Impact and Implications of Regret in Longitudinal Cancer Treatment Decisions

Abstract

The purpose of this research is to determine how post-decisional regret changes optimal treatment decisions for patients suffering from early-stage cancer. Regret studies in medicine have repeatedly demonstrated that a substantial percentage of cancer patients would have made different treatment decisions after experiencing the outcomes of their chosen therapies. Representing this regret process as a mathematical model could provide insight into factors that determine where such decision instability exists, and how patients and the multidisciplinary team that works to cure their cancer might proceed with this new information. However, the behavior of certain regret dynamics needed to inform such a model are not universally agreed upon by healthcare professionals. Many previous studies have produced inconsistent results as to the frequency of regret for specific cancers, as well as the trajectory of regret longitudinally: whether regret fades, remains stable, or increases following initial treatment. This information is critical to accurately representing the temporal survivorship experience in regret models. To address these and other related regret issues, this research used a combination of a systematic review, a dynamic programming model, and an approximate dynamic programming model to investigate longitudinal decision regret trajectories for cancers in general, prostate cancer, and an increasingly common type of head-and-neck cancer, respectively. This investigation produced several important insights. The systematic review showed that regret remains stable over time, suggesting that normal affect-regulation processes have little influence on longitudinal regret in cancer. However, many of the contributing studies were designed with time horizons that did not account for late-emerging effects that could have an impact on regret. The prostate cancer model demonstrated specific and consistent areas of decision instability that surfaced as the weight of a regret factor was increased; the alignment of this volatility corresponded with the region of preferences that most patients express at the time of diagnosis. The head-and-neck cancer model examined the choice between robotic surgery and radiotherapy for initial treatment of human papillomavirus-associated oropharyngeal cancer of an unknown primary and identified several outcome variables that play a vital role in an informed clinical decision. Yet these same variables have wide-ranging reported values across the breadth of medical literature that change treatment recommendations; improper application of this data in decision-making would result in suboptimal care, increasing the likelihood of regret. These findings demonstrate the importance of expectation management for patient outcomes throughout the survivorship process, as well as the role that individual institutional experience plays in therapy selection for these diseases. This research will also contribute to the shaping and refinement of protocols for future clinical trials and other investigations into the decision regret phenomenon.