OBJECTIVE: Previous studies have shown a relationship between serum prostate-specific antigen (PSA) level and prostate tumour volume. Reports based on selected case series have also indicated that serum PSA may be used for staging, although a varying prevalence of metastasizing tumours complicates the interpretation of these studies. In order to determine the accuracy of the serum level of PSA in predicting the presence of metastases we performed a prospective cohort study of a geographically defined population of men with prostate cancer.
METHODS: Serum level of PSA and the results of investigations for regional lymph node and distant metastases were recorded for all 8328 men with prostate cancer registered in the Swedish National Prostate Cancer Register 1996-1997.
RESULTS: The prevalence of lymph node metastases among men who had undergone lymph node exploration was 4%, 16% and 33% for well, moderately and poorly differentiated tumours. The corresponding prevalence of distant metastases was 12%, 30% and 48%. With serum PSA <20 ng/ml as a cut-off point the negative likelihood ratios for well and moderately differentiated tumours were found to be 0.47 and 0.45 for lymph node metastases and 0.24 and 0.18 for distant metastases, resulting in post-test probabilities >92% for the exclusion of metastases. In men with poorly differentiated tumours, the negative likelihood ratio would need to be even lower to safely exclude disseminated disease.
CONCLUSION: For well to moderately differentiated tumours, further investigations to assess the presence of metastases may be omitted with no great risk for understaging if serum PSA <20 ng/ml.