The Friedewald equation for estimating low-density lipoprotein cholesterol has been used in clinical practice for more than 40 years. Though this equation has limitations, none of the previously proposed methods to estimate LDL-C have proven superior. Researchers at Johns Hopkins University and the University of Illinois recently developed a new means of estimating LDL-C, and their findings are the subject of this issue of Strategies.

Researchers at the National Institutes of Health in 1972 first proposed what is now known as the Friedewald equation as a simple way to estimate LDL-C using only measurements of serum total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), and triglycerides, without performing more elaborate and costly testing to directly measure LDL-C. Over the ensuing decades, this equation, LDL-C=TC – HDL-C – (triglycerides/5), has been widely adopted in clinical practice and in research, even though it has limitations.

Shortcomings of the Friedewald equation include that it is inaccurate in patients with very high levels of triglycerides >400 mg/dL, and in those with type III hyperlipidemia. The Friedewald equation also is meant to be used only in samples from fasting patients, as it does not account for cholesterol in chylomicrons that form after meals, nor does it reflect cholesterol on lipoprotein (a). A final knock against the Friedewald equation is that it depends on accurate measurement of HDL-C, yet most clinical laboratories over the past decade have switched to fully automated direct HDL-C assays, which can be affected by certain disease states and can have substantial biases in comparison to the reference direct HDL-C assay method, beta quantification.

Even with these drawbacks, alternative equations for estimating LDL-C proposed over the years have not proven to be any better than the Friedewald equation. Yet the need to provide an estimate of LDL-C that is as robust as possible is arguably even more important today than it has been in the past, according to Seth Martin, MD, a Pollin fellow at Johns Hopkins University in Baltimore and lead author of a study that investigated a new method to estimate LDL-C (JAMA 2013;310:2061–8).

“There’s been the suggestion for a long time that in certain situations the Friedewald equation didn’t work as well as one would want, such as in patients with low LDL-C and high triglycerides. There’s also been an increase in the population of patients with metabolic syndrome who have high triglycerides,” said Martin. “As clinicians, we don’t want to do well for most people; we want to know when we have an individual patient that we can trust the LDL-C value we’re making decisions on.” Martin added that treatment recommendations aimed at lowering cardiovascular disease risk traditionally have focused on lowering LDL-C, another reason to have solid estimates of LDL-C.

He and his colleagues set about developing a new way to estimate LDL-C by accessing de-identified lipid profile results performed on 1,350,908 samples by Atherotech Diagnostics Laboratory. TC, LDL-C, very low density lipoprotein-C (VLDL-C), and HDL-C all were measured using an ultracentrifugation assay; triglycerides were directly measured. The researchers randomly assigned two-thirds of the results to a derivation data set; they used the remaining third to internally validate the new LDL-C estimation method.

The authors explored how age, sex, and lipid profile characteristics contributed to variance in the ratio of triglycerides to VLDL-C by performing multiple linear regression analyses and by analyzing the distribution of triglycerides to VLDL-C ratios in comparison to triglyceride and non-HDL-C concentrations. The median triglyceride:VLDL-C ratio was 5.2, and the first to 99th percentile range was 3.1–9.9. The researchers determined that more than 50% of the range of this variation was explained by triglyceride levels. Adding non-HDL-C accounted for about two-thirds of the variance, while adding age and sex did not materially improve it.

The researchers generated 16 tables of median triglyceride:VLDL-C ratios using from 10 to 2,000 cells, and settled on 180 cells in the final model. “We found that the model didn’t perform significantly better when we used the 360 or larger tables,” explained Martin. “If you look at the 180-cell table, the change in the factors is very small from one cell to the other, at least in the bulk of the table. They go from 5.1, 5.2, 5.3, etcetera in small increments. It’s only in the extreme corners of the table where there are larger jumps.” He added that the outer cells of the table have data from fewer patients in each cell, thereby reducing how robustly they reflect the relationship among the factors in estimating LDL-C.

The authors subsequently developed an automated calculator of the 180-cell table so that other researchers can test the model. This calculator is available online at Martin emphasized that the method needs to be validated by other researchers.

The authors of an editorial that accompanied the study called the model “an elegant way to estimate LDL-C.” The authors, J. Michael Gaziano, MD, MPH, and Thomas Gaziano, MD, MSc, also suggested that if validated, the model could “create a simple, but important, method to more accurately estimate LDL-C levels that could be easily implemented at little or no cost at the time that the data are presented to the clinician. This approach could reduce the need for more expensive direct LDL-C measurements….” (JAMA 2013;310:2043–4).

However, another researcher, Alan Remaley, MD, PhD, is not as confident about the generalizability of this method. “The trouble is, there are eight different assays for direct HDL-C measurement and in my own published research I’ve compared these assays with the gold standard, beta quantification, and they often don’t work well. They have positive and negative biases. So what you would end up with is, eight different HDL-C numbers depending on which assay you used. So you would need eight different 180-cell tables that the authors have proposed.” Remaley is a senior investigator in the lipoprotein metabolism section at the National Institutes of Health.

Remaley added that while clinical labs uniformly have switched to direct HDL-C assays, only about half measure LDL-C directly, and still rely on an equation to estimate LDL-C. He also has published research evaluating four different equations for calculating LDL-C using eight different direct HDL-C assays. This study found that of the four evaluated, Friedewald performed the best in estimating LDL-C, but its accuracy varied considerably depending on the direct HDL-C assay used (Clin Chem Acta 2013;423:135–40). In a supplemental table, Martin and his co-authors looked at how well seven equations for estimating LDL-C did in classifying LDL-C based on clinical practice guidelines and found that none performed as well as their proposed equation.

Martin emphasized that he and his colleagues welcome further analysis of their calculator by other research teams. “We really look forward to other people testing our methods or getting in touch if they want to collaborate with us or if there is any way we can be helpful,” he said. “There’s still a lot of work to be done and it will take a lot of expertise from a lot of stakeholders to do this as best as possible.”