An evaluation of class effect

The results of Zhou's retrospective observational study¹ conflict with the evidence obtained from well-designed clinical trials. Retrospective analyses of administrative databases may indicate an association between 2 variables, but one must not infer that a causal relation exists.

Limitations in the study design complicate interpretation of the results. Equivalence trials are designed to confirm the absence of a meaningful difference between treatments where a margin of clinical equivalence is pre-specified, which was not the case here. If equivalence trials are not designed and analyzed appropriately, they often have intrinsic biases tending toward the conclusion of no difference.^2,3

Despite adjustments for many confounding variables, the study did not capture a key independent risk factor that affects baseline cardiovascular risk, namely total cholesterol or low-density lipoprotein cholesterol blood concentration.⁴ Finally, the rate of switches for non-atorvastatin users was high, resulting in “contamination” of other statin groups with atorvastatin users. This was not appropriately accounted for in the analyses. Switching may occur not only because of worsening clinical status, but also because of higher baseline cholesterol, which confers a higher cardiovascular risk.

Zhou highlights the care gap observed between 1997 and 2001: 67% of elderly subjects did not receive lipid-lowering therapy after myocardial infarction, and most of the remaining patients received low starting doses of statins. It is encouraging, however, that persistence rates were high, which is important in optimizing care after myocardial infarction.