Outcomes in surgical management of sinonasal malignancy—A single comprehensive cancer center experience

Abstract

Background:

Sinonasal malignancies are a complex and diverse group of tumors. Over the past five decades, treatment advances have changed the management paradigms for these tumors. Our aim was to analyze the outcomes of patients from a comprehensive cancer center.

Materials and Methods:

We retrospectively assessed 400 patients with sinonasal malignancies treated with surgery at our center between 1973 and 2015. Multiple variables were reviewed to assess the influence on 5-year outcomes.

Results:

The median age was 56 years (IQR 46.8–68). Two hundred and fifty-nine (65%) were males and 141 (35%) were females. Overall survival (OS) and disease-specific survival (DSS) improved in the last analyzed decade. Orbital invasion, advanced pT-classification and pN-classification, and melanoma histology were associated with poorer outcomes.

Conclusion:

Treatment outcomes for patients with sinonasal malignancy have improved over time. This is likely multifactorial with advances in surgical technique, adjuvant treatment, and patient selection. pT-classification, pN-classification, orbital invasion, and histology are predictive of survival.

1 |. INTRODUCTION

Sinonasal tumors are a rare and diverse group of malignancies. Complicated anatomy and a large variety of histology render these tumors challenging to manage. Tumors of the sinonasal cavity tend to progress with mild non-specific symptoms, invading adjacent tissues and extending to the skull base and orbit. Treatment of these tumors has evolved in the last 50 years with studies failing to demonstrate improvement in outcomes.^1,2 Despite their diversity, the mainstay of treatment for most remains surgery followed by adjuvant treatment. In addition to the gold standard of open surgery, over time there has been an evolution in surgical technique to include combined open and endoscopic and endoscopic procedures alone, which utilize advances in imaging, instrumentation, technique, and intraoperative navigation.^3–6 The evolution of microsurgical techniques has rendered reconstruction feasible even following the most extensive defects.^7–9 The role of adjuvant and neoadjuvant treatment has increased with the development of new chemotherapy, targeted therapy, immunotherapy drugs as well as refinements of radiotherapy including the introduction of IMRT and proton therapy.¹⁰ We have analyzed the outcomes of patients treated surgically for sinonasal malignancy in a single comprehensive cancer center over five decades to evaluate the impact of the specific patient, tumor, and surgical technique factors on oncologic outcome.

2 |. METHODS

Following approval of the Memorial Sloan Kettering Cancer Center Institutional Review Board, we retrospectively identified 400 patients with sinonasal malignant tumors treated with surgery at our center between 1973 and 2015. The clinical and histologic characteristics of the cohort can be found in Table 1. Patients were pathologically staged according to the eighth edition of the American Joint Committee on Cancer (AJCC). Statistical analysis was performed using R (version 3.6.2; R Foundation for Statistical Computing, Vienna, Austria). The follow-up interval was calculated in months from the date of initial curative surgery at our institution. Interventions received before the main tumor treatment performed in our center were classified as a prior treatment. Several variables including sex, age, type of tumor, comorbidity, prior treatment, skull base resection, intracranial invasion, surgical margins, histology, pT, pN, pM, adjuvant therapy, and a decade of surgery were reviewed to assess influence on outcomes. The resection of the skull base structures (bone, dura, or the brain) was classified as a skull base resection case. Tumor invasion to the above-mentioned structures as specified in the pathology report was classified as bone, dura, and brain invasion. Orbital invasion was classified as periosteum/bone or intraorbital contents depending on the structures invaded by the tumor. We divided the cohort into four separate groups according to decade of surgery (1976–1985, 1986 1995, 1996–2005, and 2006–2015). The Kaplan–Meier method was used to compare outcomes of interest, and a log-rank test was performed to compare the groups. Multivariable analysis of these factors was performed using the Cox proportional hazard regression model. A p-value of less than 0.05 was considered statistically significant.

TABLE 1.

