Quantifying Patient Preferences for Treatment Outcomes in AML: A Discrete-Choice Experiment

Background: Treatment of acute myeloid leukemia (AML) remains a significant challenge: induction chemotherapy is associated with substantial toxicities, and most patients will eventually die related to their disease. While inducing complete remission (CR) has historically been the primary aim of most treatment regimens, it is unknown to what extent patients would be willing to trade-off the chance at remission for other outcomes such as a reduced toxicity profile. Clinicians routinely make such decisions for patients by selecting less intense regimens for elderly patients or those with significant comorbidities, or when remission may not be clinically relevant or likely achievable. Patient preferences for treatment outcomes often inform shared decision making at the bedside and are increasingly recognized by the FDA and industry as a critical aspect to incorporate into drug development. As part of a patient-focused drug development initiative led by The Leukemia & Lymphoma Society (LLS), we sought to quantify patient preferences for treatment outcomes in AML.

Methods: In collaboration with diverse stakeholders (including patients, clinicians, industry representatives, and the LLS staff), a survey instrument was developed to quantify patient preferences for treatment outcomes in AML. A discrete-choice experiment (DCE), in which participants made choices between 9 pairs of hypothetical treatments, was utilized with five attributes: event-free survival, complete remission, time in hospital, short-term side effects, and long-term side effects. All attributes were tiered to three clinically relevant levels. A national survey was conducted and analyzed using a conditional logistic regression. Elicited preference weights, reflecting the desirability for individual attributes, were aggregated into preference estimates. Sub-group analysis compared outcomes based on remission status.

Results: The survey was sent to 896 patients with AML who were identified through the LLS database. Three-hundred and twenty-two patients participated in the survey; 294 completed the DCE. Most patients were white (89.4%), married (74.7%), college-educated (74.4%), and privately-insured (76.9%). The majority of patients had received an allogeneic stem cell transplant (63.8%) and nearly all (95.0%) were in remission. The mean time since diagnosis was 8.0 years (range = 1 - 40). Mean age was 56.8 years (Standard Deviation = 12.4). Based on the DCE, CR was identified by patients to be the most important attribute (Preference estimates, 10% increase in CR = 1.05, Standard Error (SE) = 0.06), followed by severity of long-term side effects (1-step increase = -0.52, SE = 0.05), event-free survival (6 months increase = 0.29, SE = 0.02), severity of short-term side effects (1-step increase = -0.33, SE = 0.04), and time in hospital (1 month increase = -0.12, SE = 0.03). These preference estimates suggest that patients were willing to accept a 1-step increase in severity of short-term side effects for a 3% increase in chance at CR. Patients would be willing to accept a 5% decrease in the chance at CR for a 1-step decrease in long-term side effects. Patients equivalently valued an additional 3 more months in the hospital to a 3% increase in chance of CR. They valued a 10% increase in chance of CR as approximately the same as 22 months of event-free survival. No differences were seen between patients currently in remission and those not in remission. Patients found the questions easy to understand (79%), easy to answer (68%), and relevant to them (72%).

Conclusions: In addition to demonstrating the feasibility of a DCE in eliciting patient preferences and creating the first large dataset of its kind, this study illustrates that AML patients are willing to make trade-offs regarding treatment outcomes. While patients most highly value the chance at CR, they were willing to forgo small increases in the probability of remission to improve other outcomes, especially long-term side effects. Demonstrating patients' willingness to make such trade-offs becomes critical as incremental gains in CR with novel treatments may come with significantly increased toxicities. These data will be important to inform shared decision making, drug development and approval. Prospective studies using DCE may be helpful to guide individual treatment recommendations.

Figure.

View largeDownload slide

Figure.

View largeDownload slide

Close modal