Patterns of symptom change in major depression: Classification and clustering of long term courses

To evaluate treatment effects in depression, it is important to monitor change during treatment and also to follow up for a reasonably long time. Describing the variability of symptom change trajectories is useful to better predict long-term status and to improve interventions. Outcome data (N__complete = 518, 4 time points, 1 year of observation time) from a large naturalistic multi-center study on the effects of inpatient and day hospital treatment of unipolar depression were used to identify clusters of symptom trajectories. Common outcome classifications and statistical methods of longitudinal cluster analysis were applied. However, common outcome classifications (in terms of e.g. remission, relapse or recurrence) were not exhaustive, as 49.3% of the trajectories could not be allocated to its classes. Longitudinal cluster analysis reveals 7 clusters (fast response, slow response, retarded response, temporary or persistent relapse, recurrence, and nonresponse). Nonresponse at the end of treatment was a predictor of poor outcome at long term follow up. The classification of patterns of symptom change in depression should be extended. Longitudinal cluster analysis seems a valid option to analyze outcome trajectories over time if a limited number of time points of measurement are available.