The National Cancer Data Base Report on Cancer of the Head and Neck

Background  The National Cancer Data Base (NCDB), a large sample of cancer cases accrued from hospital-based cancer registries, is sponsored by the Commission on Cancer of the American College of Surgeons and the American Cancer Society. The NCDB permits a detailed analysis of case-mix, treatment, and outcome variables.

Objective  To provide an overview of the contemporary status of the subset of patients with head and neck cancer in the United States.

Methods  The NCDB, which obtains data from US as well as Canadian and Puerto Rican hospitals, accrued 4583455 cases of cancer between 1985 and 1994. Of these cases, 301350 (6.6%) originated in the head and neck. We address 295022 cases of head and neck cancer limited to the 50 United States and District of Columbia. Cases were segregated into an earlier group (1985-1989) to permit 5-year follow-up and into a later group (1990-1994) to analyze a more contemporary group. Comparison between both periods permits identification of trends.

Results  The largest proportion of cases arose in the larynx (20.9%) and oral cavity, including lip (17.6%) and thyroid gland (15.8%). Squamous cell carcinoma (55.8%) was the most common histological finding, followed by adenocarcinoma (19.4%) and lymphoma (15.1%). Income level (low), race (African American), and tumor grade (poorly differentiated) were most notably associated with advanced stage. Treatment was most commonly surgery alone (32.4%), combined surgery with irradiation (25.0%), and irradiation alone (18.9%). Overall 5-year, disease-specific survival was 64.0%. Cancer of the lip demonstrated the best survival (91.1%) and cancer of the hypopharynx the worst survival (31.4%).

Conclusions  This NCDB analysis of cancer of the head and neck provides a contemporary overview of head and neck cancer in the United States. It also serves to introduce a series of NCDB articles that address specific anatomical sites and histological types through separate, detailed analysis.

THE NATIONAL Cancer Data Base (NCDB) is a large sample of cancer cases accrued from hospital-based cancer registries in the United States. This database is jointly sponsored by the American College of Surgeons' Commission on Cancer (COC) and the American Cancer Society. It is designed to provide descriptive information about the demographic, management, and outcome variables characterizing cancers involving all ethnic groups in all 50 states. National cancer registries have been functioning for many years in other countries. The most highly developed registries currently exist in European countries with small populations that include Sweden, Norway, and Denmark.^1-6 In 1988 the United States established its national clinical cancer registry with the creation of the NCDB.^7,8

Although there is no current system for gathering incidence data for the entire United States, a population-based registry termed the Surveillance, Epidemiology, and End Results program⁹ exists to provide estimates of cancer-related incidence and mortality. This program, which was mandated by the National Cancer Act of 1971, currently surveys 14 distinct population groups representing approximately 14% of the population.¹⁰

The Surveillance, Epidemiology, and End Results and the NCDB programs are separate cancer data systems that are designed for different purposes and rely on different methodologies.¹¹ The Surveillance, Epidemiology, and End Results program is a population-based registry that is intended to accurately sample a measurable segment of the US population. The NCDB is a hospital-based registry that monitors patterns from a much larger patient base derived from community hospitals, teaching hospitals, and cancer centers.⁸

The goal of the NCDB is to improve cancer management through analysis of data characterizing a large proportion of all cases of cancer in the United States. To help achieve this goal, the NCDB has established the objective to collect 80% of all US incident cancers by the year 2000. The first call for data by the NCDB yielded an estimated 24% of all cancer cases diagnosed in 1985. This sample represented 232577 cases reported from 501 hospitals. The number of participating NCDB hospitals and cases accrued for the year 1994 increased to 1227 hospitals reporting 689714 cases, reflecting an estimated 57% of all cancer cases. This increase in reporting over the past 9 years has paralleled an increase in the number of hospital cancer registries that have become computerized.

An additional impetus to increased reporting of cases is a mandate established by the COC. Since 1997, all hospitals participating in the COC's approvals program are required to submit data to the NCDB. Continued expansion in the accrual of cases is anticipated as more hospitals fulfill these requirements.

