Azza Shoaibi1,2, Gowtham A Rao3,4, Bo Cai3, John Rawl5, Kathlyn Sue Haddock6, James R Hébert2,3,4. 1. Department of Health Sciences, Medical University of South Carolina, Columbia, South Carolina. 2. South Carolina Statewide Cancer Prevention and Control Program, University of South Carolina, Columbia, South Carolina. 3. Department of Epidemiology and Biostatistics, Arnold School of Public Health, University of South Carolina, Columbia, South Carolina. 4. Department of Family and Preventive Medicine, School of Medicine, University of South Carolina, Columbia, South Carolina. 5. Columbia Urological Associates, P.A., Columbia, South Carolina. 6. Research Department, Veterans Affairs Medical Center, WJB Dorn VA Hospital, Columbia, South Carolina.
Abstract
PURPOSE: To investigate if a prostate specific antigen (PSA)-derived growth curve can predict the occurrence of high-risk prostate cancer (PrCA). METHODS: Data from 38,340 men randomized to the PrCA screening arm in the prostate, lung, colorectal, and ovarian cancer screening trial (PLCO) were used to develop a PSA growth curve model to estimate PSA rate of change. The model was then used to predict high-risk PrCA in clinical data available from 680,390 veterans seeking routine care. The PSA growth curve was modeled using non-linear mixed regression and the PSA rate was estimated by taking the 1st derivative of the growth curve equation at 1 year prior to diagnosis/exit. RESULTS: In the PLCO, PrCA incidence was 8.1%; ≈19% of whom had high-risk PrCA. Overall, a PSA rate threshold of 0.37 ng/ml/year had the best combination of sensitivity (97.2%) and specificity (97.3%) for detecting high-risk PrCA. In the VA data; 7,347 men were diagnosed with PrCA; of these 4,315 (58.7%) were diagnosed with high-risk PrCA. The PLCO optimal threshold of 0.37 ng/ml/year produced sensitivity = 95.5% and specificity = 85.2%. An optimal threshold of 0.99 ng/ml/year in AA produced sensitivity = 89.1% and specificity = 80.0%. PSA rate was a better predictor than the single last PSA value. CONCLUSIONS: PSA growth curves predicted high-risk PrCA in the PLCO data. Fitting the same algorithm in the VA data produced lower specificity. Although encouraging, this finding underlines the need for further research to prospectively test the algorithm, especially for African-American men, the population group at highest risk of aggressive PrCA. Prostate 77:173-184, 2017.
RCT Entities:
PURPOSE: To investigate if a prostate specific antigen (PSA)-derived growth curve can predict the occurrence of high-risk prostate cancer (PrCA). METHODS: Data from 38,340 men randomized to the PrCA screening arm in the prostate, lung, colorectal, and ovarian cancer screening trial (PLCO) were used to develop a PSA growth curve model to estimate PSA rate of change. The model was then used to predict high-risk PrCA in clinical data available from 680,390 veterans seeking routine care. The PSA growth curve was modeled using non-linear mixed regression and the PSA rate was estimated by taking the 1st derivative of the growth curve equation at 1 year prior to diagnosis/exit. RESULTS: In the PLCO, PrCA incidence was 8.1%; ≈19% of whom had high-risk PrCA. Overall, a PSA rate threshold of 0.37 ng/ml/year had the best combination of sensitivity (97.2%) and specificity (97.3%) for detecting high-risk PrCA. In the VA data; 7,347 men were diagnosed with PrCA; of these 4,315 (58.7%) were diagnosed with high-risk PrCA. The PLCO optimal threshold of 0.37 ng/ml/year produced sensitivity = 95.5% and specificity = 85.2%. An optimal threshold of 0.99 ng/ml/year in AA produced sensitivity = 89.1% and specificity = 80.0%. PSA rate was a better predictor than the single last PSA value. CONCLUSIONS:PSA growth curves predicted high-risk PrCA in the PLCO data. Fitting the same algorithm in the VA data produced lower specificity. Although encouraging, this finding underlines the need for further research to prospectively test the algorithm, especially for African-American men, the population group at highest risk of aggressive PrCA. Prostate 77:173-184, 2017.
Authors: Daria M McMahon; James B Burch; James R Hébert; James W Hardin; Jiajia Zhang; Michael D Wirth; Shawn D Youngstedt; Nitin Shivappa; Steven J Jacobsen; Bette Caan; Stephen K Van Den Eeden Journal: Ann Epidemiol Date: 2018-11-02 Impact factor: 3.797