Genetic-phenotypic discordance reveals divergent cardiovascular risk at extremes of HDL
E Shemesh, K Pang, P Meyre, A Man, J Graf, S Wang, Y Li, G Da Rocha, M Shapiro, G PareAbstract
Background
A U-shaped relationship between HDL-C and cardiovascular outcomes has been described, yet the underlying mechanism remains unclear. Extreme HDL-C levels reflect distinct determinants — lifelong genetic predisposition vs. acquired exposures.
Purpose
To characterize associations between HDL-C extremes, CVD outcomes and mortality, and to determine whether genetic-phenotypic discordance identifies individuals at distinct risk beyond measured HDL-C alone.
Methods
ASCVD-free White participants from UK Biobank (UKB; n=393,766) and All of Us (AoU; n=89,140) were analyzed. Incident ASCVD, all-cause, and cause-specific mortality were assessed using multivariable Cox models; extreme HDL-C was defined as low (<1mmol/L) or high (≥2.3mmol/L). A 247-variant polygenic score stratified participants into quintiles; Generalized additive models (GAM) visualized risk across the full HDL-C spectrum. Discordance groups were defined at each extreme: acquired low HDL (Q5, low measured) vs. inherited low HDL (Q1, low measured); acquired high HDL (Q1, high measured) vs. inherited high HDL (Q5, high measured). Discordance comparisons used age and sex-adjusted Cox models.
Results
Extremely low HDL-C was associated with higher ASCVD (UKB HR 1.33, 95%CI 1.27–1.39; AoU HR 2.10, 1.91–2.32) and mortality (UKB HR 1.25, 1.20–1.30; AoU HR 1.84, 1.58–2.13). Extremely high HDL-C was not associated with excess ASCVD but was associated with higher mortality in UKB (HR 1.28, 1.19–1.38), driven by COPD mortality (HR 3.06, 2.23–4.21) and lung infection mortality (HR 2.62, 1.71–4.02). Genetic stratification revealed a dose-response: acquired low HDL-C had the highest ASCVD (UKB 8.2% vs 7.2%, p=7.5×10⁻⁶; AoU 5.5% vs 4.3%, p=0.003) and mortality (UKB 11.3% vs 10.4%, p=0.002; AoU 2.3% vs 1.5%, p=0.0005). Acquired low HDL-C was independently associated with higher ASCVD (UKB HR 1.11, 1.04–1.17, p=0.001; AoU HR 1.27, 1.09–1.48, p=0.003) and mortality (UKB HR 1.07, 1.02–1.13, p=0.005; AoU HR 1.55, 1.21–1.99, p=0.0005) vs inherited low HDL-C. Adding the polygenic score improved risk stratification within the low HDL-C subgroup (UKB and AoU p<0.001). At the high extreme, acquired high HDL-C showed no excess risk, whereas inherited high HDL-C showed progressively lower risk (Figure).
Conclusions
Genetic-phenotypic discordance at HDL-C extremes reveals divergent cardiovascular risk that HDL-C alone cannot capture. At the low extreme, discordance — where genetics predict high HDL but measured levels are low — identifies individuals with up to 27% greater ASCVD risk and up to 55% greater mortality risk than those whose low HDL is genetically determined, replicated in both cohorts. At the high extreme, the high-HDL mortality paradox is confined entirely to discordant individuals, while concordant high HDL shows progressively lower risk. Polygenic context improves risk stratification at HDL-C extremes beyond conventional measurement.Figure 1