High-density lipoprotein cholesterol (HDL-C) has traditionally been regarded as a protective factor against cardiovascular disease (CVD). However, recent studies have challenged this idea, suggesting that extremely high HDL-C levels may paradoxically increase mortality risk (1–3). A U-shaped or J-shaped association between HDL-C and mortality has been observed, meaning that not only low but also very high HDL-C levels are associated with increased mortality risks. A prospective, multicenter cohort study using data from the UK general population and three Emory Healthcare sites in Atlanta, Georgia, found that individuals with coronary artery disease and HDL-C levels exceeding 80 mg/dL had a 96% higher risk of all-cause mortality and a 71% higher risk of cardiovascular mortality compared to those with HDL-C levels between 40 and 60 mg/dL, even after adjusting for covariates (1). Additionally, a population-based cohort study in the United States revealed a non-linear association between HDL-C levels and all-cause, cardiovascular, and cancer mortality in diabetic patients (2).

The contentious nature of these findings, particularly when stratified by sex, adds complexity to using HDL-C as a biomarker of the risk of CVD. The divergence in findings regarding the associations between HDL-C and mortality may stem from various factors, including sex-related factors. HDL-C levels tend to be higher in females than males (4), which is thought to be influenced by biological factors such as hormones. Indeed, HDL-C decreases in females after menopause (5). Differences among ethnic groups also exist, and lifestyle, behavioral factors, and genetic factors all contribute to variations in HDL-C levels (4). Moreover, these factors likely not only affect HDL-C levels but also impact other aspects of the body. Adjusting for these factors as comprehensively as possible to confirm associations between HDL-C and mortality is important.

The landscape of research on the association between HDL-C and mortality has been significantly influenced by studies focusing on specific risk groups, including individuals with type 2 diabetes mellitus (T2DM), atherosclerotic cardiovascular disease (ASCVD), and the older population (2, 6, 7). However, evaluating the association within specific risk groups can add an additional layer of complexity to understanding the role of HDL-C in mortality. In patients with T2DM, dyslipidemia, including of HDL-C, is a common finding. The prevalent lipid abnormality pattern in T2DM, referred to as diabetic dyslipidemia, encompasses elevated triglyceride levels, a diminished concentration of HDL-C, and a transition to small dense low-density lipoprotein (8). Hypertriglyceridemia leads to an increased level of triglyceride-enriched HDL, which undergoes accelerated clearance by the kidneys, ultimately resulting in decreased concentrations of HDL in the peripheral circulation (9). Furthermore, T2DM is characterized not only by reduced levels of HDL-C, but also impaired HDL function. This deficiency in functional HDL is closely linked to changes in the metabolism and structure of HDL within the vasculature (10, 11).

To date, research investigating the association between HDL and mortality in apparently healthy individuals without T2DM is still lacking. Therefore, our study aimed to explore the association between HDL and all-cause mortality in non-diabetic Koreans, focusing on sex differences.

Our study revealed a U-shaped association between HDL-C levels and all-cause mortality in non-diabetic males. Both low HDL levels (<40 mg/dL) and very high HDL-C levels (>80 mg/dL) were linked to an increased risk of mortality compared with the reference group (40–60 mg/dL). Furthermore, the Kaplan–Meier curve for all-cause mortality by HDL-C group further emphasized that, in males, the very high HDL-C group experienced a higher incidence of mortality than the low HDL-C group (Figure 3). Conversely, in females, when compared to the reference group (50–75 mg/dL), there was a significant association with increased mortality in the lower HDL-C range (<50 mg/dL), while no significant association was observed in the higher HDL-C range (>75 mg/dL).

An increase in HDL-C levels has been associated with a decrease in ASCVD risk, leading to its classification as “good” cholesterol (12, 13). In more detail, in the ASCVD risk prediction model proposed by the Framingham Heart Study, an HDL-C level ≥ 60 mg/dL results in a reduction of 1 when performing the ASCVD risk factor calculation. This implies that maintaining HDL-C levels above this threshold has cardio-protective benefits (14). Based on this, there has been considerable interest in raising HDL levels with the anticipation of a subsequent reduction in ASCVD risk. However, the results have been somewhat disappointing. In randomized controlled trials evaluating the effects of inhibiting the cholesteryl ester transfer protein (CETP) to elevate HDL-C levels (15, 16), no reduction in CVD risk was observed and in some cases, the trial was prematurely terminated due to an increase in mortality rate (15). Additionally, numerous epidemiological studies have consistently revealed a U-shaped association between HDL and mortality (17–22). This has raised questions about whether HDL is indeed the “good cholesterol” and has underscored the necessity for reevaluation of its purported cardiovascular protective effects.

