OBJECTIVE To evaluate how serum prostate-specific antigen (PSA) levels in a population-based cohort of men with prostate cancer vary with age and intensity in the diagnostic activity and to describe the treatment selection processes associated with PSA level. MATERIAL AND METHODS All men in the Swedish National Prostate Cancer Register diagnosed during 1996-1997 were included. In 1996 the register included 19 counties, covering 61% of the Swedish male population, and in 1997 21 counties with 79% of the Swedish male population. RESULTS A total of 8328 men were registered. PSA levels were missing in 341 cases. With increasing PSA there was a shift towards more advanced and poorly differentiated tumours. PSA at diagnosis increased with age, with the exception of patients younger than 50 years who had higher PSA values. The mean logarithm of PSA correlated negatively with the percentage of localized tumours (p < 0.005) and the age-adjusted incidence (p < 0.05) in each respective county in 1997. PSA was higher in men receiving radiotherapy compared with those treated with radical prostatectomy as well as in the group treated with bilateral orchiectomy compared with those receiving GnRH-analogues. CONCLUSIONS If PSA is used as a surrogate measure of extent of tumour volume in a population of prostate cancer patients, our findings indicate that age distribution and differences in incidence (possibly due to variation in diagnostic activity) should be taken into account. In our cohort there was a selection process, probably in part guided by PSA level, when choosing type of curative or palliative treatment.