Before 1997 the COC did not require reporting of data to the NCDB. The NCDB obtained reports on a voluntary basis from those institutions that elected to contribute data from their cancer registries. As a result, the database prior to 1997 must be considered a convenience sample with the potential to be affected by a selection bias that could skew the sampling of cases. Despite these potential limitations in data collection, the large numbers of cases accrued offer demographic, management, and outcome information from a broad spectrum of treating facilities in the United States. A recent comparison between NCDB and Surveillance, Epidemiology, and End Results data identified patterns that differed only marginally in the analysis of breast, colorectal, lung, and prostate cancers evaluated for the diagnostic year 1992.¹¹

We address the subset of the NCDB cancer cases limited to the head and neck. Most cancers commonly grouped as head and neck malignancies arise from the mucosal lining of the upper aerodigestive tract and the adjacent salivary glands. Thyroid, parathyroid, sinonasal, and ocular cancers are also considered cancers of the head and neck. In addition to cancers arising in these sites, lymphomas and other less common tumors that arise from the adjacent soft tissue, bone, and cartilaginous structures in the head and neck region are considered in this article. Skin cancers that occur in the head and neck region are not included in this review. Brain malignancies are not generally classified as cancers of the head and neck and are also not considered herein.

To date, reports from the NCDB of cancers of the head and neck have been limited to the larynx¹² and thyroid gland.¹³ We present the first analysis of the NCDB providing a broad review of cancer of the head and neck.

The NCDB data are collected yearly on a voluntary basis through a computerized format from hospital-based cancer registries.⁷ The NCDB cancer registry data are coded according to schemata published in the Data Acquisition Manual,^14,15 the first through the fourth editions of the American Joint Committee on Cancer (AJCC) Manual for Staging of Cancer,^16-19 and the second edition of the International Classification of Disease for Oncology (ICD-O 2).²⁰

The head and neck cancer data set was defined by the ICD-O 2 topography codes and included the lip, oral cavity, oropharynx, nasopharynx, hypopharynx, and major salivary glands (C00.0-14.8), sinonasal tract (C30.0, 31.0-31.9), and larynx (C32.0-32.9). Although the lip is included with other subsites in the oral cavity according to AJCC staging, its behavior is sufficiently different from the remainder of the oral cavity that it was considered separately in this article. Other head and neck sites included in the head and neck data set were the middle ear (C30.1), trachea (C33.9), eye and ocular adnexa (C69.0-69.9), olfactory nerve (C72.2), thyroid gland (C73.9), parathyroid glands (C75.0), and other endocrine gland–related structures (C75.2, C75.4-75.9), excluding the pineal and pituitary glands. Additional sites isolated to the head and neck included bones, joints, and articular cartilages (C41.0-41.9), peripheral nerves and autonomic nervous system (C47.0), connective, subcutaneous, and other soft tissues (C49.0), and lymph nodes (C77.0). The nonspecific sites within the head and neck classified as "other" and "ill-defined" (C76.0) were also included.

The reporting hospitals provide only those cases that were diagnosed and/or treated at their institute as a primary cancer. Although patients are followed up longitudinally and recurrent disease is added to their record if identified, the NCDB does not collect records of patients who were identified at the reporting hospital with recurrent disease. To ensure that the database does not include more than 1 record for each patient (eg, a patient having received primary treatment at 2 different reporting hospitals), an algorithm based on patient and disease characteristics was used to identify and remove these duplicate records.

Patterns of presentation and treatment across time are investigated by dividing the years of diagnosis into an earlier period (1985-1989) and a later period (1990-1994). Case-mix characteristics and treatment are stratified by anatomical site and extent of disease, when appropriate, to provide a more detailed analysis.

