2157THE ASSOCIATION OF DIABETES MELLITUS AND HIGH-GRADEPROSTATE CANCER IN A MULTIETHNIC BIOPSY SERIES.

Daniel Moreira*, Tiffany Anderson, Leah Gerber, Jean-AlfredThomas, Madeline McKeever, Lionel Banez, JayakrishnanJayachandran, Stephen Freedland, Durham, NC

INTRODUCTION AND OBJECTIVES: Diabetes mellitus (DM)and prostate cancer are major public health problems. The associationbetween these two common entities has been the subject of muchinvestigation. Previously, we found diabetes at the time of radicalprostatectomy was associated with worse outcomes for white-obesemen, but better outcomes among black men and non-obese white men.The association of DM with prostate cancer grade has only beenevaluated in a limited number of studies that showed mixed results. Weanalyzed the association between history of DM with risk of prostatecancer and cancer grade among men undergoing prostate biopsy andanalyzed how obesity and race modify these associations.

METHODS: Retrospective analysis of 1,031 subjects from theDurham VA undergoing first prostate biopsy between 1994 and 2008with complete data available. History of DM was determined by chartreview. Patients’ characteristics at biopsy were analyzed with chi-square and ranksum. Multivariable analyses of DM and risk of prostatecancer and cancer grade (�7 vs. �7) were done with logistic regres-sion adjusting for preoperative PSA, body mass index (BMI), race, age,year of biopsy and digital rectal exam.

RESULTS: At the time of biopsy, 290 (28%) men had DM.History of DM was associated with black race (P�0.006) and higherBMI (P�0.001). Diabetes was not associated with prostate cancer riskon either univariate (P�0.682) or multivariate analysis (P�0.596).Similar results were found after stratification by race and obesity. Inmultivariable analysis, DM was significantly associated with greater riskof high-grade disease (RR�2.15, 95%CI�1.14-4.07, P�0.018). Theassociation was stronger among obese men (RR�3.76, 95%CI�1.30-10.92, P�0.015) and null in non-obese subjects (RR�1.34,95%CI�0.57-3.17, P�0.498). After further stratification by race, DMwas associated with high-grade disease only in obese white men(RR�4.82, 95%CI�1.17-19.9, P�0.029) but not in obese black men(RR�2.83, 95%CI�0.54-14.9, P�0.218). DM was not associated withrisk of low-grade disease in all men (RR�1.01, 95%CI�0.75-1.38,P�0.902) or after stratification by obesity or race.

CONCLUSIONS: Among men undergoing prostate biopsy, his-tory of DM was significantly associated with greater risk of high-gradedisease. The association was strongest among obese white subjectssuggesting the effect of DM on high-grade prostate cancer is modifiedby race and obesity, which is in agreement with our previous study ofDM and biochemical recurrence after radical prostatectomy.

Source of Funding: Department of Defense, Prostate CancerResearch Program, (SJF), and the American UrologicalAssociation Foundation/Astellas Rising Star in Urology Award(SJF).

2158COMBINING TMPRSS2:ERG AND PCA3 IN POST-DRE URINEWITH SERUM PSA TO IDENTIFY PROSTATE BIOPSYCANDIDATES: DEVELOPMENT AND VERIFICATION OF ACLINICALLY PRACTICAL ALGORITHM

Scott Tomlins, Ann Arbor, MI; Martin Sanda*, Simpa Salami, Boston,MA; Bharathi Laxman, Ann Arbor, MI; Meredith Regan, Boston, MA;Javed Siddiqui, Ann Arbor, MI; David Rickman, Douglas Scherr,Mark Rubin, New York, NY; John Wei, Arul Chinnaiyan, Ann Arbor,MI

INTRODUCTION AND OBJECTIVES: To develop a clinicallyuseful algorithm combining serum PSA with TMPRSS2:ERG fusion(T-erg) and PCA3 detected in urine collected after DRE to predictprostate cancer on subsequent biopsy.

METHODS: Serum PSA and post-DRE urine was collectedbefore prostate biopsy in a prospective multicenter EDRN cohort(N�353). In a pilot cohort (N�48) PCA3 and T-erg fusion were de-tected by RT-PCR of urine sediment; their performance in predictingprostate cancer was evaluated and a multivariable algorithm to predictprostate cancer was developed by logistic regression. Performance ofthe multiplex algorithm was verified in 305 men with T-erg and PCA3measured by GenProbe Transcription Mediated Amplification (TMA) ofpost DRE whole urine.

RESULTS: In the pilot cohort PCA3 had the highest sensitivity(93%) in predicting prostate cancer diagnosis and T-erg had the high-est specificity (87%). Combining serum PSA with urine PCA3 and T-ergin a multivariate algorithm optimized for clinical utility improved cancerprediction (model AUC � 0.88; at 80% sensitivity, specificity � 90%)over PSA alone (AUC � 0.72). The algorithm recommends biopsy ifeither T-erg or PCA3 is positive when PSA � 10ng/ml, and wheneverPSA � 10ng/ml. Performance of this algorithm was tested in 305subjects by TMA whole urine assay for T-erg and PCA3; 299 wereinformative and 45% had prostate cancer. The clinical algorithm basedon PSA, T-erg and PCA3 combined (AUC � 0.75) performed betterthan for PSA alone (AUC � 0.59) in predicting prostate cancer diag-nosis (p�0.05, Figure). At sensitivity � 82%, the multiplex algorithmspecificity was 60%, compared to 23% specificity for PSA alone (cor-responding to PSA � 3.5 ng/ml). Use of the model algorithm wouldhave avoided unnecessary biopsy in 32% of men in the validationcohort.

CONCLUSIONS: A clinically practical algorithm that recom-mends biopsy for all patients with PSA �10ng/ml, while restrictingbiopsy among those with PSA �10ng/ml to men who have either T-ergor PCA3 present in post-DRE urine, outperformed PSA in predictingprostate cancer. This algorithm would eliminate nearly than one third ofprostate biopsies without substantive compromise of cancer detection.Prospective multi-center validation is warranted.

*MS, JW & AC share senior authorship

Source of Funding: NIH-EDRN

2159THE USE OF CANINES FOR PROSTATE CANCER DETECTION:TOWARDS A NON-INVASIVE ALTERNATIVE SCREENING TOOL

Jean-Nicolas Cornu*, Caroline Girardet, Geraldine Cancel-Tassin,Philippe Ulmer, Olivier Cussenot, Paris, France

INTRODUCTION AND OBJECTIVES: Detection of bladder andbreast cancer by trained dogs, recognizing smell of volatiles organiccompounds (VOCs) in human urine samples has been studied withpromising results. Our aim was to evaluate the effectiveness of a testfor prostate cancer detection by trained dogs on human urines sam-ples.

METHODS: Fresh urine samples, frozen for preservation, wereobtained from 66 patients referred to a urologist for elevated PSA or

Vol. 183, No. 4, Supplement, Wednesday, June 2, 2010 THE JOURNAL OF UROLOGY� e839