Trends in incidence and prognosis for head and neck cancer in the United States: A site-specific analysis of the SEER database


  • André Lopes Carvalho,

    1. Head and Neck and Otorhinolaryngology Department, Centro de Tratamento e Pesquisa, Hospital do Cancer A.C. Camargo, São Paulo, Brazil
    Search for more papers by this author
  • Inês Nobuko Nishimoto,

    1. Biostatistics Department, Centro de Tratamento e Pesquisa, Hospital do Cancer A.C. Camargo, São Paulo, Brazil
    Search for more papers by this author
  • Joseph A. Califano,

    1. Department of Otolaryngology-Head and Neck Surgery, Johns Hopkins Hospital, Baltimore, MD, USA
    Search for more papers by this author
  • Luiz Paulo Kowalski

    Corresponding author
    1. Head and Neck and Otorhinolaryngology Department, Centro de Tratamento e Pesquisa, Hospital do Cancer A.C. Camargo, São Paulo, Brazil
    • Head and Neck and Otorhinolaryngology Department, Centro de Tratamento e Pesquisa, Hospital do Cancer, A.C. Camargo, Rua Antonio Prudente, 211; São Paulo, Brazil
    Search for more papers by this author
    • Fax +55-11-3341-0326


Despite recent advances in the diagnosis and treatment of head and neck cancer, there has been little evidence of improvement in 5-year survival rates over the last few decades. To determine more accurate trends in site-specific outcomes as opposed to a more general overview of head and neck cancer patients, we analyzed the site-specific data collected in the Surveillance, Epidemiology, and End Results—SEER Public-Use Database 1973–1999. Based on the selection criteria, 96,232 cases were evaluated for trend analysis in incidence, clinical stage, treatment and 5-year survival. During the period 1973–1999, site-specific incidence rates for head and neck cancer changed significantly. Site-specific analysis of survival from 1974–1997 showed significant improvements in 5-year survival rates for cancers of the nasopharynx, oropharynx and hypopharynx (38.1% to 56.7% for nasopharynx, p < 0.001; 36.3% to 49.1% for oropharynx, p = 0.001 and 28.3% to 33.3% for hypopharynx, p = 0.015). The prognosis for early-stage salivary gland cancer during 1983–1997 and late-stage larynx cancer during 1974–1997 also demonstrated improvement (82.7% to 88.5%, p = 0.012 and 22.2% to 38.3%, p = 0.013, respectively). On the other hand, the prognosis for regional stage oral cavity cancer as well as early-stage larynx cancer patients declined during 1983–1997 (49.2% to 43.8%, p = 0.032 and 82.3% to 74.3%, p = 0.002, respectively). Site-specific changes in treatment and staging were also noted. Site-specific analysis allows for a more accurate description of incidence, staging, treatment, and prognostic trends for head and neck cancer. © 2004 Wiley-Liss, Inc.

Population-based cancer statistics in the United States (US) for 2004 is expected to yield 28,260 new cases of oral cavity and pharynx cancer and 20,260 new cases of larynx cancer, with a mortality of 7,230 and 3,830 deaths per year, respectively.1 Despite improvements in diagnosis and treatment, in the last 3 decades there have been no changes in the 5-year survival rate for larynx cancer patients and only a slight but significant improvement for oral cavity and pharynx cancer patients when comparing patients treated during 1973–1977 (5-year survival rate of 53%) to patients treated during 1993–1997 (5-year survival rate of 56%) according to results based on data from the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute.2 This slight change in survival was first noted in the 2002 publication,2 but these data can still be interpreted as indicating that “there is no change in prognosis for the last 2 decades” for head and neck cancer (HNC). However, during the last 2 decades many articles have consistently demonstrated advances in treatment, most notably the use of combination therapy (adjuvant radiotherapy after surgery, concurrent chemotherapy and radiotherapy) with improvements in survival rates.3, 4, 5, 6 In addition, improvements in radiotherapy and perioperative care, as well as prevention of chemotherapy-related complications, have resulted in increased quality of life and decreased treatment-related mortality.6

Another important factor that may influence survival includes lead-time bias. New technologies such as fiberoptic laryngoscopy, more widespread dental professional screening for oral cancer and higher-resolution radiological techniques may result in earlier stage diagnosis for some tumors, as well as more accurate staging.7, 8, 9 Such improvements in diagnosis may also occur in a site-specific manner.

