Use of survival curves in analysis of antipsychotic relapse studies.