Patients were classified by geographic regions that were organized by grouping individual states into 6 regions as previously reported.⁷ Income was inferred for each patient based on the average family income of the ZIP code of residence. To compare the level of income of patients with head and neck cancer with the income of all patients with cancer within the NCDB, 3 income groups were created. These income groups were chosen to approximate the lowest 10%, the highest 10%, and intermediate incomes for all NCDB cases.²¹ The low-income group included patients with annual incomes of less than $20000 that represented 11.2% of the NCDB data set. The high-income group included patients with annual incomes of $47000 or more, which represented 10.3% of the NCDB data set.

Extent of disease was represented by "combined stage" that reflects pathologic staging (pAJCC stage group) when it was available through the reporting cancer registrar's review of the chart, and clinical stage (cAJCC stage group) when pathologic stage was not recorded or not appropriate. Cases were broadly grouped into general histological categories according to the ICD-O 2morphologic codes. These groups were carcinoma not otherwise specified (NOS) (M8010, M8012-8022), squamous cell carcinoma (M8052-8082), verrucous carcinoma (M8051), adenocarcinoma (M8140-8580), and lymphoma (M9590-9723). All other histological codes for these head and neck sites were combined into a category labeled other.

A substantial proportion of thyroid cancers were categorized as "papillary carcinoma, NOS" (M8050), which is a subset of squamous cell carcinomas. Analysis of the database indicated that 99.6% of the papillary carcinoma, NOS cases were thyroid in origin. This classification of thyroid papillary carcinoma as a subset of squamous cell carcinoma was interpreted as a coding error. As a result, thyroid cancers coded as papillary carcinoma, NOS (M8050) were recoded as papillary adenocarcinoma, NOS (M8260).

Although distinct grading schemes are used for cancers that differ by specific morphologic and anatomical site grouping, the NCDB uses the World Health Organization's²⁰ standard grading system. This system provides 4 separate grades (and the additional category "unknown") for all cancers except lymphomas and leukemias. Through review of pathology reports, hospital-based cancer registrars assign grade 1 (well differentiated), grade 2 (moderately differentiated, moderately well differentiated, or intermediate differentiation), grade 3 (poorly differentiated), grade 4 (undifferentiated or anaplastic), or grade "unknown" to all cases. According to this scheme, when chart review reveals 2 different degrees of grading, the higher code is used. As a result, a neoplasm considered poorly differentiated with areas that are undifferentiated would be considered grade 4 (undifferentiated or anaplastic). Cases without numeric grading but described as low grade or partially well differentiated are considered grades 1 to 2, and therefore are recorded as grade 2. Cases described as medium grade are considered to be grades 2 to 3, and therefore are coded as grade 3. Moderately differentiated or relatively undifferentiated cancers are considered grade 3 and coded as such. Cases listed as high grade are considered grades 3 to 4, and therefore are recorded as grade 4.^15,22

Treatment presented in this report is limited to the first course of cancer-directed therapy used to manage the primary tumor. This initial cancer-directed therapy may include a combination of modalities and may span many weeks or several months if irradiation or multiple cycles of chemotherapy are included in the original treatment plan. Subsequent therapy to address recurrences is not included herein.

All analyses were performed using SPSS statistical software.²³ Survival analyses represent annual disease-specific rates with the date of diagnosis as the starting point and the date of death with cancer as the end point. For the purposes of this survival analysis, it was assumed that the cause of death was cancer for those patients whose status at last follow-up was recorded as dead with cancer. χ² Statistics were performed on selected contingency tables and pairwise comparisons were performed on selected survival rates. Because of the large number of patients within this sample, all resultant P<.0001. Therefore, no inferential statistics were reported. Instead, most data were presented in stratified form (eg, cross-tabulations) so that clinically relevant associations could be directly assessed.

Cancers of the head and neck represented 6.6% of all NCDB cases between 1985 and 1995 (range, 6.1%-7.6% per year) (Figure 1). From these overall counts, cases from reporting hospitals outside the United States (ie, Canada and Puerto Rico) were excluded from analysis, dropping the total number to 295022. Approximately 25% of the patients with cancer of the head and neck were accrued from smaller community hospitals, 33% from larger community hospitals, 33% from teaching and National Cancer Institute–designated institutes, and the remaining 9% from other types of hospitals (eg, free-standing cancer programs).

