New Martin-Hopkins LDL-C-Estimating Equation Simplifies Implementation for Comparable Results

A new investigation into the Martin-Hopkins equation has produced a simplified version of the LDL-C-estimating calculation to streamline care and broaden implementation.¹

The original Martin-Hopkins Equation was developed in response to the rising importance of LDL-C calculation and the continuous risk of underestimating LDL-C levels. However, the equation, first drafted in 2013, requires some laboratories to undertake extra steps in implementing it for patients. To this end, investigators at Johns Hopkins University aimed to streamline the system into something universally applicable and accessible.²

“The equation itself is a bit complex, but it can be coded into the lab IT system as a single line of code, so clinicians never need to deal with that complexity,” Seth Martin, MD, MHS, a professor in the division of cardiology at Johns Hopkins University and an investigator in the study, told HCPLive in an exclusive interview. “It makes it easier for lab directors to implement this, because they can take the line of code that currently is offered with the Friedewald equation to calculate LDL cholesterol, replace it with this new line of code, and get substantially enhanced LDL cholesterol accuracy.”

Martin and colleagues included patients from the second harvest of the Very Large Database of Lipids (VLDBL), which was collected from several US clinics from October 2015 to June 2019. These patients exhibit similar lipid distributions to those of the broader US population. Patients with missing values for total cholesterol, HDL-C, triglycerides, or LDL-C concentration were excluded. The primary analysis focused on patients with triglyceride concentrations <400 mg/dL.¹

The team measured patients’ total cholesterol, LDL-C, VLDL-C, and HDL-C concentrations using the Vertical Auto Profile (VAP) test. Triglyceride measurements utilized the Abbott ARCHITECT C-8000 system. Additionally, the team implemented the Friedewald equation, the Martin-Hopkins equation, the extended Martin-Hopkins equation, the Sampson-National Institutes of Health equation, and the modified Sampson-National Institutes of Health equation.¹

Machine-learning methods were also employed – Martin and colleagues estimated VLDL-C using total cholesterol, triglyceride, and HDL-C concentrations as independent variables, followed by calculating LDL-C concentration by subtracting VLDL-C from non-HDL-C. Of the 3 machine learning algorithms implemented in the study, MARS created a new equation for the estimation of VLDL-C.¹

Martin and colleagues then conducted a statistical analysis of the new equation, randomly assigning patients in a 2:1 ratio to either a training or test dataset. Of the initial 5,081,680 patients in the dataset, 10,856 were excluded for incomplete data, and 131,296 were removed for having triglyceride concentrations ≥400 mg/dL. Of the remaining patients, 3,292,889 were assigned to the training dataset and 1,646,639 to the test dataset. Median LDL-C concentration was 114 mg/dL (interquartile ratio [IQR], 90-141), and the median triglyceride/VLDL-C ratio was 5.1 (IQR, 4.4 to 5.9) for both groups.¹

During the test, the MARS equation showed minimal bias in estimating LDL-C concentration (median [IQR], -0.1; -2.1 to 1.8 mg/dL), comparable with the existing Martin-Hopkins equation. The median difference between the 2 was -0.5 (-1.2 to 0) mg/dL, indicating the methods’ potential comparability. Additionally, the proportion of patients correctly classified to each clinical category was almost identical for the MARS equation and the Martin-Hopkins equation, with 89.7% and 89.6% of patients correctly assigned, respectively.¹

Ultimately, the team determined that this newer version of the Martin-Hopkins equation, developed by the MARS machine learning algorithm, may provide the same quality of results while simplifying its implementation.¹

“We have extensive external validation with the original equation – this one performs similarly, and we have additional external validation included in this paper,” Martin said. “I think this is an equation that’s ready for implementation and really has robust science behind it.”

Editors’ Note: Martin reports disclosures with Amgen, Merck, Arrowhead, Verve Therapeutics, Novartis, Regeneron, and others.

References

1. Park J, Fan L, Marzinke MA, et al. Development and validation of a simplified Martin-Hopkins LDL-C equation using machine learning. JAMA Cardiology. Published online July 15, 2026. doi:10.1001/jamacardio.2026.2314

2. Johns Hopkins Medicine. Machine Learning Method Enables Easier and Wider Use of Martin-Hopkins Equation to Accurately Assess LDL Cholesterol Risk. July 15, 2026. Accessed July 16, 2026. https://www.hopkinsmedicine.org/news/newsroom/news-releases/2026/07/machine-learning-method-enables-easier-and-wider-use-of-martin-hopkins-equation-to-accurately-assess-ldl-cholesterol-risk