The study by Psaty and colleagues [3] is a retrospective, population-based, case–control study that was designed to assess the risk for acute myocardial infarction associated with different classes of antihypertensive drugs. Patients whose hypertension was treated with a calcium antagonist had an almost twofold higher risk for acute myocardial infarction compared with those who received treatment with a diuretic alone. Two possible sources of bias must be taken into account in any case–control study. First, a "selection bias" could have resulted from the fact that calcium antagonists were used for hypertensive patients whose concomitant disorders also required a calcium antagonist. Predictably, a hypertensive patient with coronary artery disease is more likely to be given treatment with a calcium antagonist than with a diuretic. As a consequence (and despite the attempt to match the groups), more hypertensive patients with coronary artery disease may have been in the calcium antagonist group (and died of acute myocardial infarction) than in the diuretic group. Second, the possibility exists that an undocumented "confounding bias" may have invalidated the case–control study. For instance, left ventricular dysfunction or ventricular arrhythmias in the patient who has had a myocardial infarction could be such a confounder for which the study was not controlled. Also of note in the study by Psaty and colleagues [3], predominantly short-acting, older formulations of calcium antagonists were used—of which two of three were not even indicated for the treatment of hypertension. Moreover, these agents require a three times daily or four times daily dosage regimen, which is unlikely to be followed by patients who have an asymptomatic lifelong disorder such as hypertension. Most importantly, calcium antagonists are a heterogeneous group of molecules, and not all of them are created equal with regard to reinfarction rates in the patients who have had myocardial infarctions [9].

The meta-analysis of Furberg and colleagues [4] has several substantial drawbacks: First, a meta-analysis that included 15 of the 16 studies was published previously by one of the authors [10]. The only difference between the recent analysis and the previous one is the attempt to show a dose-related increase in mortality with nifedipine in higher doses by adding one more study (INTACT [International Nifedipine Trial on Antiatherosclerotic Therapy]) [11]; indeed, the major case against nifedipine hinges on this single study. Including it in the meta-analysis must be considered inappropriate [6] because all 15 of the other studies dealt either with the early phase of acute myocardial infarction or with unstable angina, whereas, in the INTACT study, only 12% of patients had total coronary occlusions and one third had a previous myocardial infarction by electrocardiographic criteria. A meta-analysis "should be restricted to a trial study population with similar risks" [12], and, hence, patients with stable coronary artery disease should not be included in a population of patients who have had myocardial infarctions or have unstable angina. Second, studies used in the meta-analysis seemed to have been selected at random; no inclusion criteria were reported and some publications were omitted. Third, in some of the 16 studies analyzed, there were several misquotations or errors regarding the number of patients and dose of calcium antagonists. When the data are recalculated from the original publications, the overall significance becomes borderline (P < 0.07). Fourth, interestingly enough, in a previous paper coauthored by Furberg [13], the authors requested that because of the problem of statistical multiplicity, the "results from overviews should only be accepted if the levels of statistical significance are extreme (e.g., P < 0.001)." The present meta-analysis falls short of the authors' own ambitious goal by a factor of at least tenfold: The best P value, as listed, is 0.01. This would suggest that meta-analyses and case–control studies are much like a bouillabaisse: No matter how much fresh seafood is added, one tainted fish will spoil the pot!

Of note, both of these studies only indict older short-acting calcium antagonists, and at least some preliminary evidence attests to the safety of the more modern longer-acting preparations in treating a variety of cardiovascular disorders (Table 1).

Despite the flaws of both of these studies, we must remember that blood pressure is a surrogate end point that often, but not always, correlates with cardiovascular morbidity and mortality. Calcium antagonists do lower blood pressure, but so do congestive heart failure, anemia, shock, and even leeches! As was recently shown in the Medical Research Council study in the elderly [14], the fact that a drug reduces arterial pressure (or provides relief in a patient with angina) does not necessarily mean that it will reduce morbidity or prolong life. This notwithstanding—perhaps a devil's advocate point of view—it is clear that what we urgently need are the results from thorough randomized, controlled clinical trials based on sound pathophysiologic reasoning rather than incorrect meta-analyses or retrospective case–control studies.