Should early improvement be re-defined to better predict the maintenance of response in first-episode schizophrenia patients?

OBJECTIVE: To evaluate the predictive validity of early response in first-episode schizophrenia within a 1-year follow-up trial and to compare the resulting cutoff to the currently proposed early response definition (20% improvement by week 2). METHOD: Receiver operator characteristic (ROC) analyses were used to identify the predictive validity of the psychopathological improvement of treatment from week 1 to week 8, regarding the maintenance of response until week 52 as well as to define the most reasonable cutoff in 132 first-episode patients. The Youden Index (maximum of sensitivity and specificity) was used to compare the newly developed and the commonly used early response definition. RESULTS: Starting with week 6, a reasonable validity to predict the maintenance of response was found (area under the curve = 0.721) with the best fitting cutoff being a 51.6% PANSS total score improvement. Using this cutoff 74 patients (56%) were correctly identified to become responder and maintain response during follow-up (sensitivity: 0.747). The Youden Index was higher applying the newly developed early response cutoff featuring higher specificity compared to the commonly used early response definition. CONCLUSION: Regarding long-term treatment, it seems more appropriate to base predictions of the patient's maintenance of response not before 6 weeks of treatment.