Diffusion-weighted MRI of lymphoma: prognostic utility and implications for PET/MRI?

With the recent introduction of PET/MRI, we investigated whether diffusion-weighted imaging (DWI) can complement PET for predicting local treatment response in Hodgkin lymphoma. This retrospective study included 39 patients selected from a hospital database with a histological diagnosis of Hodgkin lymphoma undergoing whole-body MRI (supplemented by DWI) and PET/CT before and after two cycles of vincristine, etoposide, prednisolone and doxorubicin (OEPA). The pretreatment volume, MRI apparent diffusion coefficient (ADC) and PET maximum standardized uptake value (SUVmax) of the largest nodal mass were determined quantitatively for evaluation of the local response following two cycles of OEPA. Quantitative pretreatment imaging biomarkers (disease volume, ADC, SUVmax) were compared between sites with an adequate and those with an inadequate response using Fisher's exact test and Mann Whitney statistics. Multivariate models predictive of an inadequate response based on demographic/clinical features, pretreatment disease volume and SUVmax without (model 1) and with (model 2) the addition of ADC were derived and crossvalidated. The ROC area under curve (AUC) was calculated for both models using the full dataset (training) and the crossvalidation (test) data. Sites with an adequate response had a significantly lower median pretreatment ADC (1.0 x 10-3mm2s-1) than those with an inadequate response (1.26 x 10-3mm2s-1; p < 0.01). There were no significant differences in patient demographic/clinical parameters, pretreatment SUVmax or pretreatment nodal volume between sites with inadequate and adequate response. The ROC-AUCs for prediction of an inadequate response for the training and test data for model 1 were 0.90 and 0.53, and for model 2 were 0.84 and 0.71, respectively. DWI complements PET for prediction of site-specific interim response to chemotherapy. (orig.)