Does clinical judgment of baseline severity and changes in psychopathology depend on the patient population? Results of a CGI and PANSS linking analysis in a naturalistic study.

Background: Linking of the Clinical Global Impression (CGI) Scale and the Positive and Negative Syndrome Scale (PANSS) was performed within a naturalistic sample. Furthermore, these linking results were compared with those derived from randomized controlled trials to examine if the baseline severity might influence the linking results.

Methods: Biweekly PANSS and CGI ratings were performed from admission to discharge in 398 schizophrenia patients treated within a naturalistic study. Equipercentile linking was performed using the statistical program, R 2.8.1. To evaluate how the naturalistic study design would influence linkage results, a so-called study sample was computed with patients of the naturalistic study fulfilling common inclusion criteria of randomized controlled trials (n = 199). Patients not fulfilling these criteria (less ill sample) and those fulfilling the criteria (study sample) were compared using confidence intervals.

Results: We found a considerable difference between the linking of the CGI severity score and the PANSS total score comparing the less ill sample and the study sample. Being considered "mildly ill" at admission in the less ill sample corresponded to a PANSS total score of 47 points and to a PANSS total score of 67 points in the study sample. Considering the linking of the CGI improvement score and PANSS changes, similar results were found for CGI improvement ratings ranging from "very much improved" to "minimally improved".

Conclusions: Despite considerable differences, a 50% PANSS reduction was found to correspond to a clinical rating of much improved, which seems to be a suitable definition for response in clinical drug trials.