Skip to content
Michael Fulks MDJan 11, 2023 11:19:00 AM4 min read

Study Suggests Low HDL May Have Less Impact on Black Adults: True? Maybe Not

Michael Fulks, MD, comments on recent research in the Journal of the American College of Cardiology. Dr. Fulks reviews the study’s approach, findings, and potentially substantial limitations.

There has been hype on the recently-released study by Neil Zakai, et al. in the October Journal of the American College of Cardiology concluding that “low HDL-C levels were associated with increased CHD risk in White but not in Black adults. High HDL-C levels were not protective for either race.” A review of this study suggests reasons to be cautious in accepting these results.

Research Background

This sub-study is part of a long-term stroke study intentionally pulling over half of participants from the Southeast U.S. (stroke belt). Of note, 46% of the Black participants had high school or less education, compared to 30% of White participants; this raises the question of whether this is a comparison based on color or more based on socioeconomic status. This alternate interpretation is reinforced by percentage use of tobacco (17% vs. 12%), BMI average (31 vs. 28) and presence of diabetes mellitus (28% vs. 13%), all consistent with expected differences by socioeconomic status.

Further complicating interpretation is the low CHD event rate in the Black cohort, which was only two-thirds that of the White cohort with a remarkable 54% of events in Black participants occurring in females as compared to a more typical 41% for White participants. Finally, the total number of CHD events in Blacks is only 664; this means there were few events at high and low HDL levels, each of which contained just over 30% of Black participants. In contrast to the total of 1,615 total CHD events for this study, an earlier lipid study by CRL had >41,000 events (all-cause mortality) allowing separate analysis by age band, sex, etc.

Interpreting Results

Attempts were made in the current study to adjust for the myriad of potential variables impacting the number of CHD events recorded. Unadjusted race-stratified spline models show increasing risk for the White cohort as HDL falls, and similar but smaller changes for the Black cohort. After adjustment for lipids, age, sex, region, smoking, BP, BMI, DM, statin-use and BP meds, White participants show only a small increase in events at low HDL and no decrease at high HDL, while Black participants show a non-significant decrease in events at both high and low HDL with wide confidence intervals for both cohorts. Decreasing CHD risk as HDL both falls and rises from around 50 mg/dL in the most fully adjusted spline models for the Black cohort is surprising and requires explanation.

Adjusting for unrelated factors influencing the outcome of concern (CHD event) such as age and sex when trying to tease out the impact of a variable such as HDL is well supported but, in this study, related factors such as statin use and cholesterol level are both included, potentially over- or under-counting impact. In contrast, socio-economic class (for which education is a decent surrogate) – a well-known and potent CHD variable – is left out of adjusted variables. To be included, adjusted variables should work the same in the comparison populations; in this case, the CHD event rate in Black females is much higher than for White females, indicating that adjusting for sex has a different impact in White and Black cohorts. It would have been better to split rather than adjust by sex, thus presenting male and female data separately, but the paucity of events likely prevented the authors from doing so. No data on the levels of variables by sex overall or by sex and racial grouping is provided.

Increasing mortality and CHD events are typically seen at very high HDL levels, shown both in CRL and non-insurance studies, resulting in a “J” shaped mortality or event curve and that is presumed to be a result of the association between heavy alcohol use and very high HDL levels. Instead, the Black cohort in this study shows continued reduction in risk as HDL increases, while in the White cohort, risk is flat as HDL increases. Potential explanations for this may be the regional nature of the current study, lack of data, and the nature of the variable adjustments performed.

It is not just the results for the Black cohort which are surprising for this study; the results for the White cohort were also unexpected when compared to earlier larger studies. In contrast, the difference in CHD events between Black and White cohorts in this study is actually very small.


Based on issues raised above including the regional nature of the study, the small number of events precluding needed splitting instead of attempting adjustment for variables and ignoring the major difference in socio-economic class focusing only on color, this article is interesting but hardly conclusive. More dependable is the earlier, larger CRL life insurance and other non-insurance lipid studies such as those using the Framingham dataset which were of adequate size; unfortunately, these studies had few Black participants and/or lacked the ability to identify them. What, if any, adjustment should be made to current algorithms including HDL values for predicting CHD risk for Black or White patients remains uncertain. In addition to self-identification as Black or White (or stratification by socioeconomic status), changes in the US population including increasing obesity, diabetes and use of statins may have also impacted the value of HDL in prediction of future CHD events or mortality. This article may stimulate further analysis, reporting and research.


About the Author

Michael Fulks, MD, Consulting Medical Director, is board-certified in internal and insurance medicine. After leaving practice, he served as a medical director, creating or editing several underwriting manuals and preferred programs. More recently, Mike has consulted for CRL participating in its mortality research on laboratory test results, BP and build, and in the development of risk-scoring tools for laboratory and non-laboratory data.