Correlation of Initial Wellness With Long-Term Morbidity in Recurrent Major Affective Disorder Patients as a Test of Survival Analysis

Abstract

Because long-term monitoring of morbidity or wellness is often impractical, survival analysis is widely used in experimental therapeutics. In long-term applications, the method assumes that time-to-recurrence or time-to-intervention is a surrogate of long-term clinical status during sustained treatment. Because this plausible assumption has not been tested empirically, we evaluated prospectively and systematically acquired data from consenting adults with major affective disorders treated with lithium for an average of 5.4 years to compare, within-subjects, estimated long-term morbidity versus the first-wellness interval after clinical remission of a treated index episode. These measures were inversely correlated in all 450 subjects with mood disorder (rs = -0.65) and in those with bipolar I (n = 259; rs = -0.58), II (n = 150; rs = -0.65), or recurrent unipolar depression (n = 50; rs = -0.66). The wellness interval also predicted episodes per year (rs = -0.58, all P < 0.0001). These associations were sustained in multiple-regression modeling, in which a bipolar diagnosis (but not sex, age, or measures of preintake morbidity) was associated with greater long-term morbidity than in major depression. These findings provide moderate support for a basic assumption underlying the use of survival analysis for psychopharmacology research and suggest that clinical outcomes might be predictable from initial wellness intervals.