It has been widely stated that the available research data has not demonstrated a SSRI dose response for major depression. We re-evaluated the methods used to analyze the SSRI data by clarifying two key alternative definitions of dose response and their implications for enhancing analysis of currently available data as well as future study design. We differentiated "potential" dose response, which focuses exclusively on response excluding tolerability effects and asks whether differences in dose can result in significant differences in response, from "expressed" dose response, which incorporates all tolerability effects currently associated with dose (including those caused by study protocol or treatment practice) and asks whether differences in dose do result in significant differences in response. To analyze potential dose response for all studies, one should use a "dose-tolerant" sample, i.e., an ITT sample from which dropouts due to adverse events have been removed. To analyze an expressed dose response, an ITT sample is the optimum sample if the study conforms to several design specifications. In the absence of conformance to these specifications, an ITT sample may be an approximation of the appropriate sample. Given design limitations of currently available studies, a dose-tolerant sample may provide a more informative approximation of an optimal sample to be used in evaluating the expressed dose response that could be expected in the best clinical practice. Future studies of dose-response relations could be enhanced by taking into account the principles noted above, and currently available data should be reanalyzed based on these principles. This re-analysis is performed in a companion article [Baker et al. 2003, Depress Anxiety 17:1-9].
Copyright 2003 Wiley-Liss, Inc.