Glossary category

Trial design and conduct

Terms describing how a clinical study is constructed and run.

Terms

33 documents

Table 1. Terms in the trial design and conduct category.

Term	Definition
Active-controlled	A design in which the comparator is another active treatment. Establishes a difference between the two treatments and not an effect against no treatment.
AdjudicationEndpoint adjudication	Independent classification of endpoint events by a committee blinded to assignment. Reduces the risk that ascertainment differs between arms for events requiring judgement.
Allocation concealment	The prevention of foreknowledge of the next allocation by the person enrolling a participant. Distinct from blinding, which concerns knowledge after allocation, and independently associated with effect-size inflation when absent.
AttritionLoss to follow-up	Loss of participants from a trial before its primary analysis timepoint. Differential attrition between arms threatens the comparability that randomisation established.
BlindingMasking	The withholding of knowledge of assignment from participants, investigators, outcome assessors or analysts. The Institute records who was blinded rather than describing a trial as blinded without qualification.
Cardiovascular outcome trialCVOT	An event-driven trial with a composite cardiovascular endpoint, conducted for several agents in this class as a regulatory requirement to exclude excess risk. A trial designed to exclude harm is not thereby designed to establish benefit.
ComparatorControl	The treatment against which the intervention is compared. An effect estimate is always an estimate against a specific comparator, and no estimate is transferable to a different one without an assumption the Institute states.
Crossover	A design in which each participant receives more than one treatment in sequence, separated by a washout. Efficient for stable conditions and unsuitable where the condition or the effect changes over time.
Data monitoring committeeDMC · DSMB	An independent group with access to unblinded accumulating data, empowered to recommend modification or termination. Its recommendation to stop for efficacy is a source of effect-size inflation.
Dose-finding	A study comparing several doses to characterise the dose–response relationship, usually conducted in phase 2 and usually not powered for a definitive efficacy conclusion at any single dose.
Double-blind	A design in which neither participants nor investigators know the assignment. The Institute records the specific parties blinded, because the term is used inconsistently in the literature.
Early stopping	Termination of a trial before its planned completion, most often for efficacy at an interim analysis. Trials stopped early for efficacy systematically overestimate effect size, and the Institute attaches this qualification wherever such an estimate is reproduced.
Estimand	A precise statement of the quantity a trial is estimating, including how intercurrent events such as discontinuation or rescue therapy are handled. Two trials reporting the same endpoint under different estimands are not reporting the same thing.
Event-driven	A trial that continues until a pre-specified number of endpoint events has accrued rather than for a fixed duration. Follow-up duration is therefore a consequence of the event rate and not a design parameter.
Intention to treatITT	Analysis of participants in the group to which they were randomised, regardless of the treatment they received. Preserves the comparison randomisation created and generally produces a more conservative estimate.
Intercurrent event	An event occurring after randomisation that affects the interpretation of the outcome, such as treatment discontinuation, rescue medication or death. The strategy for handling each is part of the estimand.
Maintenance dose	The dose reached after titration and administered for the remainder of the treatment period. The dose to which an effect estimate attaches.
MarginNon-inferiority margin	The largest difference, specified before the trial, that would still be regarded as acceptable in a non-inferiority design. A non-inferiority conclusion is only as meaningful as the justification for its margin.
Non-inferiority	A design testing whether a treatment is not worse than a comparator by more than a pre-specified margin. A non-inferiority result does not establish superiority over placebo, and the Institute does not report it as though it did.
Open-label	A design in which assignment is known to participants and investigators. Objective endpoint measurement limits but does not eliminate the resulting risk, because co-intervention and retention may differ by arm.
Open-label extensionOLE	A period following the randomised phase in which all participants receive active treatment. Reports persistence and long-term tolerability; cannot re-establish comparative efficacy.
Per protocol	Analysis restricted to participants who adhered to the protocol. Breaks the randomised comparison and is reported by the Institute only as a secondary analysis alongside the intention-to-treat result.
Phase 1	First administration to humans, evaluating safety, tolerability and pharmacokinetics, usually in small numbers and often in healthy volunteers. Efficacy endpoints in a phase 1 study are exploratory by construction.
Phase 2	Exploratory clinical development establishing dose and preliminary efficacy. Phase 2 estimates in selected populations are systematically larger than the confirmatory estimates that follow them.
Phase 3	Confirmatory clinical development, adequately powered for a pre-specified primary endpoint in the population for which authorisation is sought.
Placebo-controlled	A design in which the comparator is an inert preparation matched to the intervention in appearance and administration. Matching is imperfect where the intervention produces a perceptible effect.
Pragmatic trial	A trial designed to estimate effectiveness under ordinary conditions rather than efficacy under ideal ones. Typically has broader eligibility, less intensive follow-up and a lower internal validity than an explanatory trial.
Randomisation	The allocation of participants to study arms by a chance process, which distributes both known and unknown prognostic factors between arms in expectation. The single design feature that most constrains bias.
Randomised withdrawalWithdrawal design	A design in which participants who have responded during an open-label lead-in are randomised to continue or to discontinue. Establishes whether an effect requires continued exposure to be maintained.
Run-in periodLead-in	A period before randomisation during which all participants receive the same treatment. A run-in that excludes non-responders or non-tolerators produces a randomised population that is not the population of interest.
Stratification	Randomisation within strata defined by prognostic factors, to ensure balance on those factors. Stratification factors should be included in the primary analysis model.
TitrationDose escalation	The stepwise escalation of dose toward a target. Tolerability in this compound class depends on the titration schedule as much as on the compound, which the Institute records wherever tolerability is compared across programmes.
Washout	A period between treatments in a crossover design intended to allow the effect of the first to dissipate. A washout shorter than approximately five half-lives leaves carry-over.