Dropout rates for intent-to-treat and per protocol analyses

PURPOSE: To describe dropout rates for the intent-to-treat and per protocol analyses from prospective clinical trials. METHODS: Review of prospective multi-center parallel studies of 100 patients or more from 1996 onwards. RESULTS: We identified 33 articles (70 treatment arms) that fit the criteria for this study. No statistical differences in dropout rates were observed among drug classes for either the intent-to-treat (P =.075) or per protocol analyses (P =.40). A difference was observed in the percent dropout rate for the intent-to-treat analyses decreasing with the length of the study (P <.0001). This finding was not observed by the number of study visits (P =.44). However, a statistically greater percent dropout rate was observed for the per protocol analyses increasing with the length of the study (P =.034) and number of study visits (P =.01). No statistical differences were observed or with increasing sample size of the study for either the intent-to-treat or per protocol analyses (P >.05). CONCLUSIONS: Known discontinuation rates for per protocol and intent-to-treat analyses may help in plan, ning sample sizes for future clinical trials. (C) 2004 by Elsevier Inc. All rights reserved.