In addition, treatment for head and neck cancer has evolved and become site-specific with treatment modalities individualized for specific anatomic sites and stages.4, 6 Therefore, analysis of head and neck cancer as a group may obscure important differences in survival trends for site-specific tumors that are dependent upon site-specific treatment advances.

For all historical periods, specific head and neck cancer sites have had very different treatment approaches and more importantly, different 5-year survival rates. For example, lip cancer has a 5-year survival rate over 80%, while hypopharynx cancer survival rates reach only about 30%.4 It is also possible that the stage and site distribution of cancer cases has changed over time. Unfortunately, conventional methods of assessing head and neck cancer survival have sampled cancer cases in aggregate, thereby potentially obscuring changes in survival rates that have occurred over time in specific sites. In a previous study,10 we observed drastic decreases in lip cancer incidence rates and significant increases in oropharynx cancer cases over the past 5 decades. Overall, we observed a decrease in the 5-year survival rate for all oral cavity and oropharynx cancers, but in site-specific analysis, we observed an increase in the 5-year survival rate for oral cavity (excluding lip cancer) and oropharynx cancer in the last decade. Similar phenomena may occur in other populations, and the SEER database provides an opportunity to analyze these effects in a large cohort.

Accordingly, we undertook a subgroup analysis of the SEER database for head and neck cancer in order to reveal an improved discrimination of historical survival trends among different HNC sites.


Our study is based on the incidence, clinical stage, diagnosis, treatment approach and 5-year relative survival data obtained from the SEER Public-Use Data 1973–1999 program of the National Cancer Institute, which covers about 10% of the US population.11 All authors have a signed authorization to access and use this data. The SEER register used in this study includes patients from 9 registries: Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, San Francisco-Oakland, Seattle-Puget Sound and Utah.12 Although the SEER database has included data from Los Angeles, San Jose-Monterey and the Alaska Native Tumor Registry since 1992, they were excluded since they do not provide data for an analysis of trends before this date.

Cases were selected according to the following inclusion criteria: invasive cancer originating from the oral cavity, pharynx, larynx and major salivary gland (carcinoma “in situ” cases excluded).

Cancer cases were classified according to the International Classification of Diseases for Oncology.13 Analysis of head and neck cancer from the database was stratified by site: lip (C00), oral cavity (C02–C05.0; C06), salivary gland (C07; C08), nasopharynx (C11), oropharynx (C01; C05.1,2; C09.0,1,9; C10.0,2,3), hypopharynx (C12; C13) and larynx (C32). Cases that could not be distinguished by specific site (oral cavity or pharynx) were grouped as “other mouth/pharynx.”

Clinical stages were grouped into the following categories: localized, regional or distant. This data is already available in the SEER database and is based on the Summary Staging Guide.14 Staging is defined as follows: 1) localized, an invasive neoplasm confined entirely to the organ of origin; 2) regional, a neoplasm that has extended a) beyond the limits of the organ of origin directly into surrounding organs or tissues, b) into regional lymph nodes by way of the lymphatic system or c) by a combination of extension and regional lymph nodes and 3) distant, a neoplasm that has spread to parts of the body remote from the primary tumor either by direct extension or by discontinuous metastasis (e.g., implantation or seeding) to distant organs or tissues or via the lymphatic system to distant lymph nodes. In the “distant stage” category, particularly for HNC, “remote spread by direct extension” means invasion to other specific site, e.g., invasion of the nasal cavity for oral cavity tumors or invasion of the oral cavity or mandible for oropharynx tumors and “distant lymph nodes” means involvement of neck levels not usually at risk; e.g., levels I for larynx tumors and usually level V for all specific sites.14 Usually the features for “remote spread by direct extension” would be included in the definition of “T4” tumors in the AJCC staging system.15 Although this classification is different from current AJCC staging criteria, we chose to use this clinical staging criteria because prior analyses of the SEER database have employed a similar staging system that has been in use since the inception of the SEER database.1, 2, 7 In addition, the AJCC staging system in the SEER database is available only from 1988 and only for some tumor sites of tumor, excluding HNC. For some cases staging information was not available. These cases were designated as “data not available” for their clinical stage category.

