Complementary value of pre-treatment apparent diffusion coefficient in rectal cancer for predicting tumor recurrence

Purpose: To assess the complementary prognostic value of pre-treatment tumor apparent diffusion coefficient (ADC) for the prediction of tumor recurrence in patients with rectal cancer. Methods: From March 2012 to March 2013, a total of 128 patients with mid/lower rectal cancer who underwent pre-treatment rectal MRI were enrolled in this retrospective study. Two radiologists in consensus evaluated conventional imaging features (C-img) in pre-treatment rectal MRI: tumor height from anal verge (<= 5 cm vs. > 5 cm), T stage (high vs. low), the presence or absence of lymph node metastasis, mesorectal fascia invasion, and extramural venous invasion. The mean tumor ADC values (Tumor(ADC)) based on high b-value (0, 1000 x 10(-3) mm(2)/s) diffusion weight images were extracted. A multivariate Cox proportional hazard (CPH) regression was performed to evaluate the association of C-img and Tumor(ADC) with the 3-year local recurrence (LR) rate. Predictive performance of two multivariate CPH models (Cimg only vs. C-img + Tumor(ADC)) was compared using Harrell's c index (HCI). Results: Tumor(ADC) (Adjusted HR, 7.830; 95% CI 3.937-15.571) and high T stage (Adjusted HR, 8.039; 95% CI 2.405-26.874) were independently associated with the 3year LR rate. The CPH model generated with T stage + Tumor(ADC) (HCI, 0.820; 95% CI 0.708-0.932) showed significantly higher HCI than that with T stage only (HCI, 0.742; 95% CI 0.594-0.889) (P = 0.009). Conclusions: In patients with mid/lower rectal cancer, integrating Tumor(ADC) to C-img increases predictive performance of the CPH model than that with C-img alone for the prediction of LR within 3 years after surgery.