s right. Your “3% risk” verdict from the AHA’s PREVENT algorithm sounds like a fact but it isn’t. Without the data expert, you either start a pill you didn’t need, or skip a test that would have caught real danger in time.

But he also picked the one example that hands his critics a free shot. Fix that, morph his suggestion into a practical decision aid, and what you get is a real shot at better-informed shared-decision-making with your doctor.

Your Takeaways

You’ll walk away with a two-minute decision chart that tells you where to start — data expert, or doctor.

And along the way you’ll discover how the right follow-up diagnostic, run through the right math, can swing that risk number hard in either direction.

Table of Contents
· Prasad’s Provocation: Not Every Health Question Needs the Same Expert
· Why the 27-Year-Old Statin Example Is Flawed
· Same Risk Score, Different Arteries
· The Data Expert’s Job Description
· So, Which Expert First? A Practical Decision Aid
· Cited References

This is a public health scientist’s educational information, not medical advice — talk to your doctor before changing any treatment.

Prasad’s Provocation: Not Every Health Question Needs the Same Expert

Imagine you are 52, you consulted the latest online CVD risk calculator to estimate your 10-year risk for a heart attack or stroke. The calculator returned a 3% estimate (borderline risk, potentially qualifying you for a statin prescription). Will you be better off asking your doctor (the disease expert) or a data expert whether to take a statin or go for some additional tests?

That’s the type of question Vinay Prasad explored a few days ago in a post on Sensible Medicine. He built his piece around a series of paired patient scenarios, and one question for each pair — who needs the disease specialist, and who will be better off with someone fit enough to digest trial data?

In his cardiology example, Patient 2 was a 27-year-old, with LDL of 105 mg/dL, normal weight, normal blood pressure, no other risk factors, trying to decide between a coronary CT, an Lp(a) test, or simply starting a statin.

Prasad’s answer was straightforward. A cardiologist’s years of bedside experience not only add nothing but may even be unfavourably biased compared to what a data expert extracts from the trial literature.

I guess you recognize the provocation in Prasad’s piece?

A representative of the medical guild challenges the guild’s claim to exclusive MD-credential-based authority over every question about health.

I fully agree with his split. I’ve been arguing a version of it since I started the Double-Check Doc newsletter. Much of what passes for medical guidance is clinical trial results translated into an average patient that fails to describe the person actually asking the question.

But Patient 2 also illustrates where Prasad’s argument becomes unnecessarily assailable, which is why my introductory hypothetical scenario needed to deviate from Patient 2.

Why the 27-Year-Old Statin Example Is Flawed

At age 27, none of the calculators (specifically PREVENT or SCORE2) spits out a risk estimate. PREVENT and SCORE2 are blind to ages below 30 and 40, respectively.

Age dominates the math so heavily that as a young adult (between 30 and 40) you need to be a severely overweight, diabetic smoker with elevated blood pressure to cross the 5% risk threshold. It doesn’t require an expert (neither of the medical nor of the data persuasion) to foretell a less-than-bright future if you don’t do anything about that profile, like RIGHT NOW.

So let’s ask Prasad’s question about a patient it applies to.

Same LDL, same clean profile, twenty-five years later: at 52 years old, a non-smoker, with normal blood pressure and the same cholesterol profile lands at a 3% 10-year risk (you can check yours at the American Heart Association’s PREVENT online calculator). That’s the “borderline” threshold where the AHA/ACC’s 2026 guidelines’ all-clear signal morphs into get more information [1].

And that’s where the data expert comes into play.

Same Risk Score, Different Arteries

Two people with identical profiles can land on those 3% with wildly different actual risks. The reason: the 3% estimate is a population average, built from those few coarse inputs (sex, lipids, blood pressure, smoking, and diabetes status) over a large group of people (hundreds of thousands, typically).

That group contained the entire range of individuals, from those whose arteries are prematurely aged vascular wrecks (remember ‘vascular age’ that I discussed in an earlier post), to others whose lifestyle choices (or a lucky hand of genetic cards) had kept their vessels in mint condition. Those differences hide in the risk scores’ blind spots. And that’s why the average hardly tells you anything unless you look further.

The AHA/ACC 2026 guidelines, which I dissected in an earlier post, tell your doctor when to look further and what to look for. But they fall short of how to translate those findings into a math-based adjustment of risk estimates and intervention strategy (which is why, in an earlier post, I gave you a downloadable tool to do just that).

So, let’s take our Patient 2 (at age 52) and illustrate the data expert’s job.

We’ll use the coronary artery calcium (CAC) score as an additional risk diagnostic. For one version of Patient 2, the CAC score comes back at “zero” (no detectable calcified plaques in his coronary arteries). The unlucky peer of Patient 2 scores >100. Both patients share an identical population-level risk number (3%), but they do not share the same disease.

What the data expert needs to apply here is Bayesian reasoning. The image below illustrates the result [2].

A Layperson’s Primer on Bayesian

Bayesian reasoning starts with your best current risk estimate, befittingly called the prior. In our case, the prior is the 3% estimate derived from PREVENT or SCORE2. Bayesian statistics then aim to update the prior with additional evidence to more precisely estimate your personal true risk beyond the score’s one-size-fits-(not)-all estimate.

The CAC score (together with many diagnostic tests) offers such additional evidence. It comes with its own predictive power: e.g., a zero CAC score translates into an X% 10-year risk prediction, whereas a score of >100 carries a substantially higher Y% predicted risk.

But no test is 100% accurate. Whether the cutoff is zero, 100, or any other arbitrarily defined number, there are always false positives (people above the cutoff who don’t suffer the predicted event) and, correspondingly, false negatives.

Long story short, we need those data to extract the likelihoods of the test results to correctly predict the future outcome (event, or no event). We then combine your prior with the CAC score’s likelihoods to get your posterior (the statistical, not the anatomical version): a personalized risk estimate that can swing your prior both ways (higher or lower risk).

At this point, my original draft contained a more detailed explanation of priors and likelihood ratios, but my wife forbade me to traumatize you any further, so I had to delete it (otherwise the likelihood of me suffering spousal emotional violence would have increased intolerably).

Now, where do you get your likelihoods for any given additional test from?

You can order them on Amazon (look out for Prime Day or Black Friday deals), and in case they aren’t legally available there, I’m running a back-alley shop in the seedier part of town, where you can buy them.

JUST KIDDING

That’s where the data expert comes in.

The Data Expert’s Job Description

He or she extracts the Bayesian-required data from published research. The 2018 paper of Budoff et al. about CAC’s predictive power is a good example [2]. I used it to derive the likelihood ratios for the CAC score in our illustration.

But aside from mastering Bayesian statistics, a data expert’s real value for you is in their ability to decide which evidence is admissible, how strongly it should move the prior, and where the update may become misleading. Their key prerequisite is the ability to find, read, and judge scientific papers and the markers they evaluate on their case-relevance.

• When study participants’ profiles are too dissimilar from the patient/person in question, calculations can be mathematically precise, but medically misleading.

• When the prior’s endpoint doesn’t match the study’s ( “ASCVD,” “major adverse cardiovascular events,” “all-cause mortality,” or “coronary events” are not interchangeable), the posterior becomes a hybrid.

And then there is the clinical boundary awareness:

The data expert never diagnoses a condition, recommends for or against a treatment, or overrides a physician. The data expert clarifies priors, evidence quality, and posterior risk. The data expert preps you for better-informed shared-decision-making with your doctor.

Does that mean “First ask the data expert, then use the answers to confer with your doctor?”

Not in every medical situation. That is Prasad’s point. You need to know when to ask the data expert first.

Subscribe now

So, Which Expert First? A Practical Decision Aid

With the chart below, I translated Prasad’s framework into a decision aid that helps you answer the question: Do I need a medical or a data expert first?

You can download the decision chart and explanatory table by clicking on the image below.

If you’re the 52-year-old in our introductory example, wondering whether the guideline applies to you, the medical expert may not be your best first address, but neither should the data expert be your only one. First ask the data expert, then use the answers to confer with your doctor. Neither one replaces the other. Together, they are the answer to “what should I do”.

The data expert’s work is what I do. Read about her eon my Double-Check Doc Substack.

Share

Leave a comment

Cited References

[1] C. P. Guidelines, “2026 ACC / AHA / AACVPR / ABC / ACPM / ADA / AGS / APhA / ASPC / NLA / PCNA Guideline on the Management of Dyslipidemia: A Report of the American College of Cardiology / American Heart Association Joint Committee on Clinical Practice Guidelines,” Circulation, 2026, doi: 10.1161/CIR.0000000000001423.

[2] M. J. Budoff et al., “Ten-year association of coronary artery calcium with atherosclerotic cardiovascular disease (ASCVD) events: the multi-ethnic study of atherosclerosis (MESA),” Eur. Heart J., vol. 39, no. 25, pp. 2401b–2408b, 2018, doi: 10.1093/eurheartj/ehy217.

Bryan Johnson is the billionaire ”Don’t Die” longevity experimenter who recently revealed the diagnosis of a potentially life-threatening condition, ironically concealed for years by his longevity strategy.

Why it matters

Bryan’s diagnostic delay exposes a fatal flaw of algorithm-driven “biohacking”: Mistaking medical signs for supplement-amendable anomalies.

Your tangible takeaways

You need neither a billionaire’s budget nor a shotgun supplement stack. The key metrics and the science-okayed self-experimentation protocols are already available for everyone. As a scientist with a 20-year research record on decelerating biological aging I have helped develop them.

Table of Contents
· Forever Young has an Expiration Date
· The Fountain of Youth Meets the Stomach Lining
· Treating the Symptom, Missing the Cause
· Project Blueprint’s Mortal Flaws
· The N-of-1 Illusion
· Show Me the Arteries, Not the Slogans
· The Vascular Verdict: My 20-Year Bio-Age Gap
· Cited References

Forever Young has an Expiration Date

With a humility typical of your next-door tech billionaire, Bryan Johnson, famous for popping a 111-pill combo of longevity-flogging supplements a day, considers Jesus his main rival (seriously).

Jesus believed in the inevitability of death; Bryan doesn’t. At least he didn’t until a few days ago, when he made it onto the front page of a “Schadenfreude”-prone media.

In case you didn’t know, Bryan is the tech billionaire who founded Braintree, sold it for 800 million bucks to PayPal, bought into biohacking’s idea of living forever (again, seriously), and just found out that even “forever” comes with an expiration date.

In his case, the expiration date’s name is stomach cancer from incurable autoimmune gastritis (AIG).

Two years ago, I introduced you to his biohacking “Project Blueprint”. I thought it prudent you picked up on his strategy of how to beat Jesus (not seriously), but also why it won’t work (very seriously).

In order to revert his biological age to 18 years by calendar age 60, Bryan designed a strategy that now contains the following key elements:

• take >40 different supplement pills (and a few prescription drugs) every day

• dine latest at 11:00 (AM)

• monitor your nightly boners with (I’m not making this up) the appropriate metrics of frequency and strength

• have a handful of medical specialists on stand-by

• go through hundreds of clinical measurements and hyped-up high-tech treatments each month

You immediately see where the label “blueprint” comes from: It is what you, I, and every one of our fellow billionaires can afford to copy. And not to worry, in case you are one of those unfortunate few who have to slum it on, say, a shrimpy CEO salary, you can still subscribe to Bryan’s “longevity stacks” of pills and powders that he peddles online. At $500+ a month, it’s a little less affordable than the Meghan Markle schlock, but, hey, these poor people need to earn some money, too, right?.

In case you suspect that I share in the Schadenfreude, I don’t. I’ll get to the “Why” in a moment.

But first, here is this week’s news:

Long story short, Bryan’s gang of medical experts just discovered that his stomach has been eating its own lining, literally, and probably for the past 2 decades (a salute to the monthly monitoring). The correct name for his condition is autoimmune gastritis (AIG).

The Fountain of Youth Meets the Stomach Lining

AIG is a chronic, incurable condition where the body’s immune system goes rogue and attacks its own stomach cells. Particularly those that produce the stomach acids and the proteins needed to absorb iron and Vitamin B12. The consequence is pernicious anemia (an insufficient supply of red blood cells or hemoglobin to carry oxygen to the body’s various tissues), and B12 insufficiency.

