WEDNESDAY, July 29, 2026 (HealthDay News) -- A simplified machine learning-based low-density lipoprotein cholesterol (LDL-C) equation provides results that are comparable to the original Martin-Hopkins equation, according to a study published online July 15 in JAMA Cardiology.
Jihwan Park, Ph.D., from Johns Hopkins University Bloomberg School of Public Health in Baltimore, and colleagues used multivariate adaptive regression splines to develop a simple machine learning-based alternative to the Martin-Hopkins equation (LDL-C-MH-MARS) and compared its performance to that of the Friedewald (LDL-C-F), Sampson-National Institutes of Health (LDL-C-S), Modified Sampson (LDL-C-MS), and Martin-Hopkins (LDL-C–MH) equations in a study involving 4,939,528 patients with complete lipid panel data.
Participants were randomly assigned to a training set and a test set (3,292,889 and 1,646,639, respectively). The researchers found that the LDL-C-MH-MARS equation had a very low median bias of −0.1 mg/dL, which was comparable with the LDL-C-MH. The median difference between these two equations was −0.5 mg/dL, supporting their comparability. The root mean square error was smallest for LDL-C-MH-MARS and LDL-C-MH (4.7 and 4.9 mg/dL, respectively), followed by LDL-C-S, LDL-C-MS, and LDL-C-F (5.8, 6.0, and 7.2 mg/dL, respectively). For LDL-C-MH-MARS and LDL-C-MH, the proportion of patients correctly classified according to clinical categories was nearly identical (89.7 and 89.6 percent, respectively), but was lower for LDL-C-S, LDL-C-MS, and LDL-C-F (86.3, 84.7, and 83.1 percent, respectively). In external validation datasets, the patterns of results were similar, with the highest accuracy seen for LDL-C-MH-MARS and LDL-C-MH.
"We've optimized the calculation of LDL cholesterol and made this equation accessible and easier for all labs to implement," coauthor Seth Martin, M.D., from Johns Hopkins Hospital, said in a statement.
Several authors disclosed ties to the biopharmaceutical industry.
