Table 2: Summary of main Phase II clinical trial designs.


One stage one arm designCompare with historical control.
Smallest sample size.
Delay the evaluation of effectiveness.
Historical control may not be valid.
Subject to population differences, time trends, evaluation bias, and so forth.

Gehan’s two stage designWith interim monitoring.
Rule out ineffective drug with minimized sample size.
No testing on agents showing some promise.
Only suitable for binary outcome.
The endpoint is different from that in following Phase III trial.

Simon’s two
stage design
The samples in two stages are optimized.
Quickly screen out agents without effectiveness while testing further agents with some promise.
Two choices: optimal versus minimax
Only suitable for binary outcome.
The endpoint is different from that in the following Phase III trial.

Bayesian Phase II designFlexible monitoring schedule.
More efficient and robust.
Intensive computation.
Relies heavily on statistician during trial.

Randomized Phase II designUse of randomization.
Reliable control and less bias.
More similar to Phase III trial.
Sample size increases.
Length of trial increases.
Cost increases.

Phase II pick the winner designEfficient and effective way of comparing two or multiple experimental regimens.
Each experimental regimen compared with historical controls.
Not appropriate for comparison of adding an experimental agent to standard regimen.

Phase II
screening design
Limits the sample size required for a randomized Phase II comparison.
Good for comparison of the addition of an experimental agent to standard regimen.
No statistical comparison between the selected arms.

Phase II randomized discontinuation designGood when significant continued benefit after initial benefit implies significant benefit overall, and vice versa, or when benefit is restricted to a nonidentifiable subgroup of patients.May need a large number of patients treated at a treatment not effective for them.

Phase II/III designUse of Phase II data in Phase III trial.
Minimize delay in starting up Phase III study.
Use of concurrent control.
Useful for new drugs showing efficacy.
Large sample sizes.
Needs Phase III infrastructure developed even if it stops early.