Which Cholesterol Number Actually Predicts a Heart Attack

Most men have had a cholesterol test and been told a number. Far fewer have been told which number, or that the one printed largest on the report is not the one that best predicts a heart attack. The gap between those two facts has narrowed in the guidelines over the last two years, and it is worth understanding before your next appointment.

What is the difference between apoB and LDL cholesterol?

An artery is damaged by lipoprotein particles crossing into the arterial wall and lodging there. What matters is how many particles arrive, not how much cargo each is carrying.

Every atherogenic particle in your bloodstream carries exactly one molecule of apolipoprotein B. One particle, one apoB. So a measurement of apoB is a direct count of the particles capable of causing atherosclerosis, covering LDL, VLDL, IDL, chylomicron remnants and Lp(a) together.

LDL-C measures something different: the total mass of cholesterol being carried inside LDL particles. If your particles happen to be small and cholesterol-poor, you can have a great many of them while your LDL-C looks reassuring.

What happens when apoB and LDL-C disagree?

Across 293,876 UK Biobank participants, apoB correlated with LDL-C at r = 0.96. That is very high, and it is why LDL-C has worked as well as it has for decades. But a correlation of 0.96 still leaves room for a substantial minority to sit somewhere unexpected.

The concrete version: at a fixed LDL-C of 130 mg/dL, the 95 percent range of apoB values spans 85.8 to 108.8 mg/dL. One LDL-C reading maps to a wide band of actual particle counts.

A study of 375,544 people not on lipid-lowering therapy sorted them by whether their apoB and LDL-C percentile ranks agreed. Just under 80 percent were concordant. About 10.3 percent had discordantly high apoB, and 9.9 percent discordantly low. So roughly one adult in five is a person for whom the two numbers tell different stories.

In that group, outcomes followed apoB. Men with discordantly high apoB had a hazard ratio of 1.06 (1.01 to 1.11) for major adverse cardiac events and 1.09 (1.04 to 1.14) for all-cause mortality. Those with discordantly low apoB sat at 0.91 (0.85 to 0.97) for events.

These are modest hazard ratios, and they should be read as modest. The point is not that discordance transforms your risk. It is that when the two measures disagree, the particle count is the one that tracks what happens next, and a man being reassured by a normal LDL-C while his apoB sits in the top decile is being reassured about the wrong thing.

Which cholesterol number actually causes heart attacks?

Three methods with quite different weaknesses have been applied to this question, and they converge.

Mutually adjusted observation. An analysis of 389,529 UK Biobank participants entered apoB, non-HDL-C and triglycerides into the same model so each had to earn its keep against the others. ApoB survived at a hazard ratio of 1.27 (1.15 to 1.40) per standard deviation. Non-HDL-C collapsed to 1.09 (0.97 to 1.21), which crosses one. Triglycerides went to exactly 1.00.

Residual analysis. A 2024 study asked what is left of each marker once the other is accounted for. The residual portion of apoB still predicted events at HR 1.06 (1.04 to 1.07). The residual portion of LDL-C, once apoB was in the model, came in at 0.99 and did not reach significance.

Mendelian randomisation. This uses inherited genetic variants as a natural randomisation, which sidesteps most confounding. In a multivariable analysis, apoB retained an odds ratio of 1.92 (1.31 to 2.81) for coronary heart disease after mutual adjustment, while LDL-C fell to 0.85 with a p-value of 0.44. A separate genetic analysis across 30 lipoprotein measures and 113,937 coronary cases gave apoB a marginal inclusion probability of 0.868 against 0.075 for LDL-C, and found no other lipid measure significant.

The cleanest single demonstration comes from comparing genetic variants that lower triglycerides with variants that lower LDL-C. Per 10 mg/dL reduction in apoB, both routes produced almost exactly the same reduction in coronary risk. The benefit tracked the particle count, not which kind of cargo was removed.

Remnant cholesterol belongs in the same story rather than competing with it. In 41,928 Danes followed for a median of 12 years, those in the top five percent for remnant cholesterol had a myocardial infarction hazard ratio of 1.58 (1.27 to 1.98), and adding it reclassified up to 40 percent of the people who went on to have one.

What the guidelines have done with this is cautious. As summarised by the American College of Cardiology and American Heart Association, the 2026 dyslipidaemia guideline treats apoB as a Class 2a measurement, useful when triglycerides sit between 150 and 499 mg/dL and there is a discrepancy with LDL-C, and for assessing residual risk in diabetes and established cardiovascular disease. The 2019 European guideline goes slightly further, making apoB and non-HDL-C measurement a Class I recommendation in patients with elevated triglycerides.

What is Lp(a) and should you test it?

Lipoprotein(a) is an LDL-like particle with an extra protein attached. Its level is set almost entirely by genes, it is stable across a lifetime, and it is not meaningfully changed by diet, exercise or statins.

Somewhere between 20 and 30 percent of people worldwide have an elevated level. Most of them do not know.