Demographic data were segregated into 118292 cases from the earlier period (1985-1989) and 176730 cases from the more current period (1990-1994). Age distribution remained stable across the years, with the 60- to 69-year-old cohort representing the largest percentage of cases (27.0%) (Table 1). Males outnumbered females at a ratio of approximately 1.5:1 that remained stable across the periods studied. The proportion of patients considered non-Hispanic African American increased from 8.0% to 8.8% from the earlier to the later period. There was a concurrent decrease from 4.4% to 4.0% in patients labeled "other/unknown." Reporting from the 6 regions of the United States was lowest from the Mountain region (16692 cases) and highest from the Midwest (71377 cases).

Income groupings were designed to broadly classify patients according to an approximation of the lowest and highest 10% of annual family incomes in the United States. Patients with cancer of the head and neck were disproportionately overrepresented in the lower-income group (13.8%) and underrepresented in the higher-income group (9.6%). The income grouping of patients varied substantially according to the anatomical site of the cancer (Table 2). Except for those patients with nasopharyngeal cancer, more than 15% of all patients with pharyngeal cancer were considered low income. The sites with the greatest proportion of patients in the high-income category were the thyroid gland (12.8%), other (10.6%), and major salivary glands (9.8%).

The largest number of tumors arose in the larynx (20.8%) and was followed in decreasing order by the oral cavity, including lip (17.6%), thyroid gland (15.8%), and the oropharynx (12.3%) (Table 3). The major salivary glands, which include the submandibular, sublingual, and parotid glands, were the site of origin in 4.5% of cases. The predominant histological type was squamous cell carcinoma, which represented 55.8% of all cases. Adenocarcinomas demonstrated the largest proportional change in histological types across the 2 periods; the increase from 18.2% to 20.3% represented an 11.5% proportional change in these 2 percentages. This increase paralleled a similar increase in the proportion of thyroid cancers. A smaller 4.8% increase in the proportion of cases that were lymphoma was noted between 1985-1989 (14.7%) and 1990-1994 (15.4%). A larger proportionate increase was noted for cases of lymphoma arising from the mucosal lining of the upper aerodigestive tract and adjacent salivary glands (data not shown).

Among those cases that were staged, the distribution remained stable across the 2 periods. Most cases were stage I (36.2%). A marked decrease in the proportion of cases without known stage occurred between 1985-1989 (40.8%) and 1990-1994 (18.3%). There was virtually no change across these years in the proportion of cases with unknown grade. Degree of differentiation was recorded as unknown in 44.7% of 1985-1989 cases and in 44.2% of 1990-1994 cases.

The most common initial management strategies used were surgery alone (32.4%), surgery combined with irradiation (25.0%), and irradiation alone (18.9%). In general, there was no substantial change between 1985-1989 and 1990-1994 in type of treatment. However, a notable increase in combined radiotherapy and chemotherapy was identified across the 2 periods. This combined modality approach accounted for 5.5% of patients treated in the earlier period and 6.3% of patients treated in the more current period, reflecting a proportionate increase of 14.5%.

Most cancers arising from mucosal surfaces of the upper aerodigestive tract (lip, oral cavity, pharynx, and larynx) were squamous cell carcinoma (Table 4). Adenocarcinoma was the most common histological type among thyroid gland (92.0%) and major salivary gland (55.4%) malignancies. Lymphoma comprised 79.8% of the cases originating in the site termed other, which includes lymph nodes of the head, face, and neck. The proportion of lymphoma cases arising at designated sites was greatest for eye and ocular adnexa (18.0%), major salivary glands (13.9%), sinonasal (12.0%), and nasopharynx (10.2%). Although lymphoma represented only 6.8% of cases in the oropharynx, this site was second only to other in the absolute number of lymphoma cases.

