Patterns of Failure and Outcome in Esthesioneuroblastoma

Objectives  To evaluate the results of standardized treatment of esthesioneuroblastoma (ENB) during a 17-year period and to identify pertinent factors for clinical outcome.

Design  Review of clinical and radiographic data and retrospectively staging ENB according to 3 staging systems: Kadish, Biller, and Dulguerov and Calcaterra.

Setting  Hospital do Cancer I–Instituto Nacional de Cancer, Rio de Janeiro, Brazil.

Patients  Thirty-six patients with histologically confirmed ENB treated between January 1, 1983, and December 31, 2000; 35 fulfilled study inclusion criteria.

Interventions  Treatment included gross tumor resection through a transfacial approach with postoperative radiotherapy (RT) in 11 patients, craniofacial resection (CFR) and postoperative RT in 7, exclusive RT in 14, CFR alone in 1, and a combination of chemotherapy and RT in 2. Histopathological slides were reviewed and graded using the Hyams staging system. Analysis of prognostic factors was performed.

Main Outcome Measures  Evaluation of survival rates using the Kaplan-Meier method. Analysis of prognostic factors carried out with the Fisher exact test and the log-rank test.

Results  Analysis of survival showed that the Kadish classification best predicted disease-free survival (P = .046). The presence of regional and distant metastases adversely affected prognosis (P<.001 and P = .01, respectively). Craniofacial resection plus postoperative RT provided a better 5-year disease-free survival rate (86%) compared with the other therapeutic options used (P = .05). The 5-year disease-specific survival rate was 64% and 43% for the low- and high-grade tumors, respectively (P = .20). Disease-free survival for this cohort of 35 patient was 46% and 24% at 5 and 10 years, respectively. Overall survival was 55% and 46% at 5 and 10 years of follow-up, respectively.

Conclusions  The development of cervical nodal metastases and distant metastases had a significant adverse impact on prognosis. The value of the Kadish staging system was confirmed in our study, significantly correlating with prognosis. Tumor grade according to the Hyams staging system also seems to be an important factor in determining prognosis for tumor recurrence and survival. Aggressive multimodality therapeutic strategies, particularly CFR and adjuvant RT, yielded the best treatment outcome.

ESTHESIONEUROBLASTOMA (ENB) is an uncommon malignant tumor originating from the olfactory neuroepithelium of the upper nasal vault first described by Berger and Luc¹ in 1924. This tumor constitutes 3% of all intranasal neoplasms and can be seen in all ages, with a peak in the second and sixth decades of life and with equal distribution between the sexes.^2,3 As a result of its anatomic origin high in the nasal cavity, most patients have nonspecific symptoms, precluding early diagnosis and leading to the development of locally advanced disease that involves the paranasal sinuses and anterior cranial fossa through the cribriform plate in a large number of patients.

The study of ENB patients is complicated by several factors, the most important being (1) the lack of a uniform, effective staging system; (2) the relatively small number of cases reported in heterogeneous series; and (3) the fact that most series cover a long period during which significant changes in the management of ENB occur. Treatment modalities have included extracranial surgical excision, craniofacial surgical resection (CFR), and radiotherapy (RT) and chemotherapy used alone or in combination.^4-9 The most significant therapeutic advance in the treatment of ENB was made with the introduction of CFR, which improved disease-free survival significantly, particularly when compared with extracranial surgical excision.⁵ Since then, different therapy protocols, including CFR with preoperative or postoperative RT, CFR with a combination of RT and chemotherapy, chemotherapy with RT, and chemotherapy alone, have been applied.^4,6,9-16 Despite the adoption of aggressive therapies, local and regional recurrence and distant metastases occur, often after extended periods of follow-up. The ideal treatment modality for this disease has yet to be fully agreed on. The objective of this article is to evaluate the results of standardized treatment of ENB in a single institution during a 17-year period and identify possible factors of prognostic impact to predict treatment failures and identify the optimal therapeutic option.

Between January 1, 1983, and December 31, 2000, 36 patients with histologically confirmed ENB were treated at the Hospital do Cancer I–Instituto Nacional de Cancer in Rio de Janeiro, Brazil. One patient, who underwent RT (2000 rad [20 Gy]) as the only treatment provided, was considered inadequately treated and, therefore, excluded from this study. Patients included 20 men and 15 women between 8 and 77 years of age (median age, 43 years). Clinical records and radiographic studies of the patients were reviewed. Documented data of the initial symptoms, head and neck examination findings, surgical techniques and approaches, radiation treatment protocols (exclusive or adjuvant), and chemotherapy treatment regimens were analyzed.

