Background: With remission now the treatment goal, antidepressant trial duration has increased. However, most patients do not remit and are exposed to prolonged, ineffective treatment.
Methods: Conditional probabilities and signal detection methodology were contrasted in early detection of nonremitters in three comparator, 12-week antidepressant trials conducted in late- or mid-life depression.
Results: The mid- and late-life samples did not differ in rates or time-to-onset of remission or accuracy of early identification of nonremission. Using conditional probabilities, there were marked differences in predictive power depending on the remission criterion. With signal detection methods, sensitivity and specificity of early identification were uniform across the differing medication conditions, remission criteria, and the three studies. By week 6, > or = 60% of ultimate nonremitters were identified, while maintaining a false positive rate < or = 20%.
Conclusions: The goals of providing maximal opportunity to achieve remission, while minimizing exposure to ineffective treatment can be satisfied by use of duration adaptive designs (DAD). While use of conditional probabilities has been the traditional method for early detection of nonremitters, this approach is inferior to use of signal detection methods. The findings also contradicted the widely held view that patients with late-life depression require longer treatment duration.