General discussion 515 13 the desired (patient) outcomes [764]. This should take into account, among others, the proposed place of the experimental diagnostic in the clinical pathway. That is, if the experimental diagnostic intends to replace an existing (e.g., outdated) test, assessing diagnostic accuracy based on existing evidence may suffice and patient outcomes may be measured after implementation in clinical practice. If the experimental diagnostic is intended as a new triage or additional test in the clinical pathway, like with [18F]FDG-PET/CT in the preoperative workup of indeterminate thyroid nodules, randomized trials are considered preferable or necessary, respectively, to assess clinical utility prior to implementation in clinical practice [766]. In addition, considering that our health care system is already showing cracks under the pressure of the increasing health care volumes and staffing shortages, the added time investments, volumes, and costs of the randomized diagnostic and therapeutic interventions should be balanced against the expected merits of clinical utility data [731-733]. Finally, the feasibility of the proposed trial design should be balanced against its expected efficacy and the validity of its results. This includes methodological considerations concerning the preferred moment of randomization prior to the diagnostic intervention or prior to the subsequent treatment as well as the ethical balance between the experimental and control diagnostic interventions and treatments (i.e., patients cannot be randomized to a known inferior diagnostic technique for the sake of research) [33, 34, 765, 766]. In the following sections, I would like to discuss a number of important methodological and ethical challenges that we encountered during the EfFECTS trial and that are important to consider in diagnostic RCTs. Selection bias: a result of the test-treatment design and its complexity? Even though the RCT design is meant to offer the most generalizability and the lowest degree of bias, new forms of bias may be introduced in test-treatment trials that occur less frequently in other trial designs [35]. In the EfFECTS trial, we may have encountered ascertainment bias, a form of selection bias that may occur when some members of a target population (e.g., patients with a Bethesda III/ IV thyroid nodule) are systematically more likely to be included in a study than others. This may have had at least two underlying reasons. First of all, most eligible patients with a Bethesda III/IV nodule were only interested in trial participation because it offered them an opt-out of the diagnostic thyroid surgery that had been advised to them in accordance with the current guidelines [467]. As compared to the target population, this may have resulted in an overrepresentation of patients who preferred active surveillance over diagnostic surgery. This may have influenced HRQoL estimates, for example by a more negative HRQoL perception following diagnostic thyroid surgery if the patient had sincerely
RkJQdWJsaXNoZXIy MTk4NDMw