Predictive MINT Pathologic Risk Score for Adjuvant Chemotherapy in Resected Cholangiocarcinoma: A Propensity Score-Matched Multicenter Study in Thailand.

Purpose: This study aims to clarify the benefit of adjuvant chemotherapy (AC) in resectable cholangiocarcinoma (CCA) and develop a predictive risk score for treatment selection.

Methods: Patients with resected CCA undergoing curative surgery, with or without AC, were identified from three centers in Thailand. Patients with R2 resection and 30 days postoperative death were excluded. Using the largest center as the discovery cohort, we generated propensity score matching (PSM). A predictive model for overall survival (OS) was identified, and a predictive risk score was developed from the PSM discovery cohort, classifying patients into high- and low-risk groups. The proposed risk score was validated in the other two centers.

Results: In the discovery cohort, 493 patients were identified. After PSM, 328 patients were categorized into surgery (n = 164) and surgery + AC (n = 164) groups. The baseline characteristics in the PSM discovery cohort were well-balanced. In the validation cohort (n = 83), patients with positive lymph node 1 received AC more frequently than those under observation (47% v 18%; P = .02). A MINT pathologic risk score was developed from multivariate analysis for OS. The score includes margin, perineural invasion, pathologic nodal status, and pathologic tumor size. In the PSM discovery cohort, for the low-risk score group, the surgery group had significantly longer OS compared with the surgery + AC group (49.4 v 31.5 months; hazard ratio [HR], 1.78 [95% CI, 1.11 to 2.86]; P = .016). Conversely, for the high-risk score group, the surgery + AC group had better OS than the surgery group (18.8 v 8 months; HR, 0.60 [95% CI, 0.46 to 0.79]; P < .001). The results were comparable in the validation cohort.

Conclusion: Patients with resected CCA with a high-risk MINT pathologic risk score were likely to benefit from AC, whereas those with a low-risk score were not. Further validation in a larger prospective cohort is warranted.