Although radiotherapy is an effective treatment of NPC, it also has the potential to cause secondary malignancies. The incidence of RIOSM is on the rise due to improved early discovery and treatments of NPC [1–3]. In the English-language literature, one study reported a cumulative occurrence rate of 0.03%–0.8% of RIOSM after radiotherapy in patients with NPC [2]. Liu et al. [1] estimated that the occurrence rate of RIOSM in patients with NPC was nearly 0.037%; of the 15 cases they studied, five tumors arose in the maxilla, seven in the mandible, and three in the junction of the nasal cavity and para-nasal sinuses. In our study, we considered RIOSM arising in the maxilla (33 cases) and mandible (12 cases) only, with a maxilla-versus-mandible ratio of 2.75:1, corresponding to an estimated cumulative occurrence rate of 0.084%.

Radiotherapy is the most common treatment of NPC, and the irradiation fields may extend from the skull base down to the lower neck region. As a late complication of radiotherapy, osteosarcoma can arise in the maxilla and mandible. Some researchers have suggested that RIOSM is most likely to occur after exposure to doses <30 Gy [6, 7]. In our study, the mean dose that patients received was 68 Gy, which was much higher than the mean dose of 45 Gy (range 25–110 Gy) that was reported in a previous study of RIOS [8]. Since the maxilla and mandible were not as routine conservation tissue it is difficult to calculate the exact radiation dose on the bones in NPC radiotherapy. Moreover, many uncertainties can affect the calculation of radiation dose, such as scatter dose; a previous study pointed out that a high scatter dose would cause scattering low dose ratio, which induced malignancies [9]. As radiotherapy technology advances, ionizing radiation dose may affect the occurrence rate of RIOSM. However, we could not obtain exact data on the radiation dose that would increase the occurrence rate of secondary malignancies.

Intensity-modulated radiotherapy (IMRT) has enabled target volumes to be more precisely, but IMRT actually increases the volume of normal tissue subjected to low-dose ionizing radiation [10]. Some researchers have suggested that medium-to-low-dose radiation may induce carcinogenesis more effectively than high-dose radiation [5, 10]. Moreover, some studies have found that concurrent chemotherapy significantly improved treatment outcomes in NPC patients, although others have suggested that chemotherapy may increase the risk of radiation-induced sarcoma [11, 12]. We made two speculations: first, whether IMRT indeed increases the occurrence rate of secondary malignancies; and second, whether IMRT did not increase the risk of secondary cancer. These speculations should be tested in future studies with a longer follow-up and a larger number of patients by analyzing clinicopathologic data. As more patients with NPC are treated with irradiation and now survive longer, the incidence of RIOSM is likely to increase [13, 14].

Previous studies reported latencies ranging from 5.0 to 30.0 years (mean 12.9 years) [15]. Similarly, in the present retrospective study, the median latency of RIOSM was 8.0 years (range 3.0–34.0 years). The factors that influence latency are unknown. In this cohort, patients who received radiation doses >68 Gy had a significantly shorter latency than patients who received ≤68 Gy (P = 0.005). However, the radiation source (orthovoltage, cobalt-60, or megavoltage X-rays) and patient sex or age had no significant effect on latency.

RIOSM is an aggressive tumor with a very poor prognosis. Most studies have reported low 5-year OS rates, ranging from 10% to 30% [16–21]. In the most recent study reported by Tabone et al. [22], the 8-year OS and disease-free survival rates for 23 patients with RIOSM were 50% and 41%, respectively, suggesting that the prognosis of patients with RIOSM has improved over time. However, in our cohort, the 1-, 2-, and 3-year actuarial OS rates for patients with RIOSM were 53.3%, 35.6% and 13.5%, respectively. Moreover, all patients died within 5 years after RIOSM diagnosis. Other studies have reported that the 5-year OS rate of patients with primary maxillofacial osteosarcoma was 44%–70% [6, 7, 13, 23, 24]. RISOM results in worse outcomes compared with stage-matched osteogenic sarcomas of the jaw. Thiagarajan et al. [25] suggested that the poor outcomes were due to the following reasons: (1) delayed diagnosis in previously irradiated tissue; (2) compromised resection margins, due to proximity of the tumor to critical structures; (3) limited treatment options in a maximally irradiated field (i.e., technical difficulties of operating within an irradiated field and difficulties with irradiation to the field with surrounding normal tissues, which have been treated to near tolerance); (4) poor tumor sensitivity to chemotherapy; (5) the high-grade nature of the vast majority of RIOS; and (6) host immunosuppression caused by a combination of tumor-related factors and previous treatment [1, 26, 27]. This study clearly indicates that the prognosis of patients with RIOS in the maxilla and mandible is poorer than patients with primary osteosarcoma of the maxilla and mandible.

