Topic

Trial and study design

Study design decides how far a comparison between treatments can be trusted to be free of bias. Randomised controlled trials, with safeguards such as allocation concealment, sit beside observational approaches like cohort and case-control designs. Adaptive and group sequential designs allow a trial to be changed or stopped early after an interim analysis.

Concepts in this topic

• Active Control TrialAn active control trial is a randomised trial testing a new treatment against a known effective treatment, not placebo, for superiority or non-inferiority.
• Adaptive Trial DesignAdaptive trial design is a trial plan whose prespecified rules let interim data change the trial, such as stopping early, while keeping inference valid.
• Administrative CensoringAdministrative censoring is right censoring that occurs when a trial's planned end of follow-up or data cut-off arrives before a patient has the event.
• Allocation ConcealmentAllocation concealment hides the next assignment in a randomised trial from recruiters until allocation, protecting the treatment effects that inform HTA.
• Basket TrialA basket trial tests one targeted treatment in several diseases, usually tumour types, that share the same biomarker, with each disease forming a basket.
• Bayesian Adaptive DesignBayesian adaptive design is a clinical trial design in which pre-set rules on posterior or predictive probabilities trigger changes such as early stopping.
• Block RandomizationBlock randomisation assigns participants in small blocks with set numbers per arm in random order, keeping arms balanced throughout trials that inform HTA.
• Case-Control DesignA case-control design selects people according to whether they have an outcome and compares their prior exposure histories with those of appropriately sampled controls from the population that produced the cases.
• Clinical TrialA prospective research study in which participants receive one or more interventions to evaluate their effects on health-related outcomes.
• Cluster RandomizationCluster randomization assigns intact groups, rather than individual participants, to study strategies and requires design and analysis that account for similarities among members of each group.
• Cohort DesignA cohort design follows a defined group of people from an established starting point to observe subsequent outcomes and, when relevant, compare those outcomes across baseline exposures or treatment strategies.
• Cross-Sectional DesignA cross-sectional design is an observational design that measures exposure and outcome in a population at one point in time, often to estimate prevalence.
• Crossover DesignA crossover design assigns each participant to receive two or more interventions in a planned sequence, so the same person contributes outcome data under multiple treatments.
• Crossover TrialA trial in which participants sequentially receive two or more treatments under comparison, with each participant's response compared against their own.
• Dependent CensoringDependent censoring occurs when the chance of being censored is linked to a patient's risk of the event, so standard survival estimates can be biased.
• Electronic Health RecordA digital, longitudinal record of a person’s health information maintained for care and, where permitted, shared or reused across authorised settings and purposes.
• Equivalence TrialA trial designed to show a new treatment's effect is neither substantially better nor worse than an existing treatment, within a set margin.
• Factorial DesignA trial design evaluating two or more interventions simultaneously by assigning participants to every possible combination, revealing each effect and any interaction.
• Futility AnalysisA futility analysis assesses interim trial evidence against a prespecified rule to judge whether continuing is sufficiently likely to meet the study's final objective.
• Group Sequential DesignA group sequential design is a trial design with prespecified interim analyses and stopping boundaries that permit early decisions while controlling the relevant overall error rates.
• Immortal TimeA period of follow-up during which, due to how exposure groups were defined, a participant could not possibly have experienced the outcome.
• Independent CensoringAn assumption that, within the information used for a survival analysis, censoring does not convey additional information about a person’s unobserved event time.
• Intention-to-Treat AnalysisIntention-to-treat analysis compares outcomes according to participants’ original randomised assignments, irrespective of subsequent treatment receipt or adherence, while explicitly addressing missing outcome data.
• Interim AnalysisA planned analysis of trial data conducted before final completion, typically to assess safety, efficacy, or futility and inform continuation decisions.
• Left TruncationLeft truncation (delayed entry) occurs when people join a survival analysis risk set after time zero and are observed only if still event free at entry.
• Longitudinal DesignA longitudinal design observes the same individuals or other study units at multiple time points to characterize within-unit change, temporal sequence, and differences in trajectories over time.
• Loss to Follow-UpLoss to follow-up occurs when trial participants stop providing data before follow-up ends, so later outcomes, costs and QALYs can be missing.
• Minimal Clinically Important DifferenceThe minimal clinically important difference (MCID) is the smallest change in an outcome score that patients perceive as important, beneficial or harmful.
• N-of-1 TrialA trial conducted within a single participant, repeatedly alternating experimental treatment and control in random, blinded sequence to determine its individual effect.
• Non-Inferiority TrialA trial designed to show a new treatment is not meaningfully worse than an active comparator, within a predefined acceptable margin.
• Observational StudyA study that observes exposures and outcomes without investigator assignment of the exposure under investigation, using a specified design to describe patterns or assess associations and, under additional assumptions, causal effects.
• Quasi-Experimental DesignA research design estimating an intervention's causal effect without random assignment, instead using features such as a natural experiment or matched comparison group.
• Randomised Controlled TrialA trial design in which participants are randomly assigned to an experimental intervention or a comparator, the gold standard for causal evidence.
• RandomizationThe process of assigning trial participants to treatment groups by chance, producing groups comparable on both known and unknown influencing factors.
• ScreeningThe systematic offer of a test or assessment to people without recognised symptoms to identify possible disease or elevated risk for further action.
• Sequential SamplingA data collection approach in which the decision to continue or stop sampling is made based on accumulating data, rather than a fixed sample size.
• Surrogate EndpointA trial outcome used as a substitute for a direct clinical benefit measure, such as survival, based on its expected predictive relationship to that outcome.
• TruncationA situation in survival or longitudinal analysis where certain observations are systematically excluded based on a variable's value, such as required registration.