Combining tumor response and personalized risk assessment: Potential for adaptation of concurrent chemotherapy in locoregionally advanced nasopharyngeal carcinoma in the intensity-modulated radiotherapy era

Abstract

Background and purposeIn the intensity-modulated radiotherapy (IMRT) era, the role of concurrent chemoradiotherapy (CCRT) after induction chemotherapy (IC) in locoregionally advanced nasopharyngeal carcinoma (LANPC) is undetermined, while concerns exist about CCRT-associated excessive toxicity. We aimed to combine tumor response and risk assessment to guide decisions about concurrent chemotherapy. Materials and methodsFrom April 2009 to December 2015, 744 LANPC patients treated with CCRT/IMRT after IC were included. Matching techniques were performed for treatment effect evaluation. Tumor response to IC was used for patient stratification. A nomogram was built based on multivariable Cox regression analysis to predict overall survival (OS). ResultsAfter IC, 508 patients (68.3%) had favorable tumor response (complete or partial response), among whom IC + CCRT achieved significantly superior 5-year disease-free survival and OS than IC + IMRT (82.2% vs. 72.5%, P = 0.025; 89.2% vs. 79.9%, P = 0.025). However, no significant difference was found in patients with unfavorable response (both P > 0.05). For favorable responders, a nomogram was built integrating age, smoking, T category, N category, pretreatment Epstein-Barr virus DNA and treatment modality. The concordance index was 0.713 and calibration was good. The nomogram determined three risk groups with distinct OS. High-risk patients benefited from CCRT after IC regarding disease-free survival, OS and distant metastasis-free survival, whereas low- and intermediate-risk patients did not. ConclusionsFor LANPC patients with unfavorable response to IC, subsequent CCRT seems inadequate, rendering intensification necessary. For favorable responders with low risk, IC + IMRT represents a reasonable de-intensification approach, although confirmation by prospective data is needed.