The American Joint Committee on Cancer (AJCC)/International Union Against Cancer (UICC) staging system for liver cancer is based on data exclusively derived from hepatocellular carcinoma (HCC) patients and thus may be inappropriate for patients with intrahepatic cholangiocarcinoma (ICC). We sought to empirically derive an ICC staging system from population-based data on patients with ICC. The Surveillance, Epidemiology, and End Results (SEER) database was used to identify 598 patients who underwent surgery for ICC between 1988 and 2004. The discriminative abilities of the AJCC/UICC liver cancer and two Japanese ICC staging systems were evaluated. Independent predictors of survival were identified using Cox proportional hazards models. A staging system for ICC was then derived based on these analyses. The AJCC/UICC T classification system failed to adequately stratify the T2 and T3 cohorts due to tumor size >5 cm not being a relevant prognostic factor [hazard ratio (HR) 0.97, 95% confidence interval (CI) 0.72-1.30]. In contrast, presence of multiple lesions (HR 1.42, 95% CI 1.01-2.01) or vascular invasion (HR 1.53, 95% CI 1.10-2.12) predicted adverse prognosis. Based on these findings, an ICC staging system was developed that omits tumor size. This system showed no loss of prognostic discrimination compared with the AJCC/UICC system and significant superiority over the Japanese systems. We conclude that the AJCC/UICC liver cancer staging system fails to stratify ICC patients adequately and inappropriately includes tumor size. We propose a staging system specifically developed for ICC based on number of tumors, vascular invasion, lymph node status, and presence of metastatic disease.