What VESALIUS Really Tells Us About the New Lipid Guidelines
Andreas Vesalius, the 16th century Flemish anatomist, founded modern medicine on a simple but radical principle: look carefully at what is actually in front of you, not what received authority tells you should be there. His De humani corporis fabrica overturned a thousand years of Galenic doctrine by the straightforward act of opening bodies and reporting honestly what he found. It is more than fitting then, that a landmark cardiovascular trial bearing his name should demand the same discipline — examine what the data actually shows, rather than what we want it to show. The VESALIUS-CV trial was a carefully constructed positive trial which is used by lipid guideline chair Roger Blumenthal to contextualize the 2026 ACC/AHA lipid guidelines in an accompanying editorial. The central thesis appears to be to treat more patients more aggressively. This deserves careful examination.
Where This All Started: The Foundational Trials
The story of lipid lowering and cardiovascular outcomes begins in the 1990s with trials enrolling the patients most likely to benefit — high-risk individuals who had already suffered a heart attack, with no background statin therapy and LDL cholesterol levels in the 150–190 mg/dL range. The 4S trial is the archetype. Simvastatin versus placebo in 4,444 post-MI patients with a median LDL of 188 mg/dL and no prior statin therapy. The results were unambiguous: 30 fewer deaths per 1,000 patients treated over five years, with both coronary mortality and all-cause mortality formally and significantly reduced. LIPID, CARE, and WOSCOPS followed, each confirming the basic finding in similar populations.
These trials established something important: lowering LDL in very high-risk patients with high baseline LDL saves lives. Not just prevents heart attacks — actually prevents death. That finding, earned in the right population with the right endpoints, is the scientific foundation everything since has been built upon.
As the years passed, trials moved downstream into lower-risk populations — patients without prior heart attacks, patients with diabetes, patients identified by risk scores rather than by prior events. The benefits persisted on a relative basis. The Cholesterol Treatment Trialists’ meta-regression showed roughly a 22% relative risk reduction in major vascular events per 1 mmol/L of LDL lowering regardless of baseline LDL level or patient risk category. But absolute risk reduction — the number of deaths or heart attacks actually prevented — shrank substantially as baseline event rates fell. A treatment that prevents 30 deaths per 1,000 in post-MI patients with LDL of 188 prevents far fewer deaths in lower-risk populations with lower baseline LDL. The relative effect is the same. The absolute benefit is not.
This is not a controversial observation. It’s just math. But that observation of declining absolute risks as baseline LDLs got lower and patients enrolled became lower risk was glossed over as trials shifted toward composite endpoints — combinations of death, heart attack, stroke, and revascularization — specifically because lower-risk populations don’t generate enough hard mortality events to power trials around death alone. Composites inflate event rates to make trials of a certain size feasible. But this also makes results look more impressive than the underlying mortality data suggests. The 3-point composite outcome of MI/stroke/cardiac death is positive! Sounds really great until you take a quick look at Table 2 and see that almost all the benefit is from fewer stents placed in one arm. Fewer stents is an outcome patients and clinicians should care about, but it’s certainly not the same as reducing death.
A New Mechanism, A New Hope
PCSK9 inhibitors represent a fundamentally different approach to lipid lowering than statins. Statins work by blocking HMG-CoA reductase, the rate-limiting enzyme in the liver’s cholesterol synthesis pathway, which causes hepatocytes to upregulate LDL receptors and clear more LDL from the bloodstream. The LDL reduction is real but partial — typically 40–55% from baseline.
PCSK9 inhibitors work upstream from that. PCSK9 is a protein that degrades LDL receptors after they have cleared LDL from circulation — essentially a molecular brake on the liver’s ability to remove LDL. By blocking PCSK9 with a monoclonal antibody, evolocumab and alirocumab allow LDL receptors to recycle rather than degrade, dramatically amplifying the liver’s LDL-clearing capacity. The result is LDL reductions of 55–65% on top of statin therapy — levels far beyond what statins alone can achieve. Median achieved LDL in VESALIUS evolocumab patients was 45 mg/dL at 48 weeks. That is roughly half the LDL level most patients on high-intensity statins can reach.
