Key Facts

The RCT is the most rigorous method for determining whether a cause-effect relationship exists between an intervention and outcome. Proper conduct requires attention to randomisation, blinding, sample size, and analysis.

Study Conduct

1. Define research question (PICO: Population, Intervention, Comparator, Outcome)
2. Calculate sample size (power calculation)
3. Recruit participants (inclusion/exclusion criteria)
4. Randomisation (simple, block, stratified, minimisation)
5. Allocation concealment
6. Intervention delivery
7. Blinding of participants, clinicians, outcome assessors
8. Follow-up and outcome measurement
9. Analysis (ITT primary; per-protocol secondary)
10. Reporting (CONSORT)

Types of RCT

• Parallel group: most common; two or more groups run simultaneously
• Crossover: each participant receives both treatments in sequence; washout period between
• Factorial: tests two or more interventions simultaneously (e.g. ISIS-2: aspirin × streptokinase)
• Cluster: randomises groups (e.g. GP practices) rather than individuals
• Non-inferiority/equivalence: tests whether new treatment is not worse than standard
• Adaptive: allows modification of trial design based on interim results

Randomisation Methods

• Simple: coin flip/random number generator
• Block: ensures balanced groups at regular intervals
• Stratified: stratifies by important prognostic factors before randomisation
• Minimisation: algorithm-based allocation balancing multiple factors

Interpreting RCT Results

• Primary outcome: pre-specified main outcome of interest
• Secondary outcomes: additional outcomes; hypothesis-generating
• Absolute risk reduction (ARR): difference in event rates between groups
• Relative risk reduction (RRR): proportional reduction in risk
• NNT: number needed to treat to prevent one event (1/ARR)
• Confidence interval: range within which true effect likely lies (95% CI); if CI for RR crosses 1.0, not statistically significant
• P-value: probability of observing result if null hypothesis true; p<0.05 conventionally significant

Threats to Validity

• Selection bias (inadequate randomisation/allocation concealment)
• Performance bias (lack of blinding)
• Attrition bias (differential loss to follow-up)
• Detection bias (unblinded outcome assessment)
• Reporting bias (selective outcome reporting)

Critical Appraisal of RCTs (CASP)

1. Was the trial focused on a clear question?
2. Was allocation randomised and concealed?
3. Were groups similar at baseline?
4. Were participants/clinicians/assessors blinded?
5. Were all participants accounted for (follow-up rate)?
6. Were outcomes measured in a standard, valid way?
7. Were results reported for all pre-specified outcomes?
8. Was the treatment effect size clinically important?
9. Were confidence intervals precise?
10. Can results be applied to your population?

Statistical Concepts

• Type I error (α): false positive (rejecting true null hypothesis); conventionally set at 0.05
• Type II error (β): false negative (failing to reject false null hypothesis); conventionally 0.1-0.2
• Power: 1 − β; probability of detecting a true difference
• Intention-to-treat: includes all randomised participants; conservative estimate
• Per-protocol: includes only compliant participants; may overestimate effect

Landmark UK Medical Trials

• MRC Streptomycin Trial (1948): first properly randomised clinical trial
• ISIS-2 (1988): aspirin + streptokinase in acute MI; factorial design
• HOPE (2000): ramipril in high-risk CVD patients
• UKPDS (1998): intensive glucose control in T2DM
• PARADIGM-HF (2014): sacubitril/valsartan in heart failure
• DAPA-HF (2019): dapagliflozin in heart failure
• RECOVERY (2020): dexamethasone in COVID-19

Ethical Considerations

• Equipoise: genuine uncertainty about which treatment is better
• Informed consent: participants fully informed of risks, benefits, alternatives
• Data Safety Monitoring Board (DSMB): independent review of interim safety data
• Stopping rules: pre-defined criteria for early termination (harm or overwhelming benefit)
• Research Ethics Committee: approval required before recruitment
• Clinical trial registration: mandatory prospective registration (e.g. ClinicalTrials.gov, ISRCTN)

• RCTs provide the most reliable evidence for treatment decisions when properly conducted
• Even well-designed RCTs may have limited external validity if populations are highly selected
• Publication bias means positive trials are more likely published; funnel plots and Egger test assess this
• Fragility index: number of patients whose outcome would need to change to make result non-significant
• Replication of findings across multiple RCTs increases confidence in treatment effects

CONSORT Flow Diagram Components

NNT Calculation Example