Prostate-Specific Antigen Trends Predict the Probability of Prostate Cancer in a Very Large U.S. Veterans Affairs Cohort

• Published in: Frontiers in oncology Vol. 8; p. 296
• Main Authors: Karnes, R Jeffrey, MacKintosh, F Roy, Morrell, Christopher H, Rawson, Lori, Sprenkle, Preston C, Kattan, Michael W, Colicchia, Michele, Neville, Thomas B
• Format: Journal Article
• Language: English
• Published: Switzerland Frontiers Media S.A 06.08.2018
• Subjects: Oncology; prostate cancer; prostate cancer screening; prostate-specific antigen; PSA; PSA trend; screening; prostate cancer; screening; PSA trend; prostate cancer biopsy; prostate cancer diagnosis; prostate cancer screening; PSA; prostate-specific antigen
• ISSN: 2234-943X; 2234-943X
• DOI: 10.3389/fonc.2018.00296

If prostate-specific antigen (PSA) trends help identify elevated prostate cancer (PCa) risk, they might provide early warning of progressing cancer for further evaluation and justify annual testing. Our objective was to determine whether PSA trends predict PCa likelihood. A biopsy cohort of 361,657 men was obtained from a Veterans Affairs database (1999-2012). PSA trends were estimated for the 310,458 men with at least 2 PSA tests prior to biopsy. Cancer tumors may grow exponentially with cells doubling periodically. We hypothesized that PSA from prostate cancer grows exponentially above a no cancer baseline. We estimated PSA trends on that basis along with five descriptive variables: last PSA before biopsy, growth rate in PSA from cancer above a baseline, PSA variability around the trend, number of PSA tests, and time span of tests. PSA variability is a new variable that measures percentage deviations of PSA tests from estimated trends with 0% variability for a smoothly increasing trend. Logistic regression models were used to estimate relationships between the probability of PCa at biopsy and the trend variables and age. All five PSA trend variables and age were significant predictors of prostate cancer at biopsy ( < 0.0001). An overall logistic regression model achieved an AUC of 0.67 for men with at least 4 tests over at least 3 years, which was a substantial improvement over a single PSA (AUC 0.58). High probability of PCa was associated with low PSA variability (smooth trends), high PSA, high growth rate, many tests over a long time-span and older age. For example, at 4.0 PSA the probability of cancer is 32% for 1 PSA test and increases to 68% for 8 tests over 7 years with smooth, fast growth (0% variability and 50% exponential growth). Our results show that smooth, fast exponential growth in PSA above a baseline predicts an increased probability of PCa. The probability increases as smooth (low variability) trends are observed for more tests over a longer time span, which makes annual testing worth considering. Worrisome PSA trends might be used to trigger further evaluation and continued monitoring of the trends-even at low PSA levels.