Skip to main content

Box 2 Definitions of key technical terms

From: The adaptive designs CONSORT extension (ACE) statement: a checklist with explanation and elaboration guideline for reporting randomised trials that use an adaptive design

Validity—The ability to provide correct statistical inference to establish effects of study interventions and produce accurate estimates of effects (point estimates and uncertainty), to give results that are convincing to the broader audience (science community and consumers of research findings).
Integrity—Relates to minimisation of operational bias, maintenance of data confidentiality, and ensuring consistency in trial conduct (before and after adaptations) for credibility, interpretability, and persuasiveness of trial results.
Pre-planned adaptations or adaptive features—Pre-planned or prespecified changes or modifications to be made to aspects of an ongoing trial, which are specified at the design stage or at least before seeing accumulating trial data by treatment group, and are documented for audit trail (such as in the protocol).
Unplanned changes—Ad hoc modifications to aspects of an ongoing trial.
Type of AD—The main category used to classify a trial design by its pre-planned adaptive features or adaptations. Some ADs can fall into more than one main category of trial adaptation (see Table 1).
Adaptive decision-making criteria—Elements of decision-making rules describing whether, how, and when the proposed trial adaptations will be used during the trial. It involves pre-specifying a set of actions guiding how decisions about implementing the trial adaptations are made given interim observed data (decision rules). It also involves pre-specifying limits or parameters to trigger trial adaptations (decision boundaries). For example, stopping boundaries that relate to pre-specified limits regarding decisions to stop the trial or treatment arm(s) early.
Interim analysis—A statistical analysis or review of accumulating data from an ongoing trial (interim data) to inform trial adaptations (before the final analysis), which may or may not involve treatment group comparisons.
Binding rules—Decision rules that must be adhered to for the design to control the false positive error rate.
Non-binding rules—Optional decision rules that can be overruled without negative effects on control of the false positive error rate.
Statistical properties or operating characteristics—Relates to behaviour of the trial design. These may include statistical power, false positive error rate, bias in estimation of treatment effect(s), or probability of each adaptation taking place.
Simulation—A computational procedure performed using a computer program to evaluate statistical properties of the design by generating pseudo data according to the design, under a number of scenarios and repeated a large number of times.
Fixed (non-adaptive) design—A clinical trial that is designed with an expected fixed sample size without any scope for pre-planned changes (adaptations) of any study design feature.
Bias—The systematic tendency for the treatment effect estimates to deviate from their “true values”; including the statistical properties (such as error rates) to deviate from what is expected in theory (such as pre-specified nominal error rate).
Operational bias—Occurs when knowledge of key trial-related information influences changes to the conduct of that trial in a manner that biases the conclusions made regarding the benefits and/or harms of study treatments.
Statistical bias—Bias introduced to the study results or conclusions by the design: for example, as a result of changes to aspects of the trial or multiple analyses of accumulating data from an ongoing trial.
Subpopulation(s)—Subset(s) of the trial population that can be classified by characteristics of participants that are thought to be associated with treatment response (such as genetic markers or biomarkers).
Adaptation outcome(s)—Outcome(s) used to guide trial adaptation(s); they may be different from the primary outcome(s).