The diagnosis of ENB was based on histopathological features, including microscopic findings of uniform small round blue cells with scanty cytoplasm and hyperchromatic round nuclei with inconspicuous nucleoli. Features such as the presence of neurofibrils, rosettes, pseudorosettes, and neurosecretory granules were also included. Immunohistochemical analysis was performed in 7 cases to confirm the diagnosis of ENB.

Available clinical and radiographic data on admission to our institution were used to retrospectively stage the ENB of patients who were initially seen with primary disease using the staging systems of Kadish et al¹⁷ (patients with tumors limited to the nasal cavity were classified as having stage A disease; tumors involving the nasal cavity and extending into paranasal sinuses were stage B; and tumors spreading beyond the nasal cavity and paranasal sinuses indicated stage C), Biller et al⁴ (T1 indicates tumor involving the nasal cavity and paranasal sinuses, excluding the sphenoid, with or without erosion of the bone of the anterior cranial fossa; T2, tumor extending into the orbit or protruding into the anterior cranial fossa; T3, tumor involving the brain that is resectable with margins; and T4, unresectable tumor), and Dulguerov and Calcaterra¹⁸ (T1 indicates tumor involving the nasal cavity and/or paranasal sinuses, excluding the sphenoid, sparing most superior ethmoidal cells; T2, tumor involving the nasal cavity and/or paranasal sinuses, including the sphenoid, with extension to or erosion of the cribriform plate; T3, tumor extending into the orbit or protruding into the anterior cranial fossa; and T4, tumor involving the brain).

Treatment included gross tumor resection through a transfacial approach with postoperative RT in 11 patients, CFR and postoperative RT in 7, RT alone after biopsy in 14, CFR alone in 1, and a combination of chemotherapy and RT in 2. Radiation therapy was administered using standard techniques. Megavoltage external-beam radiation was delivered using a 3-field technique with an anterior port combined with wedged lateral fields to provide a homogeneous dose distribution. Postoperative RT was given to 18 patients, with doses ranging from 4000 to 7000 rad (40-70 Gy) (median, 4800 rad [48 Gy]). Radiotherapy alone was given to 14 patients, with doses ranging from 5000 to 7000 rad (50-70 Gy) (median, 5500 rad [55 Gy]), and associated with chemotherapy in 2 patients, with doses of 6000 rad (60 Gy) each. Dose per fraction was 180 or 200 rad (1.8-2.0 Gy), with a median number of fractions of 27. The patient whose entire neck was treated received 5000 rad (50 Gy). The duration of RT ranged from 28 to 62 days (median, 46 days). Two patients received neoadjuvant chemotherapy that included cisplatin and vincristine sulfate in one and cisplatin, vincristine, and doxorubicin hydrochloride in the other. Response to therapy was defined by head and neck examination and/or computed tomography (CT) or magnetic resonance imaging (MRI). Complications were described according to their potential to lengthen the patient's hospital stay in major and minor complications.¹⁹

For the purpose of this study, all cases were reviewed with attention to the histopathological grade of tumors, according to the grading system proposed by Hyams²⁰ (Table 1). A head and neck pathologist (R.A.A.) who was blinded to the demographic characteristics, treatment, and survival data of patients performed the revision of the histopathological slides.

The major statistical end points of this study were relapse-free survival and overall survival. Response to therapy was defined by head and neck examination and CT or MRI and classified as complete response or partial response. Survival was calculated according to the Kaplan-Meier product-limit method.²¹ Analysis of prognostic factors (epidemiological, clinical, histopathological, and therapeutic) was performed by the use of a computerized statistical software package (SPSS Inc, Chicago, Ill). Statistical analysis was performed with the Fisher exact test and the generalized Wilcoxon log-rank test. P values are 2-tailed and are considered statistically significant when less than .05.

This study evaluates 36 cases treated at the Hospital do Cancer I–Instituto Nacional de Cancer during a 17-year span. The average follow-up time was 5.3 years. Five patients (1%) were initially treated at outside institutions and had undergone gross partial tumor resection before definitive treatment in our institute.

