Tumor Treatment Response-Guided Adaptive Dose De-escalation/Escalation Strategy for Concurrent Chemoradiation Therapy of Nasopharyngeal Carcinoma.

Abstract

A novel tumor dose response-guided adaptive chemoradiation therapy process for nasopharyngeal carcinoma (NPC) was constructed and evaluated prospectively and retrospectively. In this process, tumor voxel dose response matrices (DRMs) quantified using 2-point fluorodeoxyglucose (FDG) positron emission tomography/magnetic resonance images acquired before and after an induction chemotherapy (IC) were used to guide quantitatively the individual patient dose de-escalation/escalation for patients with NPC in the post-IC concurrent chemoradiation therapy (CCRT). Two fluorodeoxyglucose-positron emission tomography/magnetic resonance images were obtained before and after IC for each of 21 patients with stage III and IV NPC who underwent IC + CCRT. The changes of tumor voxel standardized uptake value (SUV₀) on the pretreatment baseline image and the baseline voxel volume were tracked on the post-IC image using a deformable image registration tool and used to construct the tumor voxel DRM as the surrogate of tumor in situ SF₂ at the voxel level. The equivalent dose in the IC treatment was determined, assuming that the IC treatment achieved the same mean DRM as the one achieved using a dose in the CCRT treatment alone. Therefore, the equivalent dose should not be extrapolated with respect to any other clinical endpoints. The tumor voxel SUV₀ and DRM were used to predict the treatment outcome and retrospectively create the expected treatment dose for the post-IC CCRT treatment. For a few patients who were expected to receive a large dose escalation, new treatment plans targeting the highly resistant tumors were generated to assess the clinical feasibility of an adaptive dose fractionation painting schema. The equivalent dose in 2 Gy per fraction for 3 cycles IC (gemcitabine + cisplatin) was approximately 40 Gy. The mean and coefficient variation of SUV₀ and DRM for all the primary tumors (gross tumor volume [GTV]-nasopharynx [GTVnx] group) were 5.98 (62%) and 0.42 (72%), respectively, which were significantly larger than 5.22 (59%) and 0.37 (55%) for the positive nodes (GTVnd) group, respectively (P < .001). Both GTVnx and GTVnd groups exhibited significantly larger intratumoral variations on SUV₀ compared with the intertumoral variations (55%, 54%) versus (29%, 24%), meanwhile, similar intratumoral/intertumoral variations on DRM (47%, 39%) versus (54%, 39%). Individual tumor local control probabilities (TCPs) for the 21 patients calculated using the individual tumor voxel (SUV₀, DRM) were from 0.54 to 1.0. To achieve TCP = 0.99, 18 of 21 patients could have their CCRT treatment dose be de-escalated for one of GTVnx and GTVnd or both. In addition, 6 of 21 patients needed the CCRT treatment dose for at least 1 GTV to be escalated to 78 to 126 Gy (equivalent dose in 2 Gy per fraction); 2 of the 6 needed to escalate the treatment dose for both the primary and positive node GTVs. Adaptive plans on the 2 highly resistant tumors demonstrated that the dose fractionation painting was clinically feasible with regard to the normal tissue tolerance and clinical delivery technology, which could improve their tumor local control from TCP = 0.54 and 0.60 in the standard IC + CCRT treatment to TCP = 0.95 and 0.94 with the adaptive dose escalation, respectively. NPC, although relatively sensitive to the chemoradiotherapy, demonstrated very large spatial heterogeneities in both the tumor cell density and dose response within the individual tumors. The spatial tumor dose response assessed using the 2-point fluorodeoxyglucose-positron emission tomography imaging before and after the IC treatment can be used to design quantitatively the individual treatment dose required in the CCRT. The CCRT treatment dose could be de-escalated for >50% of the GTVs (GTVnx and GTVnd groups), meanwhile, escalated for about 16% of the GTVs with the local hypofractionation to the highly dose resistance regions.