Objective: To evaluate the usefulness of serum alpha-fetoprotein (AFP) in detecting the recurrence of hepatocellular carcinoma (HCC) after curative hepatectomy. Methods: From January 1995 to February 2000, 283 patients who underwent curative resection of HCC had their post-operative serum AFP levels serially monitored, and the AFP levels at the time of diagnosis of recurrence were analyzed for those patients who developed recurrence. The patients were divided into those with normal (i.e. < 20 ng/ml, n = 115) and those with elevated preoperative serum AFP levels (n = 168). Results: Of the 283 patients, 151 (53.4%) developed HCC recurrence. In patients with normal preoperative serum AFP levels, the sensitivities of the postoperative serum AFP level in detecting HCC recurrence were 36.7%, 8.3% and 1.7%, respectively, using AFP cut-off levels of 20 ng/ml, 100 ng/ml and 400 ng/ml. The corresponding specificities were 86.3%, 96.1%, and 98.0%, respectively, using the three cut-off levels. In patients who had a raised preoperative serum AFP level ≥ 20 ng/ml, the sensitivities of the AFP tests were 83.1%, 55.1% and 32.6%, respectively, while the specificities were 73.4%, 86.1% and 92.7%, respectively, in detecting recurrence using the three cut-off levels. Conclusions: Serum AFP monitoring, in conjunction with imaging studies, is useful in monitoring recurrent HCC after hepatectomy. For patients with a raised preoperative serum AFP level, a postoperative serum AFP of over 20 ng/ml appears to be the best cut-off level for the detection of HCC recurrence with both satisfactory sensitivity and specificity. For patients with normal preoperative serum AFP levels, the sensitivity of postoperative serum AFP monitoring in the detection of recurrence is low, and hence the role of serum AFP monitoring is limited. Imaging studies are important for this group of patients in surveillance for recurrent HCC.