Partially Randomized Preference Trial Design

Structure

In a PRPT comparing two treatments, there are potentially four separate groups of patients, each receiving one of the two treatments, usually an established treatment and a new treatment. Patients are informed about the two treatments being compared and asked if they have a strong preference. Patients who have strong preferences for one treatment over another are allocated to a group in which they can have their preferred [Page 1010]method of treatment. In Figure 1, patients who are particularly motivated toward the new treatment are allocated to Group 1, and patients who would prefer to use an established method of treatment are allocated to Group 2. Patients who have no strong preference and are equally prepared to use either treatment are said to be in equipoise and, with their consent, have one or the other treatment type assigned randomly and are thereby allocated to either Group 3 or Group 4.

Figure 1 The Structure of a Partially Randomized Preference Trial (PRPT)

Variations on the Structure

Three-group trials may result if no one has a strong preference for one of the treatments. A very small number of patients in one of the preference groups (Groups 1 or 2 in Figure 1) may not be analyzable statistically but will still serve an important purpose in removing those with preferences from Groups 3 and 4, where randomization to a nonpreferred treatment would lead to disappointment.

If all the patients who are recruited into the PRPT have a strong preference, a feasibility study will result where all patients chose their treatment, and if no recruit has a strong preference, an RCT will result. However, if an RCT results from patients not having a preference for a particular treatment offered, it will differ from that of many conventional RCTs as motivational factors will not distort outcomes. In a conventional RCT, some patients may agree to randomization with the hope of obtaining a new and/or otherwise inaccessible therapy. Patients who think they would prefer a new treatment (those who are allocated to Group 1 in a PRPT) may have been included in an RCT if the new treatment was unavailable outside the trial and participation in the trial was their only way of obtaining their preferred method of treatment. However, RCT participants are asked to accept any of the health care options being compared. Inclusion of patients preferring the new treatment over a standard treatment will bias the RCT sample in favor of the new treatment; those with preferences for the standard treatment are more likely to decline to participate in an RCT as they can usually obtain their preferred treatment outside the trial. The RCT sample recruited will be randomized to two groups. When preferences for the new treatment are marked, one group will contain participants who are pleased to receive the new treatment whereas the other group will contain individuals who are disappointed that they have not been allocated to the [Page 1011]new treatment (as demonstrated empirically by Feine and colleagues in 1998 and discussed by Bradley in a commentary the following year). When the majority of recruits have a preference for the new treatment, an RCT creates groups that have been allocated at random but differ in respect to motivation to use the treatment assigned. The control group of participants who have been randomized to the standard treatment will contain patients who are disappointed with the treatment allocation and therefore will be more likely to drop out of the trial. They may also be less likely to follow the treatment recommendations and do less well with the treatment than they would have done if motivated to use that treatment. If disappointed patients drop out of the control group, outcomes of the control treatment will be artificially improved, minimizing any advantages of the new treatment. However, if such poorly motivated patients remain in the control group, outcomes will be worsened, thereby exaggerating the advantages of the new treatment.

...