Using medication history to measure confounding by indication in assessingcalcium channel blockers and other antihypertensive therapy

Reported findings of elevated risk of adverse events associated with calciu m channel blocker use in hypertensives may be due partly to unmeasured conf ounding by indication. To determine if such confounding occurs, we conducte d a retrospective cohort analysis of 77 196 Pennsylvania Medicaid recipient s aged 18 to 61 who were treated with antihypertensive medication between 1 990 and 1992. All diagnoses and dispensed prescriptions during the year pri or to study entry were examined. Prior recipients of multitherapy (n = 18 7 63) were more likely to have had previously diagnosed risk factors (OR = 1. 31 [95% CI, 1.30-1.331) than subjects with prior monotherapy (n=11141). New initiators (n = 47292) were less likely to have had previously diagnosed r isk factors (OR = 0.48 (95% CI, 0.47-0.49)) than previous users (n = 29904) . The likelihood of being prescribed calcium channel blocker rather than ot her monotherapy was significantly higher for subjects diagnosed during the previous week with arteriosclerotic cardiovascular disease OR = 7.78 (95% C I, 2.72-22.28), P < 0.0001; angina OR = 2.92 (95% CI, 1.77-4.83), (P < 0.00 01); diabetes OR = 1.49 (95% CI, 1.07-2.06), (P = 0.0004); and hypertension OR = 1.57 (95% CI 1.35-1.82), (P < 0.0001). Calcium channel blockers were selectively prescribed to subjects at elevated risk of adverse events. Conf ounding by indication was found in this large indigent population. Unmeasur ed confounding may contribute to overestimated relative risk of adverse eve nts associated with calcium channel blocker therapy vs diuretics or beta-bl ockers. At least 1 full year of subjects' medical and prescription drug his tory should be examined prior to study entry; shorter observation periods w ill miss clinically relevant information about risk.