Perineural (A) and intraneural (B) invasion of adenoid cystic carcinoma (ACC) and perineural and intraneural invasion of squamous cell carcinoma (SCC) (C) (hematoxylin-eosin, original magnification ×100 [A], ×40 [B], and ×20 [C]; N indicates nerve; arrowheads, tumor); propensity of neural invasion (NI) of ACC, SCC, malignant melanoma (MM), adenocarcinoma (AC), undifferentiated carcinoma (Undif), sinonasal undifferentiated carcinoma (SNUC), and other types of tumors (Others) (D); and diameter of nerves invaded by cancer (E).
Gil Z, Carlson DL, Gupta A, Lee N, Hoppe B, Shah JP, Kraus DH. Patterns and Incidence of Neural Invasion in Patients With Cancers of the Paranasal Sinuses. Arch Otolaryngol Head Neck Surg. 2009;135(2):173-179. doi:10.1001/archoto.2008.525
To characterize the incidence and pattern of neural invasion (NI) in patients with cancers of the paranasal sinuses and anterior skull base.
A tertiary referral cancer center.
The study included 208 patients with cancer of the paranasal sinuses. Patients with brain invasion or neurogenic tumors were excluded.
Main Outcome Measure
Analysis of clinical and pathologic data on patients with cancer of the paranasal sinuses.
Forty-one specimens (20%) had evidence of NI. Sinonasal undifferentiated, adenoid cystic, and squamous cell carcinoma had a high propensity for NI, whereas melanoma and sarcoma rarely invaded nerves. Intraneural invasion was found in 32% of these cases, and 34% invaded more than 1 cm distal to the tumor. Neural invasion was associated with a high rate of positive margins, maxillary origin, and previous surgical treatment (P < .04) but not with stage, orbital invasion, or dural invasion. Patients with NI were more likely to undergo adjuvant radiotherapy (P = .003), which significantly improved survival in patients with minor salivary gland carcinomas (P = .04). Multivariate analysis showed that pathologic evidence of NI was not an independent predictor of outcome.
Paranasal carcinomas have high propensity for NI, whereas melanoma and sarcoma rarely invade nerves. Patterns of NI include both perineural and intraneural invasion. Neural invasion is associated with positive margins, maxillary origin, and previous surgery.
The dissemination of cancer cells along nerves is a frequent pathologic finding among a variety of cancers, including head and neck, upper gastrointestinal, pancreatic, and prostate carcinomas.1 Tumors that have a considerable propensity to disseminate along nerves are known as neurotropic cancers. In the head and neck, the most common tumors with a predilection to invade nerves are adenoid cystic carcinomas (ACCs), followed by squamous cell carcinomas (SCCs).2,3 Tumors involving the paranasal sinuses and skull base represent a group of different histologic entities with distinct clinical implications. These tumors are associated with lower survival rates than other head and neck cancers owing to a delay in diagnosis and a proximity to vital structures (eg, dura, brain orbit and central nerves), which complicates surgical treatment.
In patients with paranasal carcinomas, a poor prognosis is associated with high-grade tumor, intracranial extent, and positive surgical margins.4 There are several controversies related to the impact of neural invasion (NI) on the survival of patients with head and neck cancers. Evidence of NI in patients with head and neck carcinomas was found to be a significant prognostic factor by some authors, while others found no impact on survival or an adverse prognosis only when there was clinical evidence of nerve involvement.5,6 Although NI of salivary glands and cutaneous carcinomas is well described in the surgical and oncologic literature, there is a lack of clinical and pathologic data on tumors involving the paranasal sinuses and skull base. Because of the infrequency of this disease, little is known about the propensity of paranasal sinus cancers to invade nerves and the impact on outcome. Most of the reports on NI of paranasal cancers are based on cohorts of patients with tumors involving various anatomical locations in the head and neck; therefore, analysis of outcomes on specific tumors or sites has not been possible.