Treatment was grouped into the following categories: 1) surgery, primary treatment based on surgery; 2) radiotherapy, primary treatment based on radiotherapy with or without chemotherapy and 3) combined treatment, surgery plus radiotherapy (pre or postoperatively). Treatment information from the SEER database was not available for patients treated in 1998 and 1999; thus, these years were excluded from treatment-related analyses. As with staging, for some cases, information regarding the treatment was not available. These cases were designated as “data not available” for their treatment category.

Because trend analysis in the American Cancer Society publications begins with cases diagnosed in 1974,1, 2 cases diagnosed in 1973 were excluded from our analysis. Patients with “data not available” for clinical stage and/or treatment were also excluded from multivariable analysis.

Statistical analysis

The incidence rates calculated per 100,000 population and age-adjusted to the 2000 U.S. standard population were obtained from the SEER statistic program (SEER*Stat 5.0.20), excluding cases diagnosed in the years 1973 and 2000.16 Otherwise, data from SEER Public-Use Data 1973–1999 were obtained an ASCII format and transferred to STATA format were subsequent analyses were performed using the Statistical Package STATA release 7.0.

The time period of follow-up for censored cases or cancer-specific death was obtained from the database. Five-year survival rates were estimated by the Kaplan-Meier method and the significance of differences among survival curves was calculated by the log-rank test. Multivariable analysis was performed using the Cox proportional hazards model to estimate the hazard ratio (HR) with a confidence interval of 95% for cancer-specific death.

Linear regression models were performed to conduct trend analysis. When the coefficient of correlation was positive the trend was considered as increased, and when negative the trend was considered as decreased. Trend analyses were conducted for 3-year period groups as well as for 2 different time period intervals: 1) from 1974 to 1997 (the entire period) and 2) from 1983 to 1997 (the latter 15 years).

For all statistical tests, significance is 2-sided and achieved when p-values < 0.05.


We identified 96,232 cases from the SEER database that fulfilled all selection criteria. The distribution of cases according to the year of diagnosis, SEER registry region, tumor site, SEER staging group and treatment is shown in the Table I. Histology of squamous cell carcinoma was reported in 92.9% of the cases, excluding tumors of the major salivary gland.

Table I. Distribution of Cases According to the Clinical-Demographic Variables
Variable CategoryFrequency (%)
  • 1

    Excluding 1998–1999.

 Male68,821 (71.5)
 Female27,411 (28.5)
Age at diagnosis 
 ≤458,472 (8.8)
 46–5517,198 (17.9)
 56–6528,771 (29.9)
 66–7526,280 (27.3)
 >7515,511 (16.1)
Interval (year diagnosis) 
 732,536 (2.6)
 74–769,416 (9.8)
 77–7910,283 (10.7)
 80–8210,800 (11.2)
 83–8511,160 (11.6)
 86–8811,023 (11.5)
 89–9111,152 (11.6)
 92–9411,317 (11.8)
 95–9711,277 (11.7)
 98–997,268 (7.5)
Tumor site (group) 
 Lip9,163 (9.5)
 Oral cavity23,667 (24.6)
 Salivary gland6,274 (6.5)
 Oropharynx18,962 (19.7)
 Nasopharynx3,967 (4.1)
 Hypopharynx6,391 (6.6)
 Larynx27,287 (28.5)
 Other mouth and pharynx521 (0.5)
SEER Staging group 
 Local39,955 (41.5)
 Regional37,519 (39.0)
 Distant9,754 (10.1)
 Data not available9,004 (9.4)
 Radiation26,272 (29.5)
 Surgery + radiation23,786 (26.8)
 Surgery30,460 (34.2)
 Data not available8,446 (9.5)
SEER Registry Regions 
 San Francisco-Oakland16,801 (17.5)
 Connecticut15,815 (16.4)
 Metropolitan Detroit18,966 (19.7)
 Hawaii3,798 (4.0)
 Iowa12,869 (13.4)
 New Mexico4,725 (4.9)
 Seattle (Puget Sound)12,145 (12.6)
 Utah4,219 (4.4)
 Metropolitan Atlanta6,894 (7.1)

In order to ensure that our trend analysis was comparable to that shown in the official analysis from the SEER database by the American Cancer Society,2 we compared our results with those previously published and our findings resemble previous conclusions regarding significance in trends (Table II).

