Context Epidemiological evidence that the risk of coronary heart disease is inversely associated with the level of high-density lipoprotein cholesterol (HDL-C) has motivated several phase III programmes with cholesteryl ester transfer protein (CETP) inhibitors.

Objectives To assess alternative methods to predict clinical response of CETP inhibitors.

Methods Meta-regression analysis on raising HDL-C drugs (statins, fibrates, niacin) in randomised controlled trials.

Results 51 trials in secondary prevention with a total of 167 311 patients for a follow-up >1 year where HDL-C was measured at baseline and during treatment. The meta-regression analysis showed no significant association between change in HDL-C (treatment vs comparator) and log risk ratio (RR) of clinical endpoint (non-fatal myocardial infarction or cardiac death). CETP inhibitors data are consistent with this finding (RR: 1.03; P5–P95: 0.99–1.21). A prespecified sensitivity analysis by drug class suggested that the strength of relationship might differ between pharmacological groups. A significant association for both statins (p<0.02, log RR=−0.169–0.0499*HDL-C change, R²=0.21) and niacin (p=0.02, log RR=1.07–0.185*HDL-C change, R²=0.61) but not fibrates (p=0.18, log RR=−0.367+0.077*HDL-C change, R²=0.40) was shown. However, the association was no longer detectable after adjustment for low-density lipoprotein cholesterol for statins or exclusion of open trials for niacin.

Conclusions Meta-regression suggested that CETP inhibitors might not influence coronary risk. The relation between change in HDL-C level and clinical endpoint may be drug dependent, which limits the use of HDL-C as a surrogate marker of coronary events. Other markers of HDL function may be more relevant.