Expression of the free beta-subunit of human chorionic gonadotropin (hCGbeta) in malignant tumors is frequently associated with aggressive disease. We have shown previously that the pretreatment serum concentration of hCGbeta is an independent prognostic variable in patients with renal cell carcinoma (RCC). We now compare the serum levels with the expression of hCGbeta antigen and mRNA in tumor tissue and studied whether these are associated with the clinical outcome. Serum samples were collected before surgery from patients with RCC (n = 256) and from 84 apparently healthy controls. HCGbeta in serum was measured by a time-resolved immunofluorometric assay. Tissue expression was detected by immunohistochemical staining of a tissue microarray (TMA) comprising 229 samples, and in selected cases by reverse transcription polymerase chain reaction (RT-PCR) of hCGbeta mRNA (n = 20) from tumor tissue. The prognostic value of hCGbeta in serum and tissue and the association with usual clinicopathological variables was analyzed by the Kaplan-Meier method, the log-rank test, Cox multiple hazard regression, Mann-Whitney U-test or Kruskal-Wallis test. The serum concentrations of hCGbeta were increased in 27% of the RCC patients and patients with increased hCGbeta levels had significantly shorter survival time than those with levels below the median (cut-off 1.2 pmol l(-1), p = 0.0044). HCGbeta antigen was detected in 15% (35 of 229) of the tumors by immunohistochemistry, and hCGbeta mRNA in 8 of 20 samples (40%) by RT-PCR. Tissue positivity for hCGbeta antigen was not associated significantly with mRNA expression, serum concentrations of hCGbeta or survival. In multivariate analysis tumor stage, grade, size and serum hCGbeta were independent prognostic variables. The serum concentration of hCGbeta is an independent prognostic variable in RCC. Tissue expression of hCGbeta detected by immunohistochemistry occurs in 15% of RCCs but it is not significantly associated with prognosis. Expression at the mRNA level seems to be associated with other predictors of adverse outcome.