OBJECTIVES: To provide a simple and reliable clinical prediction for an individual patient's overall risk of cancer at biopsy by deriving an easily implemented test based on a generalizable model. Four variables are analyzed for inclusion in the model: prostate-specific antigen (PSA) level, digital rectal examination (DRE) results, race, and age. METHODS: Two populations were used to develop and validate the test: a model (n = 633) and an independent, geographically separate, external population (n = 766). Pathology records for patients who underwent prostate biopsy between 1991 and 1995 were reviewed and screened for the presence of PSA and DRE results. Records where age and race could be determined were extracted. Multiple logistic regression was used with an iterative approach to optimize each test factor. The Wald chi-square test, receiver operating characteristic (ROC) curve, and Hosmer-Lemingshaw test were used to evaluate the model's predictive capability in the two populations. RESULTS: The model and external populations were significantly different for racial mix, PSA level, age, and biopsy detection rate, providing diverse populations to validate the test. Within a combined model, PSA, DRE, race, and age all demonstrated independent capability to predict cancer at biopsy. Predictive power of the overall test was high within the model population (ROC 80.8%), with minimal loss of power in the external population. The test demonstrated no significant lack of fit in either population. CONCLUSIONS: Within a combined test, PSA, DRE, race, and age all contribute significantly to prediction of prostate cancer at biopsy in an individual patient. The test depicts individual risk in an easily understood, visually provocative manner and should assist the clinician and patient in reaching a decision as to whether biopsy is appropriate.
OBJECTIVES: To provide a simple and reliable clinical prediction for an individual patient's overall risk of cancer at biopsy by deriving an easily implemented test based on a generalizable model. Four variables are analyzed for inclusion in the model: prostate-specific antigen (PSA) level, digital rectal examination (DRE) results, race, and age. METHODS: Two populations were used to develop and validate the test: a model (n = 633) and an independent, geographically separate, external population (n = 766). Pathology records for patients who underwent prostate biopsy between 1991 and 1995 were reviewed and screened for the presence of PSA and DRE results. Records where age and race could be determined were extracted. Multiple logistic regression was used with an iterative approach to optimize each test factor. The Wald chi-square test, receiver operating characteristic (ROC) curve, and Hosmer-Lemingshaw test were used to evaluate the model's predictive capability in the two populations. RESULTS: The model and external populations were significantly different for racial mix, PSA level, age, and biopsy detection rate, providing diverse populations to validate the test. Within a combined model, PSA, DRE, race, and age all demonstrated independent capability to predict cancer at biopsy. Predictive power of the overall test was high within the model population (ROC 80.8%), with minimal loss of power in the external population. The test demonstrated no significant lack of fit in either population. CONCLUSIONS: Within a combined test, PSA, DRE, race, and age all contribute significantly to prediction of prostate cancer at biopsy in an individual patient. The test depicts individual risk in an easily understood, visually provocative manner and should assist the clinician and patient in reaching a decision as to whether biopsy is appropriate.
Authors: Paul J Newcombe; Brian H Reck; Jielin Sun; Greg T Platek; Claudio Verzilli; A Karim Kader; Seong-Tae Kim; Fang-Chi Hsu; Zheng Zhang; S Lilly Zheng; Vincent E Mooser; Lynn D Condreay; Colin F Spraggs; John C Whittaker; Roger S Rittmaster; Jianfeng Xu Journal: Genet Epidemiol Date: 2012-01 Impact factor: 2.135
Authors: David J Hernandez; Misop Han; Elizabeth B Humphreys; Leslie A Mangold; Samir S Taneja; Stacy J Childs; Georg Bartsch; Alan W Partin Journal: BJU Int Date: 2008-10-24 Impact factor: 5.588