Management of nonsinonasal neuroendocrine carcinomas of the head and neck

Abstract

BACKGROUND

Nonsinonasal neuroendocrine carcinomas (NSNEC) of the head and neck are rare and pose a diagnostic and management challenge. The authors undertook a retrospective study to gain insights into the spectrum of clinicopathologic characteristics, patterns of failure, and optimal management of patients with this disease.

METHODS

The authors treated 23 adults with pathologically proven, nonmetastatic, primary NSNEC from 1984 to 2001. The majority (13 patients) had laryngeal origin with the following American Joint Committee on Cancer stage distribution: Stage I disease in 1 patient, Stage II disease in 2 patients, Stage III disease in 6 patients, and Stage IV disease in 14 patients. Nine patients underwent definitive surgery with or without postoperative radiation, and 14 patients received definitive radiotherapy. The median definitive radiation dose was 66 grays (Gy) (range, 44–72 Gy) using conventional fractionation. Fourteen patients received chemotherapy, with two to four cycles of induction platinum plus etoposide used most commonly.

RESULTS

The median follow-up time for surviving patients was 40 months (range, 15–89 months). The actuarial 2-year and 5-year overall survival (OS) rates were 53% and 33%, respectively; and the disease-free survival (DFS) rates were 41% and 25%, respectively. Both the 2-year OS rate (68% vs. 30%; P = 0.002) and the 2-year DFS rate (55% vs. 17%; P = 0.004) were improved with chemotherapy compared with local therapy alone. Seventy-five percent of patients with measurable disease had complete clinical responses to induction chemotherapy. There was 100% complete clinical response of tumor after radiotherapy. The actuarial 2-year local failure rate was 23%. Chemotherapy did not reduce local failure (P = 0.91). There was no regional failure. The 2-year and 5-year distant metastasis rates were 54% and 71%, respectively. The 2-year rates of metastases without and with chemotherapy were 79% and 39%, respectively (P = 0.006). The 2-year and 5-year rates of intracranial metastases were 25% and 44%, respectively, and the 2-year and 5-year rates of isolated brain metastases were 21% and 41%, respectively.

CONCLUSIONS

Based on these results, the authors' treatment strategy for patients with NSNEC is sequential chemotherapy and radiation. They recommend full-dose radiotherapy alone for patients with NSNEC who achieve a complete clinical response to induction chemotherapy. Newer chemotherapeutic regimens or additional adjuvant chemotherapy should be investigated for patients with NSNEC given the high rate of distant failure. Due to the very high rate of brain metastases among patients in the current study, the authors now consider incorporating prophylactic cranial irradiation into primary radiotherapy for individual patients who have complete clinical responses to induction chemotherapy. Cancer 2003. © 2003 American Cancer Society.

Primary neuroendocrine carcinomas are uncommon head and neck malignancies. They present with a varied histopathologic spectrum in sinonasal and nonsinonasal head and neck subsites. The sinonasal carcinomas are more diverse, with four major histologic phenotypes: esthesioneuroblastoma, sinonasal undifferentiated carcinoma, neuroendocrine carcinoma, and small cell undifferentiated carcinoma.1 These tumors occur with enough frequency that specific treatment strategies have emerged.2–5 The nonsinonasal neuroendocrine carcinomas (NSNECs), however, are represented predominantly by small cell undifferentiated carcinomas, followed by moderately differentiated (atypical carcinoid) carcinomas and well-differentiated (typical carcinoid) carcinomas. The NSNECs are rare enough that they are represented in the literature primarily by sporadic case reports.6 We contend that the differences in tumor types and treatment strategies for sinonasal and nonsinonasal sites justify a separate analysis. We reviewed our institutional experience in patients with NSNEC to determine the optimal management parameters for this disease.

MATERIALS AND METHODS

Patient Population

We reviewed the medical records of 23 adults who were treated for primary NSNEC of the head and neck at The University of Texas M. D. Anderson Cancer Center (UTMDACC; Houston, Texas) between 1984 and 2001. The patient population was identified through a search of the Tumor Registry database maintained by the Department of Medical Informatics. All patients had newly diagnosed, nonmetastatic tumors arising from nonsinonasal head and neck subsites and were treated with curative intent. Patients who were seen at UTMDACC for consultation only or for treatment of recurrent disease were excluded. Histopathologic slides prepared from archived blocks were reviewed, and immunohistochemical and electron microscopic findings were reevaluated by an experienced head and neck pathologist for every patient. This retrospective review received Institutional Review Board approval, and patient data were maintained confidentially throughout the study.

