Predicting the Outcome of the Random Prostate Biopsy R. Porter, Christopher Authors: Christopher R. Porter, Colin O'Donnell, E. David Crawford, Eduard J. Gamito, Jason Kim and Ashutosh Tewari Prostate Cancer affects one in eight American men, a rate that is remarkably similar to the breast cancer rate in women. Fortunately, men can be screened for prostate cancer, which is currently listed as the second leading cause of cancer related deaths. Screening for prostate cancer requires a physical examination, including a digital rectal exam of the prostate (DRE), and a blood test to look for an elevated Prostate Specific Antigen (PSA). While there is still some controversy as to the possible benefits of prostate cancer screening, recent death rates for the disease have shown a significant decline since the time of screening (1). At the present time both the American Medical Association (AMA) and the American Urologic Association (AUA) recommend prostate screening. Prostate screening can produce significant anxiety among men. This is due, in part, to the uncertain nature of the outcome. While abnormalities in the physical exam (DRE) and the PSA may tell the physician that more tests need to be done to look for prostate cancer, their ability to predict the chance of a patient actually having prostate cancer is limited. The authors recognize the level of patient anxiety and understand the pitfalls of trying to give the patient a sense of his particular chance of having prostate cancer at the time of biopsy. For this reason the authors have collected information from over 300 men undergoing prostate biopsy. Our goal was to construct a mathematical model that can offer the patient an accurate estimate of his chances of having a positive prostate biopsy. For the physician to diagnose prostate cancer a biopsy of the prostate must be performed. This test is done on patients that have either an abnormal PSA or DRE, or on patients that have had previous biopsies, with findings that lead the physician to recommend another biopsy. Biopsy of the prostate requires guidance by transrectal ultrasound (TRUS). Recently it has been found that between 10 and 12 biopsy samples should be taken per patient, so that the doctor has the best chance of finding the cancer. Despite taking more samples at the time of biopsy it is estimated that about 30% of the time cancer is missed on the first biopsy (2). This means that many men will require second prostate biopsies, at the discretion of their physician, and further patient anxiety may result. In fact, it has been estimated that more than one million prostate biopsies are performed annually in an attempt to diagnose prostate cancer. The authors performed 319 biopsies on men with either elevated PSA or abnormal examinations. The patients represented a racially diverse group (78% African American), and consisted of 106 men (33%) who had prior biopsies. Ninety-two percent (92%) of men had 10-12 biopsies performed. The average PSA among the men was 12.6ng/ml and 39% of men had abnormal rectal exams. Overall the positive biopsy rate was 39%. The authors looked at many different possible parameters in 319 men that could be a factor in the patient having a positive biopsy, these are listed in Table 1. PSA was evaluated with respect to the amount of PSA that was present as a free form, Free PSA (not bound to a carrier protein in the blood), and as PSA Density, which is a way of examining the relative level of PSA with respect to the size (volume) of the prostate. Table 1: Pre-biopsy Parameters Prospectively Examined. ------------------------------------------------------------------------------- Age | PSA | ------------------------------------------------------------------------------- Race | Free PSA | ------------------------------------------------------------------------------- Family History of Prostate Cancer | PSA Density | ------------------------------------------------------------------------------- AUA Symptom Score* | DRE | ------------------------------------------------------------------------------- Prostate Biopsy History | Ultrasound Findings (suspicious lesion) | ------------------------------------------------------------------------------- Potency | Calcium on Ultrasound | ------------------------------------------------------------------------------- * AUA, American Urological Symptom Score After analysis the most important factors that were associated with positive biopsy (prostate cancer) were PSA, prostate gland volume, DRE, ultrasound findings, and a history of prior biopsy. These factors were then used in a mathematical model to predict the likelihood of an individual having biopsy proven prostate cancer. An Artificial Neural Network (ANN) was constructed to predict prostate biopsy outcome. Artificial Neural Networks are computerized (software) models that are based loosely on the human brain's decision-making mechanism. ANN's can actually learn from experience and can solve complex problems. For this reason they have been used by the military and engineering for complex problem solving. They are recognized as a powerful analytic tool and have been endorsed for use in Prostate cancer by the American Joint Committee on Cancer (3). The ANN was constructed using the same five important factors (PSA, DRE, prostate volume, TRUS findings, history of prior biopsy). The ANN was tested (validated) against an independent group of patients that were selected randomly and set aside for testing purposes; the ANN predicted the outcome of a positive biopsy with an accuracy of 77% (Area Under the Receiver Operator Curve, 0.77). It is anticipated that patients and physicians alike will benefit from these types of predictive models. Further research is underway to refine the mathematical models with hope of improving their accuracy. It is anticipated that these types of models will be available on the World Wide Web for physician and patient use (4). Summary One in eight men are at risk for developing prostate cancer and many more will require biopsy to look for the disease. We have carefully looked at more than 300 men undergoing prostate biopsy, and have developed a mathematical model to predict an individual's chance of having a prostate biopsy. This model has been shown to be accurate 77% of the time. It is hoped that models such as these may benefit the patient and physician in decisions regarding prostate cancer evaluation. INSTITUTIONS: 1. Department of Urology, State University of New York, Stony Brook, NY. Christopher R. Porter, Jason Kim 2. ANNs in CaP Project, Denver, CO. Colin O'Donnell, E. David Crawford, Eduard J. Gamito 3. Department of Urology, Josephine Cancer Center, Detroit, MI. Ashutosh Tewari SUPPORT: Funded by the Institute for Clinical Research, at the Veterans Affairs Medical Center, Washington, DC. Additional Authors: Works Cited: 1. Data Source: US Mortality Public Use Data Tapes 1960-1997, US Mortality Volumes 1930-1959, National Center for Health Statistics, Centers for Disease Control and Prevention, 2000 2. Rabbani F, Stroumbakis N, Kava B, Cookson M and Fair W. Incidence and Significance of False- Negative Sextant Prostate Biopsies. J Urology, 159, 1247-1250, 1998. 3. Iczkowski K A, Bostwick D G, Prostate Biopsy 1999: Strategies and Significance of Pathological Findings. Seminars in Urologic Oncology, 17, 4, 177-186, 1999. 4. www.prostatecalculator.org Article Links: