Randomised Controlled Trials (Population Health Perspective)

Randomised controlled trials (RCTs) are intervention studies — the highest form of analytic study design. Unlike observational studies where participants self-select into exposure groups, in an RCT participants are randomly allocated to the intervention or control group. This single feature — randomisation — is what distinguishes RCTs from cohort studies and is what enables causal inference.

Randomisation (random allocation) ensures that, if enough participants are enrolled, both known and unknown potential confounders are distributed equally between the intervention and control arms. This means that at the end of the trial, differences in outcomes between groups can be attributed to the intervention itself rather than to differences in baseline characteristics.

It is important to distinguish random allocation from random selection. Random selection refers to how participants are chosen from the source population to be included in the study — it determines the generalisability (external validity) of findings. Random allocation refers to how enrolled participants are assigned to intervention or control groups — it controls confounding and determines internal validity. Many RCTs use random allocation without random selection, meaning results may have high internal validity but limited generalisability.

Cluster randomisation is a variant in which groups of participants (clusters such as schools, GP practices, or villages) are randomised rather than individuals — appropriate when the intervention is applied at the group level.

RCTs use two principal analysis strategies. Intention-to-treat (ITT) analysis analyses all participants as randomised, regardless of adherence — this preserves the benefits of randomisation and estimates effectiveness (performance in real-world conditions). Per-protocol analysis analyses only those who fully adhered to their assigned treatment — this sacrifices the benefits of randomisation but estimates efficacy (performance under ideal conditions).

Sources of bias in RCTs include: lack of allocation concealment (allows recruiter to predict assignment, introducing selection bias); lack of blinding (participants or assessors knowing allocation introduces performance or detection bias); loss to follow-up; and non-adherence.

RCT strengths: maximises internal validity; can calculate incidence and measures of association; establishes temporal sequence; best design for causal inference. Limitations: not all exposures are suitable for randomisation (ethical constraints, clinical equipoise); resource-intensive; participants are often highly selected, limiting generalisability.