S246. THE EFFECT OF INCORRECT SCALE ADMINISTRATION ON DATA QUALITY IN SCHIZOPHRENIA CLINICAL TRIALS

BACKGROUND: The availability of date and time stamps of interview start times on eCOA systems permits monitoring and documentation of the order of administration of scales and instruments at each visit. The Clinical Global Impression Scale (CGI) is a holistic instrument that synthesizes all information available from the subject, caregivers, medical notes, etc. and is therefore typically required to be rated last at a particular visit. In the current retrospective analysis of double blind eCOA collected schizophrenia data pooled from multiple clinical trials we assessed the percent of visits where CGI was not administered last among other efficacy assessments. Additionally we assessed whether the inappropriate administration order of Clinical Global Impression Scale was associated with discrepancies in the data and change from baseline between the CGI and the primary efficacy outcome. METHODS: Available eCOA data were pooled from schizophrenia double blind, placebo controlled clinical trials. Within the data, we identified those visits where the CGI was not administered last among other efficacy assessments (inappropriate administration order). Within each trial we identified as discordant those data where the actual primary efficacy score or its change from baseline differed by at least two standard deviations from the expected score after linear regression. For this procedure the CGI- S score used as a predictor and the primary efficacy outcome as dependent variable. Logistic regression was then used on pooled data to explore whether the incorrect order of CGI administration increased the odds of discordant ratings between the CGI and primary outcome measure. RESULTS: The dataset consisted of 5,784 paired ratings of the CGI and the primary efficacy outcome. Among these, a total of 4,628 visits allowed to calculate change from baseline. Inappropriate order of CGI administration was identified in a total of 443 visits (7.7% of all visits). Discrepancies between CGI-S and the primary efficacy outcome were identified in 292 visits (5.1% of data) and discrepancies between change from baseline in CGI-S and in the primary efficacy outcome were observed in 249 cases (5.3% of data). The presence of incorrectly administered CGI increased the odds of raw score discrepancy 1.6x (95%CI 1.1–2.4), and the odds of discrepancy in change from baseline 1.8x (95%CI 1.2–2.7), both significant with p <0.01. DISCUSSION: Our data indicate a relatively large percent of visits suffer from an incorrect scale administration order. We have previously explored a number of sources of possible noise in schizophrenia clinical trials. Current analyses identified a significant effect of incorrect scale administration on the presence of between scale discordances. Such findings indicate that order of CGI administration should be mandated in schizophrenia clinical trials and enforced by eCOA platforms. Additionally, violations to correct scale administration should be monitored through analytic programs and acted upon in case of repeated occurrences at the rater or site level.