Table II. Comparison of 5-Year Survival Rates Between the Published Results2 and the Results Obtained in Our Study
Tumor site (group)Published results2 Five-year survival rate (%) by year of diagnosisp-value1Results obtained Five-year survival rate (%) by year of diagnosisp-value1
  • 1

    p-value refers to the difference between 74–76 and 92–97.

Oral cavity/Pharynx53%53%56%<0.0554.3%54.4%56.3%0.049

Our first observation regarding trend analysis based on the SEER database was an overall decrease in HNC incidence per 100,000 US population, depending largely on the specific tumor site. However, the slopes of the decreases are neither similar nor significant for all sites (Fig. 1a). These variations in trends affected remarkably in different ways the proportion of each specific tumor site among all HNC cases during the given period (Fig. 1b), leading us to believe that each site contributes to the overall survival in a variable manner over the time. Furthermore, there is significant variability in 5-year survival rates depending on tumor site, from 29.0% for hypopharynx cancer patients to 89.6% for lip (p<0.001) (Fig. 2). Given the excellent overall prognosis for lip cancer, the decrease in overall and relative incidence rate of this tumor site would be expected to contribute to an overall worsening of the prognosis for HNC in aggregate.

Figure 1.

Trends in incidence according to tumor site during the given period. (a) Trends in incidence according to the tumor site (rates are per 100,000 and age-adjusted to the 2000 U.S. Standard population). (b) Trends in proportion according to the tumor site (among the HNC cases).

Figure 2.

Overall survival according to the tumor site.

The clinical stage at diagnosis also showed significant variation. For all sites combined, a decrease in the diagnosis at a local stage and an increase in the diagnosis at a more advanced stage with regional disease was seen (Fig. 3).

Figure 3.

Trends in the rate o clinical stage (by SEER staging system) during the given period.

In addition, significant changes in treatment approaches over the study period (1974–1997) were noted, demonstrating an increased use of combined treatment (surgery plus radiotherapy; p<0.001), in contrast to a decrease in the use of surgery (p<0.001) or radiotherapy (p=0.028) (Fig. 4).

Figure 4.

Trends in the rate of treatment approach during the given period.

Site specific trends


There was a significant decrease in the incidence rate of lip cancer in the US population during 1974–1999 (from 2.7 to 1.1 per 100,000 population, p<0.05) (Fig. 1a). Also among all HNC cases, lip cancer accounted for 12.7% of the cases during 1974–1976 but for only 7.2% during 1998–1999 (p<0.001) (Fig. 1b). However, the rate of patients in each clinical stage at diagnosis did not change significantly, and most of the cases were diagnosed in early stages (nearby 80%) (Table III). Despite the absence of clinical stage differences at the time of presentation, the proportion of initial treatment modalities changed from 1974 to 1997. During this period more than 80% of the cases were treated by surgery alone and less than 4% were treated by primary radiotherapy. However the number of cases treated by primary radiotherapy declined from 4% during 1974–1976 to 2% during 1995–1997 (p=0.020) (Table IV). As expected from the lack of change in stage distribution and minimal change in treatment modality, lip cancer maintained its 5-year survival rate of around 90% during the entire study period. The survival rates, according to the SEER staging system, were 91.4% for local stage, 82.6% for regional stage and 52.2% for distant stage (Table V).

Table III. Trends in Clinical Stage (by SEER Staging System) According to the Year of Diagnosis
Tumor site groupRate of cases among HNC cases (%) by year of diagnosisp-value for trend1
  • 1

    p-value for trend calculated by linear regression model: (+) increased, (−) decreased, NS [no significant change (p > 0.05)].

