Combined endorectal and phased-array MRI in the prediction of pelvic lymph node metastasis in prostate cancer

The objectives of our study were to evaluate the accuracy of combined endorectal and phased-array MRI in detecting pelvic lymph node metastasis (LNM) in patients with prostate cancer and to determine whether radiologists' predictions of LNM improve with the incorporation of Partin nomogram or MRI findings (or both) regarding extracapsular extension or seminal vesicle invasion.

Between May 1999 and September 2003, 411 consecutive patients with clinically localized prostate cancer underwent MRI before surgery. Serum prostate-specific antigen (PSA) level, Gleason grade, clinical stage, greatest percentage of cancer and percentage of positive cores in all biopsy cores, presence of perineural invasion on biopsy, and likelihood of LNM based on the Partin tables (2001 version) were recorded. MRI studies were interpreted prospectively, but the risks of LNM, extracapsular extension, and seminal vesicle invasion were scored retrospectively on the basis of the MRI reports. Surgical pathology constituted the standard of reference. The accuracy of LNM prediction was assessed using areas under receiver operating characteristic curves (AUCs) and univariate and multivariate logistic regression analyses. For multivariate models, the jackknife method was used for bias correction. A p value below 0.05 denoted statistical significance.

At surgical pathology, LNM was present in 22 (5%) of 411 patients. MRI was an independent statistically significant predictor of LNM (p = 0.002), with positive and negative predictive values of 50% and 96.36%, respectively. On multivariate analysis, prediction of lymph node status using the model that included all MRI variables (extracapsular extension, seminal vesicle invasion, and LNM) along with the Partin nomogram results had a significantly greater AUC than the univariate model that included only MRI LNM findings (AUC = 0.892 vs 0.633, respectively; p < 0.01).

Incorporation of the Partin nomogram results and MRI findings regarding both extracapsular extension and seminal vesicle invasion improves the MR prediction of LNM in patients with prostate cancer.