Table 1.
Features of alternative trial designs.
| Design | ||||||
|---|---|---|---|---|---|---|
| Parallel group | Two-stage | Fully randomised | Partially randomised | Zelen: Single consent | Zelen: Double consent | |
| Treatment preferences identified? | No | For a random subset of patients (In choice arm) | For all patients | For all patients | Not directly | Not directly |
| Which patients are allowed to choose treatment? | None | In choice arm | None | Patients with a stated preference | Patients in the single consent arm can agree to randomised treatment A or not | Patients in both consent arms can agree to randomised treatment (A or B) or not |
| Is treatment effect estimable? | Yes | Yes | Yes, overall and within preference subgroups | Yes, within preference subgroups | Yes. Unbiased if treatments are concealed, and consent is independent of patient preferences; biased if treatments are revealed | Yes. Unbiased if treatments are concealed, and consent is independent of patient preferences; biased if treatments are revealed |
| Selection effect estimable? | No | Yes | Yes, potentially biased | Yes, potentially biased | No | Yes if treatments concealed |
| Preference effect estimable? | No | Yes | Yes | No | No | Yes if treatments concealed |