The hope when PCSK9 inhibitors arrived was that this degree of LDL lowering — combined with a population already on statins — would translate into dramatic reductions in cardiovascular events and mortality. It would be the logical extension of the statin story: if lowering LDL by 40% is good, lowering it by 70% should be much better.
The First PCSK9 Trials: FOURIER and ODYSSEY
That hypothesis was tested first in the highest-risk patients — those who had already had a heart attack, stroke, or peripheral vascular event and were already on maximally tolerated statin therapy. FOURIER enrolled 27,564 patients with established ASCVD and a median baseline LDL of 92 mg/dL. ODYSSEY OUTCOMES enrolled 18,924 patients with recent acute coronary syndrome and median baseline LDL of 87 mg/dL.
Both trials showed statistically significant reductions in composite MACE endpoints. FOURIER reduced 5-point MACE by 15% relative to placebo. ODYSSEY reduced 4-point MACE by 15%. The drugs worked as LDL-lowering agents, and the LDL lowering translated into fewer composite events.
But the effect on mortality was disappointing. In FOURIER, cardiovascular mortality was essentially unchanged — HR 0.98, indistinguishable from placebo, at a median follow-up of only 2.2 years. All-cause mortality showed a numerical excess in the evolocumab arm, though not significant. In ODYSSEY, alirocumab did show a reduction in all-cause mortality — HR 0.85, significant — but only in a pre-specified subgroup with baseline LDL above 100 mg/dL. In patients with baseline LDL already below 100, the mortality benefit was attenuated.
These are patients who had already had heart attacks. Their LDL was already at 87–92 mg/dL on statins. The composite MACE benefit was real, but the mortality benefit ranged from modest to absent. The absolute risk reductions were small. These were reported as positive trials, and professional societies and guidelines quickly changed to reflect a “lower is better” philosophy, papering over the attenuating risk reduction as baseline LDL levels got lower. This most definitely was not anything like the statin 4S trial results.
It is in this context that trialists naturally moved from studying PCSK9 inhibitors in patients who have already had heart attacks and strokes to higher-risk patients who had not yet had an ischemic event. The problem for the trial designers is a tall one — how exactly does one design a trial to extend usage of PCSK9 inhibitors to lower-risk patients than FOURIER and ODYSSEY and still demonstrate a positive result? For those of us clinicians in the peanut gallery, VESALIUS gives us a master class in how to do just that.
VESALIUS: Extending PCSK9 Inhibition Upstream
VESALIUS-CV enrolled 12,257 patients with a specific profile: high cardiovascular risk but no prior myocardial infarction or stroke. They could qualify if they had coronary artery disease without prior MI, cerebrovascular disease without prior stroke, peripheral artery disease, or high-risk diabetes — defined as diabetes lasting at least 10 years, treated with daily insulin, or complicated by microvascular disease. All patients had to have at least one additional high-risk criterion such as age over 65, active smoking, or very elevated lipid levels. You were not eligible for the trial unless your LDL was above 90. Ninety-two percent were on lipid-lowering therapy at baseline and had a median LDL at enrollment of 122 mg/dL. (Recall FOURIER and ODYSSEY had a baseline LDL of approximately 90 mg/dL.)
There is nothing nefarious about these choices — there are plenty of patients just like those enrolled in VESALIUS that I am very interested in understanding the benefits of LDL lowering therapy for. But you can clearly see that trial designers are making sure to enrich this patient population with patients most likely to benefit. These are not average-risk primary prevention patients. They have atherosclerosis or high-risk diabetes. Many have coronary disease — just not a documented MI. They are distinctly different from FOURIER and ODYSSEY patients who had already survived an acute ischemic event because they sit earlier on the atherosclerotic disease continuum, with lower baseline event rates and — critically — much higher baseline LDLs.
