To test in first-episode schizophrenia (FES) patients whether early treatment improvement accurately predicts subsequent treatment outcome and to develop a valid definition for early improvement.

188 first episode schizophrenia inpatients were assessed with the Positive and Negative Syndrome Scale (PANSS) at admission and biweekly until discharge within a randomized, double-blind trial. 93 patients were treated with risperidone, 95 with haloperidol. Response was defined as a 40% improvement in PANSS total score from admission to discharge, remission according to the recently proposed consensus criteria and early improvement as an improvement in the PANSS total subscore after the second week. Receiver operator characteristic (ROC) analyses were used to evaluate predictive validity of early improvement and to find the most reasonable cutoff definition according to specificity and sensitivity. A confidence interval for the optimum cutoff was furthermore estimated.

ROC analyses revealed reasonable predictability for response (AUC=0.707) and remission (AUC=0.692) concerning early improvement as a predictor for response/remission. No significant differences were found between the two treatment groups. Early improvement defined as a 46% PANSS total score reduction for response and a PANSS total score reduction of 50% for remission obtained the best results in terms of sensitivity and specificity.

We were able to confirm that early improvement is a valid predictor of subsequent treatment response and remission. FES patients should at least improve up to 30% in the PANSS total score at week 2 to achieve treatment response and remission.