Various factors can contribute to elevated HDL-C levels, including genetic variants such as CETP deficiency and mutations in ABCA1 or APOA1, as well as lifestyle factors like moderate-to-high alcohol consumption, regular physical activity, and low triglyceride levels (23–25). In our study, individuals in the highest HDL-C quartile had lower BMI, smaller waist circumference, and higher rates of alcohol intake and exercise (Table 1). Additional linear regression analyses confirmed significant associations between HDL-C levels and BMI, waist circumference, triglycerides, and alcohol intake in both sexes (Supplementary Table 2). However, elevations in HDL-C from these factors do not necessarily indicate improved HDL function and may instead be associated with dysfunctional or pro-inflammatory HDL particles (26), which could help explain the increased mortality risk observed at extremely high HDL-C levels.

The mechanisms underlying the paradoxical increase in mortality observed at extremely high levels of HDL-C remain incompletely understood. However, several plausible explanations can be considered, including the loss of the protective functions of HDL, genetic influences, and disruptions in cholesterol transport dynamics. The conventional understanding of the cardioprotective role of HDL-C primarily centers on reverse cholesterol transport, wherein HDL facilitates the efflux of cholesterol from peripheral tissues and atherosclerotic plaques to the liver for excretion, thereby reducing the risk of ASCVD (27). However, recent evidence suggests that excessively high HDL-C levels may be associated with dysfunctional HDL particles that lose their protective properties and may even exert harmful effects (26). HDL particles are structurally diverse, differing in size, density, and apolipoprotein content, which influences their functional capacity (4). In an oxidative or inflammatory environment, HDL can become pro-inflammatory and pro-atherogenic, impairing endothelial function and promoting vascular dysfunction (28). Additionally, oxidative modifications of apolipoprotein A-I, the major protein component of HDL, have been shown to reduce cholesterol efflux efficiency, limiting the anti-atherogenic properties of HDL (15). Thus, in individuals with extremely high HDL-C levels, it is possible that HDL particles undergo structural and functional alterations that negate their vascular protective effects, thereby increasing mortality risk.

In addition to functional impairments, genetic factors regulating HDL-C metabolism may contribute to the increased mortality risk associated with very high HDL-C. Several genetic variants have been identified that lead to markedly elevated HDL-C concentrations and may exert systemic effects beyond simply increasing HDL-C levels (25). Indeed, some of these variants have been associated with an increased risk of cardiovascular disease despite high HDL-C levels, as demonstrated in previous studies (29–32). Thus, the associations observed in this study might be partially explained by underlying genetic determinants that predispose individuals with extremely high HDL-C to increased mortality risks. Furthermore, the reverse remnant cholesterol transport hypothesis offers additional insight into the non-linear relationship between HDL-C and mortality. According to this hypothesis, free cholesterol from triglyceride-rich lipoproteins is transferred to HDL for removal. However, previous research has shown that this transfer process becomes less efficient at both very low and very high HDL-C levels, leading to impaired cholesterol clearance (33). As a result, cholesterol-rich remnant particles may accumulate, contributing to a higher risk of cardiovascular and all-cause mortality. This may help explain the U-shaped relationship observed between HDL-C levels and mortality.

While our primary focus was on all-cause mortality, we also examined CV-related and cancer-related mortality to assess whether the association with HDL-C levels varied by cause of death. However, no significant associations were observed, particularly among individuals with higher HDL-C levels. These findings suggest that the impact of HDL-C on mortality may be mediated by multiple underlying mechanisms rather than a single dominant cause. Given prior research suggesting that very high HDL-C may be more strongly associated with non-cardiovascular mortality (17, 34), future studies with larger datasets and more detailed cause-of-death classifications are needed to better elucidate these associations.