The symptoms at presentation were varied, with the most common symptoms being nasal obstruction (23 patients) and epistaxis (21 patients) (Table 2). The delay between the appearance of the first symptom and the beginning of treatment ranged from 0 to 30 months, with an average of 4.5 months.

The patients' tumors were staged on the basis of clinical and radiological presentation in accordance with the staging systems described in the "Methods" section. The most common stages, according to the classification of Kadish et al,¹⁷ Biller et al,⁴ and Dulguerov and Calcaterra,¹⁸ were, respectively, B (46%; n = 16), T1 (51%; n = 18), and T1 (40%; n = 14). The distribution of patients according to the various staging systems was as follows: Kadish A (14%; n = 5), B (46%; n = 16), and C (40%; n = 14); Biller T1 (51%; n = 18), T2 (46%; n = 16), and T3 (3%; n = 1); and Dulguerov T1 (40%; n = 14), T2 (23%; n = 8), T3 (34%; n = 12), and T4 (3%; n = 1).

The primary tumor most commonly invaded inferiorly into the nasal fossa and paranasal sinuses in all except 1 patient with gross invasion of the brain. Table 3 describes the frequency of tumor extension to the surrounding structures, with the maxillary sinus being the most common site of tumor involvement.

At the time of first evaluation at our institute, 3 (9%) of 35 cases had cervical nodal disease and 2 (6%) of 35 had distant metastases. From all the patients that were initially seen only with local disease, 2 (6%) of 32 developed cervical metastases and 2 (6%) of 33 developed distant metastases. In 2 cases cervical metastases were detected before distant metastases occurred, and in 2 cases cervical metastases were detected at the same time as distant metastases. The sites of distant metastases included the lungs in 2 cases and brain and liver in 1 case each. Overall, 5 (14%) of 35 patients had cervical metastases and 4 (11%) of 35 had distant metastases during the disease.

Surgery was part of the management plan in 19 (54%) of 35 patients. Most of the patients (11 of 19) were treated with extracranial surgical excision of their tumor with postoperative RT. In all of these patients, gross eradication of disease was achieved. Seven (64%) of these 11 patients have had no evidence of recurrence, with a median follow-up of 5.4 years (minimum, 9 months; maximum, 5.8 years). Of the remaining patients, most had local tumor recurrence at 12 months (3 of 4), and 1 had locoregional recurrence at 21 months.

Eight of the 19 surgically treated patients underwent CFR performed by the skull base surgery team of our institute, which consists of head and neck surgeons in collaboration with surgeons from the neurosurgery and plastic/microsurgery departments. Most of the patients (7 of 8) received postoperative RT. All but 1 patient, who underwent CFR alone (he refused postoperative RT), had clear margins of resection after histopathological examination. Six patients (86%) who underwent CFR plus RT have had no evidence of recurrence, with a median follow-up of 5 years (minimum, 1.5 years; maximum, 10.3 years). The remaining patient died of distant metastases to the lungs. The patient who underwent CFR alone died of other causes but with gross residual local disease.

Radiation therapy was a modality of treatment used for the management of the primary site in 34 (97%) of 35 patients. Usually, it was combined with surgical resection as an adjuvant treatment, as described previously in this article. In 14 (41%) of these patients, RT was used as the initial or exclusive modality of treatment. Five patients (36%) who underwent RT alone died of the disease at the end of the first year of treatment, including 2 patients with cervical nodal metastases in combination with local residual disease and 1 patient with distant metastases in combination with the primary tumor treated with palliative intent. Five patients (36%) who underwent RT alone survived 1 more year, including 3 with cervical nodal metastases treated with palliative intent, with local or locoregional residual disease. One patient (7%) had relapse at 4 years, and 1 patient (7%) had relapse at 5 years of follow-up, both with locoregional disease. Finally, 2 patients (14%) remained free of disease at 7 and 8.3 years of follow-up.