Staging evaluation for these patients included history and physical examination, screening laboratory studies, chest X-ray, contrast-enhanced computed tomography (CT) scan or magnetic resonance image of the head and neck, and biopsy of primary or lymph node disease in all patients. Abnormalities on chest X-rays were evaluated further (with chest CT or biopsy) as necessary to exclude metastatic disease. Tumors were restaged according to the American Joint Committee on Cancer (AJCC) staging manual,7 based on documented clinical, imaging, and pathologic findings.

Pathologic Analysis

Combined cytomorphologic and immunohistochemical features of neuroendocrine differentiation formed the basis for diagnosis. All tumors that had small cells (well differentiated, moderately differentiated, and undifferentiated) with positive staining for keratin, neuron-specific enolase, and chromogranin were included in this study. There was 1 well-differentiated carcinoma (typical carcinoid), 1 moderately differentiated carcinoma (atypical carcinoid), and 19 small cell undifferentiated carcinomas. Two tumors were hybrids, with a component of squamous cell carcinoma within an extensive background of small cell carcinoma. Tumors that potentially could be confused8 with NSNEC (including paraganglioma,9 medullary carcinoma,10 basaloid squamous cell carcinoma,11 melanoma,12 pituitary adenoma/carcinoma,13 or Merkel cell carcinoma14) were excluded from the current analysis.

Statistical Analysis

Estimated rates of local failure (LF), distant failure, disease-free survival (DFS), and overall survival (OS) were calculated using the Kaplan–Meier method. Survival estimates were calculated from date of diagnosis. Clinical and pathologic variables were assessed using the Mantel log-rank test for univariate analysis, and the Cox proportional hazards model was used for multivariate analysis. Retrospective categorization of late treatment toxicity was determined according to the National Cancer Institute Common Toxicity Criteria Version 2.0 (which includes the Radiation Therapy Oncology Group/European Organization for Research and Treatment of Cancer Late Radiation Morbidity Scoring Schema).15

RESULTS

Patient Characteristics

The median patient age at diagnosis was 64 years (range, 38–86 years), and males were diagnosed as commonly as females (Table 1). Eighty-three percent of patients were current or former smokers, with a median 50 pack-years (range, 0–160 pack-years) of tobacco use reported. The majority (13 patients) had laryngeal primary tumors. Most patients presented with locoregionally advanced disease. The distribution according to the AJCC staging system was as follows: Stage I in 1 patient, Stage II in 2 patients, Stage III in 6 patients, and Stage IV in 14 patients.

Table 1. Patient Characteristics

Characteristic	No. of patients (%)
AJCC: American Joint Committee on Cancer.
Gender
Male	12 (52)
Female	11 (48)
Primary site
Larynx	13 (56)
Supraglottic	11 (—)
Glottic	1 (—)
Subglottic	1 (—)
Oropharynx	3 (13)
Oral cavity	1 (4)
Hypopharynx	2 (9)
Nasopharynx	2 (9)
Parotid gland	2 (9)
Tumor status
T1	4 (17)
T2	5 (22)
T3	6 (26)
T4	5 (22)
Tx	3 (13)
Lymph node status
N0	5 (22)
N1	4 (11)
N2	10 (44)
N3	4 (17)
AJCC stage
I	1 (4)
II	2 (9)
III	6 (26)
IV	14 (61)

Treatment Characteristics

Prescribed therapy varied during the long study interval, as is expected for a rare disease. Nine patients had definitive surgery (with or without postoperative radiotherapy [RT]), and 14 patients had definitive RT. The median definitive radiation dose was 66 grays (Gy) (range, 44–72 Gy) using conventional fractionation. RT treatment volumes included the primary tumor and bilateral cervical/supraclavicular lymph nodes for nearly all patients; contralateral lymph nodes were excluded for highly selected patients with well-lateralized primary tumors (e.g., parotid gland). No patient received prophylactic cranial irradiation (PCI).

