Owing to encouraging disease control results and the capacity for delivering a high radiation dose to the target while sparing adjacent organs, intensity-modulated radiotherapy (IMRT) is the standard treatment of nasopharyngeal carcinoma (NPC) [1, 2]. With modern treatment, the prognostic differentiation between T1 disease and T3 disease is narrowing [3]. However, favorable prognosis remains a challenge for patients with T4 disease (characterized according to the American Joint Committee on Cancer [AJCC] staging system [4]), particularly those with intracranial extension. Because the tumor is located near neurological structures, such as the brain stem, temporal lobes, optic chiasm, and optic nerves, inadequate dose coverage of the target volumes often occurs. Ng et al. [5] reported that a total dose higher than 66.5 Gy was tumoricidal. However, other researchers observed that dose escalation increases the occurrence of temporal lobe necrosis [6] and that the delivery of a high radiation dose to even a small volume in the temporal lobe was unsafe [7]. Addressing adequate dose coverage and normal tissue complication probabilities poses an obvious dilemma. If dose constraints for the brain stem, optic chiasm, and optic nerves are prioritized, even the minimum tumoricidal dose to the entire target volume is not always achievable with the maximum tolerated dose of 54 Gy [8].

Given the chemosensitive nature of NPC, the addition of chemotherapy to IMRT showed promising results for disease control and survival. With the advent of docetaxel, additional combinations of chemotherapy regimens are possible. For the treatment of head and neck carcinoma, adding docetaxel to the classic regimen of cisplatin plus 5-fluorouracil is superior to cisplatin plus 5-fluorouracil [9]. Furthermore, Du et al. [10] reported promising outcomes with good compliance and well-tolerated toxicities of induction and adjuvant chemotherapy using cisplatin, 5-fluorouracil, and docetaxel (TPF) for the treatment of locoregionally advanced NPC.

When a more effective chemotherapy regimen is available, we assumed that IMRT, despite some dosimetric inadequacy, may still mitigate the probability of devastating complications and other adverse effects without compromising disease control. In this study, by imposing strict dose constraints for neurological structures, we aimed to explore the effect of dosimetric inadequacy on local control and treatment-induced toxicities for T4 NPC patients with intracranial extension when treating with IMRT plus the TPF regimen.

In the present study, we explored the efficacy and toxicities of IMRT with strict OAR dose constraints plus TPF chemotherapy for T4 NPC patients with intracranial extension. Though dosimetric inadequacy occurred in D_min, the results showed a satisfactory outcome in terms of local response rate and 3-year LFFS rate, and showed low rates of treatment discontinuation, as well as few late toxicities of the central nervous system. Our results further suggested no significant differences for LFFS among patients with mild, medium, or deep involvement (P = 0.633), though those with deep involvement tended to have worse local control.

In the present study, the mean D_min to the GTVp was 55.2 Gy (range 48.3–67.3 Gy), which is lower than the 58.9 Gy (range 29.4–72.1 Gy) reported by Cao et al. [16]; however, we reported a 3-year LFFS rate of 87.4%, which is higher than 82.1% reported by Cao et al. [16]. In addition, no neurological dysfunction was observed in our study. This indicated that the dosimetric inadequacy caused by inadequate GTVp coverage might minimize toxicities without compromising local control. In consistent with our results, Lin et al. [17] reported similar 4-year LFFS rates for NPC patients treated with large-target-volume IMRT and reduced-target-volume IMRT, but late toxicities of the central nervous system were less in the latter group (P < 0.001). In the NPC-2003-GPOH/DCOG study, children and young adults with stage III/IV NPC received chemoradiotherapy and interferon-beta [18]. During a median follow-up of 30 months, only 6.8% (3/44) of patients with a T4N2 tumor experienced treatment failure. For all patients, the prescribed dose was 54.0–59.4 Gy [18]. It further indicated that, for T4 NPC patients with intracranial extension, treatment with IMRT at dosimetric inadequacy to some extent plus TPF chemotherapy may be feasible.

In locally advanced NPC, survival benefits were observed when chemotherapy was added to IMRT [19]. In recent years, induction chemotherapy has been used in clinical practice with potential advantages in terms of shrinking the primary tumor, which, for patients with intracranial invasion, results in wider margins to normal tissues [20]. As reported in a multicenter, phase III randomized controlled trial (NCT01245959), the addition of induction chemotherapy with the TPF regimen to concurrent chemoradiotherapy improved failure-free survival with acceptable acute toxicity [21]. Moreover, in a meta-analysis, adjuvant chemotherapy was shown to confer the highest survival benefit [22]. In the present study, the TPF regimen, given its superiority in the treatment of head and neck squamous cell carcinoma, was administered in both induction and adjuvant chemotherapy. Sun et al. [23] contended that the advantages of IMRT may “counteract” the effect of concurrent chemoradiotherapy in improving the local control rate. Therefore, according to our previous experience [10], we chose not to administer concurrent chemotherapy to reduce radiation-induced acute toxicities.

