Importance: The prognostic value of lymph node (LN) assessment after liver resection for hilar cholangiocarcinoma (HC) is still controversial, and the number of LNs required to be removed to obtain adequate staging is not well defined. Objectives: To evaluate the LN status in patients after liver resection for HC and to clarify which prognostic factor (the number of positive LNs or the LN ratio [LNR]) was most accurate for staging and what minimum number of retrieved LNs was required for adequate staging. Design, Setting, and Participants: Retrospective multicenter study of patients who underwent resection for HC between January 1, 1992, and December 31, 2007, at 8 hepatobiliary Italian centers. The last follow-up was assessed in July 2014. Main Outcome and Measures: Differences in overall survival (OS) according to the LN status were analyzed. The OS results were defined as actual because all included patients completed a 5-year follow-up. Results: One-hundred seventy-five patients with 1133 retrieved LNs were analyzed. The mean (SD) age of the cohort was 63 (10) years, and 42.9%(75 of 175) were female. The median number of LNs examined per patient was 6.5. Forty percent (70 of 175) had LN metastasis. An LNR exceeding 0.20 was associated with significantly lower 5-year OS than an LNR of 0.20 or less (10.6%vs 24.4%; odds ratio, 2.434; 95%CI, 1.020-5.810; P = .04). On multivariable analysis, the LNR was the only independent prognostic factor for OS but was influenced by the total number of retrieved LNs. The LNR was greater than 0.20 in all patients (30 of 30) with 1 to 4 retrieved LNs and in 52.5%(21 of 40) of patients with at least 5 retrieved LNs. Five-year OS in patients with 1 to 5 retrieved LNs was significantly lower than that in those with 6 to 7 retrieved LNs and those with at least 8 retrieved LNs (34.2%, 64.5%, and 62.7%, respectively; P = .047). Five-year OS did not significantly improve when the number of retrieved LNs was greater than 6. These results were confirmed in a receiver operating characteristic curve analysis performed among N0R0 patients, in whom 5 retrieved LNs was the most accurate cutoff to predict 5-year actual OS (area under the curve, 0.624; P = .004). Conclusions and Relevance: An LNR exceeding 0.20 was the only independent prognostic factor for OS in N1 patients after liver resection for HC. However, the LNR was influenced by the total number of retrieved LNs, and removal of more than 5 LNs was the minimum number of LNs required for adequate staging.