CAUSAL ESTIMATION OF TIME-VARYING TREATMENT EFFECTS IN OBSERVATIONAL STUDIES - APPLICATION TO DEPRESSIVE DISORDER

Clinicians recognize three phases of the treatment of major depression : an acute phase to control disabling symptoms, a continuation phase t o avoid relapses of a single episode, and a preventive phase to avoid recurrences of new episodes over time. With no directly measurable tra ce of the underlying pathological process, the distinction is based ar bitrarily on the passage of time in remission. The clinician who has s uccessfully treated a patient with antidepressant medications in the a cute phase has a critical clinical decision to make for the continuati on and preventive phases: whether to continue to prescribe the medicat ion, for how long, and at what dose. This decision, like most clinical decisions in psychiatry, is not yet completely determined by the resu lts of randomized clinical trials. Only a handful of such trials have been completed, covering just a fraction of the possible maintenance s trategies (defined by treatment drop times). For many reasons, observa tional studies of the outcome of naturally occurring treatment choices play an important supporting role, helping to extend the reach of com pleted studies and to design new studies. Causal inference from observ ational studies has usually been considered in the context of a decisi on among a few fixed alternatives at a single time. The particular cau sal effect of interest in the maintenance of remission dictates that t reatment be studied over remission time. This challenges the causal an alysis of the observational study. We present issues arising from asse ssing temporal treatment effects due to nonrandomized treatment assign ment over time. We use data from a large observational study of the co urse of affective illness, to illustrate an approach to this problem.