People with AIG are more likely to develop stomach cancer, though that is not a guaranteed consequence, and most patients can enjoy a normal lifespan (which, on the other hand, will be a little shorter than “forever”). So, “mortal blow” is as realistic as a certain president’s claim of victory in the Iran war.

Why do I not share in the Schadenfreude? For two reasons:

First, the media is laughably off-track about the root cause of Bryan’s AIG.

Second, in my earlier post’s prediction about why Blueprint won’t work, I missed the blindingly obvious flaw that now potentially threatens his marketing message:

Treating the Symptom, Missing the Cause

The key insight is NOT that Bryan’s 40+ daily supplements ruined his stomach lining. That is the media’s cheap-shot-insinuation born out of either dim-wittedness or duplicity, both having become journalists’ career-ensuring qualities in the legacy media.

You can find the list of supplements (and their then prices) in my earlier post, or on his website, “Don’t Die.” The latter trots out a minute-by-minute description of his daily routine. A fascinating read for anyone in need of reassurance that living another 2000 years may not be as exciting as, well, kicking the bucket a little earlier.

The key insight is also NOT Bryan’s version: that his supplement routine has postponed the inevitable (symptomatic AIG), which he would otherwise have suffered much earlier. Quite on the contrary:

The key insight is that his massive daily dose of Proferrin and B12 has masked the anemia that should have otherwise raised his medical experts’ eyebrows. His supplements compensated for an AIG-driven anemia, hiding the condition long enough until his stomach lining became so compromised that its deteriorating ability to absorb iron showed up in his blood, no matter how much iron he pumped in via his supplements.

Yes, AIG often takes years before a doctor suspects it. But Bryan’s autoimmune hypothyroidism (Hashimoto’s thyroiditis), diagnosed at age 21, should have triggered regular monitoring of his blood iron levels.

Why? Because one in three Hashimoto patients will eventually develop AIG [1]. That’s why doctors are advised to keep Hashimoto patients’ iron levels on their radar [2]. Instead, faced with chronic iron deficiency, Bryan’s purely data-driven “Blueprint” dictated a simple solution: consume more iron.

His well-paid medical team obviously bought into this overly simplistic strategy: treating a symptom while missing what caused it — the stomach’s organ failure.

Northeastern Global News

There is a word for “struggled to explain”: Incompetence.

Anyways, I hadn’t foreseen such a prosaic scenario as a legitimate reason for the Blueprint’s potential failure. Mea culpa!

Project Blueprint’s Mortal Flaws

My bet was, and still is, on two other conceptual flaws of the life extension dreams that our benevolent billionaire overlords so selflessly fund.

They ignore two fundamental laws, one of physics and the other of complexity science: Entropy and emergence. I discussed both in detail in my earlier post about Bryan.

So, here it is in brief:

First, entropy: the 2nd law of thermodynamics simply states that every system, without exception, gravitates to achieving maximum entropy. For biological systems, that’s another word for death. While we have built-in repair mechanisms that throw off some entropy, these systems, too, adhere to the one-way law of entropy (I illustrated the typically coma-inducing subject of entropy with an easy-to-understand example in my earlier Bryan Johnson post.)

Second, emergence: it’s the hallmark of any complex system. Emergence describes the unpredictability of a complex system’s behaviors when you interfere with any of its nodes. The human system is hyper-astronomically complex. On the positive side, with complexity also comes robustness, the ability to survive disturbances. But by attempting to micromanage his biology through a 40+ polypharmacy, the attempt to stall aging may also engineer the opposite outcome.

And Bryan is firing some heavy artillery at his biology. Here is his list of prescription drugs:

1. Armour Thyroid 60 mg,

2. levothyroxine 100 mcg,

3. minoxidil 3.75 mg,

4. metformin 500 mg in experimental cycles,

5. tadalafil 5 mg,

6. Jardiance 10 mg,

7. acarbose 200 mg,

8. Repatha 140 mg every other week.

9. Candesartan 8 mg

No contest to #1&2. His thyroid condition necessitates these drugs.

But hand this prescription list to your doctor, without telling them where it comes from, and ask them to draft a clinical silhouette of the patient. They would likely summarize along the following line:

A high-risk, multi-morbid elderly male suffering from advanced metabolic syndrome, severe cardiovascular disease, and profound endocrinopathy:

• Advanced metabolic syndrome & type 2 diabetes mellitus (T2DM). Acarbose + Jardiance: a combination therapy for Type 2 Diabetes Mellitus.

• High-risk cardiovascular disease (CVD). Repatha and Candesartan: Repatha is a PCSK9 inhibitor, aggressively lowering cholesterol (and, thanks to its cost, the content of your wallet); Candesartan, an angiotensin receptor blocker (ARB), treats hypertension and heart failure.

• Age-Related Urologic and Dermatologic Decline. That’s the sterilized description for erectile dysfunction (Tadalafil) and baldness (Minoxidil).

Of course, Bryan doesn’t suffer any of these conditions. He takes the “Blueprint” drug combo in the hope of reversing his biological age. And that’s where we face a slight problem: these drugs made it on the pharmacy shelves because clinical trials demonstrated their effectiveness in treating their on-label conditions. However, there exists not a single trial that proves their effectiveness in preventing them or, more to the point, reversing biological aging.

But Bryan doesn’t need that evidence. He performs what I have been advocating for years: An N-of-1 trial. At least he thinks so. I don’t.

The N-of-1 Illusion

The “N” in science publications stands for the number of participants in a clinical trial. N-of-1 simply means a trial of “one” person. That’s, btw, an FDA-recognized standard of single-case experiments. The Northeastern University’s article about Bryan’s latest revelation slips that term in as if Bryan’s Blueprint experiment deserved that label. It doesn’t. Because nothing in Bryan’s “Blueprint”, and I emphasize “nothing”, meets the FDA standards of N-of-1.

First, in a true N-of-1 trial, phases of intervention alternate with phases of no intervention (the so-called ABAB design). That allows specific statistical methods to detect whether the outcomes of the A-phases differ sufficiently enough form the B-phases to suggest an effect.

Second, you don’t enter an intervention phase with a basket of interventions (e.g., 40+ different supplements). You try to investigate one single intervention at a time. You want to isolate the effect of one drug, behavior, or supplement on the outcome in question.

Bryan’s experiment fails N-of-1 on both accounts. There is no alternation of interventions, and there is no isolation of interventions. It’s the “Blueprint” scattershot all the time.

Show Me the Arteries, Not the Slogans

Bryan still has 12 years to prove his claim: achieving bio-age 18 at calendar age 60. He and his marketing mavericks have been spinning the proof like Putin has been spinning the proof of a quick victory over Ukraine (for the past 5 years).

Check for yourself.

The spin doctors are strong on soundbites and short on substantiating metrics. That shouldn’t be a problem for us, given what our politician overlords have conditioned us to consider “normal” (and don’t think for a moment that I’m referring only to the current White House occupant).

In my sphere of work, however, I prefer measuring every longevity claim against the one under-the-hood metric that more reliably tells you your biological functional age than anything else: vascular age.

If you have read some of my earlier posts, you’ll be familiar with the rationale for vascular function as THE metric for biological age (Sydenham’s 17th century quote is as valid today as it was 300 years ago).

Specifically in my posts “The End of “Just Exercise”, and “Is Enviable Longevity Your True North? Vascular Age Should Be Your Navi” I present not only the scientific evidence and the metric, but also a downloadable file for you to help you calculate your vascular age.

The Vascular Verdict: My 20-Year Bio-Age Gap

I’m 69, 20 years older than Bryan. The image below illustrates my rate of vascular aging. At roughly 0.8 (years per calendar year), I’m biologically in my early to mid-50s. A $200 consumer device delivers the pulse wave velocity (PWV) measurement that my team and I have translated into a rate-of-aging algorithm for our research and for my clients.

I systematically test and isolate my lifestyle habits (from modifying exercise routines to dietary interventions), always one intervention at a time, to see how each affects my vascular aging process. Compare that to Bryan’s scattershot strategy. Even if he were to succeed in his 18-at-60 effort (which I doubt very much), he will have no idea which of his supplements or habits did the trick. Or, in the case of failure, which ones may have checkmated his longevity game?

Contrary to him, I know what keeps me younger. And don’t think for a moment that I’m a self-chastising diet Calvinist. Quite the contrary.

15 years younger (biologically) than calendar age is achievable. From what I have seen in our research and among my clients, a 20-year gap is probably the limit. Whether there is a limit in the opposite direction, I’m not sure. In one of our clinical data gathering exercises, a 50-year-old sedentary, obese truck driver with a history of heart attack and stroke had an arterial age closer to 80, whereas a 90-year-old spry guy was biologically as old as I am now by calendar years.

I outlined the DIY-type N-of-1 protocol in two earlier posts (both with downloadable aids) one illustrating N-of-1 with the layperson-accessible metric of vascular function and age, and the other using blood pressure.

But if you still think that Jesus Christ’s main rival has a point with his supplement combo, you’ll find all its components in my first post about Bryan. You then only need to wait for another 2000 years to see whether the Blueprint’s miracles are a match for walking on water or turning the latter into wine.

Share

Subscribe now

Cited References

[1] S. Kulnigg-Dabsch, “Autoimmune gastritis,” Wien. Med. Wochenschr. 1946, vol. 166, no. 13, pp. 424–430, 2016, doi: 10.1007/s10354–016–0515–5.

[2] J. Kaur and I. Jialal, “Hashimoto Thyroiditis,” in StatPearls, Treasure Island (FL): StatPearls Publishing, 2026. Accessed: Jul. 09, 2026. [Online]. Available: http://www.ncbi.nlm.nih.gov/books/NBK459262/

Standard cardiovascular risk calculators have a massive blind spot for women approaching menopause. They ignore the toughest and most telling stress test a woman can go through: pregnancy.

Why it matters

Cardiovascular disease kills seven times more women than breast cancer, yet medical practice often reduces a woman’s heart risk to a simple cholesterol number. Because risk scores ignore her reproductive history, that number may either exaggerate or underestimate her real risk,

This isn’t just a women’s issue; it is one for any man who wants the significant woman in his life to survive medicine’s male-centric cardiovascular care.

Your tangible takeaways

• The 5-Question Doctor Script: A printable “cheat sheet” to bring to your next appointment, designed to stop your physician from taking medical shortcuts with their prescription pad.

• The AI Interrogation Room: Direct access to my custom-built interactive research assistant, loaded with 30+ peer-reviewed studies, so you can privately interrogate the evidence and double-check the science yourself.

• The “Treat, Test, or Trial” Framework: A decision matrix to help you navigate the normal post-menopausal cholesterol bump without accepting or rejecting a statin blindly.

Table of Contents:

• Irma’s Regrets

• The Awareness Gap: Why Women Fear the Wrong Loss

• Same Disease, Different Timeline

• Pregnancy Was the Stress Test Your Risk Score Doesn’t Not See

• Menopause Changed the Cholesterol. Did It Change the Risk?

• Treat, Test, or Trial?

• The Clinical Reality Check

• The Doctor Script: Five Questions Before a Postmenopausal Statin Decision

• Dive Deeper: Interrogate the Evidence Yourself

• Irma’s Window Closed. Yours May Not Have.

• Cited References

A Mastectomy Couldn't Save Irma’s Heart

At 67, Irma still possessed the sharp mind that kept her on the board of a national non-profit org. She red-teamed her son’s (my) career moves (not always the brightest ones), and she was “damn sure” that her double mastectomy decision (in her 50s) preempted what she had always feared most: breast cancer.

At 69, she could still walk and talk, but a stroke had eroded her mental acuity, her confidence in her health priorities, and her ability to manage a full day without help. The window to prevent this theft of healthy, enjoyable years opens long before the emergency (and closes with it). Unfortunately, that window still remains hidden for too many women for two reasons:

• An awareness gap: seven times more women die from cardiovascular disease (CVD) than from breast cancer.

• A health care system inertia geared to a male-centric view of CVD

This post is about the female-oriented prevention of atherosclerotic cardiovascular disease and its downstream drama (heart attacks, strokes, vascular dementia, and heart and kidney failure). It is about prepping you for the most momentous discussion you’ll have with your doctor once your menses are drying up for good. Because that’s the time when the risk factors (e.g., cholesterol, blood pressure, body fat) often creep up, too.

