Reliability of monitoring the clinical dementia rating in multicenter clinical trials

Context: The Clinical Dementia Rating (CDR) is quickly becoming a criterion standard in multicenter clinical trials in Alzheimer disease. An abbreviated version, with formal monitoring for consistency across sites and raters, is currently used in the Alzheimer's Disease Cooperative Study (ADCS). Objective: To demonstrate the degree of agreement on CDR scoring of clinical monitors working independently from ADCS-CDR worksheets. Design: Three members of the ADCS who are experienced and highly trained with respect to the CDR independently reviewed the ADCS-CDR worksheets of 15 subjects, assigning box and global CDR scores according to the prescribed algorithm. Setting: The ratings were assigned during a single, 3-hour session in a closed room. Participants: Two clinical monitors and one project director/clinical monitor supervisor. Main Outcome Measures: Percent agreement, Kendall's tau-b, and Cohen's kappa were used to assess the degree of agreement of the raters with the previously established gold standard assessment on global and box scores for the 15 subjects. Results: Raters, blinded to patient groupings, were in agreement with the Gold Standard global CDR assessment on 87% of ratings. Kappa values indicated good (K = 0.66, orientation and judgment & problem solving boxes) to excellent (K = 0.83, global CDR) agreement. Conclusions: The ADCS-CDR worksheets were reliably and consistently scored by clinical monitors, who may be considered proxy gold standards for CDR assessment.