 Data not available9.
Oral cavity          
 Local40.642.041.841.743.242.845.046.745.9<0.001 (+)
 Distant8. (−)
 Data not available12.49.610.310.610.411.512.69.57.4NS
Salivary gland          
 Local56.050.447.948.649.644.445.245.644.90.004 (−)
 Data not available6.
 Local21.425.726.025.022.522.721.419.716.50.025 (−)
 Regional45.548.347.750.452.653.457.859.662.9<0.001 (+)
 Distant20.716.717.215.114.312.311.511.113.10.001 (−)
 Data not available12.
 Local19.022.219.915.916.910.811.09.813.20.003 (−)
 Regional43.643.239.146.949.456.160.764.464.7<0.001 (+)
 Distant26.123.427.517.518.116.212.912.210.3<0.001 (−)
 Data not available11.311.213.419.815.616.815.413.611.8NS
 Local23.019.317.514.514.111.810.515.69.30.003 (−)
 Regional42.853.352.359.562.662.370.162.865.80.002 (+)
 Distant25.721.723.917.015.517.312.714.718.60.020 (−)
 Data not available8.
 Local53.354.453.948.648.550.650.351.945.60.042 (−)
 Distant6. (+)
 Data not available9.
Other mouth/pharynx          
 Regional28.051.446.357.755.963.458.660.069.20.004 (+)
 Distant42.029.728.415.415.312.712.94.419.20.009 (−)
 Data not available22.013.514.925.025.415.518.626.77.7NS
Table IV. Trends in Treatment Approach According to the Tumor Site and Year of Diagnosis
Tumor site groupRate of cases among HNC cases (%) by year of diagnosisp-value for trend1
  • 1

    p-value for trend calculated by linear regression model: (+) increased, (−) decreased, NS [no significant change (p > 0.05)].

 Surgery + Radiotherapy3.
 Radiotherapy4. (−)
 Data not available8.411.
Oral cavity         
 Surgery + Radiotherapy20.221.722.621.526.226.928.128.7<0.001 (+)
 Radiotherapy27.222.619.817.714.215.313.313.1<0.001 (−)
 Data not available7. (+)
Salivary gland         
 Surgery58.955.148.745.242.138.737.633.8<0.001 (−)
 Surgery + Radiotherapy29.733.339.541.947.049.348.454.4<0.001 (+)
 Data not available5. (+)
 Surgery + Radiotherapy21.<0.001 (+)
 Radiotherapy54.249.845.041.640.138.437.438.0<0.001 (−)
 Data not available8. (+)
 Surgery + Radiotherapy16.312.214.415.414.214.813.916.5NS
 Data not available11.68.613.614.013.412.78.39.8NS
 Surgery + Radiotherapy25.034.638.533.339.140.135.632.2NS
 Data not available9.410.011.814.313.912.211.114.2NS
 Surgery27.527.428.825.019.919.617.515.8<0.001 (−)
 Surgery + Radiotherapy24.128.328. (+)
 Data not available8.
Other mouth/pharynx         
 Surgery + Radiotherapy22.021.623.925.017.025.434.322.2NS
 Radiotherapy56.047.343.332.733.240.832.933.30.015 (−)
 Data not available10.010.811.915.435.615.518.622.2NS
Table V. Trends in 5-Years Survival Rates According to the Tumor Site and Clinical Stage
Tumor site groupOverall5-year survival rate (%) by year of diagnosisp-value for trend1
  • 1

    p-value for trend calculated by linear regression model: (+) increased, (−) decreased, NS [no significant change (p>0.05)].