In a health care system that fails to see you as something other than a bloke with boobs, you get the short end of the prevention stick.

In 2021 the Lancet Women and Cardiovascular Disease Commission phrased that scientifically more palatable [1].

These are the main issues the Commission admonished:

• Male-centric risk scores

• Lack of women-focused CVD competence among PCPs (8 of 10, according to their own admission)

• Underrepresentation of women in clinical trials

If you are a woman reading this, you probably want to enter the examination room with the knowledge and the questions necessary to prevent the conversation from being reduced to a cholesterol number. Not every increase automatically justifies a lifelong drug prescription. And when it’s low, you might think you are in the clear. Both reactions can be wrong.

If you are a man reading this, don’t think of it as a women ’s-only post. It is about how current clinical prevention practice can (and often does) fail your mother, your wife, your sister, or your daughter. This post is about how you can help them avoid this scenario (no matter your gender). After all, her breast may be her business, but her heart is yours too.

A caveat (and disclaimer) upfront: This post is for informational and educational purposes only. It is not intended to be a substitute for professional medical advice, diagnosis, or treatment. Do not disregard professional medical advice or delay in seeking it because of something you have read here. This post is also not about recommending for or against statins. It is about getting you to better-informed joint-decision making.

The Awareness Gap: Why Women Fear the Wrong Loss

Screening invitations, mammography campaigns, and celebrity stories have given breast cancer its legitimately high visibility. But here comes the dissonance: while cardiovascular disease (CVD) kills 7.5 times more women than breast cancer (and more than all cancers combined) [2] [3], one in two women is unaware that CVD is their prime killer [4].

This awareness gap is just the less worrisome part of the story. While half of all PCPs ranked breast health and breast cancer as an “extreme concern” for their female patients (rating it a 5 on a 5-point scale), only 39% ranked cardiovascular disease as an extreme concern, and even behind weight issues [4].

This mismatch matters for two reasons:

• When CVD doesn’t take a life, it often leaves the rest of it a lot less enjoyable.

• Preventable atherosclerosis often stays silent and undetected until its consequences aren’t preventable anymore.

After a transient ischaemic attack, you might hesitate before getting behind the wheel. After a heart attack, the grandmother who used to take her grandchildren hiking now plans her trips around breathlessness-rescuing park benches. Silent or small strokes can steal speech or your ability to independently manage the routines that make an adult life self-directed.

While the risk factors for all these consequences may well be the same for men and women, their timelines aren’t. And that’s where the risk scores and the health care system fail women.

Same Disease, Different Timeline

The Lancet Commission’s disparity claim is valid: women are under-recognized, under-diagnosed, under-treated, and underrepresented in trials.

But that’s not the whole story. The male-female disparity doesn’t emerge from a fundamentally different disease process.

LDL/ApoB particles promote plaque build-up regardless of the presence or absence of the Y chromosome. The same goes for endothelial dysfunction and inflammation (the initiators and accelerators of atherosclerotic processes). And the perfect-storm scenario required to trigger a heart attack or stroke is the same in men and women (I described these processes in earlier posts, here and here).

Since the biology of ASCVD is the same in men and women, does the Lancet Commission’s claim about male-centric risk scores still hold water? Yes and no.

No, because the two major risk scores (PREVENT-ASCVD for the U.S., and SCORE 2 for Europe) have since progressed to sex-specific algorithms.

Yes, because the scores’ sex-specificity does not extend to their variables. Age, cholesterol, blood pressure, smoking and diabetes status, are all the same for both sexes. What’s different is the variables’ multipliers (the coefficients, in statistics lingo). They (a) put to rest the idea of women as the weaker gender, and (b) tell you what to keep an eye on before and after menopause.

Forget about the “weaker gender” idea! For both sexes, calendar age is the undisputed “heavyweight champion” in the risk predictor universe [5]. You have probably heard that, on average, women suffer their first CVD event roughly 10 years later than men [6]. Let’s double-check that claim.

For the chart below, I have used the SCORE2 algorithms for low-risk European populations and applied the same hypothetical configuration of blood pressure, cholesterol, and smoking status to both sexes.

You will notice two things:

• Women reach the same model-estimated 10-year CVD risk with a roughly 5-year gap.

• That gap shrinks with progressing age, reaching zero past age 75.

So, how do we reconcile the 4-5 year progressively shrinking gap with the “10-year” delay of events that keeps popping up in the press? Simple, it’s a matter of mathematical modelling with fixed variables (what I did to create the chart) versus unadjusted epidemiological reality (the 10-year gap).

When age-matched with men, premenopausal women typically enjoy a healthier cardiometabolic profile—lower blood pressure, higher HDL, and lower LDL. By artificially forcing the 40-year-old woman to have the same elevated risk factors as a 40-year-old man, I stripped away that physiological advantage. But even then…

given an identical risk factor profile, women reach the same absolute risk for an event (SCORE2’s fatal and non-fatal heart attack or stroke) 4-5 years later than men.

That’s not a sign of a weaker gender. That’s a sign of slower aging. Here is why:

Gerontology defines aging as [7]:

In the context of ASCVD, we are talking about vascular function. The greater its decline, the greater a person’s vascular age, and, hence, the greater the risk of an event.

(that quote that is as relevant and true today as it was when Thomas Sydenham first uttered it in the 17th century, conceiving one of medicine’s oldest, still valid truisms)

Since vascular function has become measurable, the concepts of Early Vascular Aging (EVA, a bad thing) and Supernormal Vascular Aging (SUPERNOVA, a good thing) have entered our clinical risk prediction repertoire [8]. Adding vascular age (VA) to the standard scores’ risk prediction models can significantly improve the scores’ accuracy [9]. I have explained vascular aging (VA) in earlier posts, here and here, and how you can even measure your vascular age at home now.

Why am I harping on vascular function and age?

For two reasons:

First, the female and male biological clocks really run at different speeds (as the above chart illustrates).

Second, sex-specific life-course events (those not considered in the risk scores) can accelerate a woman’s biological clock beyond what her risk scores suggest (the subject of the next section).

It helps you shift your perimenopausal conversation …

from “Does my risk score suggest medical intervention?”

to “Am I at higher or lower risk than the score suggests?”

After all, a postmenopausal LDL rise is not automatically pathological, but also a relatively low number is not automatically benign either.

Your Pregnancy Was the Stress Test Your Risk Score Doesn’t See

A woman’s reproductive cycle begins with menarche, ends with menopause, and is potentially punctuated by one or more pregnancies. Both early menarche (defined as first ever menses before age 12, 11, or 10, depending on whom you ask [10]) and premature menopause (cessation of regular periods before age 40) correlate with postmenopausal ASCVD events [11].

But no matter how strong the association, the litmus test is whether these ‘risk factors’ improve or modify the standard risk scores’ estimates. Both fail that test [12]. So do pregnancy complications. Among them, gestational diabetes, gestational hypertension (of which preeclampsia is the more severe version), preterm births, and low birth weights (small-for-gestational-age infants) are the ones we would plausibly expect to swing the risk scores’ estimates into a higher bracket. Yet they don’t.

Why did I say “plausibly”? Think about it this way:

Pregnancy is probably the most demanding cardiovascular stress test a woman will ever take. Blood volume expands. Cardiac output rises. The kidneys, the vasculature, the blood-pressure control system, the coagulation system, and the placenta all have to coordinate under sustained and increasing load. For months.

If preeclampsia, gestational diabetes, or preterm delivery are the result, the stress test didn’t go so well. It may have revealed something important, because in all these conditions, vascular dysfunction plays a central role [13]. So, why do all those conditions fail to affect standard risk scores?

Because they hide in the risk scores’ blind spots. By design. Here is how:

Pregnancy complications strongly correlate with later elevations of the scores’ traditional risk factors—hypertension, hyperglycemia, and hypercholesterolemia [14].T he female cohorts used to derive the standard risk scores (like SCORE2) mostly excluded women below the age of 40.

But at age 40+, complications of an earlier pregnancy may already have dialled up her traditional risk factors. That makes the mathematical separation of non-traditional from traditional risk factors nearly impossible.

The consequence: When two perimenopausal women have the same, say, 7% 10-year risk, these 7% may underestimate the actual risk of the woman with a history of pregnancy complications, but overestimate the risk of the other woman unaffected by those complications. So before accepting a risk score at face value, every woman’s reproductive history should be checked on potential “risk enhancers”.

Menopause Changed the Cholesterol. Did It Change the Risk?

Remember: this post is about the primary prevention of atherosclerotic cardiovascular disease: plaque-driven heart attacks, ischaemic strokes, and their upstream loss of vascular function.

So at this point, your doctor’s mind moves from risk score to cholesterol number, and potentially from there to statin prescription.

But not so fast. During the menopausal transition, women’s cholesterol numbers typically move up. Reason enough for the sensation-savvy press to draw the picture of a menopause-triggered dramatic increase in risk for heart attacks and strokes. The image below is from the lay-reader-facing Escardio publication of the ESC (European Society of Cardiology).

Unfortunately, the quoted article refers to a poster presentation that we can’t inspect for its methods and results. So we have to believe what Escardio tells us (which is frustratingly little): An 8.3% increase in LDL from pre-to post-menopause (which made it into the title as the “large adverse changes in blood fat profile” .

I never quote from poster presentations or meeting abstracts. Since I can’t check the methods and the results, I have no reason to trust the conclusions. So, below are the results of a properly published longitudinal study of 2659 women with an average of 4 cholesterol measurements taken from pre-menopause to late post-menopause.

What you see is a 4% rise in total cholesterol. Now run the 4% rise through the SCORE2 algorithm for a non-smoking, non-diabetic woman aged 60 with a systolic blood pressure of 140 mmHg. You’ll find no difference in predicted 10-year risk (3.6%). You can run the analysis yourself on the MedCalc site.

If you double the 4% to the Escardio-reported 8%, the 10-year predicted risk of 3.6% marginally increases to 3.7%. Dramatic? You be the judge.

So the correct question is not: “Did menopause raise my cholesterol?”

The correct question is: “Did my cholesterol, together with my vascular history, raise my absolute ASCVD risk enough to treat, test, or trial?”

Treat, Test, or Trial?

“Treat” means (to your doctor) that your risk score is already high enough to justify lipid-lowering therapy. The guidelines tell him/her what “high enough” means. I dissected the AHA/ACC’s latest 2026 guideline incarnation in detail in an earlier post to illustrate why you might want to look at them differently.

Don’t get me wrong, I’m not advocating for or against statins, I’m simply giving you the numbers for you to align with your personal tolerance for risk. The guidelines recommend statin treatment latest after crossing the 5% 10-year risk threshold.

Here is what 5% means: you have a corresponding 95% chance of not suffering the predicted event over the next 10 years. Or in other words, of 100 people with exactly the same risk score as you, 95 will make it through the next 10 years without any event (it’s all in my earlier post).

Since statin treatment typically reduces your relative risk by, on average, 25% (1.25% of the 5%), taking the drug will increase your chance of staying event-free from 95% to 96.25%. That is, 96 instead of 95 people will not suffer an event (or 4 instead of 5 will suffer an event). If you find that justifies a lifelong drug prescription, by all means, go and get treated; and I don’t mean that in any sarcastic way.

On the other hand, if your risk is truly high, you need to know that there are options to the healthcare providers’ statin reflex. I discussed those options in the preceding post about the lipid-lowering ladder that your doctor might not have the time to discuss with you.

“Test” means looking beyond the LDL number. Apo(B), Lp(a), coronary artery calcium (CAC), pulse wave velocity (PWV), and carotid stenosis can all swing your risk score one way or the other. That’s why, in two of my earlier posts (here and here), I discussed how these lab tests’ results move the needle on your risk score (including a downloadable file that does the math for you).

If “treat” and “test” still leave you in doubt about the benefit of surrendering to a statin prescription, or if statins are out of the question for you for reasons of intolerable side effects, the final option is…

“Trial”: that is, trialing a lifestyle intervention. After all, lifestyle choices, especially dietary choices, are often the culprits behind elevated ASCVD risk. There is one caveat to trialing, though. A vague promise of “I’ll eat better” isn’t a trial. It’s a delay tactic fueled by influencers, supplement manufacturers, and health bloggers angling for likes, followers, and affiliate dollars.

