The LDL/ApoB Debate: Statins, Saturated Fat, and the Science Nobody Agrees On

Heart disease kills roughly 1 in 5 Americans. The internet debate about preventing it focuses almost entirely on one variable: LDL cholesterol. Should it be as low as possible? Are statins overprescribed? Is saturated fat really the enemy?

These are important questions. But they're one variable in a ten-variable equation, and arguing about LDL in isolation misses most of what actually determines whether someone has a heart attack.

Cardiovascular Risk Factors Beyond LDL

Before diving into the LDL debate, here's the full system. Every section that follows maps back to this table.

The INTERHEART study (Yusuf et al., Lancet, 2004) quantified this across 52 countries and ~29,000 participants. Nine modifiable risk factors accounted for over 90% of the population-attributable risk of a first heart attack. The top contributors: abnormal lipids (49%), smoking (36%), psychosocial stress (33%), abdominal obesity (20%), and hypertension (18%). (These are population-attributable risk percentages, which can sum to over 100% because risk factors overlap in the same individuals.) These interact multiplicatively, not additively. Having both high blood pressure and high LDL is substantially worse than the sum of those individual risks.

This post walks through all of it: the lipid science, the statin debate, what exercise and blood pressure contribute, who profits from which narrative, and a framework for interpreting your own bloodwork.

What LDL-C and ApoB actually measure

Much of the cholesterol debate comes from people talking past each other using different metrics.

LDL-C (LDL Cholesterol) measures the total mass of cholesterol carried inside all your LDL particles. Think of it as weighing the cargo on all the boats. It's typically calculated from your standard lipid panel, not directly measured.

ApoB (Apolipoprotein B) counts the number of potentially dangerous particles. Every LDL particle, every VLDL remnant, and every Lp(a) particle carries exactly one ApoB molecule. Think of it as counting the boats. This matters because each particle is one opportunity for arterial wall penetration, and particle count appears to predict risk better than cargo weight.

About 20-30% of the population has discordant LDL-C and ApoB (one is high, the other isn't). In people with insulin resistance, LDL-C is often "normal" while ApoB is elevated (many small particles, each carrying less cholesterol). In lean people on low-carb diets, LDL-C can be very high while ApoB may be less elevated (fewer, larger particles). When LDL-C and ApoB disagree, cardiovascular risk tracks with ApoB (Sniderman et al., The Lancet, 2011). The field has largely converged on ApoB as the better metric, but most doctors still only check LDL-C.

Evidence for LDL/ApoB Causality

There are roughly two camps on this, and neither one has much patience for the other.

Camp A (Peter Attia, mainstream lipidology) says ApoB is the primary causal driver of atherosclerosis and wants aggressive early treatment. Camp B (carnivore/keto communities, some contrarian cardiologists) says the fixation on LDL misses the forest for the trees, that metabolic health is the real driver, and that statins are overprescribed.

After spending a significant amount of time in the primary literature, I think the evidence is more nuanced than either camp typically presents.

Supporting Evidence for Causation

Genetics: Mendelian randomization (using naturally-occurring gene variants as natural experiments) shows that people born with genetically lower LDL have dramatically less heart disease. Ference et al. (JACC, 2012) analyzed 312,321 participants and found that each ~39 mg/dL genetically lower LDL was associated with a 54.5% reduction in coronary disease. Because gene allocation is random at conception, this largely eliminates the confounding that plagues observational studies.

Natural dose-response: People with familial hypercholesterolemia (FH, about 1 in 250 people) have very high LDL from birth. Without treatment, roughly 50% of men have a coronary event by age 50. The homozygous form (much rarer) produces extreme LDL, and untreated patients rarely survive past 30.

Multiple drug classes converge: Statins, ezetimibe, PCSK9 inhibitors, and bempedoic acid all lower LDL through completely different mechanisms. Each reduces cardiovascular events roughly in proportion to the LDL reduction achieved. If statins worked only through anti-inflammatory "side effects" and not LDL lowering, these other drugs should not work. But they do.

Clear mechanism: ApoB-containing particles cross the arterial wall lining, get trapped, undergo oxidation, and are engulfed by immune cells (macrophages) that form foam cells, driving the inflammatory cascade that builds plaque. This has been demonstrated in animal models, human tissue samples, and imaging studies.

