CA 19-9 has been used with questionable accuracy to aid diagnosis of cholangiocarcinoma complicating primary sclerosing cholangitis. We aimed to characterize the test properties of CA 19-9 and of a change in CA 19-9 over time in predicting cholangiocarcinoma. Charts of 208 patients were reviewed. Fourteen patients had cholangiocarcinoma. Median CA 19-9 was higher with cholangiocarcinoma (15 vs. 290 U/ml, p < 0.0001). A cutoff of 129 U/ml provided: sensitivity 78.6%, specificity 98.5%, adjusted positive predictive value 56.6% and negative predictive value 99.4%. The median change over time was 664 U/ml in cholangiocarcinoma compared to 6.7 U/ml in primary sclerosing cholangitis alone (p < 0.0001). A cutoff of 63.2 U/ml for change in CA 19-9 provided: sensitivity 90%, specificity 98% and positive predictive value 42%. Only 2 patients with cholangiocarcinoma were the candidates for curative therapy. In conclusion, the positive predictive value of an elevated CA 19-9 was 56.6%; only advanced cases were detected by this method.