Prediction of Response to Very Low Dose Radiotherapy (4 Gy) in Mature B-Cell Lymphomas Using Metabolic Tumor Volume

Abstract

Numerous studies established the role of palliative very low dose radiotherapy (VLDRT) of only 4 Gy in 2 fractions for indolent non-Hodgkin's lymphoma (NHL). While objective response rates to VLDRT are excellent, we have no clear biomarkers of depth and durability of response. Radiosensitivity to VLDRT is not clearly linked to clinical features, such as tumor grade, histology, and PET SUV. Our group has recently demonstrated that pre-treatment tumor size may predict response to VLDRT; however, this remains an imperfect predictor. Radiomic features like baseline metabolic tumor volume (MTV) and total lesion glycolysis (TLG) stratify patient response to chemotherapy for some NHLs. Therefore, we aimed to determine if these features could predict response to VLDRT in indolent NHL.Between 2005 and 2018, 250 patients with follicular or marginal zone lymphoma were locally irradiated to 299 sites using 2 Gy x2. Response was assessed within 1.5-6 months of VLDRT (n = 231), and local failure (LF) was assessed over a median follow-up of 29 months. Out of the 299 sites, 254 had a PET/CT scan performed prior to radiotherapy. Scans were imported into MIM, and pre-radiotherapy disease was contoured with liver for normalization. PET parameters including MTV and TLG were extracted. Lesion size was analyzed using 4 cm and 6 cm as predetermined cutoffs. Lesion response was analyzed using multivariate logistic LASSO regression to account for predictor multicollinearity and area under the curve (AUC) analysis. To analyze time to LF, we performed Cox proportional hazards multivariate regression using ridge regression. PET parameters were divided in quartiles, and for significant predictors, optimal cutoffs were determined using maximally selected rank statistics. Estimates are shown with 95% confidence intervals.A complete response (CR) was seen in 156 (68%) lesions and LF was seen in 81 (27%) lesions after VLDRT. Using size alone had similar predictive value for response (AUC: 0.64 (0.59,0.70)) compared to MTV (AUC:0.70 (0.62,0.77)) and TLG (AUC:0.68 (0.61,0.75)). For lesions less than 6 cm, size, MTV and TLG had comparable predictive value for response (AUC: 0.58 (0.51,0.64), AUC:0.65 (0.56,0.75), and AUC:0.66 (0.57,0.75), respectively); however, through multivariate modeling we determined that TLG was not as robust as MTV, and MTV > 75th percentile was an improved predictor of no CR versus size ≥4 (OR: 0.26 (0.11,0.57) vs 0.77(0.30,2.08), respectively) with no interaction noted. Likewise, in a multivariate model of LF, MTV was shown to be a better predictor than size for lesions < 6 cm with an MTV ≥45 associated with worse LF (HR: 1.84 (1.14, 2.00)) compared to size > 4 cm (HR:1.31 (0.80, 2.15)).In stratifying response and local control for patients receiving VLDRT for indolent lymphomas, MTV may aid in patient selection. It remains to be seen whether higher order PET radiomic parameters can further improve the accuracy of PET-based radiotherapy outcome prediction.