Patient demographics, tumor characteristics, and treatment details

Characteristics	No. of patients (%)
Age (IQR)	56.0 (IQR 46.8–68.0)
Sex
Male	259 (65)
Female	141 (35)
Comorbidity
None	257 (64)
Present	137 (34)
No data	6 (2)
Type of tumor
Primary	340 (85)
Recurrent	60 (15)
Tumor site
Nasal cavity	175 (44)
Ethmoid sinus	113 (28)
Maxillary sinus	112 (28)
Histology
Adenocarcinoma	42 (11)
Esthesioneuroblastoma	48 (12)
Melanoma	56 (14)
Salivary gland	40 (10)
Sarcoma	42 (11)
Squamous cell carcinoma	131 (33)
Basal cell carcinoma	1 (<1)
Other	40 (10)
Orbital invasion
None	305 (76)
Periosteum/bone	95 (24)
Intraorbital contents	52 (13)
Intracranial extension
None	274 (69)
Bone invasion	90 (23)
Dural invasion	27 (7)
Brain invasion	9 (2)
Skin invasion
None	389 (97)
Present	11 (3)
T classification (AJCC eighth edition)
T1	28 (7)
T2	43 (11)
T3	84 (21)
T4	230 (58)
Unknown	15 (4)
N classification (AJCC eighth edition)
N−	373 (93)
N+	18 (5)
Unknown	9 (2)
M classification (AJCC eighth edition)
0	391 (98)
1	1 (<1)
Unknown	8 (2)
Type of treatment
Prior treatment
No	239 (60)
Yes	161 (40)
Previous surgery
No	15 (4)
Yes	146 (37)
Previous radiotherapy
No	116 (29)
Yes	45 (11)
Previous chemotherapy
No	140 (35)
Yes	21 (5)
Neoadjuvant treatment
No	382 (96)
Yes	18 (4)
Neoadjuvant radiotherapy
No	18 (4)
Yes	1 (<1)
Neoadjuvant chemotherapy
No	0
Yes	18 (4)
Adjuvant treatment
Yes	271 (68)
No	129 (32)
Adjuvant radiotherapy
Yes	260 (65)
No	11 (3)
Adjuvant chemotherapy
Yes	47 (12)
No	224 (56)
Details of the surgical treatment and tumor extension
Type of approach
Open	360 (90)
Endoscopic	24 (6)
Combined	16 (4)
Resected structures
Orbit	160 (40)
Ethmoid	293 (73)
Nasal cavity	389 (97)
Maxilla	345 (86)
Skin	28 (7)
Bone/cribiform plate	258 (65)
Dura	172 (43)
Brain	21 (5)
Other	56 (14)
Skull base resection
Yes	258 (65)
No	142 (36)
Neck dissection
Yes	38 (10)
No	362 (90)
Tracheostomy
Yes	49 (12)
No	351 (88)
Reconstruction
No reconstruction	105 (26)
Skin graft	66 (17)
Free flap	96 (24)
Galeal pericranial flap	161 (40)
Alloplastic material	83 (21)
Other	34 (9)

3 |. RESULTS

The median age was 56 years (IQR 46.8–68). The patienťs clinical and pathologic characteristics are given in Table 1. Two hundred and fifty-nine (65%) were males and 141 (35%) were females. Two-thirds of patients, 257 (65%), had a documented medical comorbidity. The majority of tumors were primary (n = 340, 85%). The nasal cavity (n = 175, 44%) was the most prevalent site, followed by ethmoid, and maxillary sinus tumors.

3.1 |. Prior treatment

One hundred and sixty-one (40%) patients received prior treatment (Table 1). One hundred and forty six (37%) patients had different types of previous surgery, 101 (25%) of them were incomplete resections performed in other centers. Prior radiotherapy was given to 45 (11%) and chemotherapy 21 (5%) patients. Neoadjuvant treatment had been given to 18 (4%) patients, all of whom had chemotherapy and one had neoadjuvant radiotherapy.

