In the present study, we assessed the relationship of serum insulin levels and three surrogate markers of recurrence, T stage, PSA, and Gleason score, in men with localized prostate cancer. Participants in our study were found through urology and radiation oncology clinics, and all eligible patients were asked to take part. All patients were asymptomatic and had been initially diagnosed on the basis of rising PSA or abnormal physical examination. Histological confirmation of diagnosis was obtained for all subjects. Serum insulin levels were determined by chemoluminescent assay with a standard, commercially available instrument. Patients were divided into three previously defined risk groups: Low risk: PSA < or =10, stage < or =T2a, or Gleason grade < or =6. Medium risk: 10 <PSA < or =15, Gleason 7 or stage T2b. High risk: Gleason >7, tumour in seminal vesicle biopsy, PSA >15 or stage T2c or T3. One hundred and sixty-three men with prostate cancer were studied. There was a significant increase in serum insulin with risk group (P=0.003, one way anova). Tukey's multiple range test showed that the insulin levels of high risk patients were significantly higher than the insulin levels of medium and low risk patients (P=0.05) but the insulin levels of medium and low risk patients were not significantly different from one another. Multivariate linear regression, with insulin as the dependent variable, Gleason score, PSA, and T stage (T1, T2, T3) as the independent variables, was significant overall (P<0.001, r(2)=0.120). Increased T stage was independently correlated with increased serum insulin levels (P<0.001). Gleason score was negatively, insignificantly correlated with serum insulin level (P=0.059). The positive correlation of PSA and insulin level was not significant (P=0.097). To assure normal distribution of insulin and PSA values, the regression was repeated with log (insulin) as the dependent variable, log (PSA), T stage (T1, T2, T3), and Gleason score as independent variables. The regression was significant overall (P=0.002, r(2) =0.095). Increased T stage was independently correlated with increased log (insulin level) (P=0.026). Gleason score was negatively, insignificantly correlated with log (insulin) level (P=0.728). The positive correlation of log (PSA) and log (insulin) levels was significant (P=0.010). The relationship between increased insulin level and advanced tumour stage in prostate cancer we describe here is biologically quite plausible, since insulin is a growth factor. Further studies may document whether serum insulin levels might be a useful biomarker of prostate cancer stage.