Our approach to meet these methodological concerns was to analyze pathologic and clinical data on noncutaneous paranasal sinus, nasal cavity, and anterior skull base cancers treated at Memorial Sloan-Kettering Cancer Center, New York, New York. Neurogenic tumors, as well as tumors with brain involvement, were excluded. Our aims were (1) to characterize the propensity for and patterns of NI in these tumors, (2) to assess the patterns of invasion, and (3) to identify independent predictors of outcome. Recognition of the biologic behavior of these tumors and the clinical significance of NI may have implications for the treatment and prognosis of patients with NI and may determine subsequent surgical technique with regard to the extent of resection. To our knowledge, this study represents the first clinicopathologic analysis of patterns of NI in this group of tumors.
This retrospective study was based on a review of the hospital and outpatient clinical records of patients who underwent surgery for the excision of malignant paranasal sinus tumors at Memorial Sloan-Kettering Cancer Center between 1973 and 2005. Because our goal was to study the prognostic significance of NI, patients with brain invasion or those with tumors of neural origin (eg, esthesioneuroblastoma and malignant peripheral nerve sheath tumors) were excluded. Also, cases involving skin cancers that invaded the skull base and paranasal sinuses were not included because such tumors represent a different disease than paranasal carcinomas.
Our study cohort included 208 patients with cancers of the paranasal sinuses, nasal cavity, and skull base. The patients ranged in age from 6 to 88 years, with a median age of 53 years. Table 1 presents their demographic and clinical data. The distribution of various tumor types among these patients is depicted in Table 2. The follow-up interval ranged from 1 to 234 months, with a median of 20 months.
Neural invasion was assessed as positive when cancer cells were seen in the perineurium. The pattern of NI was analyzed according to the presence of the following parameters: perineural invasion, intraneural invasion, local invasion, distal invasion, and nerve diameter (≤1 mm vs >1 mm or <2.5 mm vs ≥2.5 mm). Perineural invasion was defined as invasion of cells in the perineural space, without intrafascicular infiltration. Intraneural invasion was defined as infiltration of cancer cells between the nerve fascicles. Local invasion was defined as invasion of nerves 1 cm or less from the main tumor, whereas distal invasion was defined as invasion of cancer cells more than 1 cm along the nerve outside the main tumor. Tumor staging was performed according to the 2006 American Joint Committee on Cancer TNM classification for paranasal sinus carcinomas.
Overall survival (OS) and disease-specific survival (DSS) rates were calculated using the Kaplan-Meier method, and univariate comparisons between groups were performed using the Wilcoxon test. A P value of .05 or less was considered significant, and significant factors were entered into multivariate analysis using the Cox proportional hazards model. Nonparametric qualitative and quantitative comparisons were performed using the Fisher exact test (StatCalc 2.0; University of Louisiana, Lafayette).
Our first goal was to find out how frequently cancers of the paranasal sinuses invade nerves. The pathologic specimens of 208 malignant tumors of the paranasal sinuses and skull base were analyzed. Of these tumors, 41 had evidence of NI (20%). Table 2 shows the incidence of NI among the different histologic types. The highest propensity for NI was found in sinonasal undifferentiated carcinoma (60%) and ACC (55%). Neural invasion was rare in sarcomas, and none of the mucosal melanomas had NI. Squamous cell carcinoma had a moderate tendency to invade nerves (22%). The Figure shows the relative tendency for NI among the various tumors. Then, we investigated the pattern of invasion among tumors with NI.
Table 3 shows that two-thirds of the tumors had perineural invasion and one-third had intraneural invasion. The Figure (A through C) shows examples of tumors with perineural invasion, intraneural invasion, and both. Adenocarcinoma and sinonasal undifferentiated carcinoma (SNUC) invaded only along the perineural space, while all undifferentiated, mucoepidermoid, and myoepithelial carcinomas, as well as carcinoma ex pleomorphic adenoma, had intraneural invasion. Approximately one-third of the ACCs and SCCs had intraneural invasion.
Tumors also varied in their ability to migrate along nerves. Two-thirds of the tumors invaded nerves locally (at the tumor center), whereas one-third showed distal NI (Table 3). The highest propensity of distal NI was found in minor salivary gland carcinomas (25%-100%). Tumors tended to migrate unilaterally toward the nerve ganglion. The diameter of nerves varied from 1 to 21 mm, with a median size of 2 mm. The Figure shows the distribution of the diameters of the nerves among tumors with NI.