3.2 |. Details of surgery and pathology

Surgical details are provided in Table 1. Craniofacial resections were performed in 258 (67%) patients. An open craniofacial approach to resect the tumor was performed in 360 (90%), endoscopic or combined open and endoscopic in 40 (10%) patients. Ninety percent (n = 36) of endoscopic cases were performed in the last analyzed decade (Table S1). Contents of the nasal cavity, maxilla, and ethmoid were the structures resected in the majority of cases. Dural resections were carried out in 172 (43%) patients, and brain parenchyma was resected in 21 (5%) patients. Different types of reconstruction were used in 295 (73%) patients; free flaps were used in 96 (24%) cases. Squamous cell carcinoma (SCC) was the most prevalent tumor histology 131 patients (33%) followed by melanoma (14%), esthesioneuroblastoma (12%), sarcoma (11%), adenocarcinoma (11%), salivary gland tumors (10%), and other rare types (10%).

3.3 |. Adjuvant therapy

Two-thirds of patients received adjuvant treatment with postoperative radiotherapy (PORT) being the most utilized treatment (Table 1). PORT was given in 260 patients out of 271 cases with adjuvant treatment. In nine cases, radiotherapy was given as re-irradiation to previously treated patients. Chemotherapy was given in 47 patients and two patients received immunotherapy.

3.4 |. Outcomes

The median follow-up of the cohort was 66 months (IQR 22.8–118.2). The 5-year overall survival (OS) and disease-specific survival (DSS) were 63.9% and 67.2%, respectively (Figure 1A,B). Patients with esthesioneuroblastoma histology had the best survival (OS: 87.4%, DSS: 89.3%) and patients with melanoma had the worst OS and DSS (OS: 31%, DSS: 36%) (Figures 2A and 3A). Local recurrence occurred in 136 (34%) patients with a median time to recurrence of 38 months (IQR 8.6–88.2). The 5-year local recurrence-free probability (LRFP) was 65.4%. Esthesioneuroblastoma (5 years LRFP—82%) was the least likely and melanoma (5 years LRFP—46.3%) was the most likely histology type to develop a local recurrence. Five-year OS, DSS, LRFP, and regional recurrence-free probability (RRFP) Kaplan–Meier survival curves for the whole dataset are shown in Figure 1. Histology types, pT classification, intracranial extension, and orbital invasion were statistically associated with OS and DSS (Figures 2B–D and 3B–D).

FIGURE 1.

OS, DSS, LRFP, and RRFP Kaplan–Meier curves for the whole dataset

FIGURE 2.

OS stratified by histology, intracranial invasion, orbital invasion, and pT classification

FIGURE 3.

DSS stratified by histology, intracranial invasion, orbital invasion, and pT classification

Furthermore, we stratified patients depending on the decade of the surgery to four decades and compared OS and DSS improvement (Figure 4A–C). Both OS and DSS improved in the last decade from 2006 to 2015 with 5-year OS 68.8% and DSS 75.1%. To investigate the impact of histology on outcomes, we analyzed the distribution of histology types in each decade. Melanoma was less prevalent in the first two decades (I decade - 8%, II decade – 7%), with increase in the III decade (17%), followed by a decrease in the IV decade (15%). Esthesioneuroblastoma cases on the other hand were more prevalent in the early two decades and less prevalent in the last two decades, I decade—12%, II decade—19%, III decade—10%, and IV decade—11% (Table S1).

FIGURE 4.

OS and DSS stratified by decades of surgery

3.5 |. Multivariable analysis of the main outcomes

On multivariable analysis of local recurrence, orbital invasion, melanoma histology, adjuvant treatment, and pT classification were significant after controlling for other important factors (Table S2). The presence of skull base resection compared to other surgical variables was not significant in multivariable analysis for LRFP.