A trial needs a defined and evidence-based intervention, a start and end date, and repeat measurements. It should target the LDL pathway. Three months is often enough to see whether lipids are moving in the right direction. If they are, the trial has earned continuation, and probably the respect of your doctor. If they are not, the conversation returns to treatment. I have presented one evidence-based intervention in my earlier post about a daily lipid-lowering anchor meal strategy (breakfast).

It asks your doctor to apply the guidelines to your unique vascular biography rather than a single cholesterol number.

And that is the entire point of female-oriented ASCVD prevention: a better audit before any decision becomes lifelong.

The Clinical Reality Check

Before finalizing this framework, I wanted to field-test it against the realities of a high-level clinical practice. I reached out to Dr. Arash Bereliani, a preventive cardiologist, Clinical Associate Professor at UCLA, and author of the book “What About Her Heart? A New Perspective on Women's Health.”

I asked Dr. Bereliani to audit this question:

How do you as a specialist handle a sudden menopausal lipid spike?

His clinical protocol converges with the “Treat, Test, or Trial” framework. He noted that the menopausal transition can produce a 10 to 15 percent rise in LDL-C that looks alarming on a standard lab report but does not automatically require pharmacotherapy.

Dr. Bereliani:

“In that intermediate zone, I prefer ApoB over LDL-C because it reflects particle number directly. Two women with identical LDL-C can have very different atherogenic burdens.”

He emphasized that variables like Lp(a), a history of preeclampsia, or premature menopause should not simply be used as homemade risk multipliers to justify an immediate statin. Instead, they are high-definition signals that should drive deeper evaluation—like a CAC scan to re-risk the patient—before converting a menopause-accompanying cholesterol shift into a lifelong drug decision.

That convergence of opinions shows that the “female ASCVD risk audit” is not just my (an epidemiologist’s) desk exercise; it is the data a preventive cardiologist focused on women’s health requires before writing a prescription.

So, how do you force this level of precision in a 15-minute consultation? You walk in with the following five questions.

Upgrade to a paid subscription below to unlock your printable 5-question doctor script and gain exclusive access to my custom AI research assistant to interrogate the science yourself.

The Doctor Script: Five Questions Before a Postmenopausal Statin Decision

Doctor, before we decide on long-term lipid-lowering medication, I want to make sure my ASCVD risk is not being reduced to a single cholesterol number. I have five specific questions.

1. The absolute risk check: Which risk score are we using for me — PREVENT-ASCVD, SCORE2, SCORE2-OP, or another tool — and what is my estimated absolute 10-year risk of a heart attack or ischemic stroke?

2. The female vascular timeline: My reproductive history includes / does not include pregnancy complications such as preeclampsia, gestational hypertension, gestational diabetes, preterm delivery, or early/premature menopause. Does that history make you think my risk score may be underestimating or overestimating my real risk?

3. The particle and plaque audit: Before we decide, would ApoB, non-HDL-C, or Lp(a) help clarify my atherogenic particle burden? And if the decision remains borderline, would a CAC scan or other plaque assessment change the recommendation?

4. The absolute benefit: If I start a statin or another lipid-lowering drug, what absolute risk reduction should I realistically expect over the next 10 years? Are we talking about a large benefit, or something more like reducing my risk from 5% to about 4%?

5. The treat, test, or trial decision: Based on my full vascular biography — risk score, lipids, ApoB/Lp(a), blood pressure, metabolic markers, kidney markers, reproductive history, vascular aging, and any imaging that would actually change management — are we ready to treat, do we first need to test, or is a structured 8–12 week lifestyle trial with repeat labs reasonable?

Download the script here.

Dive Deeper: Interrogate the Evidence Yourself

To ensure complete transparency, I have compiled the 30+ peer-reviewed scientific papers and guidelines that form the backbone of this article into an interactive source library. For this post it’s open to everyone, as a sample of what Laboratory members get with every deep dive.

You can access the library here.

Think of this as your personal research assistant, trained exclusively on the papers I used for this post. Ask any question that this post might not have answered (I’m sure there are lots). If they don’t cover your question, the library will tell you so, and it will refer you back to me.

Your privacy: your questions aren’t visible to me or to other readers, but they do run through your own Google account. Describe your situation in numbers and facts only: no names, dates of birth or uploaded lab reports.

Rest assured, your privacy is fully protected: your conversation with this assistant is strictly confidential and completely invisible to me and all other readers/users.

Want the evidence applied to your own lab results? Start with the free fit check on lutzkraushaar.com.

Irma’s Window Closed. Yours May Not Have.

Irma spent the years following her stroke in a nursing home, where she passed away at age 76. Her pregnancy complications, the reason why I remained her only child, had never come up during her perimenopausal ob/gyn consultations in the mid 1970s.

So, download the doctor script and maybe head over to interrogate the interactive AI research assistant.

Ask any question that this post might not have answered (I’m sure there are lots). If the notebook can’t extract the answer from its sources, it will not hallucinate (I made sure of that), it will tell you so, and it will refer you back to me.

Cited References

[1] B. Vogel et al., “The Lancet women and cardiovascular disease Commission: reducing the global burden by 2030,” The Lancet, vol. 397, no. 10292, pp. 2385–2438, 2021, doi: 10.1016/S0140-6736(21)00684-X.

[2] C. Mattiuzzi and G. Lippi, “Leading Causes of US Deaths in the 2022,” J. Clin. Med., vol. 13, no. 23, Nov. 2024, doi: 10.3390/jcm13237088.

[3] R. L. Siegel, A. N. Giaquinto, and A. Jemal, “Cancer statistics, 2024”, doi: 10.3322/caac.21820.

[4] C. N. Bairey Merz et al., “Knowledge, Attitudes, and Beliefs Regarding Cardiovascular Disease in Women: The Women’s Heart Alliance,” J. Am. Coll. Cardiol., vol. 70, no. 2, pp. 123–132, 2017, doi: 10.1016/j.jacc.2017.05.024.

[5] V. De Martin Topranin, A. Wiig-Fisketjøn, E. Botten, H. Dalen, M. Langaas, and A. Bye, “Sex-specific cardiovascular disease risk prediction using statistical learning and explainable artificial intelligence: the HUNT Study,” Eur. J. Prev. Cardiol., p. zwaf135, Mar. 2025, doi: 10.1093/eurjpc/zwaf135.

[6] T. Ashcroft, M. de Bakker, A. Anand, N. Sattar, J. A. Maybin, and D. M. Kimenai, “Tailoring cardiovascular risk prediction to females,” J. Endocrinol., vol. 265, no. 3, p. e250084, Jul. 2025, doi: 10.1530/JOE-25-0084.

[7] A. A. Cohen et al., “Lack of consensus on an aging biology paradigm? A global survey reveals an agreement to disagree, and the need for an interdisciplinary framework,” Mech. Ageing Dev., vol. 191, p. 111316, Oct. 2020, doi: 10.1016/j.mad.2020.111316.

[8] S. Laurent, P. Boutouyrie, P. G. Cunha, P. Lacolley, and P. M. Nilsson, “Concept of Extremes in Vascular Aging: From Early Vascular Aging to Supernormal Vascular Aging,” Hypertension, vol. 74, no. 2, pp. 218–228, 2019, doi: 10.1161/HYPERTENSIONAHA.119.12655.

[9] R. M. Bruno et al., “Early and Supernormal Vascular Aging: Clinical Characteristics and Association With Incident Cardiovascular Events,” Hypertension, vol. 76, no. 5, pp. 1616–1624, 2020, doi: 10.1161/HYPERTENSIONAHA.120.14971.

[10] A. E. Lacroix, H. Gondal, K. R. Shumway, and M. D. Langaker, “Physiology, Menarche,” in StatPearls, Treasure Island (FL): StatPearls Publishing, 2026. Accessed: Jul. 02, 2026. [Online]. Available: http://www.ncbi.nlm.nih.gov/books/NBK470216/

[11] D. Zhu et al., “Age at natural menopause and risk of incident cardiovascular disease: a pooled analysis of individual patient data,” Lancet Public Health, vol. 4, no. 11, pp. e553–e564, Nov. 2019, doi: 10.1016/S2468-2667(19)30155-0.

[12] L. Tschiderer, L. Seekircher, P. Willeit, and S. A. E. Peters, “Assessment of Cardiovascular Risk in Women: Progress so Far and Progress to Come,” Int. J. Womens Health, vol. 15, pp. 191–212, Feb. 2023, doi: 10.2147/IJWH.S364012.

[13] A. D. Lane‐Cordova et al., “Life‐Course Reproductive History and Cardiovascular Risk Profile in Late Mid‐Life: The CARDIA Study,” J. Am. Heart Assoc. Cardiovasc. Cerebrovasc. Dis., vol. 9, no. 10, p. e014859, May 2020, doi: 10.1161/JAHA.119.014859.

[14] S. Vasanthakumar et al., “A Review of Traditional and Non-Traditional Cardiovascular Risk Factors in Women,” Int. J. Womens Health, vol. 18, p. 574244, Feb. 2026, doi: 10.2147/IJWH.S574244.

Your next statin discussion with your doctor shouldn’t be a binary drug-vs-no-drug decision. Statins are now just one of 6 LDL-lowering drug options, aside from precision lifestyle changes. The lipid-lowering ladder is your guide to having a truly informed joint decision-making session about your options.

Why it matters

Your doctor hardly has the time to explain all those options in a routine 15-minute consultation. The lipid-lowering ladder brings you up to speed before you enter the examination room. The side effect: Your doctor respecting you as an equal partner in designing your health journey.

Your takeaways

Aside from getting you up-to-date on the players in the cholesterol-lowering theatre, this post gives you a practical decision kit for that conversation:

• a decision chart for climbing the lipid-lowering ladder
• a doctor-discussion script for asking the right questions
• a prep sheet to organize your numbers, risks, preferences, and fallback options before the visit

When Lifestyle-First Runs Out of Road

In the previous post, I showed you how “I want to try lifestyle first” can become more than a hazy aspiration to eat better, avoid butter, and hope for a friendlier lipid panel next time.

It means rolling up your sleeves and showing the doc exactly why their insistence on a statin could be wrong—using hard evidence from an actual lifestyle experiment (that’s why we built a lipid-lowering breakfast smoothie based on hard science, not TikTok fiction).

If your next lipid panel still leaves you on the wrong side of risk, you’ll need to explore the next rungs in what I call the lipid-lowering ladder of decision-making. Statins are just one of that ladder’s 7 rungs. If that number comes as a surprise to you, it’s because statins were just the first drugs out of the cholesterol gate and the only game in town for decades. That’s no longer the case.

Once LDL-lowering is medically justified, your real question is no longer:

“Statin or no statin?” rather than:

“Which lipid-lowering strategy gives me the best profile of risk, benefit, cost, and tolerability?”

The extra rungs on that lipid-lowering ladder include

• Bile acid sequestrants, turning your smoothie’s bile acid trap into a drug

• Ezetimibe, the intestinal cholesterol-absorption blocker.

• bempedoic acid, marketed as a statin-intolerance escape hatch

• PCSK9 inhibitors, injectable drugs that can nosedive LDL

• Inclisiran, a different flavour of PCSK9 inhibition.

More options create more ways to make crappy choices. Particularly if you do not know which problem you are trying to solve.

• Lowering LDL because your absolute cardiovascular risk is high?

• Lowering ApoB because your particle burden is the issue?

• Working around muscle symptoms or other side effects?

• Trying to keep costs sane?

• Buying time for a structured lifestyle experiment?

So this post is not about “statins good” vs. “statins bad”. You have seen enough of those. This post is about the 7-rung lipid-lowering ladder as a novel decision framework. It is not a recommendation to take or reject any of its drugs.

This post is for you to enter into an informed joint-decision-making discussion with your doctor when he/she lacks the time to bring you up to speed on the statin alternatives.

Now, before we climb that ladder,…

A disclaimer upfront: this article does not recommend for or against taking statins or any of their discussed alternatives. It is not about telling you to accept or decline a prescription. It is about helping you to move from confusion to confidence, making a better-informed shared decision under medical uncertainty.

The False Binary: Statin or Nothing

The public debate about statins mostly boils down to a binary battle between two camps (each promoting its message through reasonably credentialed doctors with plus-sized egos nourished by social media fandom).

