Ventricular arrhythmias and nonsedating antihistamines

De Abajo & Rodriguez carried out a cohort study with a nested-case control analysis to assess and quantify the risk of ventricular dysrhythmia associated with the use of five nonsedating antihistamine drugs [1]. Two observations from this study merit further discussion. First, although the study reports a low incidence rate of such events (1.9/10 000 patient/years; 95%CI: 1.0–3.6) after nonsedating antihistamine use, this rate is surprisingly four times higher than the rate during nonuse. Second, the relative risk (RR) observed for terfenadine (RR = 2.0) is remarkably low when compared with astemizole (RR = 17.8) or cetirizine (RR = 7.1). This is somewhat unexpected since previous large-scale epidemiological studies found that terfenadine users had risks similar to that of users of sedating antihistamines and ibuprofen [2, 3], while an analysis of WHO spontaneous adverse drug reports suggested that terfenadine may carry similar or larger risks of serious ventricular dysrhythmia than other antihistamines [4]. These unusual findings may be related to unidentified methodological limitations of the study design that could have biased the estimates. First, depletion of susceptibles may have occurred because the reference group of ‘non use’ was formed with the time period following the use of nonsedating antihistamine drugs. Thus, because only the first ventricular dysrhythmia event was considered, subjects for whom this cardiac event occurred during the initial drug exposure were ineligible for inclusion in the reference ‘nonuse’ group, which was thereby depleted of these possibly high-risk subjects. Consequently, if some subjects were more susceptible to develop ventricular dysrhythmias, the study design forced them into the exposed group. This will necessarily underestimate the incidence rate in the nonuse reference group, thereby artificially increasing the overall RR and the RR for individual nonsedating antihistamines. Second, since terfenadine prescriptions decreased by 52% during the follow-up period (Jan 92-Sept 96), major changes in prescribing habits were taking place. This figure suggests that a large number of subjects initially started on terfenadine were switched to other agents sometime during the follow-up period. If the reason for switching was related to the risk of ventricular dysrhythmias, with switchers being more at risk, disproportionately higher relative risks will be observed with other agents. Finally, it was noted that total prescriptions of nonsedating antihistamines decreased roughly from 10 000 to 7000 per month. Thus, we can deduce that the use of nonprescription antihistamine drugs during the follow-up period, which was not considered in the study, may have been increasing to compensate the decrease in prescriptions. Consequently, the rate of ventricular dysrhythmia in the reference ‘nonuse’ group will likely increase over the span of the study. Whether this phenomenon affects the rate ratios in any way was not assessed. Based on these limitations in study design and data analysis, the relative risks reported (1) may be biased. Any future epidemiological study conducted to confirm these findings should: (1) include a concurrent reference group composed of genuine nonusers of nonsedating antihistamines, (2) assess whether depletion of susceptibles was present, (3) consider changes in prescription patterns over time in the data analysis and (4) assess the effect of calendar time when estimating the relative risks to account, at least in part, for time trends in the unavailable nonprescription exposures. Received 30 September 1999, accepted 23 December 1999.