The purpose of this investigation was to evaluate the potential dosimetric benefits of a two-phase adaptive intensity-modulated radiotherapy (IMRT) protocol for patients with locally advanced nasopharyngeal cancer (NPC). A total of 17 patients with locally advanced NPC treated with IMRT had a second computed tomography (CT) scan after 17 fractions in order to apply and continue the treatment with an adapted plan after 20 fractions. To simulate the situation without adaptation, a hybrid plan was generated by applying the optimization parameters of the original treatment plan to the anatomy of the second CT scan. The dose–volume histograms (DVHs) and dose statistics of the hybrid plan and the adapted plan were compared. The mean volume of the ipsilateral and contralateral parotid gland decreased by 6.1 cm3 (30.5%) and 5.4 cm3 (24.3%), respectively. Compared with the hybrid plan, the adapted plan provided a higher dose to the target volumes with better homogeneity, and a lower dose to the organs at risk (OARs). The Dmin of all planning target volumes (PTVs) increased. The Dmax of the spinal cord and brainstem were lower in 94% of the patients (1.6–5.9 Gy, P < 0.001 and 2.1–9.9 Gy, P < 0.001, respectively). The Dmean of the contralateral parotid decreased in 70% of the patients (range, 0.2–4.4 Gy). We could not find a relationship between dose variability and weight loss. Our two-phase adaptive IMRT protocol improves dosimetric results in terms of target volumes and OARs in patients with locally advanced NPC.