Counterarguments and Open Questions

The elderly paradox: Ravnskov et al. (BMJ Open, 2016) reviewed 19 studies involving 68,094 elderly participants. In 16 of 28 cohorts (92% of participants), higher LDL was inversely associated with all-cause mortality. There are reasonable explanations (survivorship bias, illness-driven cholesterol drops), but the finding persists across multiple datasets and creates tension with "lower is always better."

Saturated fat is murkier than expected: Dietary saturated fat raises LDL modestly. But meta-analyses on whether this translates to cardiovascular harm are not as clear as guidelines suggest. Siri-Tarino et al. (2010): no significant association between saturated fat and coronary disease in 347,747 subjects. Hooper et al. (Cochrane, 2020): reducing saturated fat lowered CVD events by 21%, but had no effect on CVD mortality or all-cause mortality. The critical nuance: what replaces the saturated fat matters more than the saturated fat itself.

Lean mass hyper-responders: Dave Feldman identified lean, metabolically healthy people on keto/carnivore diets who develop very high LDL-C (200-500+) while maintaining low triglycerides and high HDL. The formal KETO trial (JACC Advances, August 2024) used CT coronary angiography to measure actual plaque in these individuals. The result: no significant difference in plaque burden despite LDL-C averaging 272 mg/dL. It's a small, cross-sectional study (a snapshot of plaque at one point in time, not tracking change over years), so it cannot tell us whether plaque was accumulating. Not definitive, but a genuinely surprising finding that warrants follow-up.

Inflammation has its own lane: The CANTOS trial (Ridker et al., NEJM, 2017) gave 10,061 post-heart attack patients an anti-inflammatory drug (canakinumab, which targets IL-1beta) that had zero effect on LDL. The 150 mg dose reduced major cardiovascular events by 15%. This proved that reducing inflammation, completely independently of LDL, reduces heart attacks. The colchicine trials (COLCOT, LoDoCo2) confirmed the same concept with a cheap generic drug.

How to tell causation from correlation

The history of medicine is littered with confident causal claims that collapsed under scrutiny. One example worth knowing:

Hormone replacement therapy: The Nurses' Health Study found that HRT users had roughly 50% lower coronary risk. By the 1990s, guidelines recommended HRT for heart protection. Then the Women's Health Initiative (2002) randomized 16,608 women and found HRT actually increased coronary disease, breast cancer, and stroke. The entire observational signal was explained by healthy user bias: women who chose HRT were wealthier, more active, and had better healthcare. This exact bias structure applies to observational studies of people with lower cholesterol.

Austin Bradford Hill proposed nine criteria (1965) for evaluating causal claims. The LDL/ApoB hypothesis scores well on most: strong consistency across populations, clear dose-response (both in genetics and drug trials), temporal sequence (plaque precedes events by decades), biological plausibility, and experimental support from multiple drug classes. It scores weaker on specificity (about 50% of heart attack patients have "normal" LDL, though "normal" in a population where most adults have LDL above optimal levels is a low bar, making this statistic less surprising than it sounds) and coherence (the elderly paradox, the French paradox). The overall picture: the evidence for "ApoB-containing particles cause atherosclerosis" is stronger than for "the LDL-C number on your lab report determines your risk." The causal agent appears to be the particle itself, not the cholesterol cargo it carries.

Relative Risk, Absolute Risk, and All-Cause Mortality

Both sides cherry-pick the framing here. Pro-statin messaging leads with relative risk ("30% reduction in heart attacks"). Statin skeptics lead with all-cause mortality ("no mortality benefit in primary prevention"). Both are technically correct, and both are misleading when presented without the other.

What the statin trials actually show

NS = not statistically significant. NNT = number needed to treat to prevent one event. Lower NNT = more effective.
*HOPE-3: "NS" refers to mortality only. The CV event composite was significant (rosuvastatin HR 0.76, p=0.002).

The general picture is that in secondary prevention (people who already have heart disease), statins are among the most effective drugs in medicine. An NNT of 30 for preventing death over 5 years is excellent. In primary prevention (healthy people with elevated lipids), the relative risk reductions look similar, but the absolute benefit is much smaller because baseline risk is lower.

