Extreme HDL-C associated increased mortality is explained by confounding due to progressively increasing alcohol-induced multisystem morbidity
B A Ference, D L Wicks, A Hsieh, J J P KasteleinAbstract
Background
The reported U-shaped association between HDL-C and mortality has raised concern that potently increasing HDL-C therapeutically could increase cardiovascular (CV) or all-cause mortality (ACM).
Purpose
To determine whether extremely high HDL-C is independently associated with major adverse cardiovascular events (MACE), CV-mortality, or ACM after adequately controlling for confounding by alcohol consumption and alcohol-related morbidity.
Methods
We analyzed 429,892 UK Biobank participants (198,510 men; 231,382 women; 12.4 million person-years of follow-up) who experienced 36,205 MACE and 34,843 deaths. Participants were divided into a reference group (sex-specific median HDL-C) and groups with HDL-C between 90–100, 100–110, 110–140, and >140 mg/dL. Groups were compared by alcohol use quantity and frequency, gamma-glutamyl transferase (GGT), systolic blood pressure (SBP), heart rate (HR), BMI, weight, and apoB. The association between HDL-C above 100 mg/dL and risk of MACE, CV-mortality, and ACM was evaluated using time-to-event analysis with age as the time scale. Confounding by alcohol use and alcohol-related morbidity was controlled using a) Mahalanobis distance nearest-neighbor matching of participants with extreme HDL-C to two participants with average HDL-C with similar alcohol use and morbidity; b) Mendelian randomization (MR); and c) meta-analysis of CETP inhibitor trials.
Results
A total of 428 (0.5%) men and 3,295 (2%) women had HDL-C above 100 mg/dL. Participants with extreme HDL-C had progressively higher alcohol use, alcohol frequency, GGT, SBP, and HR; and progressively lower BMI, weight, and apoB—characteristic of the metabolic signature of more severe alcohol-induced multisystem morbidity (TABLE). Participants with HDL-C >100 mg/dL had lower risk of MACE and CV mortality in men and women, and lower ACM in women (HR 0.90, p=0.05). Men with HDL-C >100 mg/dL had higher unadjusted ACM (HR 2.87, p<0.005), persisting after conventional adjustment for alcohol use alone (HR 1.95, p<0.005). However, this excess risk was absent (HR 0.98, p=0.88) in matched analyses that adequately controlled for increasing alcohol-induced multisystem morbidity (TABLE). MR analyses and the CETP inhibitor trial meta-analysis confirmed that extreme HDL-C was not associated with increased mortality (HR 0.93; 95%CI: 0.85–1.02) (FIGURE).
Conclusions
Extremely high HDL-C is not independently associated with increased mortality. Prior reports of excess mortality with extreme HDL-C can be explained by a steep, dose-dependent gradient of confounding by alcohol-induced multisystem morbidity that has not been previously identified or adequately controlled. Therefore, therapeutically increasing HDL-C, including by CETP inhibition, should be safe.