In his thoughtful editorial on the recent controversy over calcium channel blockers [1], Dr. Messerli correctly points out that the research methods used in two recent studies on this topic [2, 3] are prone to various biases. Although we agree with the general comments made in the editorial, we make three points about the case–control study by Psaty and colleagues [2].

Dr. Messerli states that this study may have been subject to selection bias and that hypertensive patients with coronary artery disease may have been more likely to be treated with calcium antagonists than with diuretics. Psaty and colleagues, however, present results that address this concern: Among persons with cardiovascular disease, 52.5% were receiving diuretics and only 37.4% were receiving calcium channel blockers. Furthermore, all persons with known cardiovascular disease were excluded from the study's principal analysis.

A second form of bias that can arise in case–control studies is recall bias, in which differential recollection of previous exposures can lead to spurious associations between disease and exposure variables. Although recall bias may have affected the recording of potential confounding variables, the exposure variable of medication use was determined from computerized pharmacy data, not from patient recall.

Finally, Psaty and colleagues did their study on a well-defined patient sample—enrollees in the Group Health Cooperative of Puget Sound. Through careful screening of computerized data, researchers could identify 95% of incident myocardial infarctions. Sampling bias was thus less likely because the study cases were representative of incident disease in the underlying patient population.

Dr. Messerli is correct in stating that case–control studies are like bouillabaisse: "One tainted fish will spoil the pot." However, clinicians should recognize and appreciate the substantial strengths of the study by Psaty and coworkers and should not dismiss its findings solely on the basis of its case–control design.