The trial’s two co-primary endpoints were 3-point MACE — coronary heart disease death, MI, and ischemic stroke — and 4-point MACE, which added ischemia-driven revascularization. Both passed convincingly. Three-point MACE: HR 0.75, 95% CI 0.65–0.86, p<0.001. Four-point MACE: HR 0.81, 95% CI 0.73–0.89, p<0.001. Absolute risk reductions were 1.8% for 3-point MACE and 2.8% for 4-point MACE over 4.6 years — corresponding to numbers needed to treat of 56 and 36 respectively.
The PCSK9 inhibitor tested — evolocumab — clearly reduced clinically meaningful cardiovascular events in this population, but there are some other interesting design features of this trial that are instructive and informative.
The Hierarchy: Where the Story Gets Complicated
After the two co-primaries passed, VESALIUS tested secondary endpoints in a prespecified sequential hierarchy.
Why do hierarchical testing?
The two-word answer: multiple comparisons.
Every time you run a statistical test with a significance threshold of p<0.05 (also referred to as alpha) you’re accepting a 5% chance of a false positive. This is the risk of finding a statistically significant difference between groups that isn’t real — otherwise known as a Type I error. If you test 10 endpoints in the same trial, each at p<0.05, your chance of getting at least one spurious significant result by chance alone rises to about 40%. Test 20 things and it approaches 64%. This is the multiple comparisons problem. Without any correction, a trial could say it was going to test 20 different outcomes and call one of them significant — that had a 64% chance of being positive just by random chance — a “significant” result. This approach would be less science and more a fishing expedition.
Hierarchical testing is one way to keep the overall false positive rate at 5% across all the tests you plan to run. The logic is simple: you only get to test the next hypothesis if the previous one passed. Because each test is conditional on the prior one succeeding, you never spend more than 5% alpha total across the entire chain. No penalty needed for multiple testing as long as you follow the sequence. The chain must proceed in order and breaks permanently at the first failure. Here is how it ran:
Positions 1 through 6 were all composite endpoints containing some combination of MI, revascularization, and cardiovascular death. All six passed easily, each with p<0.001. Then came position 7: coronary heart disease death alone.
This requires a brief clarification, because the distinction between CHD death and CV death matters enormously and is frequently elided in commentary about this trial.
CHD death is death specifically attributable to coronary artery disease — fatal MI, sudden cardiac death from ischemia, death from ischemic heart failure. It is a subset of cardiovascular mortality.
CV death is broader. It encompasses CHD death but also includes fatal stroke, death from peripheral vascular disease, aortic catastrophe, and other cardiovascular causes.
All-cause mortality is the broadest and most clinically unambiguous endpoint — it counts every death regardless of attributed cause and cannot be influenced by adjudication decisions.
The hierarchy in VESALIUS placed CHD death alone as the mortality gatekeeper.
CHD death alone had 222 total events: 105 in the evolocumab group versus 117 in placebo. A difference of only 12 deaths across 12,257 patients. The hazard ratio was 0.89 — directionally consistent with benefit — but with a confidence interval of 0.68 to 1.16, comfortably crossing 1.0. The p-value was 0.39. Not significant. The chain broke.
Placing CHD death as the hierarchy’s mortality gatekeeper has logic behind it — deaths directly attributable to coronary disease are the most mechanistically plausible target for a drug that works through atherosclerosis reduction. The hierarchy is saying: if evolocumab truly prevents coronary death, prove it here first before claiming broader mortality benefit. The problem is that the trial was never adequately powered for this endpoint — only 222 events in 12,257 patients, when roughly 2,300 would have been needed for 80% power. A difference of 12 deaths broke the chain, and with it went formal testing of everything that followed:
Position 8: CV death — HR 0.79 (95% CI 0.64–0.98). Exploratory. Unadjusted p≈0.03.
Position 9: All-cause mortality — HR 0.80 (95% CI 0.70–0.91). Exploratory. Unadjusted p≈0.0004.
Position 10: Ischemic stroke — HR 0.79 (95% CI 0.62–1.01). Exploratory.
It is for this reason that the subsequent outcomes in the hierarchy are considered exploratory rather than clear evidence of a mortality reduction.
The Full Picture: What the Mortality Data Actually Show
Step back and look at the arc from 4S to VESALIUS through the lens of absolute mortality benefit — the number of deaths actually prevented per 1,000 patients treated.
