Tumor Treatment Response Assessed During the Chemo-Radiotherapy for Locally Advanced NSCLC

Abstract

To evaluate the capability of assessing intratumoral treatment response distribution with using FDG-PET/CT during the chemoradiotherapy of locally advanced NSCLC. Twelve of total 50 patients with stage III NSCLC were enrolled in the institutional protocol for concurrent chemoradiotherapy with treatment dose of 54-60 Gy in 27-30 fractions. For each patient, a pre-treatment FDG-PET/CT image (SUV0) and a mid-treatment image (SUVm) obtained within the treatment dose of 24 ∼ 46 Gy were obtained. Followed by deformable PET/CT registration between SUV0 and SUVm, the tumor voxel SUV reduction ratio was obtained to construct a tumor dose response matrix (DRM). Tumor SUVavid was also constructed by limiting tumor voxel SUVm > a given value. Spatial correlations of the tumor SUV0, SUVm, SUVavid and DRM were determined. The mean and coefficient variation (CV) of the SUV0, SUVm and DRM for all tumors were 6.56(64%), 2.82(59%) and 0.52(70%) (Table contains the individual data), which were like those on the SUVs and the mean DRM of head-neck HPV- patients reported previously, but much larger on the DRM variation. The inter-tumoral CVs of SUV0 and DRM were 17% and 43%, which were much smaller than those of the intra-tumoral CVs 61% and 55%. Meanwhile, the intra-tumoral variations on both SUV0 and DRM were much larger than those of head-neck HPV- patients. There was a weak correlation between SUV0 and SUVm with the correlation coefficient 0.32, a medium correlation of -0.51 between SUV0 and DRM; 0.58 between SUVm and DRM. It implies that the rule of tumor dose response DRM on treatment modification decision cannot be fully replaced by either SUV0 or SUVm. The spatial correlation between tumor DRM and SUVavid was 0.23 with SUVavid value > 3, which was getting weaker when increasing SUVavid value. Spatial dose response for NSCLC assessed using FDG-PET/CT feedback demonstrated high treatment resistant patterns, which had a large intra-tumoral variation. In addition, the medium correlations of DRM vs SUV0 and DRM vs SUVm imply that all these factors could be used to guide adaptive modification of NSCLC treatment.