Prediction of Prostate Cancer Recurrence Using Magnetic Resonance Imaging and Molecular Profiles

Purpose: To evaluate whether pretreatment magnetic resonance imaging (MRI)/MR spectroscopic imaging (MRSI) findings and molecular markers in surgical specimens correlate with each other and with pretreatment clinical variables (biopsy Gleason score, clinical stage, and prostate-specific antigen level) and whether they contribute incremental value in predicting prostate cancer recurrence. Experimental Design: Eighty-eight prostate cancer patients underwent MRI/MRSI before radical prostatectomy; imaging findings were scored on a scale of 1 to 7 (no tumor seen-lymph node metastasis). Ki-67, phospho-Akt, and androgen receptor expression in surgical specimens were assessed by immunohistochemistry. To examine correlations between markers and imaging scores, Spearman's correlation was used. To test whether markers and imaging scores differed by clinical stage or Gleason score, Wilcoxon's rank sum test was used. To examine time to recurrence, the methods of Kaplan-Meier were used. Cox proportional hazards models were built and their concordance indices (C-indices) were calculated to evaluate prediction of recurrence. Results: All markers correlated moderately strongly with MRI/MRSI score (all correlation coefficients >0.5). Markers and MRI/MRSI score were strongly associated with clinical stage and biopsy Gleason score (P < 0.01 for all). At last follow-up, 27 patients had recurrence. C-indices for MRI/MRSI score and all markers were associated with time to recurrence and ranged from 0.78 to 0.89. A Cox model combining all clinical predictors had a C-index of 0.89; the C-index increased to 0.95 when MRI/MRSI score was added and to 0.97 when markers were also added. Conclusions: MRI/MRSI findings and molecular markers correlated well with each other and contributed incremental value to clinical variables in predicting prostate cancer recurrence.