When Seth Shay Martin was a kid, he heard his grandfather's heart skip a beat — literally. His grandfather had a heart condition, and Martin watched him struggle with it for years. Today, Dr. Seth Martin (no relation to the name Shay) leads a team at Johns Hopkins University that's now making heart health decisions easier for millions of people.
Dr. Martin's team has created a computer program, called a machine-learning equation, that helps labs calculate a person's LDL cholesterol levels more quickly and accurately. LDL is the "bad" kind of cholesterol that can build up in arteries and lead to heart attacks or strokes. The new tool is a simplified version of an existing equation called the Martin-Hopkins equation, which labs in the United States and other countries already use.
The difference? The old equation could be tricky for some labs to set up. The new machine-learning version makes it much easier for any laboratory to use — no special steps required.
The team tested their new equation on blood samples from 4.9 million children and adults across the United States. These samples came from the Very Large Database of Lipids, a major collection of health data. The median LDL cholesterol level in these samples was 114 mg/dL.
When researchers compared their machine-learning formula to the original Martin-Hopkins equation, the results were almost identical — just 0.5 mg/dL apart. But the real power showed up when looking at high-risk patients: people with low LDL cholesterol (below 70 mg/dL) and high triglycerides (a type of fat in the blood between 200 and 399 mg/dL). In these tricky cases, the machine-learning equation correctly classified 84% of patients, compared to only 40% for an older method called the Friedewald equation.
"We've optimized the calculation of LDL cholesterol and made this equation accessible and easier for all labs to implement," Dr. Martin said.
Why does this matter? Because today's medical guidelines recommend treating people with high LDL cholesterol to prevent heart disease. If doctors underestimate someone's LDL level, they might miss the chance to start treatment that could prevent a heart attack or stroke. For patients on the borderline — where a difference of just 5 or 10 mg/dL could change whether they get medication — getting it right is critical.
"Our goal is to enable clinicians and patients to make better decisions about starting treatments that prevent heart attacks and strokes and save lives," Dr. Martin said.
The findings and computer code were published in JAMA Cardiology, a major medical journal. Now, labs anywhere can download and use the code. Dr. Martin hopes this simple tool will spread worldwide, helping doctors catch more people at risk before it's too late.
"It's these types of on-the-cusp examples that benefit most from more accurate results," he said.