SynopsisThis study compared serial severity ratings from 2 methodologically similar tricyclic antidepressant studies. Though the groups resembled each other in terms of diagnosis and baseline severity, response patterns were surprisingly different. In particular, early change was highly predictive of outcome in one group but not in the other. The results of response predictor studies must, therefore, be generalized with caution. These data, nevertheless, do allow some guidelines for predicting response from early changes and severity.