Adaptation of randomised controlled designs leading to good enough certainty of results that can accelerate treatment access to patients
| Type of design | Brief description and usage | Some methodological points of attention for HAS |
| Adaptive designs in clinical trials | Their design can be modified in several aspects prospectively (eg, sample size, stopping rules due to effectiveness, futility or safety, enrichment of subgroups that are the most susceptible to exhibit effectiveness), which can decrease the number of patients to be included compared with what was initially expected. | They require the frequent review of accrued data at multiple time points to decide or not for modifications, which may reveal information about effectiveness to stakeholders. Defining adequately the role, process and interactions of the unblind independent data monitoring committee and blind trial steering committee is therefore crucial. |
| Platform trials | A subset of adaptive designs allowing the comparison either simultaneously and/or sequentially of multiple interventions to a common or multiple control groups, with possible addition and dropping of interventions, and adaptation of the control to new standard of care. | The design of the trial must ensure the control group is contemporaneous of the intervention(s) of interest(s). Blinding of patients and investigators can be difficult to achieve as it can require the use of multiple dummies. |
| Seamless trials | A subset of adaptive designs where there is no interruption between the conduct of different phases such as between phase 2 and phase 3, allowing reuse of data from patients from phase 2 for which the investigated treatment appeared to be beneficial. | It could be argued a seamless phase 2/3 trial does not allow the phase 3 to be an independent confirmation of the results of suggested effectiveness of the phase 2. |
| Pragmatic trials | An umbrella term that can define trials which try to enhance enrolment and applicability by defining less stringent eligibility criteria that ‘traditional’ randomised controlled trials, or the partial or complete use of real-world data allowing the conduct of ‘trials within a cohort’, ‘registry-based randomised controlled trials’, ‘contactless trial’ or ‘direct-to-patient trial’ (when all data are collected using real-world data and data coming from fully remote pathways). | Relying partially or completely on sources of data that were not primarily designed for research purposes can lead to issues regarding data validity, integrity, and monitoring. The follow-up of patients may not be as standardised as de novo clinical studies. Assessment of endpoints is at risk to be unblinded and/or adjudicated in a decentralised manner. Collection of safety data can be problematic. |
| Cross-over designs | Trials where the sequence of interventions is randomised and therefore each patient receives all interventions and is its own control, leading to the enrolment of fewer patients to conclude than the parallel-group design for a given effect of interest. | They assume an identical baseline state at the start of each treatment exposure, no effect of the first treatment to which a patient is exposed during the period of exposition to the second treatment, and no interaction between treatments and periods of exposure. |
HAS, Haute Autorité de santé.