Machine Learning Update to Martin-Hopkins Equation Improves LDL-C Accuracy for Labs

Accurate assessment of low-density lipoprotein cholesterol (LDL-C) sits at the center of cardiovascular risk management, guiding decisions about when to initiate or intensify lipid-lowering therapy. A new study published in JAMA Cardiology introduces a simplified machine learning version of the Martin-Hopkins equation that matches the accuracy of the original while being easier and free for laboratories to adopt. This development presents direct implications for pharmacists interpreting lipid panels at the point of care.^1,2

Why LDL-C Accuracy Matters at the Counter

Current guidelines recommend treating patients to progressively lower LDL-C targets, making precise measurement more consequential than ever. According to senior study author Seth Martin, MD, MHS, professor and preventive cardiologist at Johns Hopkins Medicine and director of the Advanced Lipid Disorders Program and Digital Health Lab at the Johns Hopkins Ciccarone Center for the Prevention of Cardiovascular Disease, underestimation of LDL-C using older equations can produce falsely reassuring results and lead to missed treatment opportunities.^1,3

"LDL is front and center in our clinical practice, and it helps guide decisions around medicines that reduce atherosclerotic cardiovascular disease, ASCVD events, and death," Martin said in an interview with Pharmacy Times. He emphasized that the machine learning version is a fixed, transparent equation coded into the lab rather than a shifting dynamic model, ensuring consistent results year after year. This development would allow busy pharmacists and clinicians to rely on automated, trustworthy calculations.

How the Equation Performed

Researchers developed and tested the machine learning formula using blood samples from 4.9 million US children and adults drawn from the Very Large Database of Lipids, with a median LDL-C level of 114 mg/dL. More than 3.2 million samples trained the model and 1.6 million tested it, with additional reference laboratory and proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitor clinical trial datasets used for validation against ultracentrifugation, the gold-standard research measurement.^1,2

The machine learning version differed from the original Martin-Hopkins equation by just 0.5 mg/dL. Both Martin-Hopkins equations correctly classified 90% of samples into the appropriate treatment category, compared with 86% for the Sampson-NIH equation, 85% for the modified Sampson-NIH equation, and 83% for the Friedewald equation.^1,2

The High-Risk, Low-LDL-C Stress Test

The equation's advantage was most pronounced in the population Martin described as the "ultimate stress test" of LDL-C calculation: patients with low LDL-C and elevated triglycerides. Among samples with triglycerides between 200 and 399 mg/dL and LDL-C below 70 mg/dL, the machine learning equation accurately classified 84% of high-risk samples, compared with 72% for the modified Sampson-NIH equation, 61% for the Sampson-NIH equation, and 40% for the Friedewald equation.^1,2

Martin urged particular vigilance in this scenario. "If you see a lipid panel...in someone you're treating with very high risk ASCVD and their LDL levels coming back on the lower side," he said, an underestimating equation may make a patient appear at goal when additional therapy—such as ezetimibe (Zetia; Organon), bempedoic acid, or a PCSK9 inhibitor—is actually warranted.¹

What Pharmacists Should Take Away

Martin advised pharmacists to know which equation their local laboratory uses when interpreting results. For example, Quest Diagnostics uses the Martin-Hopkins equation, whereas other labs may not. He encouraged pharmacists to serve as local champions for adopting more accurate calculations where needed.

The open-access code supports implementation of the 2026 national dyslipidemia guideline, which preferentially recommends the Martin-Hopkins calculation and sets LDL-C goals of less than 100, 70, and 55 mg/dL depending on cardiovascular risk. Martin also noted the guideline incorporates non–high-density lipoprotein and apolipoprotein B goals to further confirm therapy has been optimized.^1,3

REFERENCES

1. Johns Hopkins Medicine. Machine learning method enables easier and wider use of Martin-Hopkins equation to accurately assess LDL cholesterol risk. News Release. EurekAlert! Embargoed For Release: July 15, 2026. Accessed July 14, 2026. https://www.eurekalert.org/news-releases/1135955

2. Martin S, Park J, Fan L, et al. Development and validation of a simplified Martin/Hopkins low-density lipoprotein cholesterol equation using machine learning. JAMA Cardiology. Published Online July 15, 2026.

3. Saseen J, Halpern L. 2026 dyslipidemia guidelines: earlier screening and what’s coming next. Pharmacy Times. April 2, 2026. Accessed July 14, 2026. https://www.pharmacytimes.com/view/2026-dyslipidemia-guidelines-earlier-screening-and-what-s-coming-next