All-cause mortality captures the net effect. The CTT Collaboration meta-analysis (2010, ~170,000 participants) found statins produce roughly a 10% relative reduction in all-cause mortality per ~39 mg/dL of LDL lowering. In secondary prevention, this translates to meaningful absolute benefit. In primary prevention, the absolute all-cause mortality benefit is roughly 0.5% over 5 years, and individual trials often fail to reach statistical significance.

The PCSK9 puzzle: FOURIER (2017) achieved median LDL of 30 mg/dL (far lower than any statin trial) and reduced the composite CVD endpoint by 15%, but showed no reduction in cardiovascular mortality or all-cause mortality over 2.2 years. You can crash LDL to 30 and not clearly reduce death over 2+ years. Possible explanations include short follow-up and populations already well-treated with statins, but the finding is worth sitting with.

Statins: how they work and why they also reduce inflammation

Statins block the enzyme (HMG-CoA reductase) that manufactures cholesterol inside liver cells. When cholesterol production drops, the liver compensates by making more LDL receptors on its surface, which grab more LDL particles from the bloodstream. Since the liver already clears about 70% of circulating LDL, increasing its uptake capacity has an outsized effect. Net result: LDL drops 30-55% depending on the drug and dose.

Doubling the dose only gives about 6% more LDL reduction (logarithmic curve). This is why many cardiologists prefer adding a second drug (like ezetimibe) rather than maxing out the statin dose.

Anti-Inflammatory Effects of Statins

The same metabolic pathway that produces cholesterol also produces signaling molecules involved in inflammation. When statins block the pathway, they reduce both. Specifically, statins reduce production of inflammatory proteins that recruit immune cells into arterial walls, help stabilize existing plaques (making them less likely to rupture), and increase nitric oxide production from artery walls (which is vasodilatory, anti-clotting, and anti-inflammatory).

Is this anti-inflammatory effect independent of LDL lowering? The evidence suggests yes, at least partially. In the PROVE IT trial (2004), CRP reduction and LDL reduction were only weakly correlated (r ~ 0.15), suggesting substantially independent pathways. Patients who achieved both low LDL and low CRP had the best outcomes. The JUPITER trial (2008) enrolled people with normal LDL but elevated CRP and saw a 44% reduction in major cardiovascular events, larger than expected from LDL lowering alone.

Statin side effects, brain, and stroke

Muscle symptoms: 5-10% of patients report muscle problems in clinical practice. But in blinded trials, ~90% of these symptoms occurred equally on placebo (SAMSON 2021, StatinWISE 2020). This suggests a large nocebo effect. That said, severe muscle damage (rhabdomyolysis) is rare but real (~1-3 per 100,000 patient-years).

Diabetes risk: Sattar et al. (Lancet, 2010) found a 9% increased risk of new diabetes. The NNH (number needed to harm) is roughly 255 patients treated for 4 years. The excess cases tend to occur in people already near the prediabetic threshold. Most guidelines argue the cardiovascular benefit outweighs this risk, which is probably true on average but may not be for every individual.

Cognition: The FDA added a label warning in 2012. But large randomized trials (HPS, PROSPER, HOPE-3) have consistently shown no measurable cognitive harm at the population level. Individual patients may experience reversible symptoms. A practical option: lipophilic statins (atorvastatin, simvastatin) cross the blood-brain barrier more readily; switching to a hydrophilic statin (rosuvastatin, pravastatin) may help.

Stroke: Statins clearly reduce ischemic stroke (~17% relative reduction per ~39 mg/dL LDL lowering). But the SPARCL trial (2006) found a 66% increase in hemorrhagic stroke (in absolute terms: 2.3% vs. 1.4%). The net effect on total stroke is still favorable, but for someone with hemorrhagic stroke risk factors, this matters.

Hormones: Cholesterol is the precursor for testosterone, estrogen, and cortisol. Some studies show small testosterone reductions with statin use (Corona et al., 2013). Clinical significance is unclear, especially at the very low LDL levels achieved with PCSK9 inhibitors.

