Trial adaptive feature or adaptation, motivation, and cited examples of use | Type of adaptive design (AD) and examples of underlying statistical methods |
---|---|
Changing the predetermined sample size in response to inaccurate assumptions of study design parameters to achieve the desired statistical power [34,35,36]. | Sample size re-estimation, re-assessment, or re-calculation (SSR) using aggregated interim data from all participants or interim data separated according to allocated treatment [37,38,39,40,41,42,43,44]. |
Stopping the trial early for efficacy, futility, or safety when there is sufficient evidence [45, 46]. | Group sequential design (GSD) [47, 48]; information-based GSD [49]; futility assessment using stochastic curtailment [50,51,52]. |
Evaluating multiple treatments in one trial allowing for early selection of promising treatments or dropping futile or unsafe treatments [53,54,55]. New treatments can also be added to an ongoing trial [56]. | Multi-arm multi-stage (MAMS), dose/treatment-selection, drop-the-loser, or pick-the-winner, or add arm [23, 57,58,59,60,61,62,63,64,65,66]. |
Changing the treatment allocation ratio to favour treatments indicating beneficial effects [67, 68]. | |
Investigating multiple research objectives that are traditionally examined in distinct trial phases, in one trial under a single protocol [74,75,76]. For instance, addressing learning (selecting promising treatments for further testing) and confirmatory objectives in one trial. | Operationally or inferentially seamless AD [63,64,65, 77,78,79]. |
Adjusting the trial population or selecting patients with certain characteristics that are most likely to benefit from investigative treatments [80,81,82,83]. This may involve incorporating statistical information from or adapting on a biomarker. | Population or patient enrichment or biomarker AD [84,85,86,87,88]. |
Changing the primary research hypotheses or objectives or primary endpoints [78, 89]. For example, switching from non-inferiority to superiority. | |
Switching the allocated treatment of patients to an alternative treatment influenced by ethical considerations, for instance, due to lack of benefit or safety issues. | |
Combination of at least two types of adaptations [24, 36, 89, 93,94,95,96,97,98]. | Multiple ADs such as GSD or drop-the-loser with SSR [99]; inferentially seamless phase 2/3 AD with hypotheses selection [77] or population enrichment [100]; biomarker-stratified with RAR [101]; adaptive platform trials where arms can be added or stopped early [19, 24, 102]. |