More research is needed to determine the value of using magnetic resonance imaging to predict pathology outcomes following radical prostatectomy, according to a study presented at the AUA annual meeting in Chicago.
More research is needed to determine the value of using magnetic resonance imaging (MRI) to predict pathology outcomes following radical prostatectomy, according to a study presented at the AUA annual meeting in Chicago.
While the results of the study showed minimal additional value in using multiparametric prostate MRI (mpMRI) to help predict pathologic outcomes for radical prostatectomy at the time of surgery, just three predictors were used: highest Prostate Imaging Reporting and Data System (PI-RADS) score, number of lesions, and appearance of lymph nodes. Additionally, the study relied on MRI registry data from a single center.
“We’re hoping we’ll see more of a difference by looking at lesion location as well as lesion size and by studying patients from more centers,” Karandeep Singh, MD, MMSc, assistant professor of medicine and assistant professor of learning health sciences at the University of Michigan, Ann Arbor, told Urology Times. “We are currently in the process of combining MRI registries from three large health systems and are looking at additional MRI predictors to address these limitations.”
The study was conducted by the Michigan Urological Surgery Improvement Collaborative (MUSIC), a consortium of 45 urology practices that maintains a prospective registry of men with prostate cancer. The study was funded by Blue Cross Blue Shield of Michigan as well as through a grant from the National Institute of Diabetes and Digestive and Kidney Disease.
The ability to predict pathology outcomes from radical prostatectomy can be highly useful in determining which patients would most benefit from this surgery. Such information can facilitate patient counseling and decision-making. It can also aid surgeons in preoperative planning.
The authors matched information from an MRI registry for the University of Michigan to data from the MUSIC registry and identified 392 men who underwent an mpMRI between 2015 and 2018 prior to undergoing radical prostatectomy. The team then developed random forest models to predict the presence of non-organ-confined disease (NOCD), extracapsular extension (EPE), seminal vesicle invasion (SVI), or lymph node involvement (LNI) at the time of surgery.
Next, the authors evaluated the ability to predict pathology outcomes in these men using:
The authors used a 10-fold, cross-validated area under the curve to assess model discrimination.
Next: Ability to predict lymph node involvement worsens with mpMRIAbility to predict lymph node involvement worsens with mpMRI
The results showed little difference between MRI and traditional predictors in predicting pathology outcomes. The authors also found only slight improvement in predicting NOCD, EPE, and SVO when traditional predictors and MRI predictors were combined. The ability to predict LNI, on the other hand, slightly worsened when MRI and traditional predictors were combined.
As expected, traditional predictors exhibit excellent discriminative ability for predicting pathologic outcomes, they found.
Until more than three mpMRI predictors are examined, “If you have an MRI, the only thing you might be able to provide greater information on is who to dissect,” Dr. Singh said.