The number of cases classified as carcinoma, NOS was greatest in the larynx and thyroid gland. Although the absolute number of these cases was smaller in the nasopharynx than the larynx or thyroid gland, the nasopharyx had the largest proportion (15.1%) of carcinoma, NOS cases. Verrucous carcinoma was most common in the oral cavity and larynx, where it constituted 2.0% and 1.0% of cases, respectively.

Cross-tabulations of patient and disease characteristics by stage (Table 5) indicated that a greater proportion of advanced-stage cancers (stages III and IV) occurred among the lower-income group, the geographic region of the Southeast, and African Americans. Sex also was associated with differences in stage distribution. There were more advanced-stage cancers and fewer stage I cancers in the male group. These differences must be interpreted in the context of other sex differences. Thyroid cancer, which is dominated by early-stage disease, occurs in women more frequently than in men (data not shown).

Stage I cancers were most common in the younger (<40 years) patients. Stage IV cancers were more common among the middle-age (50-69 years) group. Grade was closely associated with extent of disease when grouped as early (stages I and II) and advanced (stages III and IV) disease. The ratio of early-to-advanced stage was 2.4:1 for well-differentiated cancers and 0.5:1 for poorly differentiated cancers.

Laryngeal, thyroid, lip, salivary gland, and eye and ocular adnexa cancers were predominately localized (stages I and II), whereas cancers arising in the sinonasal and pharyngeal sites were more commonly advanced (stages III and IV). Cancers of the oral cavity were evenly divided between localized and advanced stages. Squamous cell carcinoma and carcinoma, NOS were most commonly stage IV. Adenocarcinoma, verrucous carcinoma, and lymphoma were most commonly stage I.

The distribution of treatments varied by site (Table 6). Surgery was the most common treatment for cancers of the lip (85.2%), thyroid gland (54.8%), eye and ocular adnexa (50.6%), and oral cavity (46.2%). Irradiation was the most common treatment for cancers of the nasopharynx (37.5%), larynx (33.0%), and oropharynx (27.3%). Combined surgery with irradiation was the most common treatment for major salivary glands (42.4%) and hypopharyngeal cancer (29.9%). Combined chemotherapy and irradiation was used to treat 23.5% of nasopharyngeal, 13.2% of hypopharyngeal, and 11.4% of oropharyngeal cancers. Only 3.5% of patients with laryngeal cancers were treated with combined chemotherapy and irradiation.

The overall median follow-up for all cases in the early period (1985-1989) was 37 months compared with 10 months for the later (1990-1994) period (Table 7). It is clear that survival analysis would cover a limited follow-up for patients accrued in the 1990-1994 period. As a result, survival analysis was limited to cases accrued in the 1985-1989 period. Through analysis that excludes the small number of patients with unknown status, 49.8% of patients diagnosed in the earlier period were dead at last contact (with or without known cancer). Among the patients who were alive at last follow-up, 62.5% were without evidence of recurrent cancer. The median follow-up time for those who were alive at last contact with cancer was 44 months.

Five-year, disease-specific survival for all cases of head and neck cancers diagnosed from 1985-1989 was 64.0%. When analyzed by site, patients with cancer of the lip demonstrated the highest survival rate (91.1%), whereas patients with cancer of the hypopharynx demonstrated the lowest survival rate (31.4%) (Figure 2 and Figure 3). When analyzed by histological findings (Figure 4), the highest 5-year survival was noted for patients with adenocarcinoma (87.6%) and verrucous carcinoma (78.6%). The lowest survival rate was noted for patients with carcinoma, NOS (47.4%) and squamous cell carcinoma (56.9%).

To our knowledge, this report of cancer of the head and neck represents the largest to date. Although specific cancer management decisions cannot be made from such a broad survey, our analysis provides a background for more specific forthcoming analyses of the NCDB data set addressing individual histological types and sites.

