A Novel Approach to Computing Preference Estimates for Different Treatment Pathways: An Application in Oncology.

Abstract

Patients with cancer may progress through multiple treatments with differing adverse effect profiles. Moreover, pathways may be fixed or flexible in allowing for escalation or de-escalation of treatment depending on interim outcomes. We sought to develop a methodology capable of estimating preferences for the entirety of a pathway involving a sequence of different treatments. Patients with early breast cancer completed an online discrete choice experiment to assess preferences for eight key early breast cancer attributes. Hierarchical Bayesian modeling was used to calculate attribute-level preference weights. Preference weights for hypothetical pathways were estimated by summing the respective weights for efficacy, flexible or fixed pathway, duration, administration regimen, and adverse event risk, the last two of which were time-adjusted by multiplying each weight by the proportion of time spent on a selected treatment. Increases in the risk of a serious adverse event were most influential in treatment pathway preferences, followed by increases in efficacy and decreases in overall pathway duration. Patients preferred a flexible pathway versus a fixed pathway. Pathway preference estimates fluctuated in a logically consistent manner. Switching from a flexible to a fixed pathway yielded a significantly lower pathway preference. For this same pathway, when adjuvant treatment was replaced with a treatment with a more favorable toxicity profile and shorter duration, it offset the negative impact of the more toxic neoadjuvant chemotherapy. This novel methodology accounts for patient preference throughout a sequence of treatments, allowing for comparison of preferences across complex treatment pathways.