Press "Enter" to skip to content

LDL Equation May Be Due for Accuracy Upgrade

A way to calculate LDL cholesterol levels that gives more accurate results than standard equations for people with hypertriglyceridemia or very low LDL has been reported.

The equation, which relies on β-quantification of LDL with ultracentrifugation to separate lipoprotein classes in a patient’s serum, topped the Friedewald and Martin methods for accuracy — especially for people with high triglycerides.

It was as accurate up to 800 mg/dL as the Friedewald equation was for levels under 400 mg/dL, according to Alan Remaley, MD, PhD, of the National Heart, Lung, and Blood Institute in Bethesda, Maryland, and colleagues.

While only slightly more accurate in the normal lipid range, the equation led to 35% fewer misclassifications when patients with hypertriglyceridemia in the 400-800 mg/dL range were categorized into different treatment groups according to LDL cholesterol.

The equation could “be reserved only for patients with hypertriglyceridemia and/or low LDL-C levels,” the researchers suggested in JAMA Cardiology. “From a practical standpoint, however, it may be simpler to use the new equation for all patients rather than use two different equations for calculating LDL-C levels.”

Notably, clinical laboratories that adopt the new equation would not have additional costs compared to the standard lipid panel, because the equation has no intellectual property restrictions, according to the investigators.

“It is more complicated than other equations, but the result can be automatically calculated by most laboratory information systems without any additional software changes, unlike those that may be needed to implement the 180-cell factor table used by the Martin equation,” Remaley’s group said.

Having accurate LDL readings at the very low end of the spectrum is important given that some people are now able to achieve this with PCSK9 inhibitors, the authors suggested.

“Because of the overestimation of the VLDL-C [very low-density lipoprotein cholesterol] level by the Friedewald equation and some of the other equations, these patients can have falsely low or even nonsensical negative LDL-C results, which may discourage more aggressive lipid-lowering therapy that has been shown to be beneficial in high-risk patients,” they wrote.

One way that clinicians have tried to resolve the issue of the variable composition of LDL in hypertriglyceridemia has been to take VLDL plus LDL cholesterol (non-HDL cholesterol) as the risk factor, according to William Virgil Brown, MD, professor emeritus at Emory University School of Medicine in Atlanta.

“Improved accuracy compared with the current common method of calculation (Friedewald) and with the Martin equation is addressed by this publication. However, it is a technical improvement that does not answer the key questions of whether this alteration in reported LDL-C values will improve risk estimation or its value as a target of therapy,” he wrote in an accompanying editorial.

Such key questions for Brown include, for example: “How will it compare with using non-HDL-C or the apoB measure as a risk factor or target of therapy? Will it improve the management of those treated to lower LDL-C values with new drugs that are now available? Will routine laboratories adopt this change in methods?”

To create the new LDL equation, Remaley and colleagues analyzed lipid measurements from 8,656 patients seen at the NIH Clinical Center in 1976-1999. Test results were analyzed according to the β-quantification reference method and samples randomized to training and validation data sets.

Median triglyceride level was 149 mg/dL, with 14% of samples showing triglycerides at 400 mg/dL or more.

First, the researchers worked on improving the estimation of VLDL. They created an equation for this from multiple least squares regression using triglyceride and non-HDL cholesterol as independent variables.

Quantification of VLDL was better with this equation when evaluating accuracy by root mean square error (RMSE 15.6 vs 31.7 by Friedewald and 22.1 by Martin) and the correlation coefficient R2 (0.9385 vs 0.8341 and 0.8405, respectively).

This was then incorporated into a general equation for LDL cholesterol calculation.

Equations were tested against the internal validation data set and multiple external data sets of either β-quantification LDL results (n=28,891) or direct LDL test results (n=252,888).

“Additional studies are needed in other populations, particularly those with low LDL-C levels, to more fully validate the equation and to better understand its clinical utility,” according to Remaley’s team.


This research was supported by the Intramural Research Program of the National Heart, Lung, and Blood Institute at the NIH.

Remaley disclosed no conflicts.

Brown reported receiving personal fees from Amgen, serving as a member of data monitoring committees for ongoing clinical trials sponsored by both Amgen and Esperion, and owning common stock in Akcea and Ionis.