In order to differentiate whether slight alpha-fetoprotein (AFP) increases observed in any patient are due to germ cell tumours (GCT) or to liver diseases (including hepatotoxicity of chemotherapy), we measured the binding ratio of the AFP to concanavalin A (ConA). A total of 218 serum samples were studied: 102 samples from 72 GCT patients and 116 from patients with liver diseases. Considering a cut-off value to be a ConA binding ratio of 15%, we distinguished AFP produced by GCT (> 15%) from AFP produced by tumoral and non-tumoral liver diseases (< or = 15%) with a sensitivity of 98% and specificity of 100%. The difference between mean ConA binding ratios was statistically significant (P < 0.0001). We did not distinguish AFP produced by tumoral and non-tumoral liver diseases. ConA binding ratio may be a sensitive index to distinguish whether an increase of AFP concentration as low as 15 U/ml in a GCT patient during the follow-up is produced by the tumour or by liver dysfunction (including hepatotoxicity of chemotherapy).