- 1Why randomize?Trials in Dhaka showed that a solution of salt, sugar and clean water mixed at home could replace intravenous fluids for most children with severe diarrhea, and oral rehydration therapy is now credited with saving more than 50 million lives. What random assignment does that no careful comparison of program villages with other villages can, worked from James Lind’s scurvy sailors to HPTN 052.7 min
- 2Designing a trial: the question, the participants, and the allocationThree decisions come before anything else: what question the trial will answer, who is eligible to be in it, and how participants are allocated to groups. PICO, the difference between random assignment and random selection, eligibility criteria and the population a result applies to, then simple, block, stratified and cluster randomization.7 min
- 3Blinding, and what the control group receivesWho knows which group a participant is in, and what the control group gets instead. Blinding of participants, providers, outcome assessors and analysts, including what to do when an intervention cannot be hidden, then placebo, active, standard-care and waiting-list comparators and how the choice of comparator defines the question the trial answers.6 min
- 4When is it ethical to randomize?Equipoise, how uncertainty can be specific to a setting, and why a lottery is often a fairer way to fill an oversubscribed program than whatever would have happened otherwise. Includes control groups where no services exist, the placebo-controlled AZT trials of the 1990s, and planning from the start for the people in a trial to benefit from it.6 min
- 5Running the trial: fidelity, monitoring, and reportingOnce a trial is running the work is checking that the intervention is delivered as intended, watching for harm without biasing the result, and reporting it so a reader can judge it. Implementation fidelity from the CHoBI7 trial in Bangladesh, data safety monitoring boards and O’Brien-Fleming stopping boundaries, four trials stopped early, and CONSORT.7 min
- 6After randomization: attrition, non-compliance, and spilloverRandomization gives a fair comparison at baseline, and then participants drop out, skip the treatment they were assigned, or change each other’s outcomes. Loss to follow-up and sensitivity analysis, intention-to-treat against per-protocol and CACE, spillover and SUTVA in school-based deworming in Kenya, and what randomizing clusters costs in power.8 min
- 7From a trial result to policyA significant result is where the work of turning evidence into policy starts. Stepped-wedge designs across 34 health centers in Peru, how a non-inferiority margin works, pre-specified and post-hoc subgroup analyses of aspirin for tuberculous meningitis in Vietnam, and what it took beyond the oral rehydration trials themselves to change practice.9 min