[Metaanalysis to estimate the expected drop out-rates reported in clinical trials on cataract surgery]

Abstract

Background: A realistic sample size calculation is crucial to achieve significant results in clinical trials. As an expected drop out-rate has to be included in the sample size calculation, current practice consists in the presumption of drop out-rates published in previous similar investigations. This approach may, however, result in severely over- or under-estimated sample sizes. Therefore this meta-analysis sought to aggregate the drop out-rates from published clinical trial reports on cataract surgery to derive a quantitative suggestion for the planning of future clinical trials.

Methods: The data collection was a complete review of all prospective and retrospective studies in five journals of the years 2002-2012; trial-wise recall rates of subjects at follow-up 3, 6, 12, and 24 months after recruitment were documented. The primary endpoint of the meta-analysis was the reported drop out-rates after 6 months. 95 % confidence intervals were calculated for each trial, respectively; a median drop out-rate was estimated including its 95 % confidence interval. The drop-out-rate estimates were furthermore stratified by design characteristics of the reported studies.

Results: For randomised clinical trials on cataract surgery, the median drop out-rate increased during the follow-up period of 24 months from 4 % at three months to 17 % at 24 months after recruitment; for the six-month drop out-rate a median drop-out rate of 3 % (95 % CI 0 %; 14 %) was estimated.

Conclusion: Drop out-rates in sample size calculations for clinical trials on cataract surgery were found to be over-estimated in general, ending up in the calculation of overly large patient numbers and thereby in both ethical and economic consequences. For randomised clinical trials on cataract surgery the median drop out-rate can be expected to be 5 % during a six-month follow-up and may rise up to 15 % during a 12-month trial period.