To assess the usefulness of alpha-fetoprotein (AFP) in monitoring treatment effects of transcatheter arterial embolization (TAE) in hepatocellular carcinoma (HCC) patients, a total of 31 sets of AFP levels after TAE in 21 HCC patients were analysed by linear regression between logarithmic AFP levels and days. Eleven sets of AFP data with poor linear declination were accompanied with poor TAE results except in one patient who had chronic hepatitis with acute exacerbations. Twenty sets of data with good linear declination in the first month after TAE indicated good TAE results. Seven of them showed no evidence of tumour recurrence nor elevated AFP levels within a follow-up of 6 months. The mean, standard deviation and range of half-lives of AFP in the non-recurrent group were 5.0, 1.6 and 2.9-7.2 days, respectively. The others experienced late tumour recurrence that was detected by rebound of AFP levels except one who had another non-AFP-secreting HCC. Thus, the results might be used as a reference in monitoring the treatment effects of TAE and the timing selection of repeated TAE.