The “doctor-always-knows-best” camp pontificates to you to “take the statin and stop overthinking it”. In their world, elevated LDL is the problem, and statins are the (only) answer. They treat doubt as irrational and side effects as mostly imagined.

For the “cholesterol is innocent and statins are poison” camp, rejecting the drug is a badge of independent thinking. That camp sees in every guideline and trial a pharma-contaminated exercise in massaging the statistics.

That neither camp is telling you the whole truth has been the subject of my earlier posts.

Both camps commit what statisticians call an “ecological fallacy”. They draw inferences about you from trial data averaged over groups to which you (may or may not) belong. In my earlier posts, “To Statin or Not to Statin”, “As An Epidemiologist, This Is What Ticks Me Off About Statin Prescriptions”, and the latest “Sitting on the Statin Fence,” you’ll find downloadable decision trees, a scorecard, and links to my online risk calculators to translate the US and European guidelines into your own situation.

This post is the logical evolution: If your risk justifies a prescription, what are the statins’ six serious alternatives that your doctor might not find the time to explain to you?

A caveat: this post focuses on drug monotherapies in the primary prevention theatre. Combination therapies in people with a history of cardiovascular disease events are an entirely different and immensely more complex ballgame. Discussing it would break the boundaries of a blog post, but you can always ask me.

First, Decide Whether LDL Lowering Is Worth Doing At All

Your LDL value in isolation fails to deliver the answer. 150 mg/dL LDL in a 42-year-old endurance athlete with low blood pressure, low ApoB, CAC zero, and no family history is an entirely different story from the same LDL load in a 64-year-old diabetic smoker with elevated ApoB and a stratospheric CAC score.

Same number. Different risk.

The medical associations’ risk scoring tools (AHA/ACC’s PREVENT-ASCVD algorithm & ESC’s SCORE 2) ignore most of those risk modifiers. For a valid reason. Their scores need to strike a balance between population-wide accessible screening and reasonably reliable risk estimation. The drawback: Your personal profile hides in their blind spot.

That is why my aforementioned posts help you estimate whether your cardiovascular risk is high enough to justify drug-based LDL lowering. They contain downloadable decision trees, scorecards, and links to bespoke online calculators for you to:

• Estimate your net risk considering your risk modifiers (apoB, Lp(a), CAC score, etc.), which may move your score-based risk estimate up or down.

• Calculate your achievable absolute risk reduction (as opposed to the typically touted relative risk reduction that always sounds and looks better)

• Calculate the number needed to treat (NNT, how many people with your risk profile need to take the drug to prevent a single heart attack or stroke in one of you)

• Calculate your real LDL (because your lab values are mostly estimated, not measured)

You need to look at these tools’ outputs through the lens of your personal risk tolerance. If you are happy to take a drug that improves your 10-year absolute chance of staying event-free from 94% to 96%, then by all means, take it. That’s what the often-touted 30% (relative) risk reduction actually translates to if your baseline CVD risk (cardiovascular disease) stands at 6% (the second-highest risk category, “intermediate,” as per the AHA’s PREVENT-ASCVD calculator).

If, on the other hand, those numbers compel you to try a non-pharmaceutical lifestyle change first, welcome to the first rung on the 7-rung lipid-lowering ladder.

BTW: This ladder is a decision map, not a prescription sequence. Guidelines typically treat statins (rung 4 on our ladder) as the first pharmacological recommendation. The ladder doesn’t replace that logic. It applies a physiological sequence to help you understand the different mechanisms of cholesterol reduction. And it starts, where else, in the gut.

Rung One: Lifestyle as a Real Trial, Not a Ritual.

For a lifestyle intervention to successfully lower cholesterol, it ought to take a page from the drug developers’ script: target a specific node in the cholesterol pathway to derail the body’s LDL production line.

Compared to that strategy, the “exercise more, eat better, lose weight” mantra is nothing but a shot in the dark. It may work if, just by chance, you correct one of the cholesterol-elevating mechanisms that you triggered with your lifestyle habits.

So, why not target one mechanism that works in all of us, irrespective of our lifestyle?

That’s what I did in my preceding post. It presents a breakfast-replacing smoothie designed to starve the liver of recycled LDL, forcing it to soak up more cholesterol-carrying vehicles from your circulation. The net effect is less cholesterol in the blood [1].

The liver uses cholesterol to make bile acids. The gall bladder releases them into the intestine to help digest the fats from your diet. Under normal circumstances, 95% of bile acids are reabsorbed and recycled back to the liver [2].

Viscous fibers such as psyllium and beta-glucan trap bile acids in the gut and escort them out the back door. The liver makes up for the lost recycling material by pulling cholesterol from the blood circulation.

But if the sanctity of your daily butter croissant beats the health prospect of a smoothie (or any similarly targeted dietary makeover), no worries, your next-door pharmacy sells a class of drugs that pull the same lever, only with a lot more oomph (plus the need for a prescription). Which brings us to the second rung on our lipid-lowering ladder.

Rung Two: Bile Acid Sequestrants — Turning the Bile-Acid Trap Into a Drug

If the lipid-lowering breakfast smoothie was the dietary version of the bile-acid trap, bile acid sequestrants (BAS) are the pharmaceutical equivalent.

The mechanism is the same; the effect size isn’t. Neither is the side effect profile [3].

Since bile acid sequestrants lower LDL without entering your bloodstream, you might think of them as an elegant solution: No systemic drug exposure, no potentially damaging muscle metabolism, and no liver enzyme interference.

But they aren’t elegant to live with: Constipation, bloating, nausea, and flatulence are the most frequently cited reasons for discontinuation.

Your doctor might not like them either, particularly if you are taking certain other drugs whose absorption is inhibited by the BAS (e.g., thyroid medication, warfarin, digoxin, certain blood pressure medications). You’d need to space the drugs from the BAS by several hours to avoid interference. BAS also have a track record of increasing triglycerides, particularly in people with the insulin-resistant, high-triglyceride, low-HDL phenotype.

But BAS aren’t the only option if you want to restrict the drug’s action to the gut.

Rung Three: Ezetimibe — Blocking Cholesterol at the Intestinal Gate

Ezetimibe works at the next intestinal checkpoint: a transporter (the NPC1L1 protein) anchored to the intestinal wall moves cholesterol (from your diet and from bile acids) out of the gut into the systemic circulation. Block the transporter, less cholesterol reaches the liver, and … by now you know how the liver responds to that.

Ezetimibe typically lowers LDL by about 15–20% [3]. So, if your LDL is sitting just above your desired target, Ezetimibe may be an attractive option because, compared to the BAS, it is easier on your gut than BAS, has relatively few drug interactions, and fewer side effects.

That’s what makes Ezetimibe attractive. It is less attractive when the LDL or ApoB problem is large, the risk is high, or the required reduction is substantially higher than what Ezetimibe can deliver.

The next logical rungs on the ladder belong to the drugs that dial down LDL production.

Rung Four: Statins — The Benchmark Everyone Loves to Fight About

Statins inhibit a bottleneck enzyme in the liver’s cholesterol-assembly line (HMG-CoA reductase). When that happens, the liver compensates by increasing its LDL receptor activity, pulling more LDL particles out of the bloodstream.

Depending on the type and dose of statin, LDL-C reductions range from roughly 20% at the low-intensity end to up to 65% with high-intensity therapy [4]. ApoB and non-HDL-C usually fall in parallel, which is one reason statins remain central in cardiovascular prevention.

Statins are still the benchmark on the lipid-lowering ladder because, among oral LDL-lowering drugs, they remain the most widely studied agents with the largest outcome-trial record.

They also remain the most fiercely debated drug. That is why I keep coming back to them.

In my earlier post, To Statin Or Not To Statin: What Neither Side Is Telling You, I framed the statin decision as biology poker. Whether a heart attack or stroke happens depends on a ‘perfect storm’ scenario, where everything that can go wrong does go wrong… at the same time [5]:

A vulnerable atherosclerotic plaque needs to rupture, the rupture needs to trigger the formation of a blood clot large enough to block an artery, and the body’s clot-busting mechanism (fibrinolysis) needs to be AWOL long enough that the oxygen-deprived heart muscle tissue downstream of the blockage dies.

Plaque ruptures are far more frequent than perfect storms, which is why evidence of ruptures is found in most people with atherosclerotic plaques, but only a minority of them suffer an actual infarction. We still can’t predict which plaque is going to rupture and when. But one thing is for sure: more plaques equal more tickets in the perfect-storm lottery.

A statin recommendation makes much more sense when baseline risk is high enough for the absolute risk reduction to matter. It makes less sense when the expected benefit is tiny, the risk estimate is crude (population-based), and you are being pushed into chronic medication because an age-heavy algorithm crossed a threshold.

That is why I spent an entire post (As An Epidemiologist, This Is What Ticks Me Off About Statin Prescriptions) on how to recalibrate your risk with ApoB, Lp(a), CAC, CCTA, carotid plaques, and vascular aging markers before blindly accepting a prescription.

Then, in “Sitting on the Statin Fence? Here’s How to Decide”, I showed you how to enter that discussion with your doctor based on the medical associations’ 2026 guidelines on hyperlipidemia.

One obvious caveat for all drugs is tolerability. The statin fanboys, led by the pharma-funded Cholesterol Treatment Trialists (CTT), published a paper in 2026 that basically instructs your doctor to reject your complaints about muscle symptoms (the most frequently reported side effect) as imagination [6]. I dissected the CTT’s claims in “The Statin Paradox: Why Your Doctor Thinks Your Muscle Pain is Fake”. That post also gives you an evidence-based symptom score calculator and a tool to find out whether your symptoms are a plausible statin signal or just background noise.

The important point here is: If you repeatedly develop symptoms on rechallenge, the question is justified whether that specific statin strategy really fits you.

Statins also come with the “Hotel California” problem: “you can check out any time you like, but you can never leave”.

Guidelines are very good at telling doctors when to prescribe statins. But they keep mum about deprescription. This “once on a statin, always on a statin” nimbus of inescapable lifelong prescription is a big part of patients’ hesitancy. That is why I wrote, “Are Statins the ‘Hotel California’ Drugs of Medicine?” Not as an invitation to quit medication, but as a protocol for discussing your options with your physician.

And you certainly have options, which brings us to the next rung on the ladder.

Rung Five: Bempedoic Acid — A Statin Escape Hatch With Its Own Baggage

Like statins, bempedoic acid reduces cholesterol synthesis. But it does so by inhibiting an enzyme upstream of the statin target (ATP citrate lyase, ACLY) [7]. What makes bempedoic acid immediately attractive is that, being a so-called prodrug, it has to be activated by an enzyme in the liver before it can do its job. Since skeletal muscle lacks that enzyme, the statin-associated muscle pain isn’t much of an issue with this drug.

Lowering LDL, like statins, without affecting muscle tissue sounds too good to be true. It isn’t, but it is incomplete. Bempedoic acid has a smaller effect size and (still) a much smaller evidence base. It lowers LDL by about 20%, and CVD risk by up to 15% [8]. BUT that 15% relative risk reduction translates to only 1-1.5% absolute risk reduction. These numbers come from one study, the CLEAR trial, which investigated statin-intolerant patients with a high baseline risk (>10%) for CVD events.

So, on the upside:

It is strictly liver-activated. It can lower LDL-C by an amount that sits roughly in the same neighborhood as ezetimibe. And it can be combined with ezetimibe, which makes the two-drug oral strategy more powerful than either alone. For someone who cannot tolerate statins and still needs meaningful LDL reduction, that is an option to discuss with their doctor.

On the downside:

Because it can raise uric acid, it can trigger gout and gallstones in susceptible people. In fact, while the CLEAR trial participants on bempedoic acid had a 1-1.5% lower risk of non-fatal CVD events, their risk for gout and gallstones increased by an identical amount [9]. The FDA has issued a warning that bempedoic acid should not be prescribed to people with a history of gout, high uric acid, or tendon ruptures [8].

To sum it up: bempedoic acid can be useful when statins are poorly tolerated or insufficient. But they are a poor fit when the expected absolute benefit is small, and the risk of gout, tendon rupture, or gallstones is elevated.

Up to now, our drug strategy has been focused on putting the brakes on LDL production. With the next rung, we pivot to turbo-charging LDL clearance.