For an interactive version of this chart click here.
In 4S, simvastatin prevented approximately 30 deaths per 1,000 patients over five years in post-MI patients with LDL of 188 and no prior statin therapy. That mortality reduction was formally significant and was the primary endpoint. In LIPID, approximately 14 deaths per 1,000. In HPS, approximately 13 per 1,000. As trials moved into lower-risk populations with lower baseline LDL, the absolute mortality benefit declined steadily. In FOURIER — PCSK9 inhibition in patients with established ASCVD and LDL already at 92 — CV mortality ARR was approximately 0.8 per 1,000. Essentially zero at the trial’s median follow-up of 2.2 years.
VESALIUS sits in this landscape with an exploratory all-cause mortality ARR of approximately 18 per 1,000 and CV mortality ARR of approximately 6 per 1,000. If real, they aren’t bad, but the pattern in this data is not subtle. The meta-regression literature confirms the relative benefit per mmol/L LDL reduction is stable across trials. The absolute benefit is not. The trials that drove the original mortality evidence for aggressive lipid lowering were conducted in patients with baseline LDL of 130–190 mg/dL and no background therapy. As baseline LDL has fallen — partly because patients are already on statins by the time they enroll, partly because the field has been testing benefit in lower-risk populations — the absolute mortality benefit has declined in tandem.
The Guidelines and the Editorial
For some reason the declining absolute risk reduction is barely mentioned by key opinion leaders in the field. As an example, the Blumenthal-Morris editorial that accompanied release of the recent lipid guidelines argues that VESALIUS “blurs the distinction between ASCVD risk categories” and should push guidelines toward a single LDL goal below 55 mg/dL for virtually all patients with atherosclerotic disease. This statement is befuddling given the lengths the VESALIUS trial designers went to separate out higher-risk patients.
The editorial describes the VESALIUS results as “highly significant reductions” in MACE, while noting without elaboration that there were “no between-group differences in safety events.” It reports the NNT of 56 for 3-point MACE. It does not mention that the mortality endpoints were exploratory. It does not mention the very small gap in death from ischemic heart disease — 12 deaths — that broke the hierarchical testing. It does not contextualize the absolute mortality benefit against earlier trials in higher-risk patients. And it endorses extending a drug costing roughly $5,000–6,000 per year to a broader population on the basis of composite MACE results in a trial that was neither powered nor designed to demonstrate mortality benefit in that population.
What the Data Actually Tell Us
None of this means VESALIUS was a bad trial or that evolocumab doesn’t work. It does work. The composite MACE reductions are real. For the right patient — established coronary disease with high-risk features, elevated LDL despite maximally tolerated statin therapy, high baseline event rate — a PCSK9 inhibitor is a reasonable and evidence-supported choice.
But the message the data actually support is more nuanced than the guidelines suggest. The higher the baseline LDL and the higher the patient’s underlying cardiovascular risk, the more absolute benefit they will derive from aggressive lipid lowering. This principle, demonstrated clearly in 4S and confirmed across the statin literature, remains the most reliable guide to who benefits most. Conversely, patients whose LDL is already below 100 mg/dL on background statin therapy have very different arithmetic. Their composite MACE events can be modestly reduced but their hard mortality outcomes are difficult to move. The returns diminish as you move further from the patients who anchored the original evidence base.
Andreas Vesalius built modern anatomy on the insistence that careful observation should override received doctrine. Looked at carefully, VESALIUS-CV confirms that aggressive LDL lowering works across a broad spectrum of cardiovascular risk — and also demonstrates, with reasonable precision, exactly where the limits of that benefit lie. Guidelines that acknowledge those limits do more to build trust with patients than guidelines that paper over them in pursuit of nearly universal therapy and a single universal target. A nuanced discussion may even result in more patients on lipid-lowering therapies than a conversation that reduces the discussion to a false binary.
