Long-term predictive value of serum PSA values obtained in clinical practice: Results from the Norwegian Prostate Cancer Consortium (NPCC).

Abstract

5021 Background: A baseline prostate-specific antigen (PSA) level in midlife is strongly associated with risk of lethal prostate cancer several decades later. In most prior studies, PSA was measured in archived blood from men participating in cohort studies unrelated to prostate cancer. In clinical practice, however, men commonly obtain a PSA test because of lower urinary tract symptoms. Here, we study the association between PSA values when a first PSA test was obtained as part of clinical care – and subsequent risks of prostate cancer diagnosis and death. Methods: The NPCC database contained laboratory results from 1995-2006 of the first (“baseline”) PSA test below 4.0 ng/mL in 157,836 men aged 40-69 years without prior prostate cancer (PCa) diagnosis. PSA strata were pre-defined: < 1, 1-1.99, 2-2.99, 3-3.99 ng/ml and analyses were split in two age groups based on a prior risk: 40-54 (“younger”) and 55-69 (“older”). Predictive accuracy was assessed with Harrell’s concordance (C) index. The Kaplan-Meier method was used to visualize risk of PCa death as reported on the death certificate by PSA strata for the two age groups. Results: Among 157,836 men with a PSA below 4.0 ng/mL, 83% had < 2.0 ng/mL. Risk of prostate cancer and death from prostate cancer were strongly associated with PSA-levels and age: Harrell’s C-index 0.75 and 0.72 for diagnosis of younger and older group, respectively and 0.72 for PCa death for both age groups. In the younger age group 196 deaths among 75,446 men (0.26%) were registered, in the older 943 deaths among 82,390 men (1.14%). The risk of PCa death in lowest (< 1) and highest (3-3.99) strata were 0.23% and 3.22%, respectively (189 deaths among 81,982 vs. 274 deaths among 8517 men). Conclusions: This population-based NPCC study shows as strong predictive impact of the different strata of the initial PSA measurement in routine clinical care as prior studies based on cryopreserved blood samples. [Table: see text]