Targeted scoring criteria reduce variance in global impressions y Steven D. Targum 1,2 * , Joan Busner 3 and Allan H. Young 4 1 Massachusetts General Hospital, Boston, Massachusetts, USA 2 BrainCells Inc, San Diego, California, USA 3 United Biosource Corporation, Wayne, Pennsylvania, USA 4 University of British Columbia, Vancouver, Canada Objective This study examined the confounding effect of treatment emergent physical or psychic symptoms on clinical global impression (CGI) ratings in CNS trials and examined the benefit of targeted scoring criteria on clarifying ratings and reducing scoring variance. Methods Twenty-four raters participating in an investigator meeting training session scored a series of scripted CGI scenarios that included treatment emergent symptoms. Results The addition of treatment emergent gastrointestinal (GI) symptoms or anxiety symptoms significantly changed the rating of clinical global improvement and caused a broad CGI-improvement (CGI-I) scoring variance reflecting scoring ambiguity amongst these raters. Re-rating after a presentation of well-defined criteria that addressed these scoring issues narrowed the variance and significantly improved inter-rater reliability. Conclusions It is clear that CNS trials must define scoring criteria for global ratings prior to the initiation of a study to assure ratings consistency. The actual definition of global must be study-specific and may depend upon the targeted symptoms of interest and mechanism of drug action. The targeted criteria that define global must be included in all published reports about the trial. Copyright # 2008 John Wiley & Sons, Ltd. key words — scoring criteria; CGI; CNS trials; depressive disorder INTRODUCTION The assessment of global improvement in CNS trials includes an evaluation of symptomatic, behavioral, and functional changes during treatment. The clinical global impressions scale for improvement (CGI-I) is often used in clinical trials as a measure of overall clinical change although its clinical validity is not well established (Guy, 1976; Haro et al., 2003; Kadouri et al., 2007; Spearing et al., 1997; Spielmans and McFall, 2006). As a sequential measure of clinical improvement or worsening from baseline, the proper- ties of the CGI-I scale presume that the clinical changes will be internally consistent in order to make a clinically valid comparison (baseline to specific interval). In reality, patients often experience treat- ment-emergent physical or psychic symptoms that were not present at baseline and may obfuscate the assessment of change. The etiology of these treatment- emergent symptoms may be related to the specific disease, or the drug treatment, or may be co-morbid symptoms that are entirely unrelated to either the disease or the tested drugs. Scoring guidelines often suggest that raters simply score what they observe and not attribute these symptoms to any specific etiology unless it is absolutely clear (e.g., flu symptoms in the presence of obvious flu) (Beneke and Rasmus, 1992; human psychopharmacology Hum. Psychopharmacol Clin Exp 2008; 23: 629–633. Published online 29 July 2008 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/hup.966 *Correspondence to: Dr S. D. Targum, 222 Berkeley Street, suite 1650, Boston, MA 02116, USA. Tel: 617-357-7474. Fax: 617-357- 7476. E-mail: sdtargum@yahoo.com y None of the authors have a proprietary interest in the CGI which is in the public domain. Dr Targum is chief medical officer at Brain- Cells Inc., sponsor of the study and has equity interests in United Biosource Corporation, a rater training company. Dr Busner is an employee of United Biosource Corporation. Dr Young has no potential conflicts of interest. The sponsor (BrainCells Inc) had no role in the design, collection, analysis, or interpretation of the data. Copyright # 2008 John Wiley & Sons, Ltd. Received 30 April 2008 Accepted 26 June 2008