OBJECTIVES: Acute myeloid leukemia (AML) is a progressive blood cancer with few treatment options. We sought to estimate heterogeneity in treatment preferences of patients with AML and to inform a community led patient-focused drug development initiative.

METHODS: The study was guided by two advisory committees of patients, caregivers, and AML experts. We designed a paired-comparison DCE across five attributes (event free survival (EFS), complete remission (CR), length of hospital stay, and severity of short-term side effects and long-term side effects. We used a three-block D-efficient experimental design with 9 choice tasks per block. Results were analyzed using latent class models with continuously coded attribute levels. Logistic regression was used to identify factors related to starting the DCE and to describe latent classes.

RESULTS: 322 patients with AML participated in the study (36% response rate). 28 people (9%) did not start the DCE. They were less likely to have experienced an allogeneic transplant (p = 0.03) and diarrhea (p=0.03). Latent class analysis identified two groups. Class 1 (37%) assigned most importance to a 10% increase in CR (preference estimate: 3.1) compared to other attributes (preference estimates between -.6 and 0.3). Class 2 (63%) assigned relatively less importance to CR (preference estimate: 0.7) compared to the other attributes (preference estimates between -.7 and 0.3). Participants in Class 1 were more likely to (strongly) agree that they had self-control (p=0.01), to have suffered from fatigue (p = 0.02) or organ failure (p<.01), and to have had an allogeneic transplant (p<0.01).

CONCLUSIONS: The identified preference heterogeneity highlights the need to consider attributes in addition to CR in AML treatment decisions. Those who did and did not start the DCE were mostly similar in observed characteristics suggesting that the preferences of people that did not complete the DCE might still have been reflected.