CCBs and cancer.

Abstract

Epidemiologic drug safety studies require, at a minimum, information on the study of drug exposures that is adequately complete, accurate, and relevant to a reasonable hypothesis [1]. Isolated data on exposure to study drugs many years prior to the outcome of interest is not suf~cient to establish a valid basis for estimating the relation of drugs such as calcium channel blockers (CCBs) to an illness such as cancer. A credibly interpretable study of the relation between drugs and cancer requires that a thorough, continuous history of drug exposure be available for many years prior to diagnosis of the disease. The initial reports purporting to show an increased risk of cancer in CCB users (relative risk [RR] 1.72) by Pahor et al. [2,3] were based on a small follow-up study of elderly subjects. Only a single recording of exposure to study drugs in 1982 some 6–10 years prior to the development of cancer was used to de~ne exposure to CCBs and other study drugs. No information on duration of use after 1982 was provided, nor was there critical information on changes in the use of antihypertensive drugs (AHDs) subsequent to 1982. Because it is well documented that there are large changes in AHD use over time [4,5], misclassi~cation of exposure in the Pahor study is doubtless very large. Such misclassi~cation of exposure inevitably leads to uninterpretable results [1]. More recently, a small follow-up study by Fitzpatrick et al. [6] was interpreted to show a positive association between CCBs and breast cancer (RR 5 2.57). Once again, the information on drug exposure was incomplete and arbitrarily de~ned. A study by Hardell et al. [7] was interpreted as showing a positive association between CCBs (and eight other drugs) and colon cancer. However, information on drug exposure was obtained years after the cancer was diagnosed, and such information is highly subject to recall bias and error. A study by Jonas et al. [8] reported no association between CCBs and cancer, but, again, information on drug exposure was incomplete [8]. A follow-up study by Olsen et al. [9] based on indirect registry information on CCB use and cancer concluded that there was no association between CCBs and cancer. In addition to these ~ve publications, there have been two additional reported studies on this issue. Both were large, and relevant well-documented information on exposure to study drugs was available. These included a history of drug exposure just prior to the diagnosis of cancer as well as information on duration of use of study drugs. The case-control study of Jick et al. [4], based on 446 cases of cancer and 1750 controls who used antihypertensive drugs (AHDs), yielded an adjusted relative risk (RR) estimate for CCBs of 1.27 (95% con~dence interval [CI] 0.98–1.63) compared with users of beta-blockers. Of critical importance, there was no increase in risk with increasing duration of CCB use for periods up to 4 years or longer. The case-control study of Rosenberg et al. [10] was based on 9513 subjects with cancer (in both users and nonusers of AHDs) and 6492 controls. The RR estimates for CCB users, betablocker users and users of ACE inhibitors were 1.1 (95% CI 0.9, 1.3), 1.1 (95% CI 1.0, 1.3), and 1.1 (95% CI 0.9, 1.3) respectively, compared with nonusers of AHDs. Again, there was no duration effect. Both studies [4,9] provide convincing evidence that there is no causal association between CCBs and cancer. In view of the serious methodologic de~ciencies of the studies by Pahor et al. [2,3], Fitzpatrick et al. [6], and Hardell et al. [7], and the size and methodologic strengths of the studies by Jick et al. [4] and Rosenberg et al. [10], it is now possible to conclude with con~dence that CCBs do not cause cancer. Full appreciation of proper epidemiologic principles and methods on the part of those who perform and interpret drug safety studies is essential to avoid unnecessary confusion and controversy.