Predictors of medical events and of their competitive interactions in the Cardiac Insufficiency Bisoprolol Study 2 (CIBIS-2)

Aims Predictive factors for medical events and interactions between events were not reported in the Cardiac Insufficiency Bisoprolol Study 2 (CIBIS-2) . We examined the interactions among death, permanent treatment withdrawals , nonlethal cardiovascular hospitalizations and their own occurrence in a g iven patient, the treatment received, and baseline variables during CIBIS-2 . Methods and Results A Cox model for censored data was used to analyze the r elations among baseline variables, medical events, and their interactions w ith treatment. We used competitive risk analysis to examine the interaction s between successive events in a given patient during follow-up. No baselin e variable predicted the reduction of mortality with bisoprolol. Withdrawal from bisoprolol therapy was more frequent in patients whose baseline heart rate was in the lower tertile of the distribution (P = .0002) but otherwis e was not different between patients randomized to bisoprolol and to placeb o. Event history analysis revealed that bisoprolol reduced mortality (P = . 0006) and hospitalizations for nonlethal cardiovascular events (P = .003) i n patients in whom treatment was not permanently withdrawn. Analysis of sur vival curves in patients who permanently discontinued treatment showed that bisoprolol did not reduce mortality compared with placebo in this populati on (relative risk 1.03, 95% Cl 0.67-1.59; P = .88). Recurrent nonlethal eve nts were reduced by bisoprolol. Conclusion In CIBIS-2, medical events were significantly influenced by trea tment withdrawal. Patients who derive benefit from bisoprolol therapy are t hose in whom treatment is not permanently withdrawn.