Fourteen patients received chemotherapy in addition to local therapy. Nine patients had induction chemotherapy followed by definitive RT or chemoradiation; one of those patients received additional cycles of adjuvant chemotherapy. Eight of nine patients who were treated with induction chemotherapy were evaluable for clinical disease response; one patient had no evaluable disease after generous biopsy of the primary lesion. Six of 8 evaluable patients (75%) had complete clinical responses to induction chemotherapy; the remaining 2 patients had stable disease. There was 100% tumor clearance after RT, however. Two to four cycles of platinum/etoposide comprised the most common induction regimen. Two patients were treated with definitive chemoradiation (without induction). Two patients had adjuvant chemotherapy after local treatment, and a single patient was treated with postoperative, concurrent chemoradiation.

OS and DFS

The median follow-up time for surviving patients was 40 months (range, 15–89 months). The actuarial 2-year and 5-year OS rates were 53% and 33% (Fig. 1), respectively; and the DFS rates were 41% and 25%, respectively. Both 2-year OS (68% vs. 30%; P = 0.003) and 2-year DFS (55% vs. 17%; P = 0.004) were improved with use of chemotherapy compared with local therapy alone (Fig. 2). AJCC stage (Stage I–III vs. Stage IV) and cervical lymph node status (positive vs. negative) also significantly predicted 2-year OS in a univariate analysis (Table 2). The use of chemotherapy, however, was the only significant factor (P = 0.009) in a multivariate analysis of OS that included AJCC stage, lymph node status, local treatment modality, and use of chemotherapy.

Figure 1. Overall survival (all patients).

Figure 2. Overall survival according to use of any chemotherapy (induction, concurrent, or adjuvant).

Table 2. Prognostic Factors for Survival in Patients with Nonsinonasal Neuroendocrine Carcinomas of the Head and Neck: All Patients

Factor	No. of patients	Two yr OS (%)	P value
			Univariate	Multivariate
OS: overall survival; N/A: not available; AJCC: American Joint Committee on Cancer. a Includes patients who received postoperative radiotherapy. b Includes patients who underwent planned neck dissection after radiotherapy. c Includes patients who were treated with induction, concurrent, and/or adjuvant strategies.
Age (continuous variable)	23	N/A	0.71	—
Gender
Male	12	59.9	0.82	—
Female	11	45.5	—	—
Cervical lymph node status
Negative	5	100.0	0.031	0.07
Positive	18	40.1	—	—
AJCC stage
I–III	9	85.7	0.028	0.43
IV	14	33.3	—	—
History of cigarette use
Yes	19	75.0	0.64	—
No	4	48.5	—	—
Primary tumor site
Larynx	13	59.3	0.94	—
Other	10	45.7	—	—
Definitive local treatment
Surgerya	9	40.0	0.25	0.15
Radiotherapyb	14	61.4	—	—
Chemotherapy used
Yesc	14	68.1	0.003	0.009
No	9	29.6	—	—

Patterns of Failure

Four patients had LF, yielding actuarial 2-year and 5-year LF rates of 23% and 23%, respectively. Chemotherapy did not reduce the LF rate (P = 0.91); however, the 2 nonresponders to induction chemotherapy represented 50% of the patients who had LF. Among patients who were treated definitively with RT, the prescribed total dose in the range of 44–72 Gy was not correlated with LF (P = 0.23). There was no regional cervical lymph node failure.

The 2-year and 5-year distant metastasis (DM) rates were 54% and 71%, respectively (Fig. 3), and the 2-year DM rates without and with chemotherapy were 79% and 39%, respectively (P = 0.006). All patients who were not treated with chemotherapy failed distantly by 25 months (Fig. 4). The 2-year and 5-year rates of intracranial metastases were 25% and 44%, respectively. The brain was the only site of DM at 2-year and 5-year rates of 21% and 41%, respectively (Fig. 5).

Figure 3. Freedom from distant metastases (all patients).

Figure 4. Freedom from distant metastases according to use of any chemotherapy (induction, concurrent, or adjuvant).

Figure 5. Freedom from intracranial metastases (all patients).

Late Effects of Treatment

Grade 1–2 high-frequency hearing loss was documented among patients who were treated with multiple cycles of cisplatin-based chemotherapy; this prompted a change to carboplatin for 1 patient. After locoregional RT, Grade 1–2 xerostomia (8 patients), Grade 2 hypothyroidism (1 patient), Grade 2 serous otitis (1 patient), and Grade 2 esophageal stricture (1 patient) were reported. None of the patients developed Grade 3–5 late toxicity.