Orbital invasion, pN-classification, melanoma histology, adjuvant treatment, skull base resection, and brain involvement were significant predictors of DSS on multivariable analysis (Table 2). Age was a significant predictor on multivariable analysis for OS, in addition to orbital invasion, pT-classification, pN-classification, intracranial involvement, and melanoma histology (Table S3).

TABLE 2.

Prognostic variables for disease-specific survival using Cox proportional hazard regression model

		Univariable HR (95% CI)	p	Multivariable HR (95% CI)	p
Age	Continuous, years	1.003 (0.9931–1.013)	0.562
Sex	Male	Ref	0.521
	Female	1.11 (0.8022–1.544)
Type of tumor	Primary	Ref	<0.001a	Ref	0.466
	Recurrent	1.92 (1.318–2.803)		0.72 (0.2895–1.7646)
Comorbidity	No	Ref	0.495
	Yes	0.89 (0.6419–1.239)
Prior treatment	No	Ref	0.159	Ref	0.556
	Surgery	1.32 (0.507–1.118)	<0.001a	0.87 (0.5443–1.3871)	0.495
	RT	2.33 (1.544–3.510)		1.41 (0.5283–3.7415)
Adjuvant treatment	No	Ref	0.757	Ref	0.032 a
	Yes	0.94 (0.6705–1.338)		0.61 (0.3914–0.9596)
Skull base resection	No	Ref	<0.001a	Ref	0.024 a
	Yes	2.03 (1.383–2.991)		1.81 (1.0807–3.0342)
Intracranial invasion	No	Ref	0.025a	Ref	0.764
	Bone	1.51 (1.054–2.169)	0.045a	1.07 (0.6843–1.6759)	0.104
	Dura	1.75 (1.012–3.032)	0.001a	1.65 (0.9014–3.0338)	0.024 a
	Brain	3.91 (1.700–8.976)		2.95 (1.1581–7.5112)
Histology	Non-melanoma	Ref	<0.001a	Ref	<0.001a
	Melanoma	3.01 (2.075–4.378)		5.19 (3.3016–8.1482)
pT (AJCC 8th Ed)	pT1–2	Ref	<0.001a	Ref	0.054
	pT3–4	2.54 (1.736–3.701)		1.58 (0.9918–2.5030)
pN (AJCC 8th Ed)	Nx/N0	Ref	0.004a	Ref	0.008 a
	N+	2.41 (1.331–4.347)		2.59 (1.2868–5.2388)
Margin	Negative	Ref	0.017a	Ref	0.268
	Positive	1.70 (1.265–2.296)		1.25 (0.8464–1.8221)
Orbital invasion	No	Ref	<0.001a	Ref	<0.001a
	Yes	2.56 (1.846–3.551)		2.32 (1.5675–3.4460)

^aSignificant difference on the univariable/multivariate analysis.

4 |. DISCUSSION

We report 50 years of experience in the treatment of sinonasal tumors at a single comprehensive cancer center. We describe the impact of various patient, tumor, and treatment-related factors on outcomes of these rare and diverse group of malignant diseases. The majority of patients present with advanced disease encompassing several anatomic regions and, in some cases, extending to the skull base. In general, surgery remains the primary treatment for tumors amenable to resection, but in addition to the once gold standard open surgery, there has been a shift to endoscopic or combined (open and endoscopic) resections when appropriate. At the same time, free flaps provide safe reconstruction for tumors with substantial calvarium, intracranial, intraorbital, and softtissue extension. With improved knowledge of tumor behavior, patterns of growth, aggressiveness, response to systemic agents, and surgical complications, it is now possible to better select patients suitable for surgical resection.