The 2026 dyslipidemia guidelines and the editorial endorsing their expansion represent a familiar pattern: a positive trial, enthusiastically received, composite endpoints elevated to near-equivalence with mortality, and population-level thinking overriding individual risk stratification. Everyone is better served with patients understanding they sit on a curve of benefit, and the diagnostic tools at our disposal are best used to identify where they are located on that curve. This leads to a much more informed decision on lipid-lowering therapy.
Want maximal risk reduction even if you’re a 50-year-old iron man athlete with an LDL of 105? Go on a statin. If that same individual isn’t interested in a statin, consider some testing to identify higher-risk markers like calcium scans or advanced lipid tests like apoB or Lp(a) to get a better sense of individual risk. If you aren’t interested in that, that’s not the end of the world — lipid-lowering therapy in this context is associated with pretty small reductions in cardiovascular mortality, the outcome of most interest to patients.
The bigger message to the cardiology research community is one worth sitting with. Even taking the 4-point MACE outcome in VESALIUS at face value — the most favorable composite, the softest endpoint — the treatment arm still had a 13% event rate despite achieving a median LDL of 45 mg/dL. Thirteen percent of patients on evolocumab, with LDL driven to levels that would have seemed impossibly low to the trialists of the 1990s, still had a major cardiovascular event within five years. The LDL hypothesis has yielded remarkable results — it is one of the genuine success stories of modern preventive cardiology — but the curve is flattening. We may be approaching the limit of what cholesterol lowering alone can accomplish. The residual risk that persists despite aggressive lipid lowering is driven by inflammation, by plaque biology, by factors that circulating LDL levels do not fully capture. That is where the next generation of trials needs to go — not into ever more granular risk stratification of ever lower-risk populations to justify ever more expensive versions of the same mechanism, but toward genuinely new targets. Vesalius did not refine Galenic anatomy. He replaced it. The field may need a similar reckoning.
Interactive charts supporting this analysis available at anishkokamd.substack.com
Data sources: VESALIUS-CV (Bohula et al. NEJM 2026) · Burger et al. Atherosclerosis 2024 · CTT Collaboration Lancet 2010 · FOURIER (Sabatine et al. NEJM 2017) · ODYSSEY OUTCOMES (Schwartz et al. NEJM 2018) · 4S (Lancet 1994) · Blumenthal & Morris editorial JACC 2026
I really appreciated this review, it’s details, as well as its nuance. Many people, including guideline writers put studies into “buckets” that fit their pre-determined worldview and run with it. That simplification unfortunately flattens everything related to the study outcomes. Some people will call this a PCSK9i study. Others will call it a study of subclinical atherosclerosis. I’ve already seen people try to justify increased screening with coronary artery calcium or CCTA based on the study. And the editorial: lower LDL-C targets. I think the biggest questions to ask about this (and any study, really) are “Who were the participants and how did they get in the study?" And for this one, why do we know they have atherosclerosis? It’s clear from the study design that at some point these people had symptoms that led to testing that detected arterial disease - those are very different than people who happen to walk into the door to most doctors for primary prevention. I’m sure some were found based on screening and hopefully we’ll get that information in the future, but based on the study characteristics, I suspect most were discovered to have athero for clinical reasons, not because of screening. Nearly 60% had diabetes - for them, based on the older statin studies like CARDS and HPS (with shockingly similar LDL-C values to this study), event reduction with LDL-C reduction is far from new information. Another peculiar point is the high LDL cholesterol which on average was 122 mg/dL on a lot of therapy (~73% on "high intensity"), suggesting that many of these people had FH or long-standing untreated severe hypercholesterolemia, and thus were a higher risk of population than their numbers suggest. Given that they only needed to be on stable therapy for two weeks, it’s quite possible that they only recently had their LDL cholesterol reduced into the study. These sorts of things happen in large studies all the time by aggressive investigators and make simplifying them even more perilous. Sorry for the digression: your point about diminishing returns with treating lower LDL cholesterol and lower risk and how trial design addresses that is vital and glad you addressed it. I also read that editorial and felt “befuddled” by it.
I wish doctors would acknowledge the downside of lowering LDL and the studies showing higher LDL among people who live to be very old. What is that about? Could it be LDL is involved in our immune system function and our brain health?