Study designs to compare new colonoscopic techniques: clinical considerations, data analysis, and sample size calculations

Background and study aims: Novel imaging techniques need to be evaluated for their ability to improve the detection of polyps. Critical appraisal of reported studies reveals remarkable differences in study designs, despite their similar objectives. The aim of the current study was to compare frequently used study designs for their required sample size to detect relevant differences in polyp detection rates. Patients and methods: Three commonly reported study designs were compared: design 1, the parallel randomized design; design 2, the randomized cross-over design with direct removal of polyps; and design 3, the randomized cross-over design without direct removal of polyps. A total of five different scenarios were analyzed per study design, representing a variety of clinical settings. Each scenario was repeated 1000 times for each study design and the sample size that produced a significant result per study design in each scenario was recorded. Results: In many scenarios, study design 1 required 10 - 15 times more patients to reach the same statistical power of 80 % compared with study design 2. Further reductions in sample size could be achieved when study design 3 was used, although this design is limited by its impracticality. Conclusions: The randomized parallel design requires a larger sample size to achieve equivalent power to the cross-over design. Researchers should carefully consider whether the latter design could be used instead, as it requires a lower number of subjects, although it may be more cumbersome for patients, endoscopists, and researchers.