In the report of Drexel et al. (1), an HDL-related cluster of lipid markers was significantly associated with risk of cardiovascular disease among coronary patients with type 2 diabetes, whereas an LDL-related cluster was not significantly associated. The authors conclude from their observation that “in today’s setting of effective LDL cholesterol lowering with statins, HDL cholesterol, triglycerides, and LDL particle diameter are the predominant lipid risk factors determining the fate of coronary patients with diabetes.”
Although Drexel et al. briefly discuss the issue of sample size, their substudy of patients with type 2 diabetes appears to be severely underpowered. It would be required to be more than two times larger to provide a power of 80% and at least four times larger to provide a power of 95% at an α level of 5% (2). Clearly, the nonsignificance of the association for the LDL-related pattern does not allow the authors to conclude that there is indeed no association. It seems rather likely that the observed relative risk (RR) for the LDL-related pattern (1.362 [95% CI 0.985–1.883], P = 0.061) suggests an association. In addition, point estimates for both patterns are quite similar after reparametrization. Actually, the observed RR of 0.702 for an increase of the HDL-related pattern corresponds to an RR of 1.412 if the risk associated with a reduction of the HDL-related pattern is being modeled. Thus, both patterns have similar predictive value of future vascular events. We expect that this similarity will be even more obvious if odds ratios would be calculated in a restricted analysis by excluding those 41% of diabetic patients receiving lipid-lowering drugs instead of only medication adjustment. Thus, dyslipoproteinemia characterized by high LDL cholesterol still appears to play an important role in the management of cardiovascular risk in diabetic patients. This is also in line with recent evidence from larger cohorts, where LDL cholesterol was a significant predictor of cardiovascular disease events among 746 diabetic men (3) and 921 diabetic women (4), whereas HDL cholesterol appeared to predict risk only among men.
The question remains what combination of lipid markers actually provides the best prediction for future vascular events. Statistical methods other than factor analysis, e.g., canonical discriminant analysis, are possibly more appropriate to derive the biomarker profile with maximal predictive power. Also, non-HDL cholesterol should be included in the analysis as a marker of atherogenic cholesterol because it may be more valuable to quantify atherosclerotic risk among diabetic patients, particularly among those with elevated triglyceride levels (5).
