EP453 - Leveraging AI and Expert Review to Develop a Patient-Centered Decision Aid for Rectal Cancer Treatment

Purpose/Background: The treatment of rectal cancer has undergone a rapid, dramatic change in the last decade. Current standard of care for patients with stage II/III rectal cancer starts with total neoadjuvant therapy (TNT). Following completion of TNT, patients may be recommended total mesorectal excision (TME) or a watch-and-wait (WW) approach based on their clinical tumor response. This change in treatment paradigm from up-front removal of cancer to primarily medical therapy with the possibility of delayed surgery leaves patients and providers wondering what the future might hold.

Recent works have shown that complete pathologic response (pCR) for rectal adenocarcinoma after TNT can range from 22-38%. This has led to excitement for non-operative approaches in some patients, with over half of patients ultimately avoiding surgery with optimal TNT. Given that the indicated operation for an individual without a complete clinical response could be an abdominoperineal resection with permanent colostomy, patients often cling to the belief that they will be in the cohort who will not need surgery. In this difficult to predict, amorphous clinical landscape, shared decision-making tools have been shown to be of clinical benefit. These tools have been shown to reduce decisional conflict and improve patient understanding of the care they receive. The aim of this study was to create a shared decision making tool for the treatment of rectal cancer.

Methods/Interventions: A patient decision making tool was created to aid conversations regarding stage II or stage III rectal adenocarcinoma (Figure 1). This tool was modified with help from generative AI and Open Evidence from the published rectal cancer decision making tool from Wu et al. published in Diseases of the Colon and Rectum in 2016. The prior version of this decision-making tool did not account for a watch-and-wait approach. Revised, AI generated decision making tools were evaluated by 5 colorectal surgeons who served as expert reviewers to determine construct validity using a brief, 12-question questionnaire (Figure 2).

Results/Outcomes: The patient decision making tool was assessed by 7 colorectal surgeons who had between < 5 and >20 years of clinical experience. 100% of attendings surveyed agreed or strongly agreed that the tool added value to their clinical practice, improved consistency between providers, and supported shared decision making with patients. They also all agreed or strongly agreed that it aligned with current guidelines, could be used within routine visit time constraints, and that they would recommend this tool to colleagues. 71% of surveyed respondents also agreed or strongly that the required inputs were readily available in the patient chart or during the visit.

Conclusion/Discussion: Numerous systematic reviews have shown that decision making tools improve knowledge and foster realistic expectations, especially regarding possible surgical interventions. With the rapid evolution of rectal cancer care, patients are left adrift in a sea of information and treatment options. We present one method for decision making tool creation utilizing generative AI and expert review patient. In the next step of our study, we will prospectively assess the utility and benefit of this tool in newly diagnosed rectal cancer patients using the patient-facing clinical validity questionnaire (Figure 3). Our eventual goal is wide-spread implementation of this tool to reduce clinical uncertainty and decision-making anxiety for our rectal cancer patients.

Rectal Cancer Diagnosis and Treatment Pathway
Overview of the treatment options for rectal cancer based on tumor size and location, and brief explanations of the advantages and disadvantages of surgery vs. "watch and wait" approaches.
Qualifying for Watch and Wait
Pathway to guide patients through the "watch and wait" process including recurrence rates and likelihood of needing surgery. This portion of the tool can be used in clinic to chart the patient's progress and current status.