Twelve retrospective studies that examined a total of 1593 patients were included in the current investigation. All of these studies were descriptive cohort studies, and they provided moderately strong evidence. Given that HNMM is rare and is associated with an extremely poor prognosis, the results of this study can nonetheless provide significant guidance in making clinical treatment decisions regarding this disease.
The preferred treatment method: surgical resection
The subjects included in the current study were patients with a history of past surgical treatment; for mucosal melanoma surgery achieves better results than other therapeutic approaches with respect to treatment and local control. Moreover, surgery was the first treatment for many patients in the vast majority of examined publications. In an examination of 815 HNMM patients, Thomas et al.  found that compared with patients who received surgery alone, the relative risk ratios for patients who received simple radiotherapy or no treatment were 1.56 (95 % CI, 1.35–1.72) and 2.38 (95 % CI, 2.09–2.59), respectively. The clinical significance of these findings was that compared with the administration of surgical treatment, the administration of radiotherapy alone and the administration of no treatment increased the risk of death by 56 and 138 %, respectively. Thus, patients who underwent surgical treatment (with or without postoperative radiotherapy) had better prognoses and longer survival times than patients who did not receive surgical treatment. In a study by Andersen et al., all surviving patients had received surgical treatment; however, the non-surgical treatment approach involving chemoradiotherapy alone had little or no therapeutic effect .
Furthermore, the 2012 National Comprehensive Cancer Network (NCCN, an American organization) guidelines recommend surgery as the preferred treatment for mucosal melanoma patients; in particular, the optimal objective for surgical treatment is en bloc resection that effectively achieves tumor-free margins (NCCN Guidelines Version 1.2012, Mucosal Melanoma of the Head and Neck). However, surgical treatment is only suitable for patients with stage III–IVA cancer in the American Joint Committee on Cancer (AJCC) staging system (7th edition). In these stages, cancer is limited to the mucosal layer or exhibits moderate progression; that is, the tumor may have invaded deep soft tissues, cartilage, bone or surface skin covering the tumor, but there is no lymph node metastasis or DM .
Unfortunately, most HNMM patients are already at an advanced clinical and pathological stage of the disease when they are first diagnosed; as a result, their prognoses are far worse than the prognoses of cutaneous melanoma patients, for whom surgical treatment is also the preferred therapeutic approach [21, 22]. When cutaneous melanoma patients are first diagnosed, the percentage of patients at stages 0, I, II, III, and IV (using AJCC staging) are 14.9, 47.7, 23.1, 8.9 and 5.3 %, respectively; many patients undergo surgery or other effective interventions at the early stages of this disease. A survey of 84,836 melanoma patients indicated that 91.5 % of cutaneous melanoma patients underwent surgery, and that the 5-year survival rate for cutaneous melanoma was 80.8 % . However, for HNMM, because of complexities associated with the anatomical location of the primary tumor, it is often difficult to achieve negative safety margins during the microscopic resection of this tumor . In particular, it is very difficult to completely resect paranasal sinus mucosal melanomas that are located near the cribriform plate or the skull plate, or that have invaded the anterior skull base . Therefore, many researchers believe that it is unwise to attempt to achieve complete resection at the expense of increasing surgical risk and reducing postoperative quality of life [24, 25]. However, a lack of complete resection increases postoperative complications and thereby affects patient prognosis.
Postoperative radiotherapy is effective for local control, but does not reduce the risks of DM and death
In the present study, a random-effects model was used for the meta-analysis of the effects of postoperative radiotherapy on the local control (assessed in terms of LR) of HNMM. Relative to surgery alone, surgery combined with postoperative radiotherapy can reduce the risk of LR by 45 % (five NRSs with a total of 336 patients; HR, 0.55; 95 % CI, 0.32–0.93; p < 0.05). However, postoperative radiotherapy did not reduce the risk of postoperative death in HNMM patients (nine NRSs with a total of 1465 patients; HR, 1.07; 95 % CI, 0.95–1.02). Postoperative radiotherapy also did not reduce the risk of death in cases of SNMM (five NRSs with a total of 341 patients; HR, 1.04; 95 % CI, 0.79–1.36).