SEER Local stage           
 Oral cavity71.468.068.771.772.572.275.371.872.0NSNS
 Salivary gland85.587.783.584.682.785.685.387.288.5NS0.012 (+)
 Nasopharynx65.353.845.958.881.370.666.489.970.50.049 (+)NS
 Larynx79.279.378.081.182.379.778.477.374.3NS0.002 (−)
 Other mouth/pharynx61.875.
SEER Regional stage           
 Oral cavity45.847.545.945.949.248.044.942.543.8NS0.032 (−)
 Salivary gland56.759.859.654.256.155.957.057.154.5NSNS
 Oropharynx41. (+)0.009 (+)
 Nasopharynx50.938.542.356.747. (+)0.002 (+)
 Hypopharynx29.728.227.729.926.427.032.834.334.60.028 (+)0.020 (+)
 Other mouth/pharynx38.447.632.541.640.337.033.749.229.3NSNS
SEER Distant stage           
 Oral cavity21.816.722.016.028.618.624.922.835.2NSNS
 Salivary gland23.415.626.924.
 Larynx35.422.233.026.826.233.442. (+)NS
 Other mouth/pharynx11.614.30.021.430. (−)

Oral cavity.

There was a significant decrease in the incidence rate of oral cavity cancers in the US population during 1974–1999 (from 3.6 to 2.7 per 100,000 population, p<0.05) (Fig. 1a); however these cancers accounted for a stable rate of about 24% of the all HNC cases during 1974–1999 (p=0.218) (Fig. 1b). Changes regarding the clinical stage at diagnosis were noted. From 1974 to 1999, almost half of the cases exhibited regional or distant staging; however, cases diagnosed with local clinical stage had significantly increased (from 40.6% to 45.9%; p<0.001), while the rates of cases diagnosed with regional clinical stage were unchanged. Additionally, cases diagnosed with distant staging significantly decreased from 8% to 6% (p=0.029) (Table III). In terms of treatment, almost half of all HNC cases were treated by surgery alone, but a significant shift towards combined treatment (surgery plus radiotherapy, from 20.2% to 28.7%; p<0.001) instead of primary radiotherapy was noted (from 27.2% to 13.1%; p<0.001) (Table IV). In addition, overall 5-year survival rates showed an increasing trend from 53.2% to 56.7%, but this increase was not statistically significant (p=0.063) (Fig. 5). When survival rates were analyzed by clinical stage, a nonsignificant improvement in survival was noted for localized (from 68.0% to 72.0%; p=0.058) and distant stages (from 16.7% to 35.2%; p=0.065). However, regional stage disease showed a significant decrease in survival rates from 49.2% to 43.8% (p=0.032) when taking into account the most recent 15 years (1983–1997) of the SEER database (Table V).

Figure 5.

Trends in 5-year survival rate (%) according to the tumor site and year of diagnosis.

Salivary gland.

There was a significant increase in the incidence rate of salivary gland cancer in the US population during 1974–1999 (from 1.0 to 1.1 per 100,000 population, p<0.05) (Fig. 1a), as well as relative to total HNC cases; these cancers accounted for 6.3% in 1974–1976 compared to 8.1% of all HNC cancers in 1998–1999 (p=0.002) (Fig. 1b). Trends regarding changing clinical stage at diagnosis were also evident. There was a significant decrease in cases diagnosed with local stage (from 56.0% to 44.9%; p<0.004) but no significant change in cases diagnosed with regional or distant stages (from 27.0% to 31.2%; p=0.060 and from 10.4% to 15.6%; p=0.859, respectively) (Table III). Treatment approaches also increasingly utilized multiple modalities as combined treatment strategies increased in frequency (from 29.7% to 54.4%; p<0.001) compared to decreases in surgery alone (from 58.9% to 33.8%; p<0.001) and no change in the use of primary radiotherapy (around 6%; p=0.346) (Table IV). Despite these changes in therapeutic regimes, the overall 5-year survival rates for all cases remained stable (around 68%) (Fig. 5). Survival by stage was also virtually unchanged with salivary gland cancers diagnosed at local, regional and distant stages having 5-year survival rates of 85.5%, 56.7% and 23.4%, respectively, during 1974–1997. However, in the latter 15 years of the period examined (1983–1997), there was a modest but significant improvement in survival for local stage cancers (from 82.7% to 88.5%; p=0.012) (Table V).