Lifetime Risk and Cumulative LDL Exposure

Most risk calculators give a 10-year estimate. If you're 46 (as I am), a 10-year risk of 5% sounds manageable. But 10-year windows are misleading for anyone who plans to live past their early 60s.

Berry et al. (NEJM, 2012) followed 257,384 people and found that lifetime cardiovascular risk varies dramatically by risk factor burden. For a 45-year-old man with two or more major risk factors, lifetime risk approaches 60-70%. Even with all risk factors optimal, it's still roughly 5-8%. Lloyd-Jones et al. (Lancet, 1999, Framingham data) estimated remaining lifetime risk of coronary disease at age 40: approximately 49% for men and 32% for women.

Heart disease is the leading cause of death in the U.S., accounting for roughly 1 in 5 deaths. The 10-year window understates the problem.

Cumulative exposure: area under the curve

Ference et al. (JACC, 2012; European Heart Journal, 2017) proposed thinking about ApoB exposure as total particle-years (concentration multiplied by years of exposure). People with genetically lower LDL from birth get roughly 3x the cardiovascular risk reduction per unit of LDL lowering compared to someone who starts a statin at 55. This makes sense: every year of elevated ApoB means more particles crossing into arterial walls and more plaque accumulating.

A statin started at 40 potentially removes 30-40 years of excess particle exposure. Started at 65, maybe 15-20 years. The per-year benefit is the same, but the cumulative benefit differs enormously.

WOSCOPS Long-Term Follow-Up Data

WOSCOPS (1995) was a 5-year primary prevention statin trial. Extended follow-up (Ford et al., NEJM, 2007; Packard et al., Circulation, 2014) found that patients originally randomized to pravastatin had significantly lower mortality a decade or more after the trial ended, even though most had stopped taking the statin. The likely explanation: the treatment period slowed plaque progression during a critical window, and plaques that didn't form don't rupture later.

Exercise: the most underrated intervention

In the equation table above, exercise shows up in the "interventions" column for 7 of 10 variables. No drug comes close to that breadth. Here's what it does, in plain English:

• Artery wall health: Blood flow during exercise creates physical shear stress on artery linings, which triggers increased production of nitric oxide (a molecule that relaxes blood vessels, prevents clotting, and reduces inflammation). Regular exercise improves baseline artery function even at rest.
• Anti-inflammatory: Contracting muscles release a signaling molecule (IL-6 acting as a "myokine") that, counterintuitively, triggers an anti-inflammatory cascade. Regular exercisers have lower baseline levels of inflammatory markers like CRP. Exercise also reduces visceral fat, which is itself a source of inflammatory signals.
• Insulin sensitivity: Skeletal muscle is where most of your blood sugar goes after a meal. Exercise improves the muscle's ability to take up glucose, addressing one of the root causes of metabolic syndrome (bad lipid profiles, inflammation, and high blood pressure bundled together).
• Lipid quality: Exercise lowers triglycerides (10-20%), raises HDL modestly, and shifts LDL particles from small/dense (more dangerous) toward large/buoyant (less dangerous). Total LDL-C may not change much, but particle quality improves.
• Blood pressure: Aerobic exercise lowers systolic blood pressure by roughly 5-7 mmHg in people with hypertension. Mechanisms include reduced sympathetic nervous system activity, improved artery flexibility, and better kidney function.
• Heart remodeling: Training increases stroke volume (blood per heartbeat), lowering resting heart rate by 10-20 bpm. This reduces lifetime mechanical stress on the heart and improves autonomic balance (more "rest and digest," less "fight or flight").
• Anti-clotting: Exercise increases the body's clot-dissolving capacity and reduces platelet stickiness.

How much is enough?

Data from Arem et al., JAMA Internal Medicine, 2015 (661,137 individuals).

The biggest gain comes from going from nothing to something. Vigorous exercise (running, cycling hard, HIIT) provides roughly twice the benefit per minute compared to moderate exercise. Combined aerobic plus resistance training appears optimal.

Blood Pressure as a Compounding Risk Factor

Blood pressure rarely generates tribal internet arguments. But it probably deserves more attention than it gets, because its interaction with lipids is synergistic and its mechanisms overlap extensively with atherosclerosis.

