METHODS: Participants were 27 women and 18 men enrolled in phase 1 or 2 clinical oncology trials and randomized to 1 of 3 interview protocols corresponding to 3 "target questions" about likelihood of benefit: frequency-type ("Out of 100 patients who participate in this study, how many do you expect will have their cancer controlled as a result of the experimental therapy?"); belief-type ("How confident are you that the experimental therapy will control your cancer?"); and vague ("What is the chance that the experimental therapy will control cancer?"). In semistructured cognitive interviews, we queried participants about how they understood and answered the target question. Each participant then answered and discussed one of the other target questions.
RESULTS: Participants tended to provide higher belief-type expectations (median, 80.0) than frequency-type expectations (median, 50.0) (P=.02). High expectations were expressed even by participants who understood that there was little or no evidence that the experimental therapy would work. The most common justification for responses involved the need for a positive attitude. Other justifications included references to the participant's health and religious faith. It is noteworthy that 27% said they could not answer the target question. The majority (10/12) who answered "don't know" did so for the frequency-type question after first responding to the belief-type question. Participants' reasons for saying "don't know" included that the question is unanswerable, there was not enough information, and the participant did not recall the answer.
CONCLUSIONS: Frequency-type and belief-type questions elicited significantly different responses. Many participants' responses to questions about likelihood of benefit appeared to function as expressions of hope rather than descriptions of the participants' understanding of the clinical trials. This phenomenon makes it challenging to assess patient understanding in this context. Our findings suggest, however, that some patients provide a more candid response to a frequency-type question (ie, "don't know") if first given the opportunity to express confidence in their own outcomes.