OBJECTIVES: To develop a method to assess the impact of cost on patient preferences in chronic lymphocytic leukemia (CLL) treatment.
METHODS: Assessing the impact of cost on treatment preferences in a discrete-choice experiment (DCE) can be difficult, especially for drugs with long treatment durations. When costs are high, respondents may either ignore the cost attribute or focus only on cost, reducing the value of the study. As an alternative, we first administered a DCE that did not include cost as an attribute. After the DCE, respondents answered one additional question where cost was added as attribute to predefined Medicines A and B. Respondents were randomly assigned to the high-cost version ($400 per month difference in cost between the medicines) and the low-cost version ($75 per month difference in cost between the medicines). Using the DCE results, we computed posterior preferences for each individual conditional on the pattern of observed choices and based on Bayes’ theorem without cost and forecast the share who would choose each of the two medicines. The forecast was compared to the percent who selected Medicine A or B in the two cost questions.
RESULTS: 384 patients with self-reported CLL took a DCE survey on treatments for CLL (mean age 65 years, 23% first line, 39% relapse, 53% received financial aid to pay for treatments). Using the method described above, we forecasted that 91% of the sample would prefer the medicine with the longest PFS when cost was not included, compared to 50% and 26% who actually selected that option for the low-cost and high-cost follow-up questions, respectively.
CONCLUSIONS: The fixed follow-up questions including cost provided preference information when high costs were not feasible as a DCE attribute. Respondents were very sensitive to modest changes in treatment cost, pointing to the importance of gathering this information.