There was no change in the incidence rate of oropharynx cancers in the US population during 1974–1999 (around 1.5 per 100,000 population, p=NS) (Fig. 1a). However, relative to total HNC cases, oropharynx cancers showed a significant increase from 17.6% in the period 1974–1976 to 22.6% in 1998–1999 (p<0.001) (Fig. 1b). In terms of clinical stage at diagnosis, most patients (around 70%) were diagnosed with advanced stage disease (regional and distant stages). There was a significant decrease in locally involved tumors (from 21.4% to 16.5%; p=0.025), yet distant stage tumors decreased as well (from 20.7% to 13.1%; p=0.001). As such, a significant increase in oropharynx cancers of regional stage was observed (from 45.5% to 62.9%; p<0.001) (Table III). The frequency of surgery alone as a treatment remained unchanged during the period 1974–1997 (around 18%, p=0.501). Surgery in conjunction with radiotherapy, however, became increasingly popular with combined treatment frequencies increasing from 21.0% to 34.2% (p<0.001). Radiotherapy remained the most popular treatment modality despite its decreased use over time from 54.2% to 38.0% (p<0.001) (Table IV). These changes in therapy are accompanied by a significant increase in the 5-year survival rate from 36.3% to 49.1% (p=0.001) (Fig. 5). An improved 5-year survival rate was more pronounced for patients with regional stage tumors (from 35.2% to 50.6%; p=0.001), while local stage disease demonstrated only a modest survival increase (from 56.8% to 61.0%; p=0.186). Patients with distant stage disease demonstrated an intermediate improvement in survival (from 17.5% to 30.2%; p=0.076) (Table V).


There was no change in the incidence rate of nasopharynx cancers in the US population during 1974–1999 (around 0.6 per 100,000 population, p=NS) (Fig. 1a). However, relative total HNC cases, there was a significant increase of nasopharynx cancers cases from 3.4% to 4.7% (p=0.005) (Fig. 1b). Most patients (around 70%) presented with advanced stage disease (regional and distant). As with oropharynx cancers, a significant decrease was seen in cases diagnosed with local (from 19.0% to 13.2%, p=0.003) and distant disease (from 26.1% to 10.3%, p<0.001), while significant increases in cases diagnosed with regional disease was observed (from 43.6% to 64.7%, p<0.001) (Table III). There was no significant change in treatment approach along this period, with most cases undergoing primary radiotherapy (about 70%) (Table IV). Overall, a significant increase in 5-year survival rate was observed, from 38.1% to 56.7% (p<0.001) (Fig. 5). When patients were evaluated at on the basis of their disease stage at presentation, local stage (from 53.8% to 70.5%; p=0.049) and regional stage disease had significantly increased 5-year survival rates (from 38.5% to 59.3%; p=0.017), while patients with distant disease trended toward increased survival (from 30.6% to 44.8%; p=0.225) (Table V).


There was a significant decrease in the incidence rate of hypopharynx cancer in the US population during 1974–1999 (from 1.0 to 0.8 per 100,000 population, p<0.05) (Fig. 1a) However, as a proportion of all HNC cases, there was no change over that time period with hypopharynx cancer accounting for about 6.5% of HNC cases (p=0.290) (Fig. 1b). There was a significant decrease in cases diagnosed with local stage (from 23.0% to 9.3%; p=0.003) and distant stage (from 25.7% to 18.6%; p=0.020) disease and a significant increase in cases diagnosed with regional stage disease (from 42.8% to 65.8%; p=0.002) (Table III). No significant change in treatment modality was observed (Table IV). For all stages, significant increases in 5-year survival rates were observed from 28.3% to 33.3% (p=0.015) (Fig. 5). When looking at survival rates as a function of specific stage at presentation, only cancers staged with regional disease showed increased survival (from 28.2% to 34.6%; p=0.028), while local and distant disease showed insignificant differences (from 47.6% to 56.0%; p=0.079, and from 14.5% to 12.9%; p=0.469, respectively) (Table V).