Table 4 shows the histologic characteristics of tumors with NI. Most of the cancers showing NI had tumor necrosis, whereas bone, muscle, and lymphovascular invasion was found in only 8% to 59% of the cancers. Interestingly, neural inflammation was found only in SCCs. Skip lesions were not encountered in our study.
Contingency analysis of different variables among patients with and without NI was performed (Table 5). In all tumor types, NI was associated with a higher risk of positive margins and with tumors originating in the maxillary complex. Overall, pathologic evidence of NI was associated with a 1.8-fold increase in risk for positive margins (P<.001). Patients with NI were more likely to undergo adjuvant therapy in the form of postoperative radiation treatment.
The impact of NI on 5-year DSS and OS was calculated using the Kaplan-Meyer method and analyzed using univariate and multivariate analysis. Table 6 shows the impact of various parameters on OS in patients with carcinomas of the schneiderian epithelium (SCC, SNUC, and undifferentiated carcinoma). With a median follow-up of 17.2 months (range, 1-234 months), the OS among patients with paranasal epithelial carcinomas was 66% for those without NI and 54% for those with NI (P = .01, log-rank test). However, this difference lost significance on multivariate analysis. Similarly, the pattern of invasion, nerve size, and clinical evidence of NI were not associated with survival. The 5-year DSS rate also was not statistically significant between the 2 groups (eTable 1http://www.archoto.com).
Table 7 shows the results of our analysis of predictors of outcome in patients with minor salivary gland carcinomas. Previous radiation therapy, adjuvant radiotherapy, and orbital invasion were associated with survival on univariate analysis. Orbital invasion and adjuvant radiotherapy remained predictors of outcome after multivariate analysis. Orbital invasion decreased 5-year OS by 50%, and adjuvant radiotherapy improved it by 36% (with relative risks of 2.67 and 0.58, respectively). Among patients with minor salivary gland tumors (including ACC, adenocarcinoma, mucoepidermoid carcinoma, carcinoma ex pleomorphic adenoma, and myoepithelial carcinoma), NI was not a significant predictor of poor outcome (P = .20). On multivariate analysis, NI as well as the pattern of invasion (peripheral, central, intraneural, or perineural invasion), nerve size, and clinical evidence of neural involvement also had no impact on survival. Analysis of prognostic factors influencing the 5-year DSS are shown in eTable 2.
Head and neck tumors are notorious for their ability to invade nerves. Adenoid cystic carcinomas and SCCs are among the most common neurotrophic tumors in this group. The rate of NI varies from 20% to 80% in patients with ACC of various locations5,7 and from 2% to 50% in those with head and neck SCC.3,8 Previous studies and our results indicate that the rate of NI is determined not only by histologic type but also by the anatomical location of the tumor. Nevertheless, detection of NI also depends on the meticulousness of the examining pathologist. For example, it was shown that NI is common in oral cavity carcinomas and rare in tumors arising in the paranasal sinuses.3
The data from our clinicopathologic study demonstrate that pathologic evidence of NI is most frequent with ACCs, SNUCs, and SCCs (55%, 60%, and 30%, respectively) and is rare in sarcomas and melanomas. Neural invasion was associated with positive margins but not with tumor stage, dural invasion, or orbital invasion. On this basis, patients with NI were 1.4-fold more likely to receive radiotherapy after surgery than patients without NI.
Previous studies have shown contradictory results regarding the impact of NI on survival in cases of head and neck SCC.3,8,9 These discrepancies are most likely attributable to differences in the locations of tumors, since NI was shown to have an impact on survival in patients with oral tongue carcinomas but not in those with oropharyngeal carcinomas.10 Although involvement of a named nerve has been previously reported to be an independent prognostic factor, in our cohort nerve size and clinical evidence of neural involvement were not associated with survival. Because of the anatomical complexity of the paranasal sinuses and the large number of patients with recurrent disease, we could not reliably estimate pathologic involvement of a named nerve. There are several possible explanations for these seemingly conflicting results between our study and the ones mentioned above. First, our study represents a more uniform population of patients with tumors isolated to the paranasal sinuses, as opposed to tumors in various locations. It is important to note that in most of these studies, fewer than 20% of the patients had tumors involving the paranasal mucosa; therefore, the differences between these studies negate any possibility of reliable comparison. Second, tumors of the paranasal sinuses and skull base are associated with a dismal prognosis. It is plausible that the aggressive behavior of these tumors and their proximity to vital structures masked the importance of NI and its impact on survival in our study. Finally, because patients with NI were more likely to receive postoperative radiotherapy, it is plausible that variations in treatment may have contributed to the lack of difference in prognosis. Indeed, we found that, at least for minor salivary gland tumors, postoperative radiation therapy substantially improved overall survival by 36%. It is therefore difficult to determine whether NI is a predictor of poor prognosis, since it influenced the choice of adjuvant therapy.
