Evaluation of diffusion-weighted imaging as a predictive marker for tumor response in patients undergoing chemoradiation for postoperative recurrences of cervical cancer

Abstract

To investigate diffusion-weighted imaging (DWI) as a response biomarker in patients undergoing chemoradiation for postoperative recurrences of cervical cancer. From October 2008 to March 2011, 20 patients were included. All underwent T2-weighted (T2W) and DWI before and after chemoradiation. Gross tumor volume (GTV), lateral extent, apparent diffusion coefficient (ADC), and presence of regions of focally restricted diffusion were determined at baseline. Response to chemoradiation was categorized as either partial or complete. Receiver operator characteristics (ROC) curve identified thresholds of GTV and ADC that best predict for partial response. Univariate and multivariate analysis were performed on SPSS version 15. The median GTV was 24.5 cc (4.1-110 cc). Central and lateral disease was present in 8 and 12 patients, respectively. The median ADC was 1 × 10 -3 mm² /s (0.8-1.3 × 10⁻³ mm² /s) and 12/20 (60%) patients had focal restricted diffusion. Overall 10/20 patients had partial response. ROC analysis identified volume of 25 cc or higher [sensitivity = 80%, specificity = 80%, area under curve (AUC) = 0.76, P = 0.04] and ADC more than 1 × 10⁻³ mm² /s (sensitivity = 70%, specificity = 50%, AUC = 0.62; P = 0.34) to best predict for partial response. On univariate analysis bulky disease (77.7% vs. 27%; P = 0.03), lateral disease (66.6% vs. 25%; P = 0.08), and focal regions of restricted diffusion (66.6% vs. 25%; P = 0.06) predicted for partial response to chemoradiation. All factors continued to be significant on multivariate analysis. On restricting analysis to bulky tumors ADC greater than 0.95 × 10⁻³ mm² /s predicted partial response with high sensitivity (85.7%) and specificity (100%) (AUC 0.96; P = 0.05). On univariate analysis lateral disease (P = 0.04), high baseline ADC (P = 0.07) predicted for partial response. Baseline ADC and focal regions of ADC restriction predict for partial response with moderate sensitivity and specificity in patients with postoperative recurrences of cervical cancer and need to be validated in larger cohort.