The thresholds in common use are 50 mg/dL or 125 nmol/L for elevated, with a grey zone between roughly 30 and 50 mg/dL. The 2025 European focused update treats levels above 50 mg/dL (105 nmol/L) as a cardiovascular risk-enhancing factor, Class IIa. The 2026 American guideline, as reported by both societies, goes furthest: universal Lp(a) screening at least once in a person’s lifetime, at Class 1.

Two practical points follow from the genetics. First, once is enough. Within-person correlation over an average of 8.3 years is 0.87, so a single measurement is a lifetime answer for most people. Second, statins do not fix it. A pooled analysis found a non-significant change of minus 0.4 percent, and some evidence suggests statins nudge it slightly upward. What an elevated Lp(a) changes is not the drug you take for it, but how aggressively everything else gets treated.

For men specifically, the association is slightly stronger than in women and strongest at younger ages: for Lp(a) above 40 mg/dL against below 10, the myocardial infarction hazard ratio was 1.56 in men against 1.44 in women, and 1.80 in men under 50.

What does a statin actually do for you?

Relative risk reductions make statins sound transformative. Absolute numbers are more useful for deciding whether to take one, and the US Preventive Services Task Force published both for primary prevention, meaning people who have not yet had an event.

Read that as a real but unspectacular benefit in people at average risk, over the few years these trials ran. In men whose absolute risk is higher, the same relative reduction produces a much larger absolute one, which is the entire logic of treating by risk category rather than by cholesterol number alone.

On lower being better, the secondary prevention trials support it. Adding ezetimibe to a statin after an acute coronary syndrome reduced events from 34.7 to 32.7 percent over seven years. Adding evolocumab to a statin cut LDL-C by 59 percent and reduced the primary endpoint from 11.3 to 9.8 percent, though with no reduction in cardiovascular or total mortality. Alirocumab produced a similar 1.6 percentage point absolute reduction. These are consistent, modest, and real.

Do statins really cause muscle pain?

Roughly one man in ten who takes a statin will report muscle aches. What the blinded evidence says about the cause is one of the more striking findings in cardiology, and it deserves to be handled carefully rather than used to dismiss anyone’s experience.

An individual-participant meta-analysis pooled 19 placebo-controlled trials and 123,940 people. Muscle pain or weakness was reported by 27.1 percent of those on a statin and 26.6 percent of those on placebo, a relative risk of 1.03 (1.01 to 1.06). The excess was confined to the first year, where it amounted to about 11 extra events per 1,000 person-years. After year one there was no excess at all. The authors’ summary was that more than 90 percent of muscle symptom reports in people taking statins were not caused by the statin.

Two n-of-1 trials tested this in the people most affected: men and women who had already stopped statins because of side effects. In one, 60 participants alternated between atorvastatin, placebo and no tablet. Symptom scores were 8.0 with no tablet, 15.4 on placebo, and 16.3 on the statin. Almost all the symptom burden was reproduced by the placebo. Half of the participants restarted statins after being shown their own data. A larger trial of 200 people found a difference between statin and placebo of minus 0.11 on the symptom scale, with 65.5 percent going on to resume treatment.

The clearest demonstration is a single trial that ran in two phases. During its blinded randomised phase, muscle-related adverse events occurred at 20.3 per 1,000 patient-years on statin against 20.0 on placebo, a hazard ratio of 1.03 with p = 0.72. When the same trial continued as an open-label extension and everyone knew what they were taking, the rate ratio jumped to 1.41 (1.10 to 1.79, p = 0.006). The drug had not changed. The knowledge had.

Two things follow, and both matter. The symptoms are real and are experienced as real; a nocebo effect is not imaginary. But the probability that a given ache is caused by the drug is much lower than most people assume, which makes a supervised rechallenge, or a switch to a different statin or dosing schedule, a far more reasonable step than permanent avoidance. For men there is an additional detail: in the pooled meta-analysis the relative risk in men was exactly 1.00, against 1.09 in women.

What should you actually do?

A short list, in the order that produces the most information for the least effort.

Get a standard lipid panel, and calculate your ten-year risk. A panel alone is a number without a context. The risk calculation is the context, and current American guidance applies the PREVENT equations from age 30.

Get Lp(a) measured once. This is the single highest-value addition for most men, because it is genetically fixed, common, invisible to standard panels, and never repeated. The 2026 American guideline now recommends it universally.

Ask for apoB if your triglycerides are elevated, if you have diabetes or metabolic syndrome, or if your LDL-C looks fine but your family history does not. Those are the situations where discordance is most likely and where both guideline bodies already support measuring it.

If you are 40 or over and land in the borderline or intermediate risk band, ask about a coronary calcium scan. Current American guidance puts that threshold at 40 for men and 45 for women.

One point specific to men, and the reason this article exists on a men’s health site. In UK Biobank data, men carry higher LDL-C than women until around age 50, after which the relationship reverses and women overtake them. The decades when men have the higher atherogenic burden are their thirties and forties, which is precisely when most men have never had a lipid panel and assume there is nothing to find.

This article is for information only and is not medical advice. Do not start, stop or change a statin without speaking to the clinician who prescribed it. Talk to a qualified health professional about your own situation.