Validation of prognostic models for predicting postsurgical outcomes in intrahepatic cholangiocarcinoma patients...

Few studies to date have externally validated and directly compared conventional prognostic models that predict postsurgical outcomes in patients with intrahepatic cholangiocarcinoma (ICCA). We aimed to validate the performance of prognostic models predicting postsurgical outcomes in a multicenter cohort of patients with ICCA.

Consecutive patients with ICCA who underwent curative-intent hepatic resection for ICCA at six tertiary referral institutions between January 2009 and December 2016 were retrospectively analyzed. The predictive abilities of the American Joint Committee on Cancer TNM 8th edition, Wang nomogram, Hyder nomogram, MEGNA score, and Tsilimigras prescore and postscore models were assessed based on preoperative and postoperative clinical, pathological, and imaging data. The ability of the six prognostic models to predict early recurrence (within 1-year) and 5-year overall survival (OS) was evaluated, including their discrimination and calibration capabilities.

Among the 333 patients included (mean age [standard deviation], 62.4 [9.8] years; 206 [61.9%] men), 142 (42.6%) experienced early recurrence and 225 (67.6%) died during a median follow-up of 45.1 months. In predicting early recurrence, the Tsilimigras postscore model showed the highest area under the receiver operating characteristic curve (0.811), followed by the Wang nomogram (0.780). In predicting 5-year OS, the Wang nomogram had the highest concordance index (0.704), followed by the Tsilimigras postscore model (0.675). All six models showed good calibration, with strong agreement between predicted and observed outcomes. Risk stratification based on these models effectively differentiated OS at 1, 3, and 5 years post-surgery ( P ≤ 0.005).

The Tsilimigras postscore model and the Wang nomogram were optimal in predicting early recurrence and long-term survival, respectively, in ICCA patients who underwent curative-intent resection. These findings may help select the most appropriate prognostic model for predicting postsurgical outcomes in ICCA patients.