Chemotherapy was administered to 2 (6%) of 35 patients. One patient received RT (6000 rad [60 Gy]) and adjuvant chemotherapy, and another patient was treated with neoadjuvant and adjuvant chemotherapy in combination with RT (6000 rad [60 Gy]). Patient compliance and tolerance directed the course of chemotherapy: 1 patient received 2 cycles (adjuvant cisplatin and vincristine), and 1 patient received more than 3 cycles (neoadjuvant and adjuvant cisplatin, vincristine, and doxorubicin). There was a partial response documented with the first patient, who died of local gross residual disease and distant metastases at 1.9 years of follow-up, and a complete response with the second patient, who remains free of disease at 10.4 years of follow-up.

Neck metastases were present in 5 patients (14%). One of those 2 patients who developed neck metastases after the initial treatment of the primary tumor underwent salvage radical neck dissection type 1 and remained disease free for 1.6 years and died of distant metastases. The other patients were treated with RT and died of locoregional disease (3 of 4 patients) and of regional plus systemic disease (1 of 4 patients) within the first 2 years of follow-up. Distant metastases were present in 4 patients (11%). One patient was treated with RT plus adjuvant chemotherapy but died of uncontrolled local and systemic disease at 1.1 years of follow-up. The other patients died of local plus systemic disease (1 patient) and systemic disease (2 patients) within the first 2 years of follow-up.

Fifteen (43%) of the 35 patients developed complications during treatments. Eleven patients (73%) had postoperative complications and 4 patients (27%) had postradiation complications. Among patients who developed postoperative complications, 3 who underwent CFR plus RT developed local major complications: 1 had an early postoperative cerebrospinal fluid leak, 1 developed meningitis, and 1 had an alteration of neurological status after recovering from the anesthetic procedure (all cases were treated with conservative measures). One patient of the transfacial resection plus RT group had a major systemic complication, postoperative hypovolemic shock, which was successfully treated with blood and fluids transfusion therapy. Seven patients who underwent transfacial resection plus RT had minor complications: 4 had local wound sepsis, 2 developed nasolacrimal fistula or obstruction, and 1 had a nasocutaneous fistula. Among patients who had postradiation complications, 2 developed radiation-induced retinopathy, 1 of them also developed postradiation cataract, 1 had radiation keratopathy, and 1 had transient diplopia.

No significant difference in relapse-free survival and overall survival was found comparing the 2 age groups (≤50 vs >50 years). Sex also did not seem to have any influence (P>.05). Analysis of survival according to the 3 staging systems used showed that the Kadish et al¹⁷ classification predicted disease-free survival statistically significantly better than the Biller et al⁴ and Dulguerov and Calcaterra¹⁸ staging systems (P = .046). Comparative results for patients according to the staging system used are given in Table 4. The presence of regional and distant metastases adversely affected prognosis, reaching statistical significance (P<.001 and P = .01, respectively). The 5-year disease-specific survival rate was 61% in the 31 patients without neck metastases compared with 0% in the 5 patients with neck metastases and 58% in the 33 patients without distant metastases compared with 0% in the 3 patients with distant metastases. The type of therapeutic approach also had a statistically significant impact on disease-free survival. Comparison among the groups, according to the type of treatment, revealed an 86% 5-year disease-free survival for CFR and RT patients, 45% for transfacial resection plus RT, and 25% for RT alone (P = .05). Survival rates are shown in Figure 1. Pathological grade affected disease-free survival in our study; patients with low-grade tumors (grades 1 and 2) had a higher disease-free survival rate compared with those with high-grade tumors (grades 3 and 4). The 5-year disease-free survival rate was 64% in the 12 patients (34%) with low-grade tumors and 43% in the 23 patients (66%) with high-grade tumors (P = .20) (Figure 2).

Disease-free survival for this cohort of 35 patients was 46% and 24% at 5 and 10 years, respectively. Overall survival was 55% and 46% at 5 and 10 years of follow-up, respectively (Figure 3 and Figure 4).

Esthesioneuroblastoma is a rare and challenging malignancy of the neuroepithelium, with fewer than 1000 cases reported in the literature, most of them in small series covering long periods in which considerable changes in the management of ENB occurred.³ Because of these factors and the heterogeneous nature of many of these series, recommendations for the optimum management of ENB remains to be defined.^4-7,9,17,18

Controversy surrounds the prognostic reliability of various factors in the management of ENB, including the use of a uniform, effective staging system to serve as a guide for standardized treatment protocols and the prognostication of survival. There have been attempts to determine prognosis by demographic aspects,^9,18,22 clinical presentation,^22,23 stage,^2,9,17 tumor extention,^22,24,25 and pathological grade.^20,26

Sex and age had no significant prognostic effect in our series, in accordance to the observation of Foote et al,⁶ Resto et al,²⁵ and Morita et al.²⁷ Dulguerov and Calcaterra¹⁸ and Polin et al⁹ found that advanced age adversely affected prognosis, and Homzie and Elkon²² developed a mathematical equation to weight the negative prognostic contributions of several factors, including age.

