Final Trial Results of Repeated Diffusion-Weighted MRI and FDG-PET Imaging for the Prediction of Response to Chemoradiation Therapy in Esophageal Cancer

Abstract

To present the final results of a prospective study to evaluate imaging predictors for pathologic complete response (pathCR) to chemoradiation therapy (CRT) in esophageal cancer using repeated diffusion-weighted magnetic resonance imaging (DW-MRI) and 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET) imaging before, during, and after CRT. Between January 2014 and July 2017, 60 patients with stage II-III esophageal cancer were prospectively enrolled. DW-MRI and FDG-PET imaging were acquired before treatment (baseline), 12-14 days after initiation of CRT (interim, IM), and at first follow-up 4-6 weeks after completing CRT. The primary outcome was pathCR as assessed after subsequent surgery, with secondary outcomes evaluating locoregional and distant progression-free survival (LPFS and DPFS), and overall survival (OS) according to the imaging parameters. Logistic regression analysis was used to evaluate the association of primary tumor apparent diffusion coefficient (ADC) values, maximum and mean standardized uptake value (SUV), metabolic tumor volume (MTV), and tumor lesion glycolysis (TLG) at baseline and their relative changes (Δ) at interim and follow-up scans with pathCR. Survival analyses applied to evaluate the association of the imaging parameters with LPFS, DPFS, and OS. Surgery after CRT was performed in 29 (48%) of 60 included patients. Imaging data was eligible for analysis in 28 patients in whom pathCR was observed in 7 (25%). The change in tumor mean ADC from the baseline to the interim DW-MRI scan, at a threshold of +27.7%, was best predictive for pathCR (c-statistic=0.98). The change in TLG from the baseline to interim FDG-PET scan, at a threshold of 58% decrease, was the best FDG-PET-based predictor of pathCR (c-statistic=0.77). Multivariable logistic regression suggested no significant added value of ΔTLGIM to ΔADCIM for the prediction of pathCR. Median follow-up was 29 months (range: 5-59) overall and 36 months (range: 24-59) for survivors. In surgical patients no locoregional progression was seen, and studied DW-MRI and FDG-PET parameters were not significantly associated with DPFS or OS. For non-surgical patients with ΔADCIM above vs below +27.7%, observed locoregional recurrence rates were 1/5 (20%) vs 7/25 (28%) (NS), and distant recurrence rates were 1/5 (20%) vs 15/25 (60%) (NS). For non-surgical patients with ΔTLGIM below vs above -58%, observed locoregional recurrence rates were 2/11 (18%) vs 5/17 (29%) (NS), and distant recurrence rates were 3/11 (27%) vs 12/17 (71%) (log-rank test, p=0.04). No association between imaging parameters and OS was observed. ΔADC on DW-MRI after 2 weeks of CRT strongly predicts pathologic complete response in esophageal cancer, with superior performance to repeated FDG-PET imaging. Early response assessment using DW-MRI is potentially useful for stratification of esophageal cancer patients to treatment intensification or organ-preserving strategies.