Proposal for Risk-Adjusted Surveillance Strategies in Extrahepatic Bile Duct Cancer with Cost-Effectiveness Validation: A Real-World Multi-institutional Cohort Analysis.

PURPOSE: We developed a risk-adjusted surveillance strategy incorporating prognostic factors for disease-free survival (DFS) in patients with extrahepatic bile duct cancer after curative resection and lymph node dissection.

METHODS AND MATERIALS: After ethical approval with informed consent waived, this retrospective analysis included 1,443 patients with extrahepatic bile duct cancer treated between 2000 and 2015 (median follow-up, 35.9 months). DFS, defined as the time from surgery to recurrence or death, was analyzed using multivariate Cox proportional hazards models to identify prognostic factors. A random survival forest model adjusted these factors and generated risk-adjusted survival curves stratified by radiation therapy (RT) status. Surveillance visit points were determined at intervals corresponding to a 7% recurrence or death risk increase.

RESULTS

Adverse prognostic factors for DFS included pT3-4 stage (hazard ratio [HR], 1.220; P = .005), pN1 stage (HR, 2.208; P < .001), margin involvement by invasive carcinoma (HR, 1.670; P < .001), moderate (HR, 1.342; P = .002) or poor (HR, 1.870; P < .001) tumor differentiation, and perineural invasion (HR, 1.573; P < .001). Favorable outcomes were observed in patients with distal tumors compared with those with perihilar tumors (HR, 0.773; P < .001) and in those receiving concurrent chemoradiation therapy followed by maintenance chemotherapy (HR, 0.549; P < .001). These factors were integrated into the random survival forest model to generate risk-adjusted survival curves tailored to the RT status. The resulting surveillance strategy, based on a 7% incremental risk, included 8 follow-up visits for RT(-) patients and 9 visits for RT(+) patients over 5 years, with no visits scheduled in year 5 and detection delays of 1.9 and 1.3 months, respectively. This approach yielded incremental cost-effectiveness ratios of $41,227 and $37,469 per quality-adjusted life year for the RT(-) and RT(+) groups, respectively.

CONCLUSIONS

A risk-adjusted survival model tailored to prognostic factors and RT status enables individualized surveillance, improves recurrence detection, and reduces unnecessary visits, supporting efficient and cost-effective survivorship care.