Moderated Poster

Poster, Podium & Video Sessions

MP08-14: Development and Validation of a Nomogram for Predicting PIRADS 4-5 Lesions on Multiparametric Prostate MRI

Friday, May 12
9:30 AM - 11:30 AM
Location: BCEC: Room 160

Presentation Authors: Matthew Truong*, Thomas Frye, Dang Lam, Ji Hae Park, Bokai Wang, Changyong Feng, Gary Hollenberg, Eric Weinberg, Edward Messing, Rochester, NY

Introduction: Multiparametric MRI (mpMRI) of the prostate is gaining popularity for use in prostate cancer (PCa) detection in patients with a prior negative sextant biopsy as well as in low risk PCa patients on active surveillance. The presence of PIRADS 4 and PIRADS 5 lesions on mpMRI have the highest diagnostic yield for clinically significant PCa on subsequent MRI-ultrasound fusion biopsy. Counseling patients regarding the benefit of mpMRI is becoming an increasingly important aspect of urologic practice. Nomograms may be clinically useful to individualize decisions to perform mpMRI based on patient risk profiles.


Methods: We identified 1023 patients who underwent mpMRI of the prostate from July 2014-October 2016 at our institution. Inclusion criteria were patients who underwent mpMRI to aid PCa detection or while on active surveillance. Using clinical variables, nomogram development was performed using 883 consecutive patients who met the inclusion criteria for the study. Clinical variables assessed included age, PSA, prostate volume, and PSA density (PSAD). Multivariable logistic regression generated a nomogram incorporating age, PSA, and prostate volume for finding PIRAD 4 or 5 lesions on mp MRI. A separate nomogram using PSAD alone was generated. Internal validation of each nomogram was performed by generating an ROC, calibration, and decision analysis curves.

Results: Age, PSA, prostate volume, and PSAD were all significant predictors of PIRADS 4-5 lesions on univariable analysis (all p < 0.001). Upon internal validation, a nomogram incorporating age, PSA, and prostate volume had an AUC of 0.746 (p < 0.001). A separate nomogram using PSAD alone had an AUC of 0.729 (p < 0.001). Both nomograms had excellent calibration and high net benefit on decision curve analysis across a wide range of predicted probabilities. The two nomograms performed similarly regardless of indication for mpMRI.

Conclusions: We developed two clinical nomograms that accurately predict the probability of finding PIRADS 4-5 lesions on mpMRI, which may be useful in counseling patients undergoing prostate cancer screening or who are on active surveillance. These nomograms pose no additional cost given that age and PSA are readily available and prostate volume obtained at previous transrectal ultrasound-guided biopsy can be used as input. Externally validation should be performed to confirm the utility of this nomogram in other cohorts.

Source Of Funding: none

Matthew Truong, MD

University of Rochester Medical Center

Matthew Truong MD is a Urology Resident at the University of Rochester Medical Center in Rochester, NY. He has a special interest in prostate cancer, particularly in prostate MRI, biomarkers, and active surveillance. He is interested in applying statistical models and machine learning to improve prostate cancer detection. He attended Stanford University for college and the University of Wisconsin-Madison for medical school.

Presentation(s):

Send Email for Matthew Truong


Assets

MP08-14: Development and Validation of a Nomogram for Predicting PIRADS 4-5 Lesions on Multiparametric Prostate MRI



Attendees who have favorited this

Send Email for Development and Validation of a Nomogram for Predicting PIRADS 4-5 Lesions on Multiparametric Prostate MRI