11. Randomized Controlled Trials: Randomization, Blinding, ITT
After this lesson you can
- Explain what randomization balances and distinguish random allocation from random sampling
- Draw a trial flow diagram that keeps exclusions, non-attenders and losses to follow-up separate
- Distinguish allocation concealment from blinding and name the bias each one prevents
- Compute intention-to-treat, per-protocol and as-treated risk ratios and explain why only intention-to-treat keeps randomization
- Calculate a risk difference, number needed to treat and vaccine effectiveness, and separate efficacy from effectiveness
- Analyze a cluster randomized trial by its clusters rather than its patients
Video coming soon
Scheduled for release on November 8, 2026.
The last design lesson, presenting the randomized controlled trial as a cohort made by chance. Every idea is worked on one invented Oakhaven care-program trial of 800 residents, from the flow diagram to three competing risk ratios.
Questions and answers
Questions a learner would ask about this lesson. Think through your own answer before you open each one.
- What does randomization achieve that an observational design cannot?
Show answer
Randomization lets chance, not health or choice, decide who receives the exposure. The two arms are then alike on average in age, severity, support and motivation, and also in factors nobody measured. Any difference left between the arms is the exposure plus chance. In Oakhaven, residents who chose to attend the program were healthier to begin with, which a comparison of attenders could not untangle.
- How is allocation concealment different from blinding?
Show answer
Concealment acts before and at allocation: it hides the next assignment from whoever enrolls participants, which prevents selective enrollment and steering. Blinding acts after allocation: it hides which arm a person is in, which prevents performance and measurement bias. Every trial can conceal its allocation, but not every trial can be blinded, and a double-blind trial is not automatically concealed.
- Why are there three different risk ratios from the same Oakhaven trial?
Show answer
Intention-to-treat compares the arms as randomized: 60 of 400 against 80 of 400, risk ratio 0.75. Per-protocol keeps only the 300 who attended: 30 of 300 against 80 of 400, risk ratio 0.50. As-treated groups people by what they received: 30 of 300 against 110 of 500, about 0.45. Only intention-to-treat compares the groups chance made; the other two compare groups formed by behavior, which is tangled with prognosis.
- What does an intention-to-treat result actually estimate?
Show answer
It estimates the effect of offering the program, not of completing it. Non-attendance dilutes the effect toward the null, so it is conservative for a superiority question, and it is the policy-relevant quantity for a district. In Oakhaven the risks were 15 and 20 percent, a risk difference of 5 per 100, so 20 residents must be offered the program for one bad outcome to be prevented.
- Why is a trial that randomizes clinics analyzed by clinics?
Show answer
The unit of analysis follows the unit of allocation. With 20 clinics of 100 patients randomized 10 per arm, the comparison is 10 clinic proportions against 10, not 1,000 patients against 1,000. Patients within a clinic resemble each other, so pooling them overstates the evidence and can turn chance differences between clinics into apparent effects.
Practice quiz
- Question 1Keeping the next assignment hidden from the person enrolling participants is called:
- Question 2In the program arm as randomized, 60 of 400 residents had the outcome; in the usual care arm, 80 of 400 did. What is the intention-to-treat risk ratio?
- Question 3Using the same intention-to-treat risks of 15 and 20 percent, how many residents must be offered the program to prevent one bad outcome?
- Question 4Analyzing only the 300 residents who attended the program, with their 30 outcomes, against the usual care arm gives a risk ratio of 0.50. What is this analysis?
- Question 5Why does only the intention-to-treat analysis keep the protection of randomization?
- Question 6Under field conditions the attack rate was 10 percent among the unvaccinated and 4 percent among the vaccinated. What is the vaccine effectiveness?
- Question 7A survey draws a random sample of residents and asks about their habits. Which statement is correct?
- Question 8Twenty clinics of 100 patients are randomized, ten per arm. The comparison is between:
- Question 9A placebo comparison arm is impossible for a visible care program. What does a usual care arm leave uncontrolled?
- Question 10In the Oakhaven trial, 100 residents allocated to the program never attended, and their outcomes were known. How should they be described?
Saved in your browser only: no account, no server.
Principles of Epidemiology is education in epidemiologic reasoning, not clinical, screening, or policy guidance. Every example is set in the fictional Oakhaven Health District, and every number is invented for teaching.