High blood pressure has several converging causes: arterial stiffening with age (the elastic fibers in artery walls break down and get replaced with stiffer collagen), overactivity of the hormonal system that controls sodium and water balance (the renin-angiotensin system), overactive sympathetic nervous system (common in obesity and chronic stress), and endothelial dysfunction (reduced nitric oxide, which is both a cause and consequence of hypertension).

Why blood pressure and lipids are worse together

High blood pressure increases the permeability of artery walls, meaning more LDL particles enter. Higher LDL concentration means more particles are available to enter. The combination creates a multiplicative effect, not just additive. The HOPE-3 trial (2016) demonstrated this: in intermediate-risk patients, rosuvastatin alone reduced cardiovascular events, blood pressure lowering alone did not reach significance, but the combination produced the largest benefit.

How much does treatment help?

The SPRINT trial (2015) randomized 9,361 high-risk patients to aggressive blood pressure targets (systolic below 120) vs. standard (below 140). Result: 25% reduction in major cardiovascular events and 27% reduction in all-cause mortality. The Blood Pressure Lowering Treatment Trialists' meta-analysis (2021, 348,854 participants) found that a 5 mmHg systolic reduction produced about a 10% reduction in major cardiovascular events, consistently across starting blood pressure levels.

What else lowers risk?

Lowering a lab number is not the same as preventing a heart attack. The question is always: does the intervention reduce hard clinical endpoints (heart attacks, strokes, deaths)?

Lp(a) and Inherited Cardiovascular Risk

Lipoprotein(a), or Lp(a), is an LDL-like particle with an extra protein attached that also promotes blood clotting. About 20% of people have elevated levels (>50 mg/dL or >100 nmol/L), which increases cardiovascular risk by roughly 2-3x, independent of LDL.

Lp(a) levels are ~90% genetically determined. Diet, exercise, and lifestyle have minimal effect. Mendelian randomization confirms it as a causal risk factor (Clarke et al., NEJM, 2009). Statins do not lower Lp(a) and may slightly increase it. This matters because a person with LDL of 130 and Lp(a) of 150 nmol/L has dramatically different risk from someone with the same LDL and Lp(a) of 10 nmol/L, but the standard risk calculator treats them identically.

Multiple Lp(a)-lowering drugs are now in large phase 3 trials. Pelacarsen (Lp(a)HORIZON, ~8,300 patients) was expected to report results around mid-2026. Olpasiran and lepodisiran are also in trials with results expected in 2026-2027. Phase 2 data has been striking: lepodisiran's ALPACA trial (ACC 2025) showed a single dose achieved 93.9% Lp(a) reduction sustained for over a year. Even now, testing Lp(a) at least once is worth doing to understand your baseline.

Feasibility of Eliminating Heart Disease

Given all the tools available (statins, PCSK9 inhibitors, blood pressure drugs, exercise, diet, the coming Lp(a) therapies), you might wonder whether we can make heart disease largely a thing of the past, the way we did with smallpox. My guess is probably not, but we can likely reduce it much further than current practice achieves.

Even after optimal statin therapy, most cardiovascular events still happen. FOURIER achieved LDL of 30 mg/dL and still saw a 9.8% event rate over ~2.2 years. The remaining events come from residual inflammatory risk (addressable: colchicine trials showed 23-31% MACE reduction), residual metabolic risk (addressable: the SELECT trial showed semaglutide reduced MACE by 20% in obese patients), residual Lp(a) risk (potentially addressable pending trial results), and arterial aging (partially addressable with blood pressure control and exercise, but fundamentally degenerative).

Addressing cardiovascular disease comprehensively means simultaneously managing ApoB, blood pressure, inflammation, insulin resistance, and smoking. That's five parallel interventions, and most people currently address zero or one of them proactively. The gap between what's achievable and what actually happens is enormous, and at this point the science is probably ahead of the implementation by a wide margin.

Financial Incentives in the Statin Debate

"Follow the money" is a valid heuristic, but you have to apply it consistently to both sides or it's just a weapon.

