Establishment of age-related prostate-specific antigen (PSA) reference limits in the Algerian population: A...

Age and ethnic related reference limits (Rls) for prostate-specific antigen (PSA), improve its diagnostic accuracy as a screening tool for prostate cancer (PCa). But, their establishment by clinical laboratories is seldom achieved due to the expense of the gold standard procedure. We used an alternative approach to produce our specific RIs.

Laboratory results over a 4-year period (1491 results of total PSA and 989 results of free PSA) were retrieved from our laboratory information system. An inclusion/exclusion process was applied. To estimate age-related 95th percentiles, Hoffmann, Bhattacharya and maximum likelihood methods followed by a linear regression model were applied based on mixed dataset and a quantile regression with restrictive cubic splines model were applied based on truncated dataset. To verify these age related RIs, an agreement analysis were used with respect to the percentage of free PSA (%fPSA) 26.4% positives as a surrogate for PCa.

The PSA 95th percentiles using Hoffmann, Bhattacharya, maximum likelihood and quantile regression methods respectively were 3.0, 2.6, 2.3 and 2.8 ng/mL for men 40 to 49 years old; 4.1, 3.3, 2.8 and 3.2 ng/mL for men 50 to 59 years old; 5.6, 4.2, 3.4 and 3.5 ng/mL for men 60-69 years old; 7.5, 5.4, 4.1 and 3.9 ng/mL for men 70-79 years old. The Bhattacharya, maximum likelihood and quantile regression age-related Rls reflected better % fPSA positives.

Our findings suggest that age related total PSA reference limits may be derived by retrospective analysis of intra-laboratory data using the quantile regression model or the Bhattacharya and maximum likelihood method, followed by a linear regression model.