Clinical stage was found to be an important prognostic factor in accordance with Kadish et al¹⁷ and Elkon et al.² Loubutin et al²⁸ observed that leptomeningeal infiltration by ENB carried a poor prognosis. Homzie and Elkon²² included the presence of metastases and local extension of the tumor (eg, ethmoidal, nasopharynx, orbital) as negative prognostic factors in their series. Resto et al²⁵ found that the only predictor of survival was nodal metastases. Dulguerov and Calcaterra¹⁸ determined that significant negative prognostic factors included tumor recurrence and metastases, among others.

Several staging systems have been proposed, but the most commonly used has been the Kadish staging system,¹⁷ particularly because of its simplicity and acceptable prognostic efficacy. Two other staging methods, the Biller method and the Dulguerov and Calcaterra method, have also attempted to better describe ENB in a clinically relevant manner. Uniform tumor staging is of paramount importance not only because of its correlation with prognosis but because it allows comparison of treatment results among patients with similar disease extent. This is particularly true for uncommon malignancies such as ENB, in which the relatively small number of patients and the heterogeneous nature of many of the series in the literature have made it difficult to compare the results from the various institutions hampered by the lack of consensus on the staging of this tumor.

With the objective of comparing the 3 staging systems, patients' tumors were staged according to the classification systems of Kadish et al,¹⁷ Biller et al,⁴ and Dulguerov and Calcaterra.¹⁸ The Kadish staging system was the only classification that revealed a statistically significant discrimination among stages for relapse-free survival (P = .046). Comparison of recurrence rates among the staging systems is given in Table 4. This result is in accordance with the observations of Resto et al,²⁵ Levine et al,^5,29 Polin et al,⁹ and Myamoto et al,³⁰ who recognized the prognostic importance of this classification.

Despite the fact that the Kadish staging system for ENB has been widely used in the literature, the inadequacies of this system have been pointed out by several authors.^4,6,18,31,32 Arguments have varied from the small number of patient candidates for stage A, lack of prognostic significance of stages A and B, the large spectrum of patients classified as having stage C disease, and the fact that regional disease was not included in the staging system because of its rarity.

The presence of cervical nodal metastases and distant metastases was an important prognostic factor in our study that adversely affected the disease-specific survival rate (P<.01 and P = .01, respectively). These results are in accordance with the observations of Homzie and Elkon,²² Dulguerov and Calcaterra,¹⁸ Resto et al,²⁵ and Dulguerov et al.³³

The lower rates of regional metastases (14%) in our group of patients correlate with the findings of Kadish et al¹⁷ and Elkon et al,² who reported rates lower than 15% in their series and considered elective treatment of the neck unnecessary. Higher rates of neck metastases in ENB seem to be associated with patients with locally advanced or recurrent disease (20%-44%).^25,27,34,35 The rates of distant metastases among our patients (11%) were also lower than the rates found in the literature, which ranged from 10% and 40%.^7,27,35

Hyams²⁰ introduced a pathological grading system classifying ENB into 4 grades based largely on the level of cellular differentiation. On the basis of this histological classification, it seems that well-differentiated tumors have a slower disease progression and less tendency for local recurrence. On the other hand, features designating higher grades in the Hyams system have been found to correlate with a rapid clinical progression and reduced survival.^27,30,33

Morita et al,²⁷ in their analysis of 49 ENBs treated between 1951 and 1990 at the Mayo Clinic, Rochester, Minn, found that the pathological grade of tumors was the most important prognostic factor of survival identified. McElroy et al¹³ observed that the Hyams²⁰ grading system was indicative of survival and, in addition, could be used as a predictor for the use of chemotherapy. Recently, Myamoto et al,³⁰ with the aim of determining the usefulness of the Hyams system in predicting outcome for ENB and undifferentiated sinonasal carcinoma, analyzed the medical records of 26 patients treated at their institution. They concluded that, despite the limited number of patients, their series demonstrated that both clinical staging (Kadish) and histological grading of tumors provided useful prognostic information.

