Assessing treatment effects in clinical trials with the discan metric of the Sheehan Disability Scale.

The Sheehan Disability Scale (SDS) is a patient-rated, discretized analog measure of functional disability in work, social, and family life. Its increasing use in clinical trials in psychiatry suggests a need to assess its responsiveness and interpretability. In this paper we identify and review studies in which the SDS was used as a treatment outcome measure. Our objectives are (i) to evaluate the sensitivity of the SDS to treatment effects and (ii) to examine potential thresholds or cutoff scores for remission and response. Studies for the review were retrieved from the National Library of Medicine's PubMed database (1966 to 21 March 2007) and other sources. All studies had to use the SDS, be double-blind, controlled or large open-label trials in English. Studies assessing nonpharmacological treatments, long-term trials (>12 weeks), small n trials (less than 20 patients per treatment arm) and trials for conditions other than one of the anxiety disorders, depression, or premenstrual dysphoric disorder were excluded. Extracted data included the diagnostic target of treatment, n, study design, and method of analysis. Initial, endpoint and/or mean change scores were extracted from tables, text, or extrapolated from figures. In all, 37 studies meeting the inclusion criteria were retrieved and reviewed. All of the studies treated the SDS as a numeric scale and analyzed mean change or endpoint differences with parametric statistics. Three provided additional outcome data using nonparametric response or remission criteria. Overall, the SDS performed well in discriminating between active and inactive treatments. The results indicate that the SDS is sensitive to treatment effects. To establish reliable and valid cutoff scores for remission and response, there is a need to supplement parametric analyses using mean change and endpoint differences with nonparametric analyses showing the percentage meeting specified response and remission criteria. In addition, the percentages with endpoint scores of zero should be reported.