We defined some standardised criteria for classifying incident cases of liver cancer into either Primary Liver Cancer (PLC) or Unspecified Liver Cancer (ULC), on the basis of the diagnostic procedures performed. A pilot hospital-based study (98 cases) was carried out in Verona, northern Italy, in order to assess the feasibility of the method. The same protocol was subsequently applied in a population-based study (349 cases) in Bresica, northern Italy. The percentage of cases with histological verification was 38.7 and 41.8%, respectively, with a wide variation among different hospitals. The percentage of cases we attributed to the PLC category was 78.6% in the hospital-based study and 78.8% in the population-based study. No differences in the proportion of cases attributed to PLC were found according to patients' age, sex or hospital of admission. Repeatability of the method was assessed through a cross-panel review of 198 cases, with a 91.9% interobserver agreement. Implications of this method are discussed and some suggestions for cancer registration and future research are proposed.