We have shown that ACCs and SNUCs have the highest propensity to invade nerves. Hanna et al11 recently investigated the rate of NI in patients with skull base ACCs and found that 66% of their specimens had NI. The importance of NI in ACC is controversial, and most of the previous studies found no prognostic impact on survival.12,13 An early study from M. D. Anderson Cancer Center, Houston, Texas, evaluated 160 patients with malignant minor salivary gland tumors, one-fourth with tumors originating in the paranasal sinuses.14 The authors found that cases involving NI in general, as well as those with involvement of a named nerve, were not associated with survival.14 Similar studies of ACC showed that NI was associated with an increased failure rate but had no impact on outcome.15 In contrast, Garden et al5 suggested a correlation between cancer invasion of a named nerve and outcome. However, multivariate analysis was not performed in that study. A recent study by Mendenhall et al16 analyzed the importance of clinical evidence of NI on patients with ACC of all anatomical sites and found a significant impact on survival. Similar to cases of SCC, it is possible that proximity of paranasal tumors to the skull base and other vital structures limited the impact of NI on survival.
Our histologic analysis showed that local tumor invasion to bone, muscle, lymphatics, and vessels had no association with NI. Most importantly, we found no statistical correlation between NI and the ability of the tumor to invade adjacent structures such as the orbit and dura. Furthermore, tumor stage also was not associated with NI. Taken together, these results suggest that the tendency of tumors to disseminate along nerves is not associated with their ability for local invasion and proliferation. It was previously hypothesized that specific interactions between cancer cells and nerves may contribute to neurotropism. In congruence with this concept, recent molecular studies have shown an association between expression of specific adhesion molecules and NI in head and neck cancers.17 Future studies of the tumor-nerve microenvironment are essential to reveal the nature of these interactions.
Early studies of the patterns of NI of head and neck carcinomas, as well as recent data, suggest that cancer dissemination along nerves occurs exclusively along the perineural space and rarely within the nerves' fascicles.18 In contrast, in the present study, we found that up to one-third of the cancers with NI showed invasion within the nerve bundles. A study from the University of Pittsburgh, Pittsburgh, Pennsylvania, on anterior, middle, and lateral skull base ACCs also found that 39% of the tumors showed intraneural invasion.19 On the basis of this finding, we named the process of cancer dissemination along nerves neural invasion instead of perineural invasion to describe this biologic feature.
Our work showed an association between NI and positive margins. The goal of surgical care for patients with paranasal carcinomas should be complete tumor resection with negative margins. However, intraoperative decision making should also take into consideration the potential morbidity of the procedure; therefore, we believe that negative margins should not be pursued at the cost of additional morbidity when complete removal of all gross tumor has been achieved. When NI of a major nerve is suspected during surgery, it should be confirmed histologically using frozen section, unless there is clear evidence of gross NI. Intracranial resection of the tumor may be carried out if there is evidence of macroscopic disease, but the potential morbidity of intracranial surgical exploration and its impact on quality of life should be taken into consideration.20
In our study, postoperative radiotherapy (median dose, 6000 cGy) was found to be associated with a better outcome in patients with minor salivary gland carcinomas. Previous studies on ACC of various origins showed that adjuvant radiotherapy leads to enhanced local control of disease after surgery.16 A recent study of 129 patients with ACC treated at The Cleveland Clinic, Cleveland, Ohio, reported local control benefit of postoperative radiation therapy only in cases involving advanced tumors and/or with positive margins.21 In view of the prognostic benefit of postoperative radiotherapy, the authors routinely include the base of the skull in the radiation fields in these cases.