DISCUSSION

NSNECs of the head and neck are uncommon and previously were characterized poorly. Prior studies of these tumors were limited in scope and number and included heterogeneous sites and histologies as well as patients with DM.16–20 These shortcomings were compounded by the lack of ancillary markers for the unequivocal diagnosis of neuroendocrine derivation. Our study was limited to patients with locoregionally confined disease who were diagnosed and treated since 1984; this insured that all patients were diagnosed accurately using modern immunohistochemical and electron microscopic techniques, and these were reviewed and confirmed individually. Patients received their treatment at a single, high-volume, tertiary cancer care center. There were only 23 patients in 17 years who met these strict criteria. Although this relatively small denominator makes sweeping conclusions less reliable, to our knowledge, this is the largest reported homogenous data set that can be used to direct treatment strategies based on patterns of failure without resorting to meta-analysis.21

In this series, the use of combination chemotherapy approximately doubled the 2-year OS and DFS rates and reduced by one-half the 2-year rate of DM. NSNEC is highly responsive to cisplatin/etoposide combination chemotherapy, and it was found that the use of this treatment regimen was the single most important factor in the improvement of treatment outcomes. Responses to chemotherapy typically were not durable, however, and > 80% of patients ultimately succumbed to DM (Fig. 3). It is possible that newer agents, such as camptothecin derivatives or taxanes, may produce better responses and more durable control, but this will require additional study.22

The 23% 2-year local failure rate for patients with NSNEC is approximately one-half of the rate seen in patients with similarly staged squamous carcinomas.23 In contrast, DMs represent the major pattern of failure for NSNEC, occurring at a rate nearly double that of local failure. Therefore, we recommend nonsurgical therapy for most patients using combined chemotherapy and RT. The single patient in our series who had a well-differentiated neuroendocrine carcinoma (typical carcinoid) underwent a supraglottic laryngectomy and remains free of disease at 32 months. This is consistent with the finding that surgery alone may be adequate for the very rare carcinoid and carcinoid-like tumors of the head and neck, as it is for carcinoid and carcinoid-like tumors at other body sites.24 The extreme rarity of head and neck carcinoids (fewer than 20 patients are reported in the published literature6) makes it impossible to make any different treatment recommendation for those arising in head and neck sites different from those arising at any other body site.

The local control rates in this series also were higher compared with the rates reported for patients with small cell carcinoma of the lung. This may have been due to the higher total RT doses used in the head and neck (60–70 Gy), although we could not establish clearly a dose-response relation, the higher response to induction therapy, or both. Although there is evidence that concurrent chemoradiation rather than sequential chemoradiation is a more effective method for inducing a complete clinical response in patients with small cell lung carcinoma,25–27 sequential chemoradiation appears to be an adequate and less toxic means of securing local control for most patients with NSNEC.

The high rates of isolated intracranial metastases suggest that the central nervous system may be a sanctuary site for NSNEC. PCI after patients achieve a complete response to locoregional therapy is used for other tumor systems that manifest high rates of brain metastases, and it has been demonstrated that PCI reduces the frequency of brain metastases and improves OS in patients with small cell lung carcinoma.28–30 By analogy, we now consider PCI for patients with NSNEC who have had a complete clinical response to induction therapy. Because matching whole-brain RT fields to prior head/neck fields is complex at best, it is advantageous theoretically to incorporate PCI into primary RT. This may be accomplished by treating large initial fields that include whole brain, primary tumor site, and draining regional lymph nodes to a dose of 28–30 Gy at 2 Gy per day (Fig. 6). This approach is supported by the previous successful application of PCI regimens using conventional 2-Gy fractions.31, 32

Figure 6. Hypothetical primary radiotherapy field (including prophylactic cranial irradiation) for a patient with laryngeal or hypopharyngeal neuroendocrine carcinoma.

In conclusion, we recommend treating patients who have NSNEC with induction chemotherapy followed by RT alone for complete responders. Concurrent chemoradiotherapy does not necessary appear to improve early complete response, local control, or survival, and it is a more toxic induction technique. An induction approach is most likely to deliver the planned dose intensity of systemic chemotherapy, the single most important factor driving survival in our analysis. Finally, the sequential use of chemotherapy and RT allows for the potential incorporation of PCI into definitive radiation fields. If concurrent chemoradiation is used, then PCI cannot be administered simultaneously without a significant increase in toxicity.33 If there is a less than complete response to induction therapy, however, then the problem of local control dominates, and concurrent chemoradiation or surgery should be considered; the use of PCI becomes a less relevant issue in this setting.