Because HNMM is multicentric, the clinical boundaries of this disease are blurred and unclear; infiltration occurs, particularly submucosal lymphoid infiltration involving melanoma cells , resulting in a high rate of postoperative LR (31–85 %) . Therefore, even if HNMM is diagnosed early and aggressive surgical treatment is administered, postoperative adjuvant treatment remains necessary. Among the possible adjuvant treatments, postoperative radiotherapy is one of the earliest examined adjuvant therapies; it has been used in a relatively large number of clinical cases. The purpose of radiotherapy is to reduce the postoperative invasion of residual tumors into the surrounding normal tissues. The NCCN has noted that postoperative radiotherapy is suitable in cases with extra capsular lesions, the invasion of two or more neck or parotid gland lymph nodes, a single nodule ≥3 cm and neck dissection without distant invasion or postoperative LR. The use of conventional radiotherapy doses (2 Gy per fraction, at a total dose of 60–66 Gy or 70 Gy) is recommended. In addition, secondary damage from radiotherapy is regarded as acceptable. The current study confirmed that although the adjuvant treatment approach outlined above produced good local control effects, it did not reduce the risk of death. The reasons for this phenomenon are analyzed below.
First, HNMM is a radiotherapy-resistant tumor; that is, HNMM tumors exhibit extremely strong sublethal damage repair capacities. Therefore, although conventional radiotherapy doses can achieve local control, traditional radiotherapy is ineffective in controlling the disease as a whole. For this reason, stereotactic intensity-modulated radiation therapy (IMRT), PT and CIT have recently been developed for the adjuvant treatment of HNMM. Although these novel approaches have exhibited enhanced efficacy relative to conventional radiotherapy, they remain at the clinical trial stage; thus, numerous clinical trials and prospective studies will be required to determine the long-term efficacy of these treatments.
Second, mucosal melanoma is a systemic disease that readily invades vascular and lymphoid tissues and therefore characteristically tends to produce DM. Thus, it is difficult for local radiotherapy to achieve good control of the hematogenous metastasis of this disease . Notably, many mucosal melanoma patients die from the spread of the disease rather than from LR. Rinaldo et al.  have demonstrated that in contrast to primary squamous cell carcinoma of the nasal cavity and the paranasal sinuses, which mainly exhibits posterior pharyngeal wall and mandibular lymph node metastasis, mucosal melanoma more frequently exhibits lung and brain metastasis . Consequently, although postoperative adjuvant radiotherapy has relatively strong local control effects, it is not a protective factor for HNMM because the main cause of death in certain HNMM patients is DM.
Finally, because of the rarity of HNMM, it is extremely difficult to conduct large-sample RCTs for the examination of this disease. Therefore, selection bias in the meta-analysis of cohort studies will have caused uncertainty in the results of the present study. In other words, in the included retrospective studies, most patients who received postoperative radiotherapy were late-stage cases involving failed en bloc resection, positive safety margins, lymphatic vascular invasion and/or DM. However, it was determined that for these patients, postoperative radiotherapy could still achieve local control but did not reduce the risk of death.
Value of other treatment methods in non-surgical therapy
As discussed above, surgical treatment is difficult to implement in HNMM cases because of various issues including: complications associated with the anatomical location of a tumor (particularly in the case of SNMMs located near important organs); the advanced stage of the disease upon initial diagnosis; patient refusal of surgical treatment; and the comprehensive consideration of surgical risk and postoperative quality of life. Consequently, to increase survival rates and prolong patient survival times in HNMM cases, many researchers advocate the use of radiotherapy, chemotherapy and/or biological therapy alone as alternatives to surgical treatment.
Evaluation of the therapeutic efficacy of simple radiotherapy for HNMM
In recent years, with the continuous advancement of radiotherapy techniques and the upgrading of relevant equipment, clinicians have gradually begun to increase the use of radiotherapy alone in the treatment of primary HNMM. Although the development of specific therapeutic strategies remains controversial, studies have already proven the efficacy of this approach and have demonstrated that the side effects of the radiotherapy are within manageable ranges.