The incidence of radiation-induced sarcoma of head and neck (RISHN) is increasing, with an estimated risk of up to 0.3% [2, 23]. RISHN development may be influenced by radiation dose, age at initial exposure, exposure to chemotherapeutic agents, and genetic features. RISHN is associated with poor outcomes, and surgical resection with clear margins seems to offer the best chance for cure [23]. Management of RISOM is more challenging, entailing surgery for irradiated tissue and a limited scope for further radiotherapy and chemotherapy.

Treatment for RIOSM includes surgery, radiotherapy, chemotherapy, or a combination of these strategies. Complete surgical excision seems to be necessary for the treatment of radiation-induced sarcoma; however, in patients with RIOSM, radical surgery is suitable only for early-stage tumors [1]. Although preoperative assessment can help determine the surgical boundaries, it is difficult for surgeons to judge whether the tumor has already invaded to the surrounding area. Moreover, RIOSM often occurs close to important structures such as the carotid artery and skull base; radical surgery in these regions is associated with a high risk of critical damage to important structures. The significance of chemotherapy for RIOSM is under debate: some researchers have concluded that RIOSM is insensitive to chemotherapy, whereas others have stated that chemotherapy is effective [1, 25, 28]. One study reported that a combination of surgery and chemotherapy resulted in a higher OS rate than either surgery alone or chemotherapy alone [29]. In our cohort, 39 patients underwent surgery; of these 39 patients, 15 (38.5%) had gross or microscopic positive margins. In multivariate analysis, margin status was not significantly associated with OS. However, for patients who received surgery alone, a negative margin was associated with significantly higher 1- and 2-year OS rates (19 patients; 63.1% and 47.0%, respectively) than a positive margin (10 patients; 40.2% and 20.1%, respectively; P = 0.029, P = 0.018). We suggest that radical surgery with a negative margin leads to a significantly better prognosis for patients with RIOSM. In our cohort, chemotherapy alone did not influence survival. Therefore, surgery combined with postoperative chemotherapy may be an effective strategy to improve survival for patients with RIOSM. In our cohort, since only one patient received combined surgery, chemotherapy, and radiotherapy, we cannot make any definitive conclusions about the effectiveness of combined treatment for RIOSM. In fact, it has been reported that radiotherapy provides no survival benefit for patients with RIOSM [18, 21, 30–32].

In our cohort, RIOSM tumor size had prognostic significance. In agreement with the results of other studies [3, 23–27], we found that a larger tumor was more likely to be associated with more advanced disease and resulted in a poorer treatment outcome. Furthermore, in cases with large tumors surrounded by vital tissues, it is extremely risky and difficult to perform radical surgery and achieve a negative margin.

Unlike other reports [7, 21, 33], in our study recurrence was not prognostic for disease control status. RIOSM is an aggressive sarcoma, and, in our study, the mean OS time was 14.2 months, and only 20 patients survived for more than 16 months. However, the interval between surgery and tumor recurrence ranged from 5.0 to 65.5 months (mean 16.9 months). We contend that all of the deaths in our study were due to RIOSM. However, only a small proportion of RIOSM cases in our study could be detected on clinical examination and received appropriate treatment.

The potential limitations of our study are its retrospective nature, the relatively small sample size, and the fact that it was performed at a single institution. Given the rarity of this complication, larger multi-center prospective studies should be conducted to confirm these preliminary results and further analyze the treatment outcomes.