Objectives: There is still debate about whether the quality of long-term efficacy of lithium in patients with bipolar disorders is influenced by atypical features. Extended Cox regression models allow for the use of all follow-up data on diseases with multiple episodes. The aim of the present analysis was to apply the best suited of these models to analyze the influence of atypical features on the widely used outcome measure of time to recurrence in a large multicenter cohort of lithium responders established by the International Group for the Study of Lithium Treated Patients.
Methods: A conditional extended Cox model with a random frailty term was applied to the data of 336 bipolar I and II disorder patients, all of whom were responders to lithium with treatment for up to 30 years.
Results: Differences were found in the long-term outcome, even in patients who have demonstrated a relatively good response to lithium treatment. The hazard for recurrence was negatively influenced by the presence of atypical features, mainly mood-incongruent psychotic symptoms, interepisodic residual symptomatology, and rapid cycling.
Conclusions: As a result of the findings, physicians should regularly reassess the quality of response in bipolar disorder patients with atypical features and, if necessary, modify treatment. Extended Cox regression models are well suited for evaluating long-term outcome and should be used more extensively to analyze treatment outcome in psychiatric and somatic disorders.