Rung Six: PCSK9 Inhibitors — Turbocharging LDL Clearance

The PCSK9 inhibitors’ target is LDL receptor recycling.

Your liver clears LDL particles from the blood through LDL receptors. Those receptors sit on liver cells, grab LDL particles, pull them inside, release the cargo, and then usually return to the cell surface for another round of cleanup.

The protein PCSK9 interferes with that recycling. It tags LDL receptors for degradation. Fewer recycled receptors mean fewer LDL particles are cleared from the blood.

PCSK9 inhibitors (such as evolocumab—Repatha, and alirocumab—Praluent) prevent the degradation of LDL receptors, thereby reducing LDL-C in the circulation by 45% to 70% on drug monotherapy and an additional 50% to 60% when added to a patient’s maximally tolerated statin therapy [10], [11], [12]. They also reduce ApoB and Lp(a) modestly.

While that sounds impressive, keep in mind that LDL reduction is only a surrogate marker for what we are really after: a reduction in event risk. While PCSK9 inhibitors reduce the risk of non-fatal major adverse cardiovascular events (MACEs), they (a) don’t match statins and (b) haven’t shown any reduction in the risk of dying from MACEs.

These questions are particularly interesting given the drugs’ spit-out-your-coffee price tag. The annual wholesale acquisition costs (WAC) for evolocumab and alirocumab are around US$6,000. Compare that with the statins’ price range from $50 to $150 (you’ll find the references in the table at the end of this post).

The outcome evidence is strongest in high-risk patients, especially in secondary prevention and patients whose LDL remains too high despite standard therapy.

The drawbacks: mode of delivery (by injection every two to 4 weeks) and (very high) costs

To sum it up: PCSK9 inhibitors make most sense when the LDL/ApoB problem is large, the cardiovascular risk is high, and cheaper or simpler rungs have not done enough—or are out of the question for other valid reasons.

They are a poor fit when the expected absolute benefit is small, when the problem could be solved with lower-burden tools, and when the extra LDL reduction is unlikely to translate into a meaningful reduction of MACE outcomes.

On the next and final rung of our lipid-lowering ladder, we move from inhibiting PCSK9 to dialing down its production.

Rung Seven: Inclisiran — Silencing PCSK9 at the Source

Whenever you see a generic drug’s name ending in -siran (such as inclisiran), you immediately know it is a gene-silencing RNA medication. The suffix -siran is the official USAN/WHO drug stem reserved for small interfering RNA (siRNA) therapeutics. So ‘siran’ for siRNA is an absolutely logical choice, …if you are slightly dyslexic, or under the influence of God-knows-what while doing your baptizing job.

Inclisiran’s job is to enter liver cells and reduce the production of PCSK9 in the first place. Less PCSK9 means less LDL-receptor destruction, which means more LDL particles are cleared from the blood.

The cholesterol-lowering effect is in the statin and PCSK9i ballpark (roughly 50% reduction) with injections given at baseline, again at three months, and then every six months thereafter. The costs (WAC) for the first-year treatment are close to US$10,000, and US$ 6,000 for the maintenance years.

The caveat: While Inclisiran lowers LDL-C and the risk for adverse events (AEs) to a similar extent as PCSK9 inhibitors, no benefits have yet been reported for fatal and non-fatal heart attacks and strokes [13].

If you find these results underwhelming, I fully agree with you. Particularly since all studies presenting those results have been conducted under pharma funding, which typically adds a positive sheen to the statistical outcomes.

Now that you know the ladder, it is time to put it all together into a decision flow chart about how to climb it.

The Ladder Is Not a Ranking — Here Is How to Climb It

The following decision aid is a tool to kick the tires of lipid-lowering options before buying blindly into your doctor’s statin spiel. The first question is not which drug comes next, but which problem you are trying to solve? I created it to help you prepare for evidence-based joint decision-making with your doctor.

Paid subscribers get the full decision aid behind the paywall, plus the downloadable Lipid-Lowering Ladder Decision Kit: the printable chart, doctor-discussion script, prep checklist, and comparison tables.

They also get access to the companion NotebookLM evidence notebook, where you can interrogate the 38 scientific sources behind this post instead of merely reading my interpretation of them.

Read more

What is New

If you are sitting on the statin fence, “I want to try lifestyle first” is not enough. This post turns a vague promise into an evidence-based breakfast intervention: a modular Anchor-Meal smoothie designed around viscous fiber, plant sterols, protein, unsaturated fats, polyphenols, nitrate-rich inputs, and measurable lipid outcomes.

Why it Matters

Lifestyle advice usually fails because it asks you to improve your entire diet at once. This approach targets one repeatable meal instead: breakfast. Build it correctly, use it daily for 8–12 weeks, and your next lipid panel can tell you whether your non-drug experiment moved the needle enough to matter.

Your Takeaways

• The Plumbing: How viscous fibers and plant sterols physically intercept cholesterol in your gut and prevent it from entering your circulation.

• The Rules: The five non-negotiable ingredient modules required to move the needle on your next lipid panel.

• The Blueprint: A modular matrix to build a personalized, evidence-based smoothie that will not trigger a glucose spike.

• The Interrogation Room (for paid subscribers): An interactive AI notebook for you to tailor The Blueprint to your exact metabolic constraints and biomarker profile.

Table of Contents:

Sitting on the Statin Fence
What this Smoothie Can and Cannot Do
Every Ingredient Has a Job
Module 1: Build the bile-acid trap
Module 2: Block cholesterol at the gate
Module 3: Renovate the fat-protein architecture
Module 4: Add the endothelial shield
Module 5: Prevent the smoothie from becoming dessert
The Modular Lipid-Lowering Anchor Meal Recipe
The Verdict: Test it for 12 Weeks, Then Check the Panel
Cited References

Sitting on the Statin Fence

There is this pain-in-the-butt place on the statin fence where you sit when your doctor tells you about your cholesterol level being too high, but your willingness to accept a lifelong statin prescription is too low to act. I understand that position. That’s why, in my preceding post, I gave you the tools and framework to make a better-informed joint decision with your doctor.

If yours is “I want to try lifestyle first”, this post is for you.

Because far too often, that decision is nothing more than a vague promise to eat better, buy oats, banish butter, and pray to the cholesterol Gods for a miraculously improved next blood test.

You may talk yourself into believing that this is a strategy. It isn’t. It’s a delay tactic fueled by influencers, supplement manufacturers, and health bloggers angling for likes, followers, and affiliate dollars.

If you want your diet choices to move the needle on your lipid profile, there is good news and bad news. The bad news is that your nutrition has to be re-engineered like an intervention. The good news is that the re-engineering doesn’t need to upend all your diet habits. Only one.

In an earlier post, I introduced the concept of the Anchor Meal. The one daily meal designed to get you covered against the nutrient deficiencies to which all the other daily meals (aka your regular diet) might expose you. The one daily meal that for most of us is the same every day: Breakfast! After all, hardly anybody wakes up every morning contemplating what to have for breakfast.

What if breakfast could be redesigned to help you:

• lower your cholesterol and triglycerides,

• deliver your daily dose of (almost) all required nutrients, and

• maintain your other meal preferences

That Anchor Meal is your personalized breakfast smoothie. It’s designed to replace breakfast, not just accompany it. You build it once, make it your daily routine, and have an evidence-based shot at meaningfully lowering your cholesterol.

I originally developed the Anchor Meal smoothie for decelerating vascular aging, not for dialing down cholesterol. The principle, however, applies mutatis mutandis. In fact, we’ll design a smoothie that combines both objectives.

So, let’s break down how to build it around viscous fiber, (plant) protein, unsaturated fats, polyphenols, nitrate-rich inputs, and what the minimum ingredient thresholds are to actually move the needle on your next blood test.

But before we start, I need to caution you about …

What this Smoothie Can and Cannot Do

It can help you run an evidence-based non-drug N-of-1 experiment. It can increase viscous fiber, displace saturated-fat-heavy breakfast choices, improve protein and micronutrient density, raise polyphenol and nitrate exposure, and give you a repeatable daily meal whose physiological effects can be checked against your next lipid panel.

It cannot override familial hypercholesterolemia, erase a high Lp(a), undo decades of atherosclerotic plaque, or replace medication in someone whose absolute cardiovascular risk is high enough to justify pharmacological LDL lowering.

In other words, this is not a statin substitute. It is a version of the 2026 ACC/AHA multisociety dyslipidemia guideline’s “lifestyle first” routes that deserves to be taken seriously because it is evidence-based, individualized, and sustainable.

Every Ingredient Has a Job

How to make a smoothie healthy would be the wrong question to ask. Every TikTok weirdo has their personal, and often wacky, idea of what constitutes ‘healthy’. The right question is: does the smoothie contain (a) the required functional modules at (b) doses high enough to matter?

In our case, that means every component must earn its place in one of the five modules that canonize the smoothie to a potentially protective anchor meal.

Module 1: Build the bile-acid trap

The bile acid trap is a sneaky workaround. The liver produces bile acids using cholesterol as “raw material”, and pumps them into the gall bladder. The latter releases the bile into the intestine once your food delivers fat and protein.

Most of the bile acids in the gut are reabsorbed and recycled by the liver, which dispatches them back into the gall bladder. Block that reabsorption, and the liver needs to replace the missing bile acids, by, you guessed it, pulling cholesterol from the blood into its bile production line [1].

In short:

1. The liver converts cholesterol into bile acids.

2. Bile acids are secreted into the intestine to help absorb fat.

3. Normally, most bile acids are reabsorbed and recycled back to the liver.

4. Viscous fibers interfere with that recycling by retaining bile acids in the intestinal content or binding them directly.

5. The liver then needs to synthesize more bile acids from circulating cholesterol, which lowers blood LDL-C over time.

The most effective traps for bile acids are viscous fibers such as the beta-glucans of oat and barley [2] and psyllium [3]. Their effect is not as dramatic as the effects of statins. The drugs directly dial down the liver’s bottleneck enzyme for cholesterol production (HMG-CoA reductase). It is the difference between compromising a factory’s recycling loop and shutting down the assembly line altogether.

Of course, the dose matters. A decorative spoonful of oats makes your smoothie as therapeutic as a leaf of lettuce makes your Big Mac a Mediterranean Diet meal.

For oat or barley beta-glucan, the practical target is about 3 grams of beta-glucan per day [2]. Depending on the product, that may require a meaningful portion of oat bran, barley beta-glucan powder, or another concentrated source.

The equivalent Psyllium dose is somewhere between 5 and 10 grams. While 10 grams per day is closer to what is often used in cholesterol-lowering studies, some people need to slowly escalate their intake to tolerate it better [3].

A practical consideration: viscosity is the point, but it also affects the texture of your smoothie. Add too much, and it becomes gluey. It can also delay or reduce the absorption of some oral medications if taken too close to them. So, it is generally best to take medicines at least 2 hours before or 2 hours after psyllium.

For my lipid-lowering version of the anchor meal, this module is non-negotiable. You may choose the vehicle: oat beta-glucan, barley beta-glucan, psyllium, or a combination. But if your goal is to test whether breakfast can move LDL-C, non-HDL-C, or ApoB in the right direction, the bile-acid trap has to be there. Otherwise, you are not really designing a lipid-lowering smoothie. You are just blending a healthier breakfast.

Module 2: Block cholesterol at the gate

If viscous fiber breaks the bile acid recycling loop, plant sterols and stanols directly block cholesterol entry. They do that by competing with cholesterol for the vehicles that taxi cholesterol to the gut’s wall for uptake into the circulation.

Think of sterols and stanols as botanical lookalikes of cholesterol. The taxi driver thinks it’s cholesterol, so he accepts the fare, drops the lookalike at the border, and leaves a legitimate passenger (cholesterol) stranded in the gut [4]. And “stranded in the gut” is a euphemism for “down the toilet”.

The border guards are not as blind as the taxi drivers. They recognize the ‘fake’ cholesterol and throw it back into the gut (aka the toilet destination). So, neither the cholesterol nor the sterol/stanol shows up behind the border in the circulation.

The less cholesterol enters the circulation, the more leaves with the stool, the more the liver has to compensate by pulling LDL cholesterol out of the circulation.

How large is the effect? Large enough to significantly (in statistical terms) lower blood cholesterol. The National Lipid Association boiled it down to a 7–10% reduction per 2 grams of sterols/stanols.