Douglas et al.  recommended radiotherapy as the preferred treatment approach for HNMM patients who cannot withstand surgery or refuse surgical treatment. In a retrospective study of 68 HNMM patients, 13 patients received palliative radiotherapy, 30 received radiotherapy at therapeutic doses and 25 received surgical treatment with or without postoperative radiotherapy. Radical treatment mainly using radiotherapy achieved relatively good local control effects and a 5-year DSS of 25 %. However, there have been few relevant reports on the use of radiotherapy at therapeutic doses, and these reports have involved small sample sizes. These limitations occurred because HNMM is a radiotherapy-resistant tumor with extremely effective sublethal repair capabilities. Therefore, therapeutic hypofractionated radiotherapy (HF-RT) with an α/β ratio of 4–6 Gy per fraction has been used to overcome this radiotherapy resistance. The basic dose for the treatment of uveal melanomas is 1.5–26.2 Gy, with an average of 10 Gy, and the basic dose for cutaneous melanomas is 1.6–6 Gy, with an average of 2.5 Gy; however, the dose for mucosal melanoma remains unknown . Wada et al.  performed high-dose HF-RT with doses ≥3 Gy in 31 HNMM patients. They found that compared with conventional postoperative radiotherapy, HF-RT delivered better local control and survival; in particular, HF-RT resulted in 1- and 3-year DSS rates of 73 and 33 %, respectively. Moreover, increased radiotherapy doses produced better therapeutic effects with respect to local control and survival. However, with the aforementioned radiotherapy regimen, the occasional occurrence of lethal side effects, such as mucosal ulceration and massive hemorrhage, is inevitable. Christopherson et al.  administered therapeutic doses of radiotherapy or postoperative radiotherapy to 21 patients. They found that three patients (17 %) experienced serious complications, including the need for hospitalization and/or surgical intervention, failing to complete the entire radiotherapy regimen, and death. Two of these patients received PT alone, and one patient received photon therapy and PT. The manifestations of the complications included bilateral blindness, severe mucositis and skin necrosis. Thus, numerous clinical studies involving HF-RT are still required to develop regimens with optimal therapeutic effects and ensure that the side effects of these regimens are manageable.
Given the side effects of HF-RT, local dose-escalated IMRT, another improved photon radiotherapy technique, has been used in clinical practice. In a study on the treatment of malignant tumors of the paranasal sinuses, Madani et al.  reported that IMRT was the gold standard for radiotherapy approaches for these tumors, and that IMRT could be used alone or as an adjuvant treatment. Combs et al.  used IMRT to treat eight SNMM patients and found a 5-year OS of 80 %, a 3-year local progression-free survival rate of 28.6 % and a 3-year distant progression-free survival rate of 28.6 %. In addition, during the 27 months (12–71 months) of follow-up, radiotherapy-related side effects were within acceptable ranges. No vision loss or blindness occurred, and additional therapeutic intervention was not required. However, both the aforementioned HF-RT approach and this IMRT approach produced non-ideal effects with respect to the control of DM.
At present, PT and CIT particle therapies achieve better local control of melanomas than surgery alone, conventional photon radiotherapy, or a combination of surgery and photon radiotherapy. Therefore, the use of PT or CIT particle radiotherapy alone for HNMM is attracting widespread attention. Because PT and CIT can provide precisely distributed radiation doses, these techniques can be used to specifically deliver high doses of radiation to tumors with a reduced risk of radiation exposure to normal tissues; this characteristic is particularly relevant for malignant tumors located in or near important tissues and organs [28, 35]. In contrast to the high levels of attenuation characteristic of the traditional photon beams used for radiotherapy upon penetrating soft tissues, proton beams exhibit tremendous energy deposition in targeted tumors but minimal energy deposition in the preceding or subsequent normal tissues; consequently, proton beams can be used to accurately deliver radiation to tumors with relatively little effect on the surrounding normal tissues [36, 37]. This characteristic is highly suitable for the treatment of paranasal sinus melanomas because the radiotherapy dose required for