Statistics addressing head and neck neoplasms are often misinterpreted due to groupings of dissimilar cancers. For example, the oral cavity has frequently been reported as the most common site for malignancies of the head and neck.²⁴ An artificially high number of cancers of the oral cavity has been reported as a result of grouping cancers of the salivary glands, the oropharynx, and even the hypopharynx together with cancers of the oral cavity.^25,26 The proper grouping of cases according to site as defined by the 1992 edition of the AJCC staging manual more accurately reflects a lower proportion of cases arising in the oral cavity.¹⁹ Analysis of the NCDB data using this AJCC classification identifies that laryngeal cancer is the most common head and neck malignancy in the United States and outnumbers cancers of the oral cavity (including cancers of the lip) at a ratio of 1.2:1.

It is reasonable to expect that the vast majority of cancer cases arising in the thyroid gland, pharynx, larynx, oral cavity, and salivary glands are entered into a hospital-based cancer registry on diagnosis.²⁷ In contrast, small cancers of the lip may be evaluated and treated in physician offices outside the hospital. If the analysis of the pathological findings from the biopsy or surgical excision of the lip cancer was not performed by a hospital-based pathologist, the case may not have been entered into a hospital registry. As a result, the sampling of lip cancer data may not be as complete as for the other sites of the head and neck.

It is noteworthy that cancer of the head and neck disproportionately affects the elderly and individuals in the lower-income bracket. Males are also represented to a much greater degree in the database than females. This sex disparity reflects the greater prevalence of squamous cell carcinoma among males. Although cancers of the thyroid and salivary glands affect females most commonly, the overall dominance of squamous cell carcinoma in the head and neck region is associated with a male predominance.

Changes in demographic data across the 2 periods were noted. A 10.0% proportionate increase in African Americans (from 8.0% to 8.8%) and a 21.4% proportionate increase in Hispanic patients (from 2.8% to 3.4%) occurred in the NCDB composition. These changes may partially reflect changes in the population of the United States over the course of the study. It is likely that the changes noted are partly due to a true increase in the incidence of cancer of the head and neck among these minority groups.⁹ Other studies^9,28-30 have noted the number of head and neck mucosal cancers to be disproportionately greater among African Americans than whites and to be increasing in this minority group.

An 11.5% proportionate increase in the histological group adenocarcinoma occurred between 1985-1989 (18.2%) and 1990-1994 (20.3%). This increase in adenocarcinomas paralleled a 13.2% proportionate increase in cases of thyroid cancer (from 14.5% in 1985-1989 to 16.7% in 1990-1994). An increase in thyroid cancer worldwide has been identified. Although the artifact of improved record keeping has been implicated, it is likely that other factors also account for this increase.³¹ A large number of children were subjected to therapeutic radiotherapy for conditions such as acne and adenoid enlargement before 1970. The aging of this population who had been exposed to therapeutic irradiation in childhood has been cited as a primary cause for the increase in thyroid cancers.³² The development of more advanced diagnostic methods has also been cited as instrumental in increasing the identified number of thyroid cancers.³³ The increased use of sophisticated radiographic imaging as well as expanded use of fine-needle aspirate biopsies may identify small, previously undetectable cancers.

A 17.2% proportionate increase in eye and ocular adnexal cancers occurred between 1985-1989 (2.9%) and 1990-1994 (3.4%). The largest component of this change was an increase in ocular lymphoma (data not shown). The proportion of lymphoma cases within the entire head and neck database also increased, from 14.7% in 1985-1989 to 15.4% in 1990-1994. This increase in lymphoma may result from an increase in the number of cases occurring in patients with immunodeficiency that parallels an increase in the prevalence of human immunodeficiency virus infection, transplant operations, and use of medical treatments (eg, chemotherapy) that affect the immune system.³⁴ Additionally, expanded use of immunocytochemistry, developed to permit greater precision in identifying lymphoma cases, may correctly identify patients who were previously assigned other diagnoses.³⁵ The risk of developing lymphoma increases with age. It has been proposed that some of the increase seen in the population may be associated with an increase in the proportion of the population who are elderly.