In our study, we did not observe a significant association between high HDL and mortality in women, unlike in men. Women generally have higher levels of HDL-C than men, which could be a factor contributing to the lack of a distinct association with mortality (4). Guidelines for CVD prevention also divide the criteria for low HDL-C associated with elevated CVD risk by sex, setting it at <40 mg/dL in men and <50 mg/dL in women (35, 36). We applied this division when categorizing HDL-C groups for men and women, respectively. There has been inconsistency in previous studies regarding the association between very high HDL cholesterol and mortality outcomes, particularly when stratified by sex. Several prior studies, similar to ours, found a significant positive association between high HDL and mortality in males. However, these studies also failed to find a significant association between high HDL and mortality in females (6, 17, 20, 37). However, other studies have reported increased mortality risk associated with very high HDL-C in both sexes. For example, a large-scale study that included two prospective cohort studies demonstrated a U-shaped association between both extremely high and low concentrations of HDL-C and elevated all-cause mortality risk in both males and females (38). This study, conducted in a Danish population, suggests that ethnic differences might contribute to the variance in results, considering that Asian Americans typically have lower levels of HDL-C than white people (39). Another significant study in Canada, involving a big data approach with 631,762 participants without previous cardiovascular conditions, reported that individuals with elevated HDL levels (>70 mg/dL in men, >90 mg/dL in women) had a higher risk of non-cardiovascular mortality (34). It is crucial to note the distinction from our study, which focused on all-cause mortality. Finally, research conducted on Koreans with ASCVD demonstrated that both males and females in the extremely high HDL-C (>90 mg/dL) category had 35.9 and 9.5% higher 10-year mortality risks than those in the high HDL-C group (40–90 mg/dL for males, 50–90 mg/dL for females), respectively (7). However, it is essential to recognize that this study specifically targeted individuals with ASCVD, unlike our study population.

In subgroup analyses stratified by age, the association between HDL-C levels and all-cause mortality differed across age and sex. Among males under 60 years, both low (<40 mg/dL) and high (>80 mg/dL) HDL-C levels were significantly associated with increased mortality, consistent with the overall U-shaped pattern observed in the main analysis. In contrast, among males aged 60 years or older, only low HDL-C was associated with higher mortality, while extremely high HDL-C showed no significant association. In females, low HDL-C was linked to increased mortality only in those under 60, and no significant association was observed for high HDL-C in either age group. These findings suggest that younger individuals, particularly males, may be more vulnerable to the adverse effects of extremely high HDL-C, while this relationship weakens with age. One possible explanation involves age-related changes in lipoprotein metabolism and HDL functionality. Aging alters HDL composition—reducing cholesterol and apoE content, increasing sphingomyelin, and enriching pro-inflammatory proteins such as serum amyloid A—which impairs its antioxidant capacity and paraoxonase 1 activity (40). These changes may convert HDL from an anti-atherogenic to a pro-atherogenic particle. Additionally, age-related endothelial dysfunction and chronic inflammation may amplify the harmful effects of dysfunctional HDL in younger adults, while in older individuals, the cumulative burden of comorbidities may overshadow HDL-related risk. These results underscore the importance of considering age and sex when interpreting HDL-C levels and highlight the need for individualized lipid management strategies.

Our study has certain limitations. First, although we conducted additional analyses examining CV-related and cancer-related mortality, these did not yield statistically significant findings, likely due to the limited number of events. Additionally, data collection in our study focused on Korean adults aged 40–69 years, which may limit the applicability of our findings to other countries, age ranges, and ethnic groups. Second, our study was limited to Korean adults aged 40–69 years, which may restrict the generalizability of our findings to other populations. Ethnic and regional variations in genetic predisposition, dietary habits, lifestyle factors, and healthcare systems may influence the association between HDL-C and mortality (39). Therefore, validation in diverse racial and geographic populations is essential to determine whether these associations are consistent across different ethnic backgrounds. Future large-scale, multi-ethnic cohort studies are needed to further clarify these relationships and refine risk stratification for individuals with extremely high HDL-C levels. Finally, there may have been unaccounted-for residual confounding factors. The strength of our study lies in its exploration of the association between HDL-C and mortality in a relatively healthy population, excluding the confounding influence of T2DM. Particularly noteworthy is the insight from our study that in males, when HDL-C is above 80 mg/dL, the cumulative incidence of mortality is higher than the commonly recognized risk factor of low HDL-C, emphasizing the urgency of setting upper-level thresholds. Additional research is essential to determine at what concentration HDL-C loses its protective functions. Ongoing research is imperative to unravel the complexities of HDL-C’s roles in the human body (41–43).

We demonstrated a U-shaped correlation between HDL-C levels and all-cause mortality in non-diabetic Korean males, but not females. Furthermore, we observed that in males, the very high HDL-C group (>80 mg/dL) had a higher cumulative incidence of mortality than the low HDL-C group (<40 mg/dL). Further research is urgently needed, particularly in males, to establish upper-level thresholds for HDL-C.