The pharma side: The branded lipid-lowering market is roughly $15-20 billion (PCSK9 inhibitors at ~$5-6K/year, inclisiran at ~$6.5K/year). Major statin trials were pharma-funded. This is a legitimate concern, but context matters: independent analyses (CTT meta-analyses using individual patient data) confirmed the benefits, and Mendelian randomization requires no pharma funding.

The supplement/influencer side: Paul Saladino's Heart & Soil reportedly generates tens of millions per year. The low-carb/keto supplement market is substantial. YouTube channels with 2M+ subscribers generate significant ad revenue. The "your doctor is lying" narrative attracts enormous engagement. An audience built on statin skepticism cannot easily pivot to "actually, statins are useful for many people" without losing that audience.

Cognitive biases on both sides

• Confirmation bias (both): Camp A downplays the elderly paradox and modest absolute primary prevention benefits. Camp B ignores Mendelian randomization, multi-drug convergence, and FH natural history.
• Appeal to mechanism (Camp A): Because we understand how LDL causes plaque, there's a temptation to assume the effect must be large for everyone. Mechanism doesn't dictate magnitude.
• Survivorship bias (Camp B): Pointing to 85-year-olds with high LDL proves nothing. The ones who died at 60 aren't in the dataset.
• Relative risk inflation (Camp A): Reporting a "44% reduction" (JUPITER) without noting the absolute event rate (1.36% vs 0.77% per year) is persuasion, not education.
• Genetic fallacy (Camp B): Dismissing evidence because pharma funded it. The question is whether the evidence is valid, not who paid for it.

Evaluating Popular Health Commentators

Scored on a 10-point scale based on: engagement with primary literature, acknowledgment of counter-evidence, distinction between evidence quality levels, avoidance of logical fallacies, willingness to update, and transparency about conflicts. Inherently subjective; reasonable people might shift these 1-2 points.

Scores reflect adherence to the scientific method, not whether conclusions are correct. A person can score well and still be wrong. A low score means the reasoning process is unreliable.

A Decision Framework for Bloodwork Results

Rather than listing my numbers, I want to walk through the decision framework, because the logic matters more than any specific values.

Summary

1. ApoB-containing particles cause atherosclerosis. The convergence of genetics, drug trials, FH, and mechanistic biology is too strong to dismiss. Camp B influencers who deny this are ignoring the strongest evidence in the field.
2. LDL-C is an imperfect proxy. ApoB is the better metric. If you're going to track one lipid number, make it ApoB.
3. Context matters enormously. The same ApoB level means different things depending on metabolic health, age, Lp(a), family history, and whether plaque is already present. Camp B's emphasis on metabolic health is a legitimate addition to the framework, not a replacement for it.
4. Absolute risk determines the value of intervention. A statin is a different proposition for a post-MI patient (NNT ~15-30) than for a healthy 35-year-old with elevated LDL and nothing else (NNT ~130+).
5. The 10-year risk window is misleading. Lifetime risk of cardiovascular events with even one or two risk factors is 40-60%. Cumulative ApoB exposure and the WOSCOPS legacy data suggest earlier intervention produces disproportionate long-term benefit, so the decision framework probably needs to account for the next 40+ years, not just 10.
6. Exercise is the most underrated intervention. It addresses artery health, inflammation, insulin resistance, blood pressure, lipid quality, clotting, and cardiac remodeling simultaneously. No drug matches that breadth.
7. Blood pressure matters more than it gets credit for. It interacts synergistically with lipids, and addressing one without the other leaves a lot of benefit on the table.
8. Inflammation is a parallel pathway, not a competing hypothesis. CANTOS and colchicine trials proved this. The most accurate model: retained ApoB particles plus inflammatory response together drive plaque.
9. Heart disease is a system-level problem. Most of the heat in the LDL debate comes from people arguing about one variable in what is really a multi-variable equation.
10. Follow the money on both sides. Pharma funding biases one side, and supplement sales plus audience capture bias the other. Neither camp is purely motivated by evidence.
11. Get tested. ApoB, Lp(a), hsCRP, fasting insulin, blood pressure, and a CAC score will tell you more about your situation than any influencer's YouTube video.

Key references

Not exhaustive. Additional trials and studies are cited inline throughout the post.