In young adults, the formula used to calculate LDL cholesterol can hide higher risk
Table of Contents
1.0 Summary
Most cholesterol reports do not measure LDL cholesterol directly. They calculate it from the other numbers on the panel, and three formulas are in common use. A study published in late September 2026 in Circulation: Population Health and Outcomes followed 10,550 people who had a cholesterol test between the ages of 20 and 39 for an average of 22.7 years and asked which formula best separated normal from elevated LDL. The three formulas disagreed for only 2.3% to 2.9% of people. When they did disagree, the Martin-Hopkins formula got the risk right: people it called elevated (130 mg/dL or higher) while an older formula called them normal went on to have about three to four times the rate of heart attack, stroke and related disease. The study used a single cutoff, did not compare any formula with a directly measured LDL or with apolipoprotein B, and so far is available only as an abstract.
2.0 Key Points
- The study pooled 10,550 adults aged 20 to 39 (mean age 29.8, 56.8% women) from 7 population-based cohorts in the International Childhood Cardiovascular Cohort Consortium; 376 had an atherosclerotic cardiovascular event over a mean of 22.7 years.
- An LDL of 130 mg/dL or higher predicted later events whichever formula produced it, with subdistribution hazard ratios of 1.69 for Martin-Hopkins, 1.52 for Sampson and 1.47 for Friedewald.
- The formulas placed people on opposite sides of the 130 mg/dL line only 2.3% to 2.9% of the time.
- Compared with people normal by both formulas, those normal by Friedewald but elevated by Martin-Hopkins had a hazard ratio of 3.32 (95% CI 1.94 to 5.66), and those normal by Sampson but elevated by Martin-Hopkins had 3.94 (1.80 to 8.64); the opposite disagreements showed no excess risk.
- The findings come from an abstract, rest on one cutoff, and were not checked against a directly measured LDL or apolipoprotein B.
3.0 Evidence base
LDL cholesterol is the number most people know from a cholesterol test, but most labs calculate it rather than measure it. The calculation starts from total cholesterol, HDL cholesterol and triglycerides. The oldest method, the Friedewald equation, uses one fixed factor to estimate how much cholesterol is carried in triglyceride-rich particles. The Martin-Hopkins equation swaps that fixed factor for one matched to the person’s own triglyceride and non-HDL cholesterol levels. The Sampson equation is a newer alternative built to stay accurate when triglycerides are high. For most people the three give similar answers. This study looked at the people for whom they do not.
Investigators pooled 7 population-based cohorts from the International Childhood Cardiovascular Cohort Consortium and included everyone who had an LDL result between the ages of 20 and 39 and long-term follow-up. That gave 10,550 people, 56.8% of them women, with a mean age of 29.8 at the time of the test. Each person’s LDL was calculated all three ways and classed as normal (below 130 mg/dL) or elevated (130 mg/dL or above). Over a mean of 22.7 years, 376 people had an atherosclerotic cardiovascular event. The definition included heart attack, stroke, transient ischemic attack, angina, peripheral artery disease and coronary revascularization, and events were confirmed from medical records [1].
An elevated LDL predicted later events whichever formula produced it. The study reports a subdistribution hazard ratio, a measure of relative risk that allows for competing events such as death from other causes. It was 1.69 (95% CI 1.35 to 2.12) for Martin-Hopkins, 1.52 (1.22 to 1.90) for Sampson and 1.47 (1.17 to 1.84) for Friedewald [1].
The formulas put someone on opposite sides of the 130 mg/dL line for only 2.3% to 2.9% of participants, and the results come apart in that small group. Each comparison below is against people whom both formulas called normal [1]:
- Normal by Friedewald but elevated by Martin-Hopkins: hazard ratio 3.32 (1.94 to 5.66). The reverse, elevated by Friedewald but normal by Martin-Hopkins: 1.19 (0.47 to 3.05).
- Normal by Sampson but elevated by Martin-Hopkins: 3.94 (1.80 to 8.64). The reverse: 0.86 (0.39 to 1.91).
- Normal by Friedewald but elevated by Sampson: 1.69 (0.93 to 3.05). That interval includes 1, so the difference is uncertain, and no one fell into the reverse group.
When Martin-Hopkins called a young adult’s LDL high and another formula called it normal, the higher risk followed the Martin-Hopkins result. The people Martin-Hopkins called normal showed no detectable excess risk, although those groups were small and the confidence intervals are wide [1].
4.0 Where it fits in practice
This site’s initial assessment page, /clinical/03-initial-assessment.html, names Friedewald and Martin-Hopkins as the two acceptable ways to calculate LDL. This study gives a reason to prefer Martin-Hopkins where the choice exists, at least in young adults, where one LDL result near 130 mg/dL can decide whether someone gets a closer look.
The size of the effect matters here. Fewer than 3 in 100 people were classified differently, so the formula changes the answer for very few patients. For those few, it changes it in the direction that matters, because the higher reading was the one that went with the higher risk [1].
Nothing here changes who should be treated or what LDL goal applies. Those decisions rest on overall risk, set out in /clinical/04-risk-stratification.html, and on the treatment steps in /clinical/05-treatment-pathways.html. When an LDL result sits close to a decision point, it helps to remember that a calculated LDL is an estimate. Apolipoprotein B, covered in /clinical/06-advanced-tools.html, measures the particle burden directly.
5.0 Open questions
The study did not compare any formula with LDL measured directly in the lab or with apolipoprotein B. It shows which estimate tracked future events best, not which one came closest to the true value [1].
It used one cutoff, 130 mg/dL. Whether Martin-Hopkins keeps the same edge at the lower LDL levels that matter for people already on treatment, or when LDL is treated as a continuous measure, is not addressed [1].
The groups where the formulas disagreed were small, which is why the confidence intervals are so wide. The 3.94 estimate, for example, is compatible with anything from 1.80 to 8.64 [1].
The results so far are from an abstract. It does not report how many people were in each disagreement group, how their triglyceride levels compared, or whether anyone started cholesterol-lowering treatment during follow-up [1].
This summary is provided for clinical decision support only and does not replace individualized clinical judgment. Findings are summarized from the cited primary sources; consult those sources before changing practice.
Version History
| Version | Date | Description |
|---|---|---|
| 1.0.0 | 2026-10-04 | Initial release |
References
- Meng Y et al. Multi-National Evaluation of Low-Density Lipoprotein Cholesterol Equations in Young Adulthood for Long-Term Risk of Atherosclerotic Cardiovascular Disease. Circulation. Population health and outcomes. 2026-09-30. doi:10.1161/circoutcomes.125.013678. Available at: https://doi.org/10.1161/circoutcomes.125.013678