Lack of standardization across sites and raters, poor interrater reliability, and possible scoring bias affecting the primary outcome measure contribute to a high failure rate in anxiety trials. Remote centralized raters who are blinded to protocol inclusion and exclusion criteria as well as visit number may standardize assessments across raters and eliminate scoring bias, decreasing placebo response and thereby increasing signal detection. The purpose of the primary study was to test the safety and efficacy of an anxiolytic in a double-blind, placebo-controlled (no active comparator), multicenter trial. However, there was an additional prospective objective to explore site ratings compared with remote centralized ratings in the cohort of subjects on placebo. Site raters assessed subjects 6 times over an 8-week period. The primary outcome measure was the week 8 site-rated Hamilton Anxiety Scale (HAM-A). Remote centralized raters by telephone independently rated these subjects on the HAM-A at baseline and week 6. Of the 122 subjects selected by site raters and therefore randomized, remote centralized raters would have admitted 59 (48%) and excluded 63 (52%), based on their HAM-A ratings. The mean change from baseline in HAM-A total score in the placebo group admitted to the study by site raters was 9.3, significantly higher than the 5.9 point mean change on placebo as measured by the remote centralized raters.
The data are consistent with the potential for qualification bias at baseline when rated by sites. The results make a strong case for using strategies to ensure that baseline scoring is truly independent of the pressure to enroll.