The Martin-Hopkins equation to assess low-density lipoprotein (LDL) cholesterol levels in blood samples has been used by laboratories in the U.S. and other countries to guide efforts to lower cardiovascular disease risk. Now, a simplified machine learning version of this equation has been shown in a study of millions of U.S. adult and child blood samples to match the accuracy of the original — making it broadly accessible. The findings and code published today in JAMA Cardiology.

“We've optimized the calculation of LDL cholesterol and made this equation accessible and easier for all labs to implement,” says Seth Martin, M.D., M.H.S., the senior study author and director of the Advanced Lipid Disorders Program and Digital Health Lab at the Johns Hopkins Ciccarone Center for the Prevention of Cardiovascular Disease. “Our goal is to enable clinicians and patients to make better decisions about starting treatments that prevent heart attacks and strokes, and save lives.”

Accurately assessing LDL cholesterol is more important than ever because today’s guidelines recommend treating to lower levels to reduce cardiovascular risks. However, underestimation of LDL cholesterol using some equations can lead to missed treatment opportunities, Martin explains, which is a problem that the Martin-Hopkins equation helps solve. Its major strength comes in providing the most accurate results among people with increased cardiovascular risks who have low LDL cholesterol levels and high triglycerides.

“A lipid profile with low cholesterol and high triglycerides is the ultimate stress test of the LDL cholesterol calculation,” says Martin. He explains that a 5, 10 or 20 mg/dL difference, based on various equations, could change a person’s eligibility for treatment, such as with PCSK9 inhibitors, which have been shown to significantly lower LDL cholesterol levels. “It’s these types of on-the-cusp examples that benefit most from more accurate results,” he adds.

Since the 2013 equation may require some laboratories to take extra steps to implement, the researchers created a streamlined code to make it easy for all laboratories to use. 

They started by creating and testing a machine-learning formula using blood samples from 4.9 million U.S. children and adults. These samples, which are representative of the U.S. population, had a median LDL cholesterol level of 114 mg/dL and came from the Very Large Database of Lipids. The researchers then compared the machine learning formula’s results to the original Martin-Hopkins equation results and examined its accuracy in reference to the Sampson-NIH and Friedewald equations, which are other common equations used in laboratories. To gauge accuracy, they compared calculated results with LDL cholesterol levels assessed through ultracentrifugation, a gold-standard tool used in research settings.

Overall, they found that the machine-learning version of the Martin-Hopkins equation was similar to the original equation with a minimal difference of 0.5 mg/dL. Both Martin-Hopkins equations correctly classified 90% of samples within the correct treatment category, the Sampson-NIH equation correctly classified 86%, the modified Sampson-NIH equation classified 85% and the Friedewald equation classified 83%. Most importantly, says Martin, the investigators found that the Martin-Hopkins equations were the most accurate for classifying high-risk patients with lower ranges of LDL cholesterol levels.

When it came to assessing people who had triglycerides between 200 mg/dL and 399 mg/dL and LDL cholesterol levels less than 70 mg/dL, the Martin-Hopkins machine learning equation accurately classified 84% of high-risk samples, the original Hopkins-Martin equation classified 83%, the modified Sampson-NIH equation classified 72%, the Sampson-NIH equation classified 61% and the Friedewald equation classified 40%. 

As part of this research, more than 3.2 million samples from the original lipid database were used to train the machine learning model. An additional 1.6 million samples were used to test the models. Two additional datasets, including a reference laboratory dataset and a clinical trial dataset of patients who had used PCSK9 inhibitors, were used to validate these findings in comparison to ultracentrifugation-based measurements.

Martin explains that further testing the calculation and ensuring its accuracy and reliability in populations outside of the group used to develop the model supports its ability to be generalized and widely used for clinical implementation.

A strength of this machine learning equation is the ability for laboratories to use its code across different systems, the authors note. Laboratories can substitute the current triglyceride portion used in the Friedewald equation — the first LDL cholesterol equation, created and introduced in the 1970s — for the machine learning formula in the Martin-Hopkins equation. 

“This updated equation is not only highly accurate, but it’s transparent and can be easily adopted by laboratories,” says Mark Marzinke, Ph.D., a study author who is the medical director overseeing this testing in the Johns Hopkins Hospital Core Laboratories and a professor of pathology and medicine at Johns Hopkins. “We wanted to avoid creating a ‘black box’ equation that is opaque or invisible to most users.”

The authors explain that this open-access calculation can improve implementation of the 2026 national dyslipidemia guideline, which recommends preferentially using the Martin-Hopkins calculation to assess LDL cholesterol. This gives patients and their treating clinicians greater confidence in selecting treatments and achieving guideline-recommended LDL cholesterol goals to ensure optimal cardiovascular protection. Depending on one’s level of cardiovascular risk, the guideline-recommended LDL cholesterol goals are to bring levels below 100, 70 and 55 mg/dL.

Additional authors include Jihwan Park, Leon Fan, Lori Sokoll, Alagarraju Muthukumar, Sabina Murphy, Marc Sabatine, Rachana Gurudu and Jeff Meeusen. 

The Martin-Hopkins equations do not have patents or intellectual property restrictions. 

The Very Large Database of Lipids, which was used to train and test the equation, has received funding from the David & June Trone Foundation. The FOURIER trial, which provided blood samples to validate the findings, was funded by Amgen. 

Martin has received research support from Amgen and Merck. Outside of this work, he has received consulting fees from Amgen, Arrowhead, Chroma, Heartflow, Kaneka, Merck, NewAmsterdam, Novartis, Regeneron and Verve Therapeutics. Marzinke reports receiving grants from NIH and ViiV Healthcare. 

To watch a video of Martin explaining the Martin-Hopkins equation, click here.

On the Web

Cholesterol in the Blood | Johns Hopkins Medicine

The Ciccarone Center | Johns Hopkins Medicine