Our data demonstrated that patients with advanced Hyams grade tumors had a poorer response to treatment; the 5-year disease-free survival among patients with low-grade tumors was 64% compared with 43% of those with high-grade tumors (P = .20).

Dulguerov et al³³ published a meta-analysis of publications on ENB from 1990 through 2000 with the aim of reviewing recent developments in diagnosis, staging, and treatment of these tumors. They were also able to conclude that the histopathological grading according to Hyams²⁰ and the presence of cervical lymph node metastases emerged as prognostic factors of outcome.

There is little argument that overall survival and relapse-free survival rates are greater in patients who have undergone complete oncological resection compared with patients who undergo partial resection or biopsy with adjuvant RT. Complete gross resection of the tumor alone has been advocated as an adequate treatment for initial stage, low-grade tumors.^2,4,32 More advanced tumors, however, seem to be better managed with combined therapy, usually including CFR in combination with RT and/or chemotherapy. Therefore, CFR coordinated with a skull base team has become the most commonly used surgical procedure for ENB, improving local control and achieving higher survival rates, particularly when associated with adjuvant therapies.^4,5,27

Broich et al³ reviewed the literature and found a 72.5% 5-year survival rate in patients treated with combined therapies compared with 62.5% with surgery alone and 53.8% with RT alone. Dulguerov and Calcaterra¹⁸ observed in their study that 92% of patients treated with complete gross resection of the tumor plus RT remained recurrence free, as opposed to 40% with RT alone and 14% with surgery alone. Foote et al⁶ found that, despite the high grade or advanced stage of their patients' tumors, postoperative RT improved local control rates in their study. The only exception in the literature is the report by Biller et al,⁴ who indicated that CFR alone is sufficient for ENB if clear margins of resection are obtained.

The use of RT as the only treatment for ENB was first described by Berger and Luc¹ in 1924. Esthesioneuroblastoma has been described as a radiosensitive tumor in the literature, despite the fact that recurrence rates of up to 50% are found when RT is used as the only modality of treatment.^1,2,36,37 Although patients in these series treated with RT alone were not locally controlled, a significant number of patients who underwent surgical resection with positive margins benefited from adjuvant RT.

As far as our therapeutic results are concerned, 6 (86%) of 7 patients who had undergone CFR plus RT were alive and free of disease at the 5-year follow-up compared with 7 (64%) of 11 patients who had undergone extracranial excision plus RT and 3 (23%) of 15 patients who had undergone exclusive RT. We found a statistically significant difference comparing the rates of disease-free survival of these 3 therapeutic modalities (P = .01). This is in line with the result of most of the series published in the literature.^{2,3,5-7,9,18,25,27,32,37}

Finally, the benefit of chemotherapy for the treatment of ENB in our study is difficult to assess, particularly in a small cohort of patients (2/35) and because it was not used as a single therapeutic modality. Numerous protocols have been published in the literature,^{5,9,10,12-15,17,27,29} presenting favorable results with cisplatin-based therapies^5,9,10,13,15 and suggesting its role in the treatment of advanced stage ENB.¹³ One patient with locally advanced ENB (50%), who underwent chemotherapy plus RT, is alive and free of disease at 10.4 years of follow-up.

In conclusion, the results from our series, although limited by the small number of patients, highlight the development of metastases (regional and distant) as significant prognostic factors and confirm previous observations regarding the prognostic value of the Kadish staging system.¹⁷ Tumor grade according to the Hyams²⁰ staging system also seems to be an important factor in determining prognosis for disease-free survival. Aggressive multimodality therapeutic strategies, particularly CFR and adjuvant RT, yielded the best treatment outcome.

Corresponding author and reprints: Fernando L. Dias, MD, PhD, Avenue Alexandre Ferreira, 190–Lagoa, Rio de Janeiro 22470-220, Brazil (e-mail: fdias@inca.gov.br).

Submitted for publication June 17, 2002; final revision received February 3, 2003; accepted April 15, 2003.

This study was presented at the annual meeting of the American Head and Neck Society; May 11, 2002; Boca Raton, Fla.