An increase was noted in the reporting of tumor stage during the more recent (1990-1994) period. It is important to note that assignment of stage as unknown does not necessarily identify poor record keeping. Standardized AJCC staging criteria have not yet been established for all sites within the head and neck region. The high percentage of unstaged sinonasal cases (52.0%) likely results from the lack of AJCC staging criteria for all but the maxillary sinus subsite within this region.¹⁹ The high proportion of unstaged eye and ocular adnexa cancers (48.9%) is also likely related to lack of a recommended stage grouping for carcinoma of the eyelid and conjunctiva.

It has been reported that appropriate AJCC staging is most often recorded for those cancers whose treatment is largely determined by stage.³⁶ Accurate staging has become more important in the management of squamous cell carcinoma at most sites in the head and neck with the expanded application of treatment protocols investigating use of chemotherapy and irradiation. These protocols usually have stage-dependent inclusion criteria. In contrast, management of cancers of the lip and salivary gland continues to be determined primarily by clinical and pathologic features rather than general stage groupings.³⁷ It is therefore not surprising that, among the sites with established staging criteria, patients with cancers of the lip and salivary gland represented the largest proportion of cases with unknown stage. The smaller degree of improvement noted in grading cases across the periods may reflect the ambiguity that persists regarding the importance of cancer grade. It is not widely accepted that tumor grade impacts on management decisions for most cancers of the head and neck.

The improvement in reporting stage from the earlier to the more current period has been supported by the COC. As of 1998, it was mandated that continued hospital approval from the COC is dependent on complete staging by the treating physicians of all cancers that have AJCC staging schema.³⁶ The increased involvement of the hospitals across the country in central registries, cancer treatment protocols, and other programs that require or encourage AJCC staging will likely contribute to continued improvement in staging.

The 16.3% of cases of squamous cell carcinoma recorded in the major salivary glands may represent an artificially high number. Although the NCDB designs its classification system by site of origin and not by site of metastatic disease, it is possible that a portion of the squamous cell carcinomas reported as salivary gland in origin actually represented metastases. It may be difficult to discriminate between metastatic disease to the parotid gland from an occult or previously treated skin cancer and squamous cell carcinoma arising de novo within the parotid gland. Using rigid criteria to exclude metastases to the parotid gland, a retrospective review by Gaughan et al³⁸ identified that only 1.9% of parotid malignancies were squamous cell carcinoma. Other reports³⁹ have identified squamous cell carcinoma to represent up to 9% of parotid malignancies and between 2.1% and 5.5% of submandibular malignancies.

Advanced stage at the time of diagnosis was more common among patients in the low-income group and the African American racial category (Table 5). Other studies^25,28 have shown oral and pharyngeal cancers to be not only more prevalent but also more lethal among African Americans. It has been suggested that some of these racial differences are tied to economic disadvantage. The disproportionate cancer burden of minorities may be related to limited access to resources, which, coupled with unfavorable environmental and behavioral factors, leads to a greater risk of developing cancer and failing its treatment.⁴⁰ Walker et al³⁰ have called for a conjoined effort to combine clinical, epidemiological, and molecular research to identify both the social and biological factors responsible for the racial differences seen.

Although a substantial proportion of the cases were without known grade (44.4%), the large number of cases for which grade was recorded permits evaluation of the data. A clear association between advanced stage and higher grade (lower degree of differentiation) is apparent. This finding supports the generally accepted tenet that, although the capacity to accurately assign prognosis for individual cases based on tumor grade may be limited, the behavior of cancer in a large group of patients can broadly be predicted by histological criteria.²²

A wide spectrum of treatments was used for cancers within individual sites (Table 6). It is apparent that management approaches are determined by factors other than anatomical site. Tumor histological findings, stage, and patient characteristics are additional critical determinants to management choices.

Although there was little overall change in treatment patterns across time, the use of combined chemotherapy and radiation therapy expanded from 5.5% of cases in 1985-1989 to 6.3% in 1990-1994. This 14.5% proportionate increase resulted primarily from an expanded application of this combination therapy in the treatment of laryngeal and pharyngeal carcinoma (Table 3). Protocols developed to preserve the larynx, pharynx, or tongue with chemotherapy and irradiation rather than extirpative surgery have become increasingly popular over the last decade.^41-44 Whereas management of cancers at most sites with adjuvant chemotherapy has not improved cure rates, recent reports^45-47 indicate that the addition of adjuvant chemotherapy may offer more effective treatment for nasopharyngeal carcinoma.

