Can MRI visibility of prostate tumors predict outcomes?

San Francisco-Visibility of a prostate tumor on magnetic resonance imaging (MRI) is not a significant predictor of oncologic outcome. In a comparison of oncologic outcomes between MRI-visible and MRI-invisible cancer foci, no differences in biochemical recurrence-free survival, distant metastases-free survival, and prostate cancer-specific mortality were found between the visible and invisible tumors, according to data presented at the Genitourinary Cancers Symposium in San Francisco.

Based on the finding, “My suggestion is that MRI should be used judiciously for the evaluation and management of prostate cancer, particularly when considering active surveillance or focal therapy. We need to perform systematic 12-core biopsies to make sure that those patients do not have high-grade cancers elsewhere in the prostate before instituting surveillance or focal therapy,” said lead investigator Simpa S. Salami, MD, MPH.

Read: African-Americans fare better with mCRPC treatments vs. Caucasians

In the study, a retrospective cohort of 10 patients with 26 primary prostate tumors was analyzed. Patients underwent MRI prior to prostatectomy. Fourteen (54%) of the lesions were invisible on MRI; five of which (36%) were Gleason 3+4 and the remainder of which were Gleason 6.

“It’s not uncommon for some cancers to be visible on MRI and some cancers to be invisible on MRI,” said Dr. Salami, of the University of Michigan, Ann Arbor. “The question is, when you have a patient with an MRI that shows a tumor, and that same patient has multiple areas of cancer that are not visible, can you ignore those areas that are not visible? Can you deliver focal therapy, for example, and treat the area that is visible and ignore the areas that are invisible?”

Continue to the next page for more.

To determine whether the molecular profiles were different between MRI-visible and MRI-invisible cancers, next-generation sequencing was performed on visible and invisible lesions. High-confidence prioritized prostate cancer-relevant mutations were detected in 14 of 26 (54%) tumor foci and six of 14 (43%) MRI-invisible specimens. Notable point mutations were in APC, AR, ARID1B, ATM, ATRX, BRCA2, FAT1, MAP3K1, NF1, SPEN, SPOP, and TP53.

The authors developed a 9-gene signature that could predict MRI visibility with high accuracy. The area under the curve (AUC) for predicting MRI visibility was 0.89. This signature was validated with another cohort of 16 patients collected from Cedars Sinai Medical Center. The AUC in the validation cohort was 0.88, with a specificity of 100% for predicting MRI-visible tumors.

“One interesting finding is that lesions that are visible have underexpression of genes that are involved in cellular organization and structure, meaning that cancer cells that are more disorganized are more likely to be visible on MRI,” Dr. Salami told Urology Times.

Also see: Inflammatory bowel disease may raise prostate Ca risk

A testing cohort of 375 patients was pooled from two independent case-cohort studies (Johns Hopkins Medical Institute and Mayo Clinic) to test the capacity of predicted MRI visibility to predict oncologic outcomes. All 375 patients had undergone radical prostatectomy and had DNA microarray studies performed on tumor specimens, and were followed for up to 240 months for outcomes. Over this time, there was no difference in the number of biochemical recurrence-free events in tumors predicted to be invisible on MRI versus those with predicted MRI-visible tumors (72 vs. 64). There was also no difference in the number of distant metastasis-free survival events in those with predicted MRI-invisible tumors and those with predicted MRI-visible tumors (29 vs. 26), as well as prostate cancer-specific mortality events (16 vs. 12, respectively).

“The findings suggest that lesions that are invisible on MRI cannot just be ignored,” said Dr. Salami. “They may be as important as lesions that are visible.”

Several of Dr. Salami’s co-authors have disclosures related to pharmaceutical companies; see disclosures here.