Volumetric measurement to evaluate treatment response to induction chemotherapy on MRI outperformed RECIST guideline in outcome prediction in advanced nasopharyngeal carcinoma

Abstract

Background

Treatment response evaluated by tumour size change is an important indicator for outcome prediction. Advanced nasopharyngeal carcinoma (adNPC) grows irregularly, and so the unidimensional measurement may not be accurately applied to adNPC for outcome prediction. This study aimed to evaluate values of unidimensional and volumetric measurements for treatment response to induction chemotherapy (IC) for outcome prediction in adNPC and compared the values with that of RECIST 1.1 guideline.

Materials and methods

Pre-treatment and post-IC magnetic resonance images (MRIs) from 124 patients with stage III-IVA NPC were retrospectively reviewed. Sums of the maximum unidimensional diameters (D) and volumes of the targeted tumours (primary tumour and two largest metastatic lymph nodes) on the pre- (D_pre and V_pre) and post-IC MRIs (D_post-IC and V_post-IC) and percentage changes in D (Δ D%) and V (ΔV%) between two scans were calculated and correlated with disease-free survival (DFS), locoregional recurrence-free survival (LRRFS), and distant metastases-free survival (DMFS) using Cox regression analysis. Area under the curves (AUCs) of independent measurements and RECIST groups (RECIST response and non-response groups) for predicting disease recurrence, locoregional recurrence, and distant metastases, respectively, were calculated and compared using the DeLong test.

Results

Univariable analysis showed correlations between high D_post-IC with poor DFS and DMFS (P < 0.05), but not with LRRFS (P = 0.07); high V_post-IC and low ΔV% (less decrease in volume on post-IC) with poor DFS, LRRFS, and DMFS (P < 0.05); and no correlations between D_pre, ΔD%, and V_pre and the outcomes (P > 0.05). Multivariable analysis showed that ΔV% was the only independent measurement for outcomes (P < 0.05). Compared with RECIST groups, ΔV% of 47.9% (median value) showed a higher AUC for disease recurrence (0.682 versus 0.526, P < 0.01) and for locoregional recurrence (0.782 versus 0.585, P < 0.01), but not for distant metastases (0.593 versus 0.518, P = 0.26).

Conclusions

Volumetric measurement to evaluate treatment response to IC outperformed unidimensional measurement and RECIST guideline in outcome prediction in adNPC.