Although a substantial part of the cohort had extensive tumors requiring skull base resections, the 5-year OS and DSS were 63.9% and 67.2%. Skull base resections were performed in 65% of patients and it was a significant predictor for DSS after adjusting to multiple variables, but not for OS. The absence of influence on OS might represent that skull base resections are often performed to achieve a clear surgical margin, in a tumor otherwise not invading bone/dura. Similar results were reported by Dulguerov et al, which suggests no difference in outcomes between the types of performed surgery, whether it is maxillectomy or craniofacial resection.¹¹

Intracranial invasion (dura/brain) had a significant influence on DSS and OS. Intracranial invasion was significant after controlling for histology type, pT classification, pN classification, orbital invasion, adjuvant therapy, and surgical margins. Previous studies have reported similar results with worse prognosis in patients with involvement of the dura and deeper structures.^3,12,13 Brain involvement is by far the most serious adverse feature, with a significant influence on DSS (Hazard ratio [HR] 2.95, p = 0.024) and OS (HR 2.56, p = 0.039).

Local recurrence was not influenced by dural or brain involvement in our cohort of patients. A possible explanation for this could be the fact that an extended craniofacial resection was performed in these patients, which increases the risk of surgical and general systemic complications, resulting in premature death.¹⁴ More likely, the biology of a locally advanced tumor results in the increased incidence of distant metastasis, which explains the observed significant influence on DSS.

Regardless of the approach, resection with negative margins is the goal in all surgery for sinonasal malignancy. In the case of sinonasal tumors, assessment of margins is challenging due to the complex anatomy, involvement of bony structures, invasion to soft tissues, and intracranial structures beyond the bone. The results from several large retrospective trials show the influence of surgical margins on the local recurrence rate, and even on OS and DSS.^3,11,12 However, depending on the biology of the tumor and histology type as in melanoma, surgical margins and surgical approach can have less influence on the outcomes.¹⁵ Our data suggest that surgical margin has statistical significance on LRFP and DSS on univariate analysis only, with no influence on the outcomes after controlling for other factors.

In some cases, depending on the histology type, the negative prognostic effects of compromised margins might be reduced by the use of adjuvant therapy. In our cohort of patients, adjuvant therapy was a strong positive prognostic factor for LRFP, OS, and DSS, and remained significant even after adjustment for margin status, intracranial invasion, pT classification, and orbital invasion. PORT was by far the most used treatment modality as adjuvant therapy in our cohort of patients. PORT is a widely used treatment modality of sinonasal tumors after the primary surgical resection.^16,17 The results of retrospective single- and multi-institutional trials show controversial results with no improvement of both OS and DSS over several decades, including the 80s and 90s when radiotherapy became widely used after primary sinonasal cancer surgery.^1,2,16,18

However, the more recent analysis of the national cancer database performed by Robin et al. shows marked improvement of outcomes in patients who received PORT (HR, 0.658 [p < 0.001]).¹⁹ In addition, several papers presenting results of using IMRT and proton therapy indicate better results in cancer-related outcomes and in reducing of adverse events of the therapy.^{10,16,17,20–22}

Orbital invasion was another strong predictor of outcomes in our cohort of patients. Orbital floor, medial wall of the orbit, and intra-orbital content invasion are all incorporated in the TNM staging system, corresponding to T3 or T4 classification.²³ In contrast, the large international skull base surgery study that analyzed 1307 patients found orbital invasion was not significant on multivariable analysis for OS and DSS.³ Despite our cohort included 258 patients with skull base resections, there were 142 (36%) cases that did not require craniofacial resections. For patients that did not require craniofacial resections, orbital invasion might be a more significant prognostic variable, creating these discrepancies in Cox regression models.

Tumor histology is the most critical variable across all analyzed outcomes. Esthesioneuroblastoma histology had the best outcomes, compared to all other types, reaching 89.26% and 87.4% on 5 years DSS and OS respectively. In contrast, mucosal melanoma had the worse prognosis (30.1% 5 years OS and 33.4% 5 years DSS), taking in mind that mucosal melanomas are more aggressive comparing to more prevalent cutaneous melanoma.^24,25 Other histology types such as squamous cell carcinomas, adenocarcinomas, salivary gland tumors, and sarcomas had their Kaplan Meier survival curves situated in between these two extremes with 5 years OS ranging from 57.4% to 73.9% and 5 years DSS from 58.1% to 79.5%.

