Statistical review and analysis of the relationship between serum prostate specific antigen and age

We investigated the possibility that the variations in the reported correlation coefficient (r) between age and serum prostate specific antigen (PSA) is simply a reflection of the differences in the age mix of the population studied, that is the younger the population the greater the correlation and vice versa. Also, we quantified the value of r for different age groups to evaluate finally the plausibility of its practical application on sound statistical basis.

Bivariate and multivariate regression analyses were done of all identified reports in the literature that specified the exact value of r and the percentage distribution of the population studied in 10-year age groups.

The correlation coefficient between r and the percentage of the population in different age groups was statistically significant, confirming the hypothesis that the greater the percentage of younger population in the study, the greater the value of r and vice versa. The correlation between age and PSA did change with age. It was significantly high in the fifth to sixth decades of life (range 0.5 to 0.7) [corrected] and markedly decreased in decade 7 to become low and insignificant in decade 8.

The value of the correlation between age and PSA is sufficiently high to allow for its practical application up to age 60 years. Beyond that age the correlation is so low that its practical application becomes statistically flawed. This finding seems to match the practical experience when applying age-specific PSA ranges in the diagnosis of early prostate cancer.