Govert Veldhuijzen

176 Chapter 9 informed consent. And we provided insight in managing the financial basis. Thus, we reported on all these crucial items that hampered the scalability of our platform. Important issue still partly present today is establishing a safe communication between the hospital electronic health records and the CBE platform. In place, this could eliminate the nuisance of simple data transferal work currently performed by auxiliary staff. In chapter 4 , we described the technical development of the algorithm behind the CBE platform and propose a study protocol to evaluate this new patient education system. Main strength of this article was publishing it in a video journal. This allowed us to optimally visualize the CBE platform in a stepwise fashion. The presented study protocol strength was the large sample size chosen, to support non-inferiority statements. Also, the extensive collection of data on secondary outcomes proved insightful. Limitations were that we had chosen the usual contact moments to hand out the clinically collected questionnaires by the endoscopy unit operational staff. Missing questionnaires might be the result. Nevertheless, this protocol aimed to collect all relevant information at all time points. The majority of patients would be eligible for the trial and could operate the CBE even with very basic computer skills. But in the lowest literacy category, it could be problematic to test the hypotheses. As of this, we included the possibility of face- to-face patient education in the route towards the endoscopy suite for this group. We reported the results of the trial in chapter 5 . Summarizing strengths, our trial was conducted with a large, real life sample of patients. The non-inferiority hypothesis based on both intention-to-treat en per-protocol analysis of the data and power allows robust statements on CBE efficacy. As we tested the CBE in a real-world setting using three different types of endoscopy units, the results are well generalizable to daily practice. On the other hand, this trial comes with limitations. There was a significant number of dropouts after randomisation due to inclusion failures. However, this did not result in an unequal distribution regarding baseline characteristics among the arms in both the ITT and PP population, limiting the risk of selection bias. Due to the use of patient reported questionnaires we do not have 100% data collection at all time points, although the trial protocol called for that. While this did not affect our main outcome bowel preparation (100% data collected), the lower response rate might have affected assessment of secondary outcomes such as anxiety and satisfaction. We did not collect complete medical histories of our patients, including risk factors for poor bowel preparation such as diabetes mellitus, constipation, or use of motility influencing drugs. We surmise that the effect of these risk factors on the bowel preparation efficacy in our trial is limited in view of the small difference in BBPS scores. We did not collect data on adenoma detection rate (ADR) as this was outside the remit of this clinical trial. From literature, the robust correlation between adequate BBPS and ADR suggest that BBPS is a good technical proxy parameter. 8