the treatment of these tumors is extremely similar to the doses that affect the surrounding normal tissues, such as the eyes, the optic nerves, brain tissues and the skull. Compared with photon beam radiotherapy, PT is associated with lower risks regarding the development of secondary malignant tumors [33, 34]. In an examination of 11 cases treated with PT, the initial disease control rate was 85.7 %, the 3-year survival rate was 58.0 % and the mean disease-free time to progression was 25.1 months (with a mean follow-up time of 36.7 months); there were two cases of unilateral vision loss but no blindness events . In addition, in 72 cases treated using CIT, the 5-year local control rate for the disease was 84.1 %; 94.4 % of the patients who exhibited good local control received radiotherapy with ≥3.6 Gy/fraction. However, by the end of the follow-up, DM had occurred in 40 cases; LR had not occurred in 34 of these cases (85.0 %). These results confirmed that the tumors had often already invaded their microenvironment before treatment had begun and that the examined radiotherapy approach did not produce good control of HNMM that underwent early-stage hematogenous DM . In contrast, compared with proton beams, carbon-ion beams have a better penumbra and higher relative biological effectiveness. However, Demizu et al.  demonstrated that PT and CIT did not exhibit significant differences regarding therapeutic effectiveness in the treatment of HNMM; in particular, the 2-year survival rates among patients treated with PT and CIT were 58 and 62 %, respectively (p = 0.399; Kaplan-Meier [K-M] curve analysis). PT and CIT also had no significant differences regarding local control effects; these treatments resulted in local control rates of 83 and 59 %, respectively (p = 0.569; K-M curve analysis). In addition, the radiotherapy side effects of PT and CIT did not differ significantly. Although CIT and PT particle therapies had excellent local control effects (with an effective 2-year local control rate of 78 %), their outcomes regarding OS were not ideal (a 2-year DSS of 31 %). The analysis appeared to indicate that this phenomenon could be attributed to a relatively high incidence of DM. Thus, although the toxic side effects of radiotherapy could be controlled to within tolerable levels, the effects of particle radiotherapy with respect to controlling DM were not ideal.
In summary, the value of radiotherapy in the treatment of HNMM should receive extensive attention in future studies. With the advancement of radiotherapy techniques and constant upgrading of the relevant equipment, solutions that address the issue of radiotherapy side effects may be developed in the near future. Numerous controlled clinical studies and prospective studies are still required to establish radiotherapy regimens that achieve therapeutic objectives, but produce manageable side effects. In addition, because mucosal melanoma is a systemic disease, therapeutic approaches that combine radiotherapy with chemotherapy or immunobiological therapy can be designed to compensate for the deficiencies of radiotherapy alone, with respect to controlling the high incidence of DM in cases of mucosal melanoma.
Compared with cutaneous melanoma, mucosal melanoma involves more chromosomal abnormalities and abnormal gene copies . At present, gene therapies targeting the mutation hotspots of malignant melanoma have already entered clinical trials. In particular, a phase III clinical trial has confirmed that vemurafenib, a selective BRAF inhibitor, produces a higher response rate (48 % vs 5 %) and 6-month OS (84 % vs 64 %) compared with decarbonizes, another systemic chemotherapeutic agent . In a phase II clinical trial, imatinib, a targeted therapeutic agent for melanomas designed to be effective in cases involving KIT mutations, exhibited outstanding performance; it achieved a significantly higher response rate in melanomas with KIT mutations than in wild-type melanomas (40 % vs 0 %; p = 0.005) . A recent phase II clinical trial confirmed that patients with NRAS (neuroblastoma RAS viral [v-ras] oncogene homolog) mutations could benefit from treatment with MEK1/2 inhibitors (MAP/ERK kinases 1 and 2) . The mutation rates of KIT, NRAS, and BRAF mutations among SNMM patients were 7.91 % (11/139), 12.9 % (18/139) and 2.16 % (3/139), respectively [43, 45–48]. Analysis of the above data revealed that there are relatively low rates of KIT and BRAF mutations among SNMM patients. Consequently, few SNMM patients will benefit from chemotherapeutic drugs that target these two genes.