But throwing a few nuts, seeds, or vegetables into your blender won’t do the trick. To make this module work, you need a fortified product or a plant sterol/stanol powder to deliver that dose. If the product provides 2 grams per serving, add one serving to your smoothie. If it provides less, adjust the dose. The legitimacy of this module is in its well-established evidence [4].

There is a caveat, though (as if there ever was no caveat):

Plant sterols and stanols are not for people with sitosterolemia.

That’s a rare genetic disorder. Contrary to ‘normal’ people, carriers of a particular gene have ‘blind border guards’. They let sterols/stanols pass into the circulation as if they were ‘real’ cholesterol with the added advantage of diplomatic immunity. No obstructions at all. Since the liver can’t shove those sterols/stanols into the bile acid department, they accumulate in the blood, ramping up blood cholesterol to levels that can become indistinguishable from familial hypercholesterolemia, often leading clinicians on a diagnostic wild goose chase [5].

So, unless you are one of the very rare sitosterolemia specimens, if you want the strongest LDL-focused version of the anchor smoothie, add a properly dosed plant sterol/stanol module to the bile acid trap.

Module 3: Renovate the fat-protein architecture

It is about replacing saturated with unsaturated fat, refined with complex carbohydrates, and adding more protein and whole-food matrix.

That’s where the nuts, seeds, and plant proteins come in. They deliver unsaturated fats, fiber, minerals, phytosterols, and polyphenols in a compact whole-food package. But ‘replacement’ is the magic word. Nuts instead of a croissant are a lipid intervention. Nuts on top of that croissant are just extra calories wearing a phony halo. That’s a reminder why our anchor meal smoothie should not be an add-on to your breakfast but its replacement.

Pistachios, walnuts, and almonds are the most effective tree nuts for improving cholesterol profiles [6], [7].

Chia seeds affect cholesterol metabolism through several pathways [8]:

Soy protein, specifically soy milk, may lower cholesterol by dialing down the liver’s synthesis of Apolipoprotein B (apo(B)), the structural protein that is part of all non-HDL cholesterol [9]. In my preceding post, I explain apo(B)’s role as a diagnostic tool to gauge the atherosclerotic payload of your cholesterol panel.

Module 4: Add the endothelial shield

The first three modules target lipids. They are about trapping bile acids, reducing cholesterol absorption, and renovating the fat-protein-fiber architecture of breakfast.

But your cardiovascular disease risk (CVD risk) isn’t driven by cholesterol only. It’s largely driven by how permissive the arterial walls are to being damaged by cholesterol. That’s a function of vascular aging (relative to chronological age). I have discussed this issue in the first post about the Anchor Meal, which I designed specifically to decelerate or even reverse that aging process (to a degree).

That post presented a smoothie focused on: polyphenol density, nitrate-rich vegetables, berries, cocoa, avocado, beetroot juice, potassium-rich foods, essential fatty acids, extra-virgin olive oil, and bioavailability tricks such as lemon juice, turmeric, and black pepper.

They are not cholesterol-lowering weapons in the same sense as psyllium, beta-glucan, or plant sterols. But the synergy is obvious. It’s about strengthening the endothelial shield while reducing the potentially damaging firepower of the blood lipids.

I will not rebuild the argument of the first post here. Please refer back to it for clarity about the smoothie’s vascular aging design and issues like how to optimize texture (think frozen fruits and avocados).

So, for this module, I keep the rule simple. A lipid-lowering anchor smoothie should include at least one nitrate-rich input, at least one dark berry or cocoa-based polyphenol source, and at least one potassium-supporting ingredient.

The next and final module exists because …

Module 5: Prevent the smoothie from becoming dessert

If your triglycerides are high, your waist is expanding, and your blood glucose is elevated, then the smoothie has to be designed with glycemic and caloric control in mind.

It needs enough protein to slow gastric emptying and maintain satiety. It needs viscous fiber to blunt absorption and support lipid control. And it needs low-sugar fruits, preferably berries, rather than a tropical-fruit overload.

That brings us to the actual recipe.

Paid subscribers have access to the Modular Anchor Meal Blueprint and the interactive AI Interrogation Room to tailor their breakfast smoothie to their metabolic constraints and biomarker profiles.

Example Prompts:

• “I am a 55-year-old postmenopausal woman with elevated LDL and pre-diabetes. Using the blueprint, calculate a 5-module Anchor Meal that hits the 3g beta-glucan threshold without triggering a postprandial glucose spike.”

• “Summarize the mechanism of the bile-acid trap based on the uploaded clinical literature.”

• “I cannot tolerate psyllium husk. Based on the matrix, what is my exact required dose of oat bran to achieve the same mechanical lipid-lowering effect?”

Read more

What’s new: The 2026 cholesterol guidelines rewrote your doctor’s script for heart attack and stroke prevention. New algorithms and “risk enhancers” make it easier than ever to justify lifelong lipid-lowering therapy.

Why it matters: Millions of healthy adults are now caught in a statistical “grey zone.” Population-based algorithms can classify your statistical risk, but they cannot tell you if your physical arteries actually show disease, or if a lifelong daily pill will yield a meaningful benefit for your specific biology.

Your takeaway: A step-by-step audit to asking for the right advanced testing, and an examining room script for better-informed shared decision making about your treatment plan. The goal is not to escape statins. The goal is to escape a bad decision.

The Low-Risk Illusion

Maybe you are sitting on the fence: your heart disease risk score is borderline or elevated, but you are free from ASCVD. You wonder whether to surrender to the potentially lifelong statin prescription that your doctor tries to sell you on.

Or your score is low, your doctor says you don’t need a statin (at least not yet). But you don’t trust the score because it ignores biomarkers that may reveal whether you are as safe as the calculator claims. If either one is you, this post is for you.

A warning upfront: this article does not recommend for or against statins. It is not about telling you to accept or decline the prescription (when your doctor brings it up). It is about helping you to make a truly informed shared decision with your doctor when a risk score extracted from 6 million people’s averaged data doesn’t jibe with your personal risk expectation.

Or in short:

In my preceding post, we went 20 Matryoshka dolls deep into the AHA/ACC’s 2026 123-page statin guidelines [1].

We uncovered how a change in cardiovascular risk algorithms can reclassify the same person without that person changing: High risk in 2013 under the Pooled Cohort Equations, back toward lower risk with the PREVENT equations in 2023, and now pushed toward “needs a statin” by the 2026 guideline that re-interprets what low-risk actually should mean.

Same person. Same blood. New label. It boils down to:

That’s what I meant by “The Low-Risk Illusion”. The image below summarizes the 2026 Guidelines that we addressed in the previous post.

In Part 1, we asked the question:

“Should you trust that risk algorithm?”

A justified question. The algorithm rests on inputs and proxies that may not describe you as well as your lab report suggests.

LDL-C, the chief villain of the statin story, is estimated rather than measured. So is eGFR, the new kid on the block, that ropes kidney function into the risk prediction game. It easily misclassifies people whose body size, muscle mass, or creatinine metabolism does not fit the ‘average-patient’ assumption.

That is why I gave you calculators to audit these two numbers before accepting the risk category.

Here, in part 2, we will ask the follow-up question:

“What can you do when you doubt the algorithm’s conclusions?”

(be that for or against accepting a statin prescription).

The CPR Loophole — When “Personalization” Becomes a One-Way Ratchet

The guidelines introduce something that, on first blush, sounds like a desirable deviation from one-size-fits-all medicine: the CPR framework for risk evaluation—Classify-Personalize-Reclassify.

It acknowledges that a risk equation extracted from 6 million people’s averaged data may not fit your personal profile. And who can argue with personalized medicine (not me, to be sure, since N-of-1 is my North Star, as you know from many of my earlier articles).

Yet I am still suspicious about that CPR acronym. Did its creators deliberately use it to evoke the life-saving image of cardiopulmonary resuscitation? My snarkier alter ego certainly thinks so.

As you’ll see, the one thing the guidelines’ CPR resuscitates is the flow of statin prescriptions to the 17 million US adults that the PREVENT equations had snatched away from statin eligibility [2]: The ones the PREVENT equations migrated from “borderline” or “intermediate” risk to the low-risk strata (you’ll find the numbers and illustration in my preceding post).

Here is how CPR works:

Step 1: Calculate & Classify: The doctor runs your numbers through the PREVENT calculator. The result determines in which of the AHA’s four risk buckets you land.

Step 2: Personalize: If your 10-year risk for ASCVD comes back at 5% or more, the statin prescription (or lipid-lowering therapy, LLT) is a foregone conclusion. Personalization plays out primarily in the Borderline and Low-Risk buckets, where doctors should check for “risk enhancers”.

Since risk enhancers are not part of the PREVENT equation, their presence (or absence) doesn’t change your risk score. What changes is the risk narrative.

For example: “Yes, your risk is 3%, BUT since your Lp(a) level is elevated, your actual risk is probably higher.” How much higher? The guidelines keep mum on that, which is a bit surprising because research has answered that question.

Or maybe it’s not so surprising once you discover that below-population-average Lp(a) levels (aka the absence of the “risk enhancer”) may actually downshift your risk.

But the guidelines explicitly forbid this:

“The absence of risk enhancers should not be taken to indicate lower ASCVD risk than calculated using PREVENT-ASCVD equations and does not warrant delay or avoidance of LLT.”

Well, you might disagree with this statement once you see the actual numbers in the Lp(a) section further down (and you might then also sympathize with my snarkier alter ego).

I chose the Lp(a) example because Lp(a) is the most atherogenic lipoprotein floating through your arteries. That’s why this month the AHA published their Lp(a) Discovery Project to raise awareness and testing [3].

The 2026 guidelines set up “Personalization” as a one-way risk ratchet. And it works on the low-risk bucket, too. While the CPR flow chart indicates no personalization in that group (aka no search for risk enhancers), the text says otherwise:

“Risk enhancers may also identify individuals with low estimated risk (<3%) by PREVENT-ASCVD equations who may be reasonable for consideration of LLT.”

That brings us to…

Step 3: Reclassify

That’s where you and your doctor are supposed to address “clinical or patient uncertainty”. Sounds good, but it will only be as good as you are prepared for that process. So, let’s get into those risk enhancers, starting with the new one that you might not have heard of yet: CKM.

The CKM Dragnet: When Risk Factors Get Counted Twice

The concept of the Cardiovascular-Kidney-Metabolic Syndrome is about acknowledging that heart attacks and strokes don’t just boil down to derailed fat metabolism and diabetes (the Cardio-Metabolic Syndrome). Equally insufficient is the cardio-renal syndrome’s tunnel-vision (the heart-kidney axis).

The CKM syndrome takes a refreshingly holistic view. It describes a vicious cycle of deteriorating health that you can enter at any point, but people typically kick-start it with excess and dysfunctional body fat [4]. That is, fat around organs that should be lean (liver, heart, kidney).

This visceral fat can trigger systemic inflammation, oxidative stress, and insulin resistance. A problem in one of the CKM’s three systems fuels impairments in the others. Rather than treating individual symptoms in medicine’s subspecialty silos, CKM helps to focus on early holistic interventions that protect the three systems simultaneously. At least, that’s the idea behind the diagnostic staging of CKM syndrome.

Long story short, CKM has entered the guidelines’ “risk enhancers” league. To a healthcare provider, that makes sense. To a biostatistician, it doesn’t.

All the elements of the CKM syndrome—excess body fat, high blood pressure, (pre-)diabetes, and impaired kidney function—are factored into the PREVENT algorithm already (as you can see in the image below). Yes, these markers are not exactly good prospects for your heart health. Nobody debates that. But, no, counting them twice isn’t good statistics. In the image, I simulated a 35-year-old obese man with elevated blood pressure, a less-than-sterling cholesterol profile, and sub-optimal kidney function (lower than mine at double his age).

Now, envision the following dialogue between this man and his doctor:

Doctor: “Your 30-year risk is still below 10%, but your CKM being at stage 2 increases your risk score.”

To which the patient (who happens to be a biostatistician) responds: “No, doc, it doesn’t. The risk enhancer is fully factored into the 9.6% risk already.”

In an ideal world, both would acknowledge that risk prediction for a young adult is seriously flawed because the prediction equation’s heavy-weight risk factor is age.