Despite the advantages of IMRT, late toxicities of the central nervous system remain the treatment bottleneck for T4 NPC cases [24]. As reported by Su et al. [25], for NPC patients treated with IMRT, temporal lobe injury was not observed in patients with T1-2 disease, but its occurrence rate increased significantly in patients with T4 disease (13.4%) compared with those with T3 disease (3.1%). Higher doses to OARs might account for more late toxicities. Su et al. [25] also found a 2.5% augment of temporal lobe injury per Gy of D₁ exceeding 52 Gy. Uy et al. [26] reported 1 case of radiation-related brainstem necrosis in 40 meningioma patients with D_max to the brainstem of 55.6 Gy; the necrotic area was in the high-dose coverage part of the brainstem, which may also have contributed to the one patient’s death. Therefore, one should be cautious about increasing dose constraints of the nerve system. A previous study from our institution assessed late toxicities in 80 T4 NPC patients and identified 6 cases of temporal lobe necrosis [27]. In this previous study, the mean D_max delivered to the affected side of the temporal lobe was 75.9 Gy, and most irradiation hot spots were found in the enhanced region of necrosis [27]. This suggests that, after IMRT, irradiation hot spots may have a close association with late injuries of neurological tissues. Based on the above results, we applied strict dose constraint criteria for D_max. We observed a D_max of 68.4 Gy to the temporal lobes with no temporal lobe necrosis. In addition, the D_max to the spinal cord inferior of 45.0 Gy may have contributed to no spinal cord injuries in our present patients, except 1 patient with a D_max of 45.7 Gy. No other neurological dysfunctions, e.g., brain stem injury and cranial nerve palsy, occurred in our present study, which we attribute to the strict dose constraints for neurological structures.

Ng et al. [5] found that the minimum dose to the target volumes was related to locoregional failure, and a dose of more than 66.5 Gy was reported to be tumoricidal. They further reported that most T4 diseases (and some T3 diseases) were underdosed (< 66.5 Gy) and that an underdosed GTVp of 3.4 cm³ was a prognostic factor for short LFFS [5]. However, adequate dose coverage of target volumes may lead to an unavoidably high dose of irradiation to neurological structures, causing serious late toxicities [8]. In patients with locally advanced NPC, He et al. [28] studied the distance between the primary tumor and the brainstem. For patients with distances > 4.7 versus ≤ 4.7 mm, with a mean D_min to the target volume of 63.1 Gy (range 21.7–74 Gy) versus 46.4 Gy (range 21.9–66.6 Gy), respectively, the corresponding mean D_max to the brainstem was 51.8 Gy (range 41.7–68.8 Gy) versus 54.1 Gy (range 45.1–68.4 Gy). In patients with locally advanced NPC, the proximity to neurological structures influences the radiation dose that reaches OARs and the tumor. For patients with T4 lesions (intracranial extension), this phenomenon is more common. In fact, one important factor affecting the degree of target dose coverage is the treatment planning protocol. We achieved strict dose constraints to neurological structures with maximum D_max of 45.7, 58.2, 68.4, 59.6, and 60.1 Gy to the spinal cord, brainstem, temporal lobes, optic chiasm, and optic nerves, respectively, and 0.2% and 0.04% of V_{65 Gy} (percentage volume receiving a dose of ≥ 65 Gy) to the affected and unaffected temporal lobes, respectively. The mean D_min to the GTVp was 55.2 Gy. For some series that used a less strict OAR dose constraint set, higher dose coverage of target volumes was reported. Cao et al. [16] reported a mean D_min to the GTV of 58.9 Gy. However, in their series, the maximum point doses to the spinal cord, brainstem, temporal lobes, optic chiasm, and optic nerves were 50.5, 80.3, 87.2, 89.9, and 87.6 Gy, respectively. Optimal dose coverage cannot be obtained without exceeding dose constraints for neurological structures. However, combined with effective chemotherapy, the acceptable extent of dosimetric inadequacy may be redefined with further investigation.

The limitations of this study include the relatively small sample size and short follow-up. Additionally, some injuries to the central nervous system may be asymptomatic or subtle and may not be diagnosed by clinical symptoms and physical/image examination, which may result in low occurrence rates of late injures to neurological structures. A more accurate and careful neurological examination is therefore required. For some cases, the follow-up time was less than 4 years, whereas, the peak time of temporal lobe necrosis was about 3–4 years after radiotherapy. Finally, as a single-arm study, there was no well-matched control group to be compared with. Thus, our results should be considered preliminary and require further confirmation.

To not exceed dose constraint criteria for neurological tissues, most patients who have T4 NPC with intracranial extension are underdosed. In treatment that comprised IMRT plus effective chemotherapy, we observed that slight dosimetric inadequacy led to satisfactory local control rate with few late toxicities of the central nervous system. However, the acceptable extent of dosimetric inadequacy remains unknown and requires further investigation.