Obtaining early markers of predictive response in clinical practice has become a need in the management of depression.

To detect early (at week 2) clinical markers predictive of clinical response after 6 weeks of treatment with agomelatine in an open, naturalistic study.

2,780 outpatients were treated by Agomelatine during 6 weeks for MDD. They were assessed at baseline, week 2 and week 6 for the QIDS, a VAS assessing present level of mood, sleep complains (LSEQ), social functioning (SDS), and emotional state (MAThyS). ROC curves were used to assess the best balance between sensitivity and specificity for each parameter.

The patients were 47 years old in average (+/- 12.4), with 67% of females, with, in average, a 10-year course and 2 previous episodes.

The ROC curves showed that 20% improvement of the QIDS at week-2 had the largest area under the curve (AUC = .687). the CGI-EI (AUC = .671), and a simple VAS assessing the level of mood at week-2 (AUC = .659) had results in the same range. Treatment response, improvement of social functioning (SDS) or a decrease of sleep complains (LSEQ) were not significant, confirming that early improvement of sleep is not predictive of later response.

Early improvement at week 2 with agomelatine may be used to detect later responders. This improvement can be perceived both by clinicians (QIDS or CGI-EI) or patients (level of mood according to a simple VAS). to our knowledge, this positive predictivity has not been previously observed in a naturalistic study on MDD.