The median follow-up for patients from the 1985-1989 period who were "alive without cancer" at last contact was 66 months. Patients with the confirmed status of "alive with cancer" at last contact were followed up for a median of only 44 months. This substantially shorter duration of follow-up may indicate that a large proportion of patients recorded as being alive with cancer were lost to follow-up; it is reasonable to propose that some of those lost to follow-up died with cancer present. Had they not been lost to follow-up, they likely would have then been been classified as dead, with cancer.

For patients with squamous cell carcinoma who fail treatment, recurrence is usually identified within the first 2 years following completion of initial treatment.⁴⁸ In the absence of identified recurrence after this period, the chance of control of the index cancer is sufficiently great that the focus of further cancer surveillance is shifted to the greater risk of a second primary cancer developing.⁴⁹ Although recurrence data were not included in this report, data were presented that addressed the interval from time of diagnosis to date of last contact or death (Figure 4). Most patients with squamous cell carcinoma who died with disease within the 5-year follow-up period did so within 2 years of diagnosis. The rate at which patients died differed substantially by site and by histological findings. The proportion of patients who died within the first 2 years following diagnosis was greatest for squamous cell carcinoma and carcinoma, NOS. In contrast, the survival rate for patients with adenocarcinoma was more constant across each of the intervals evaluated. These findings are consistent with reports⁵⁰ suggesting that, while 2- to 3-year follow-up is adequate to assess treatment efficacy for squamous cell carcinoma, much longer follow-up (ie, 10-15 years) is required to assess outcome for adenocarcinomas.

Squamous cell carcinoma is the predominant histological classification for cancers of the upper aerodigestive tract. The differences in survival for patients with squamous cell carcinoma follow a general pattern: the more posterior and inferior the site, the worse the prognosis. This progression holds for the lip, oral cavity, nasopharynx, oropharynx, and hypopharynx but not the larynx. Although these survival figures may partially reflect anatomical differences, it is likely that a greater impact on survival comes from differences in the type and extent of the squamous cell carcinomas at these locations.

There are, however, anatomical differences between sites that are partially responsible in determining cure rates. For example, lymphatics in the larynx are not as abundant as they are in the hypopharynx. This difference in lymphatic drainage has been used to explain the higher metastatic rate and lower cure rate for hypopharyngeal cancer.⁵¹ Anatomical differences may also compromise the ability to treat cancer adequately. The morbidity incurred from resecting a tumor confined to the anterior mobile tongue is much less than that from removing a similarly sized tumor in the base of tongue. It is more likely that speech and swallowing will be preserved with an aggressive resection of the more anteriorly located tongue lesion. To preserve these functions in the process of treating a tumor involving the base of tongue, some compromise may be accepted through less aggressive treatment, which may result in decreased chance of cure.^52,53

A more detailed analysis of the NCDB is required to identify relevant findings that help direct management of individual patients. Detailed assessments of these variables will be provided individually for specific histological types and anatomical subsites through a currently ongoing review of the NCDB. These analyses will be presented in a forthcoming series of reports.

This national survey of cancer of the head and neck using NCDB represents the largest to date. Demographics, patterns of care, and outcomes are broadly addressed in review of cancers affecting the upper aerodigestive tract and adjacent structures. More specific reports that analyze specific histological types and anatomical sites within the NCDB will follow in subsequent publications.

Reprints: Herman R. Menck, MBA, National Cancer Data Base, Commission on Cancer of the American College of Surgeons, 55 E Erie St, Chicago, IL 60610 (e-mail: menckh@aol.com).

Accepted for publication April 1, 1998.

Supported in part by a Clinical Oncology Career Development Award (95-33) by the American Cancer Society.