We report superior OS and DSS of patients treated in the most recent analyzed cohort of patients from 2006 till 2015, compared to all other decades. Many previous retrospective studies have failed to show improvement in treatment outcomes. It is postulated, in previous decades, an increased number of surgically treated cases for increasingly complex pathology was a possible explanation of stable results.¹⁶ The superior results in the treatment of surgically treated patients from 2006 till 2015 can be explained by several factors (Table S1). These include an increased understanding of the tumor biology of different histological types. Improved imaging allows for more detailed assessment of the tumor extent resulting in more accurate surgical resections. In addition, we see improvements in radiotherapy techniques, including IMRT and IMPT.

Treatment paradigms continuously evolve for sinonasal malignancy. Resectable sinonasal undifferentiated carcinomas were once universally treated with upfront surgery including craniofacial resection, but recent retrospective data from M.D. Anderson Cancer Center indicate that induction chemotherapy, followed by CRT in those that respond significantly, improves DSS.²⁶

In the case of melanoma, we see an increase of surgical cases till 2005, when the international skull base collaborative study results were published. Ganly et al. demonstrated poor outcomes for patients with melanoma, with no influence of surgical variables on the outcomes.²⁷ The introduction of immunotherapy drastically changed the way we treat cutaneous and mucosal melanomas. Despite the results published by D'Angelo et al., which demonstrated a lower response rate for mucosal melanomas, the results were still impressive for this poorly performing cohort of patients, with an overall response rate of 37.1%.²⁸ Today, we rarely carry out craniofacial resection or orbital exenterations for this cohort of patients.

A recent retrospective analysis published by Abelmeguid et al. explored organ preservation strategies for the treatment of sinonasal SCC patients.²⁹ The use of induction chemotherapy demonstrated good results in preserving function in patients with tumors invading the orbit and the skull base,.²⁹ The definitive answer to the question of benefit of induction chemotherapy for squamous cell carcinoma hopefully will come from the ongoing trial conducted by Eastern Cooperative Oncology Group-American College of Radiology Imaging Network (ECOG-ACRIN) Cancer Research Group (EA3163) study (ClinicalTrials.gov identifier NCT03493425).

Overall, we believe that proper patient selection, use of improved adjuvant, and neoadjuvant therapy have had the greatest impact on the improved OS and DSS of patients treated from 2006 to 2015.

We recognize that our study has several limitations. The retrospective nature of the study and the obvious selection bias observed cannot be fully controlled for. Furthermore, the more recent records had a better quality of the data and were more reliable. The study population is spread over decades, which encompasses several changes in management and selection of patients to surgery in addition to the fact multiple surgeons and physicians have been involved in the treatment of these patients. The study is also limited in generalizability as more recent management changes are constantly occurring, such as the evolution at our center for endoscopic and combined resections in amenable tumors, use of induction chemotherapy in certain histologies, and the use of CRT instead of RT as adjuvant treatment. We are unable to draw conclusions of evolution of every aspect of management of sinonasal malignancy as we made a decision to include only those treated definitively with surgery.

However, despite these limitations, the study analyses one of the largest, well-documented cohort of patients with sinonasal malignancies, and represents the experience of a single comprehensive cancer center with a uniform institutional philosophy.

5 |. CONCLUSSION

In conclusion, our cohort of patients with surgery for sinonasal malignancy showed improved survival in the last analyzed decade. Advances in surgical technique, adjuvant treatment, and patient selection likely had the most influence on this improvement. As has been shown in prior studies, pT-classification, pN-classification, orbital invasion, intracranial invasion, and histology impact treatment outcomes.

DATA AVAILABILITY STATEMENT

The author elects to not share data.