Onset of Improvement under Antidepressants

Conventional repeated measurements analysis of variance suggests that response to antidepressants is slow, and that differences between antidepressants and placebo do not normally reach statistical significance until the third to fourth week of treatment. This statistical "response lag" has finally led to the "delayed onset of action" hypothesis, although response-to-treatment and onset-of-action represent two principally different concepts that relate to different aspects of efficacy of antidepressant drug therapy. While efficacy in terms of response-to-treatment has its main focus on the proportion of patients in whom antidepressants induce therapeutic response, onset-of-action refers to the speed at which symptoms reduce under antidepressants in comparison to placebo. Standard drug trial procedures tend to obscure differences in onset of action across drugs (as well as across patients), because such procedures rely upon mean depression scores derived from studies which amalgamate separate depressive subgroups, rapid and slow remitters, as well as partial remitters, non-remitters and premature withdrawals.

Survival Analysis and Cure Models

In the conventional repeated mesurements approach, psychopathology scores assessed at weekly intervals are analysed as a function of time, and assessment times are represented as fixed independent variables in a (linear) regression model when testing for time differences in rates of symptom reduction. This is equivalent to a regression of the sample’s mean scores on assessement time, so that a justification of the "fast acting" claim of the newer antidepressants is, from the methodological point of view, difficult to accomplish. An alternative approach is survival analysis where time is treated as a function of symptom reduction (dependent variable). This has the advantage that variations around the prespecified, cross-sectional observation days (typically ±2 days) can be incorporated into the model, thus providing (1) a better temporal resolution with respect to the individual course of recovery from depression, (2) a way to distinguish between early and late responders, and (3) the possibility to assess the predictive value of early improvement with respect to later outcome. Additionally, the method allows one to take into account that treatment response has not been observed in a subgroup of patients due to premature withdrawal or the limited observation time.

Early Improvement

Central to all approaches to determining early onset of antidepressant action is the appropriate definition of improvement (used as indicator of a specific drug action). We distinguish between "relative improvement", defined by a percentage baseline score reduction regardless of the subsequent course of depressive illness that may include a significant deterioration; "sustained relative improvement", which requires a prespecified percentage baseline score reduction without subsequent deterioration beyond achieved improvement score; "sustained absolute improvement", which requires a prespecified percentage baseline score reduction along with the attainment of a prespecified threshold value (e.g. HAMD-17 score of 10) without subsequent deterioration.

References