The Accurate Sample Size: A Rather Daunting Task When Evaluating New Endoscopes.

Abstract

To the Editor: Núñez-Rodríguez et al1 who randomized 249 fecal occult blood test–positive patients to undergo either full-spectrum endoscopy or standard forward-viewing colonoscopy reported equivalent adenoma detection rate (ADR) between the 2 modalities. These results should be seen critically, as a fundamental caveat in the study`s design might have undermined detection of a potential difference: the inaccurate sample size that also challenged the results of a similar study from Italy.2 In order to measure ADR differences between the 2 groups in a high risk for colorectal cancer population, the investigators calculated the sample size of their parallel group design study on the basis of values extracted from a randomized crossover design trial, devised to address adenoma miss rate (AMR) between the 2 endoscopes,3 in a mixed symptomatic and screening/surveillance population: different study designs, end points, and populations. Correct sample size calculation would have been also facilitated if the historical ADR of the participating endoscopists in a similar fecal occult blood test–positive population was provided. We query whether detection of a difference of 20% would be realistic if it was known that participating endoscopists were able to achieve an ADR of > 68% using the standard technology in this high-risk population. Moreover, it is not reasonable to assume that the true prevalence of adenomas is equally distributed between the 2 arms in this small sample study. Any noted differences in ADR may not be attributable to a specific device being better than the other but to the true adenoma prevalence in the arms. Therefore, it is recommended to recruit some thousands of participants to accurately detect ADR differences in a 2-arm randomized parallel group design study.4 Indeed, we have calculated that, if investigators had recruited 1026 patients per arm, the detected difference in ADR would have reached the level of significance, with α= 5% and P= 80%. Until fully powered, parallel group design, multicenter, randomized trial results are available, conclusions with regard to the efficacy of the different endoscopic modalities should be drawn mainly from studies with substantially fewer design limitations and adequate statistical power.3,5 Georgios Tziatzios, MD Paraskevas Gkolfakis, MD Konstantinos Triantafyllou, MD, PhD Hepatogastroenterology Unit, Second Department of Internal Medicine— Propaedeutic, Research Institute and Diabetes Center, Medical School, National and Kapodistrian University of Athens ‘‘Attikon” University General Hospital Athens, Greece