Rosenberg et al.  demonstrated that the transfer of lymphokine-activated killer (LAK) cells induced by interleukin-2 (IL-2) could increase the survival rate of patients with malignant tumors. In addition, mucosal melanoma is a systemic disease that readily undergoes DM; thus, LAK immunobiological therapy can help to improve the prognosis of mucosal melanoma patients. LAK cellular therapy is an emerging treatment modality that has been applied to mucosal melanoma in recent years [49, 50–52]. The core steps in this treatment are the collection of venous blood, the in vitro incubation of these blood samples with a certain concentration of IL-2, and the subsequent transfusion of the blood back into patients; patients are then treated with sustained administration of an appropriate concentration of IL-2 to maintain LAK activity [52, 53]. Kanetaka et al.  examined postoperative treatment with IL-2 and LAK in 13 HNMM patients and found that the 5-year DSS rates of the group receiving the immunobiological treatment and the untreated group were 67 and 33 %, respectively. DM is regarded as an important cause of death among HNMM patients. Although surgery, radiotherapy and chemotherapy are the most common treatment approaches, they do not produce particularly strong effects with respect to controlling DM. Thus, although biological therapy for HNMM is uncommon, this type of treatment approach nonetheless has potential research value as an adjuvant therapy.
Moreover, in recent years, it has been confirmed that immunobiological therapy produces certain radiation-sensitizing effects on radiotherapy-resistant tumors. The local tissue hypoxia caused by the strong metabolic activity of tumor cells can induce local immunosuppressive effects, which are manifested as tumor-promoting disruptions in normal innate immunity . Tissue hypoxia can induce the local downregulation of the interferon-γ (IFN-γ) immune response pathway and thereby inhibit the activation of various mediating factors, such as CD4+/CD25+ regulatory factors (where CD represents a cluster of differentiation), IL-10 and transforming growth factor-b. A study has demonstrated that the phenotype of activated IFN + CD8+ T cells was associated with a 1.8-fold increase in tumor sensitivity to radiotherapy . Therefore, IFN-γ + CD8+ T cells serve as a bridge between immune and radiotherapy responses, and can produce targeted radiosensitization effects. Clinically available immunological adjuvants and tumor vaccines can become effective tools to simultaneously stimulate tumor immunosurveillance and enhance radiotherapy responses. Although HNMM is highly malignant, has a low incidence, and occurs through molecular mechanisms that have not yet been fully elucidated, there are nonetheless broad prospects for the potential use of biological therapies in the treatment of this disease.
Potential bias in the included NRSs
Greater potential biases exist in NRSs than in RCTs  Among these biases, selection bias is a major concern because factors that influence prognosis occurred unevenly across different groups in the selected studies. In addition, commonly observed imperfections in the NRSs remain present in the reports, and in descriptions of research protocols and the evaluations of measurements that affect prognoses and results. These flaws all contribute to uncertainties in meta-analyses. In particular, because a meta-analysis is a non-experimental observation study, completely uniform control standards could not be achieved in the current meta-analysis; one specific manifestation of this issue is that uniform quality control cannot be performed in each case. Therefore, subgroup analysis (such as separate examinations of different radiotherapy regimens, different clinical and case stages, and cases involving lymphatic vascular invasion) cannot be performed. These issues caused selection bias in evaluations regarding the efficacy of postoperative adjuvant radiotherapy, because in the included retrospective studies postoperative radiotherapy was mostly used in patients with a positive safety margin, local metastasis, DM and/or primary tumors with larger scopes.
Side effects of radiotherapy
HNMM lesions are located near various important tissues and organs, such as the optic nerves and brain tissues. Therefore, during the course of surgery and radiotherapy patients may experience various adverse effects, including optic nerve damage and mucositis, which will affect prognosis. However, the relevant data of the included publications were incomplete, and these publications did not use uniform quality control standards; consequently, the prognostic impact of radiotherapy-related side effects on prognosis was not analyzed. However, relevant reports have indicated that certain side effects affect patient prognoses and survival times.
Radiotherapy subgroup analysis
In the present study the influence of radiotherapy on prognosis, local control and metastasis was analyzed. However, recent advances in radiotherapy techniques and upgrades to radiotherapy equipment have led to the development of novel radiotherapy strategies, such as PT and CIT. In the context of HNMM, these novel strategies are superior to conventional radiotherapy because they can both control LR and improve patient survival. In addition, compared with conventional radiotherapy, these emerging radiotherapy approaches clearly produce reduced side effects. However, there are few relevant reports regarding these approaches, the extant reports involve small sample sizes, and studies of many potential novel radiotherapy regimens remain in the experimental stage. Therefore, future research will be required to address this topic.