In a recently completed study at Washington University Medical Center in St. Louis, experienced radiologists were no more accurate than younger colleagues at diagnosing prostate cancer on multiparametric MRI (mpMRI). Instead, where performance gaps showed up, the confounding variables were given factors such as lesion location, biopsy history and advanced patient age.
Senior author Eric Kim, MD, and colleagues had their study report published online Feb. 23 in Urology.
The team reviewed the cases of 459 consecutive men who received prostate mpMRI in a clinical setting prior to being sent for biopsy of suspicious lesions. Nine radiologists read the scans, and all used the ACR’s PI-RADS system to classify their findings.
Comparing the rads’ accuracy following biopsy results, Kim and colleagues found considerable variation. Sensitivity ranged from 71 percent to 100 percent, specificity ranged from 30 percent to 63 percent, positive predictive value ranged from 36 percent to 56 percent, and negative predictive value ranged from 78 percent to 100 percent.
Yet the researchers found no significant variability between individual radiologists.
In fact, they found that radiologists who had previously interpreted more than 500 prostate mpMRI exams tended to have lower sensitivity and negative predictive value with no corresponding increase in specificity or accuracy.
“As the use of prostate multiparametric MRI is relatively novel, particularly outside of a research setting, increasing experience may not provide a benefit in the accurate interpretation of these studies,” the authors observed.
They added that the standardization provided by PI-RADS “may attenuate the advantages of experience.” Here they cited prior research showing that more-experienced radiologists performed significantly better when using non-standardized reporting while performing similarly to less-experienced radiologists when using PI-RADS classification.
Kim et al. underscored that mpMRI, which combines three separate image-acquisition techniques—T2-weighted, diffusion-weighted and dynamic contrast-enhanced—has limitations in differentiating between prostate cancer and benign prostatic hyperplasia in some cases. Still, they noted, the approach has appropriately grown in reputation as a key tool for managing prostate cancer.
They acknowledged as a limitation in their study design its retrospectivity and reliance on purely clinical data.
Still, as a descriptive study of prostate mpMRI in a clinical setting, the study’s results “are likely more representative of radiology practices across the country, thus providing greater insights into factors affecting the test performance of prostate mpMRI interpretation when using PI-RADS,” they wrote.
“Ultimately, a prospective assessment of ‘non-expert’ radiologists, blinded to patient clinical data, with multiple radiologists reading the same mpMRI in an independent fashion with correlation to either biopsy or prostatectomy pathology, is necessary.”