Phase III Clinical Trials: Design and Analysis Considerations
Phase III trials are the large-scale comparative trials most of us have in mind when
we think of clinical trials. They typically feature:
Concurrent comparison with best available treatment, sometimes a placebo.
Randomized assignment to treatment.
Objective evaluation of response:
By objective endpoints if possible,
Otherwise by blinding or masking of participants and evaluators.
Usually two-sided hypotheses, not one-sided. (Why?)
Other features that may complicate the statistician’s job:
More than two treatment arms.
Unbalanced assignment to treatment.
Need to block by site or by patient characteristics.
Testing for noninferiority instead of improvement over standard.
More than one outcome measure.
An interesting historical example: the 1954 polio vaccine trial. The heart of the study was
a randomized, double-blind placebo-controlled trial in which about 200,000 children got
active Salk vaccine and 200,000 got placebo. Among those who got vaccine, 57 got polio,
total, for a rate of 28 per 100,000. Among those who got placebo, 142 got polio, for a rate
of 71 per 100,000. This is a statistically signiﬁcant diﬀerence. The rate for children whose
parents declined consent was 54 per 100,000.
This seems like a striking result. Could we have gotten by without placebo control?
The initial study plan was to have an area where 2nd graders got vaccinated but 1st and
3rd graders did not, and to use the 1st and 3rd grade as a control group. In fact this
was done, with 222,000 children in grade 2 in a diﬀerent area vaccinated, another 124,000
in grade 2 whose parents did not consent, and a total of 725,000 unvaccinated 1st and
3rd grade controls. The polio rate per 100,000 in the vaccinated 2nd graders was 25 per