And in an ideal world, the informed patient would insist on escalating the chain of evidence. They would ask their doctor to examine the risk enhancers that actually drive atherosclerotic cardiovascular disease (ASCVD) independently of age:

• first, count the atherogenic vehicles,

• then those carrying additional atherogenic explosives,

• then ask whether those vehicles have already vandalized the coronary wall.

That’s what the next section is about.

From Cholesterol Cargo to Particle Traffic

ApoB: Counting the Vehicles, Not the Cargo

ApoB, short for apolipoprotein B, is the name of a (very large) lipoprotein that is attached to every non-HDL particle (aka the lipoproteins, like LDL, that promote atherosclerosis). The neat thing about apoB is that every non-HDL vehicle carries exactly one apoB.

So, since your lab report tells you how much cholesterol floats through your arteries, measuring apoB tells you over how many vehicles that cholesterol cargo is distributed.

Now, think about it this way: the more vehicles there are moving along a road, the greater the chance that some of them will bump into and damage the guardrail. Metaphorically, that’s what’s happening in your blood vessels, too. The more cholesterol-carrying vehicles for a given total cholesterol load, the greater the chances of plaque-promoting collisions with your arterial wall. The image below illustrates the concept.

This matters because two people can have the same LDL-C but very different apoB levels. Certain conditions deplete the cholesterol load of particles but keep their numbers high. Diabetes is one of those conditions.

That’s why the LDL-C/apoB ratio has consistently shown a stronger predictive relationship with ASCVD than LDL-C [5]. But if you expect thresholds to exist that would allow your doctor to neatly recalibrate and reclassify your risk based on your apoB count, you are in for a surprise. There is no single universally accepted numeric definition of LDL-C/ApoB discordance [6].

Fortunately, a 2024 study of 12,688 adults not using statins in the US National Health and Nutrition Examination Survey tried to bridge this gap between population evidence and clinical practice [7]. I translated the findings into a discordance matrix that even your doctor might find helpful (LDL-C/ApoB discordance is not something most clinicians can compute from a lab report).

Since LDL and apoB strongly correlate (statistically), the researchers established what a “normal” or “expected” apoB level is for any given LDL level (the blue line in the image above). If your apoB is higher than expected for your LDL level, you are “positively discordant”, which actually isn’t positive at all (remember, more vehicles means more collisions with the arterial wall). Negative discordance works the opposite way.

The discordance concept becomes useful for you when the mismatch is large enough to change your clinical narrative:

Let’s say your PREVENT score is in the low or borderline range, LDL-C says “not much of a cholesterol problem here,” but ApoB screams “too many atherogenic particles are circulating,” then you might want to be more receptive to LLT, be that a lifestyle change or a drug. You also might consider a CAC scan or CCTA as described in the imaging section below.

The number of cholesterol vehicles isn’t the only marker. There exists a type of vehicle that wreaks havoc by design. It is called Lp(a), shorthand for lipoprotein A.

Lp(a): The Genetically Loaded Wrecking Ball

I dedicated an earlier post to Lp(a). So, here I’ll give you what you need to know for your risk-enhancer investigation.

Lp(a) is an LDL particle that has an apolipoprotein A particle—apo(a)—attached to its apoB. apo(a) is prothrombotic and anti-fibrinolytic. In other words, Lp(a) doesn’t work in your favor when a plaque rupture in one of your heart’s arteries precipitates a blood clot large enough to cork up that artery. The image below illustrates what you need to keep in mind when asking for an Lp(a) blood test.

Since the length of apo(a) can vary by a factor of 10, you and I might have the same number of Lp(a) vehicles cruising through our arteries, yet their summary weight can differ substantially. But it is their number, not their weight, that determines Lp(a)’s destructive potential. That’s why you should always insist on having that blood test done by a lab that measures particle count (nmol/L), not summary weight (mg/dL). There is also no fixed conversion equation between the two metrics, as I explained in my earlier Lp(a) post.

The guidelines advise clinicians to use elevated Lp(a) qualitatively during the “Personalize” step to justify initiating or intensifying lipid-lowering therapy if your Lp(a) reading comes back at ≥125 nmol/L. I’m neither a fan of “qualitatively” nor of binary decision thresholds. That’s why I created the table below. It is based on the European Atherosclerosis Society’s (EAS) ‘bucket’ system to quantify Lp(a)’s risk enhancement [8].

An example: If your Lp(a) test comes back between 70-119 nmol/L, multiply your PREVENT score by 1.17 to arrive at your Lp(a)-corrected risk estimate.

While apo(B) gives you a count of the atherogenic particles, Lp(a) identifies the genetically loaded ones.

Neither test tells you whether those particles have already left tracks in the arterial walls. That’s where the imaging techniques come in: Coronary Artery Calcium Scan (CAC scan) and Coronary Computed Tomography Angiography (CCTA).

Anatomy Beats the Algorithm: CAC & CCTA Imaging

In a previous post, I discussed these two imaging techniques in detail: their pros and cons, and who they serve best.

CAC is the preferred first imaging test for most borderline/intermediate-risk adults aged 45+. It is relatively cheap and often decisive. Scoring zero on a CAC scan dramatically drops your 10-year risk. Not forever, but the warranty period of a zero-score is typically between 3 and 5 years, depending on your age, gender, and ethnicity.

You’ll find all the numbers and graphs in that earlier post. But CAC zero is not the same as “no plaque,” specifically not in younger adults or in people with certain risk factors, like high Lp(a), that promote non-calcified plaque. To detect the latter, CCTA is the right choice.

The catch: While CAC scans miss early soft plaque, CCTA may show plaques that never would have harmed you. The consequences are unnecessary anxiety and overtreatment. These aren’t to be trivialized.

Even among the rare specimens of people with optimal risk factors (i.e., normal-weight never-smokers without elevated blood pressure, sugar, and lipids), one in five has detectable atherosclerotic plaque in their coronary arteries [9]. But with a 10-year risk score of <5%, most of those plaques won’t cause any problems.

The table below summarizes some of the main points of my earlier article.

To wrap it up: The PREVENT score gives you a population-based risk estimate, apoB and Lp(a) tell you about your actual exposure to the risk-promoting vehicles, and CAC/CCTA show whether these vehicles have left skid marks in your coronary arteries.

So far, we have worked on quantitatively personalizing your risk estimate. Now we need to do the same with the achievable benefits from statin treatment.

The Benefit Illusion: Relative Risk Is Not Your Risk

We need two numbers:

• the achievable reduction in LDL and

• how this translates into absolute risk reduction.

The table below presents the absolute LDL reductions achievable with various statins across their different doses [10].

The consensus view is that every 1 mmol/L (39 mg/dL) LDL reduction lowers your risk by 22% [11]. That sounds impressive. But it loses its shine when you translate that number into what you as the patient (or pharma customer) really want to know: absolute risk reduction (ARR) and the number needed to treat (NNT).

The table below illustrates that translation.

A 22% relative risk reduction shaves off just 1.32% of a 6% 10-year risk. And that translates into a number needed to treat of 76. That is, 76 people need to take the drug to prevent one heart attack or stroke. In other words, 75 people take the drug without realizing the advertised benefit. I have discussed the NNT in greater detail in an earlier post about statins.

Given these tables, you can now calculate the ARR and NNT for your personal situation. The results are meant to put the benefits of your doctor’s statin recommendation into perspective. The higher your risk score, the lower the NNT.

The illustration above is not my way of saying you shouldn’t take the statin. It is a way of visualizing the treatment benefits quantitatively. It is a way of moving from the confusion of the Guidelines’ mandated qualitative risk discussion to quantitative clarity.

Only with clarity can you make an evidence-based decision when the medical establishment offers deliberately little to address your uncertainty. To arrive at an informed and shared decision with your doctor based on what you now know, here is a cheat sheet of questions to raise with them.

The Examining-Room Script: Questions For Your Doctor

Now you need to prep for that 15-minute appointment with your doctor.

The idea is to walk in with questions that force the discussion out of guideline autopilot and into truly informed shared decision-making.

Here are six sample questions that you can use as a template to prepare your own.

1. The “Double-Counting” Audit (PREVENT & CKM)

“Before we use my [fill in the risk enhancers your doctor mentions, e.g., BMI, blood pressure, etc.] as ‘risk enhancers’ to justify a prescription, can we confirm that the PREVENT algorithm hasn’t already baked them into my baseline risk score? I want to ensure my risk isn’t being double-counted.”

2. The Particle Count Veto (ApoB vs. LDL-C)

“I see my LDL-C mass is flagged, but standard panels can over-or underestimate risk depending on discordance between LDL weight and particle size. Before we medicate based on mass, can we run an ApoB test to check for possible discordance? “

3. The Bi-Directional Risk Check (Lp(a))

“I understand that Lp(a) can substantially modify the risk estimate. If we test my Lp(a) and it comes back completely pristine, how would that influence your estimate of my risk?”

4. The Physical Reality Veto (CCTA)

“The PREVENT algorithm calculates the mathematical probability that I might develop plaque. Before I commit to a lifelong daily medication based on a probability, can we order a CAC or CCTA scan to see the actual burden of calcified and non-calcified plaque, and how that may influence your medication recommendation?”

5. The “Number Needed to Treat” Reality Check (ARR/NNT)

“You mentioned a relative risk reduction if I take this statin. Based on my specific baseline, the type of statin, and its prescribed dose, what is my Absolute Risk Reduction, and exactly how many people with my exact profile need to take this drug for 10 years for one person to actually benefit?”

6. The Closing Decision

“Given my measured risk, my imaging status, my ApoB, my Lp(a), my expected ARR, and my tolerance for uncertainty, would you consider the statin prescription a strong recommendation, a reasonable option, or something we can safely defer and reassess after I have modified my lifestyle?”

Now you have the science and the questions to ask. But to get the best out of that discussion, I have built the N-of-1 clinical tools you need. Below the paywall, you will find instant access to:

• The three online calculators (true LDL & eGFR, interactive ARR and NNT)

• The interactive interrogation room: an NLM notebook that answers your questions about your personal profile. It contains the guidelines, their sources and supporting papers, an interactive mind map, and helpful illustrations.

• An interactive Excel file (The Statin Risk Reduction Calculator) for you to calculate your true risk and your NNT, given your apoB, Lp(a), PWV, and other biomarkers.

Upgrade below to unlock your biological audit toolkit right away.

Read more

When I started the Double Check Doc publication, I knew there were many people like you.

People who didn’t feel like they were getting the whole story from the medical press. People who realized that the standard templates of medical care rarely fitted their personal profiles. People who wanted a source that doesn’t play favorites, doesn’t blindly regurgitate the latest study results, and doesn’t stop short of giving them the practical tools to stay fully functional for life.

That’s the type of people for whom I write, research, and develop.

I have been adding downloadable tools and calculators to my posts. They are for you to apply a post’s subject matter to your specific case and profile.

My latest tool is the article-specific, custom-built Interrogation Room—a NotebookLM engine pre-loaded with an article’s primary literature and my post. This tool is for you to interrogate not only the science and my conclusions, but to find answers specifically for your personal profile.

The first interrogation room came with my post:

I introduced my rationale for pivoting from passive reading to interactive “bionic” posts in a follow-up article:

I do this because here is what we are up against:

That medical machine trivializes your individual data profile and prioritizes population averages and consensus guidelines, often biased by pharmaceutical funding.

Fighting that machine—and developing the interactive tools for you to bypass it—requires independent, uncompromising R&D. It is a full-time clinical endeavor. Which brings me to my question:

The Diagnostic Audit: Where Do I Allocate My Resources?

The majority of you are reading the deep dives, but few of you have used the AI Interrogation Room.

I need to know why, because your answer dictates how I allocate my R&D resources.

I have seven hypotheses for why you left the AI keys on the table. Your feedback directly determines what I build—and what not—next.

Instead of leaving a public comment, select the option that best describes your decision in the anonymous, 30-second audit embedded below.

One More Thing:

As of today, paid subscribers have access to the brand new DCD Interrogation Room.

It is your research and decision engine trained on the full and up-to-date DCD archive. It’s my “brain dump” made available for you to question 24/7.

Without having to sift through all DCD articles, it allows you to:

• Evaluate medical and health claims

• Test your assumptions and beliefs

• Apply my frameworks to your personal situation

Read about, and access it from here:

Subscribe now

Leave a comment