Peripheral nerves are relatively resistant to direct invasion of cancer cells. This resistance could be the result of the structural characteristics of the peripheral nervous system. The axonal environment in the endoneurium of peripheral nerves is isolated from the extracellular matrix by a diffusion barrier called the blood-nerve barrier.22 The blood-nerve barrier consists of tight and adherence junctions between both perineural and endothelial cells in the endoneurium. It has been suggested that the blood-nerve barrier is responsible for the resistance to tumor invasion.23 The process of cancer spread into nerves consists of various stages, including migration, invasion, and tumor growth. Cancerous tissue can influence the functional and structural integrity of the nerve by compressing it without genuine invasion of the sheath. Alternatively, the tumor can perforate the nerve sheath directly and invade the perineural space, or it can disseminate in between the nerve fascicles. Biologic substances secreted in the tumor-nerve microenvironment may play a role in different stages of neural invasion. Extraneural, perineural, and intrafascicular invasion may represent different stages in tumor dissemination. Conversely, intraneural invasion may represent a more aggressive biologic behavior of tumors. Our multivariate analysis did not find significant correlation between survival and pattern of invasion. However, a larger cohort is needed to find out whether the pattern of neural invasion is associated with prognosis.
Previously, it was assumed that cancer cells spread along nerve sheaths via perilymphatic vessels or by means of the route of least resistance.1 However, this theory has been refuted for 2 main reasons: (1) it was demonstrated that the perineurium is devoid of lymphatic vessels; and (2) it was shown that specific tumors present with profound neural invasion even at an early disease stage, whereas other aggressive tumors do not infiltrate nerves even at advanced stage. An alternative hypothesis is that nerve cells express soluble or cell-to-cell signaling proteins that could initiate and sustain cancer invasion if the cancer cells have the appropriate receptors.24 These neuroneoplastic interactions may facilitate tumor progression in patients with head and neck tumors or related malignant neoplasms. Candidate factors that may contribute to neural invasion include neurotrophins, neurotransmitters, and adhesion molecules.25 Cellular proteins that were implicated in neural invasion of head and neck cancers include ICAM-5,26 laminin 5,17 and nerve cell adhesion molecule.7,27,28 Why the propensity of cranial nerve invasion differs among carcinomas (most common), sarcomas (less common), and melanoma (least common) is currently unknown. Answers to these questions could lead to a better understanding of the pathogenesis of NI by cancer.
In conclusion, our study demonstrated that, overall, ACC and SNUC have a high propensity for NI, whereas melanoma and sarcoma rarely invade nerves. Surprisingly, one-third of the tumors involving nerves showed invasion within the neural bundles, and 34% had distal NI. Neural invasion by cancer was associated with a high rate of positive margins, with maxillary origin, and with previous surgical treatment. Patients with evidence of NI were more likely to undergo adjuvant radiotherapy. Overall, NI was not an independent prognostic factor of outcome. Further studies are needed to verify the impact of NI on the survival of patients with paranasal carcinomas.
Correspondence: Ziv Gil, MD, PhD, Department of Otolaryngology–Head and Neck Surgery, Tel Aviv Sourasky Medical Center, 6 Weizmann St, Tel Aviv 64239, Israel (firstname.lastname@example.org).
Submitted for Publication: January 17, 2008; final revision received April 7, 2008; accepted May 5, 2008.
Author Contributions: Dr Gil had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Gil, Lee, Shah, and Kraus. Acquisition of data: Gil, Carlson, Gupta, and Hoppe. Analysis and interpretation of data: Gil, Carlson, and Lee. Drafting of the manuscript: Gil and Carlson. Critical revision of the manuscript for important intellectual content: Gil, Carlson, Gupta, Lee, Hoppe, Shah, and Kraus. Statistical analysis: Gil. Obtained funding: Gil. Administrative, technical, and material support: Hoppe. Study supervision: Lee, Shah, and Kraus. Pathologic analysis: Carlson.
Financial Disclosure: None reported.
Funding/Support: This study was supported in part by the Israeli Association of Otolaryngology–Head and Neck Surgery (Dr Gil).
Additional Information: eTables 1 and 2 are available at http://www.archoto.com.