There was a significant decrease in the incidence rate of larynx cancer in the US population during 1974–1999 (from 5.4 to 4.5 per 100,000 population, p<0.05) (Fig. 1a). Larynx cancer frequency as a proportion of all HNC cases also exhibited a significant decrease from 28.6% to 26.9% (p=0.049) (Fig. 1b). Regarding clinical stage at presentation, the proportion of local disease decreased significantly (53.3% to 45.6%; p=0.042), regional disease was unchanged (from 30.8% to 34.4%; p=0.543) and distant disease increased significantly (6.6% to 16.2%; p=0.002) (Table III). Almost 36% of all larynx cases were treated with primary radiotherapy during 1974–1997. Once again, the trend toward combination therapy was observed with surgery plus radiotherapy increasing in frequency from 24.1% in the initial period to 34.6% at the end of the period (p=0.013). The use of surgery alone declined from 27.5% to 15.8% (p<0.001) during the same period (Table IV). Overall, a significant decrease in 5-year survival rate was observed from 65.2% (1974–1976) to 62.0% (1995–1997; p=0.044) (Fig. 5). Regarding specific survival for clinical stage, the survival rates in the latter 15 years of the evaluated study interval revealed a significant survival decrease for cancers with local stage from 82.3% (1983–1985) to 74.3% (1995–1997; p=0.002). For regional stage cancers, the 5-year survival rate remained around 55% over the entire period (1974–1997; p=0.857), but increased survival rates were observed for patients with distant stage disease from 22.2% to 38.3% (p=0.013) (Table V).

Other mouth/pharynx.

This classification includes cancers without specific descriptions as to the exact site within the oral cavity or pharynx. There were no changes in the incidence rate of other mouth/pharynx cancers in the US population during 1974–1999 (around 0.3 per 100,000 population, p=NS) (Fig. 1a). Among all HNC cases, these cancers represent a very small and unchanging proportion of cases (less than 0.75% of the HNC cases) (Fig. 1B). Clinical stage at diagnosis shifted from distant towards regional disease (from 42.0% to 19.2%; p=0.009 and from 28.0% to 69.2%; p=0.004, respectively) (Table III). There was also a decrease in the rate of cases treated with primary radiotherapy, from 56.0% during 1974–1976 to 33.3% in 1995–1997 (p=0.015) (Table IV). For the entire study period, changing trends in survival were not observed. However, when taking only the last 15 years of the data into account, the five-year survival for patients diagnosed with distant stage disease showed a significant decline, from 30.0% in 1983–1985 to 0% in 1995–1997 (p=0.031) (Table IV).

“Data not available” patients for clinical stage and treatment.

Patients with “data not available” for clinical stage account for 9,004 HNC patients (9.4%) during the evaluated time period (Table I). The proportion of this group among the HNC cases remained constant for each category of tumor site (Table III). The overall 5-year survival rate for patients with “data not available” for clinical stage was 49.2%.

Regarding patients with “data not available” for treatment, they account for 8,446 HNC patients (9.5%) during 1974–1997 (Table I). Their proportion among HNC cases also remained constant for the majority of the tumor sites but had increased within the oral cavity cases (p=0.037), salivary gland cases (p=0.019) and oropharynx cases (p=0.029) (Table IV). The overall 5-year survival rate for patients with “data not available” for treatment was 41.2%.

Finally, in the Cox Regression model (adjusted for age, gender and SEER registry region), it was shown that site-specific differences in survival remained significant. This also showed that clinical stage and treatment approach are independent prognostic factors (Table VI). The treatment based on primary radiotherapy not only showed a decrease in survival when compared to primary surgery (HR 1.84; 95% CI 1.78–1.89) but also when compared to combined surgery and radiotherapy (HR 1.43; 95% CI 1.40–1.47). There was also a significant improvement in the survival rate during the study period, with those patients treated in the latter 15 years (1983–1997) showing better survival rates (Table VI). In addition, within the last decade (1989–1997), there has been continued improvement in survival rates with patients treated from 1995–1997 demonstrating the most improvement. Cases treated during 1995–1997 had better disease-specific survival rates when compared to patients treated during 1989–1991 (HR 0.92; 95% CI 0.88–0.96) or to patients treated in 1992–1994 (HR 0.92; 95% CI 0.88–0.97).

Table VI. Multivariate Cox Model Adjusted by Age, Gender and SEER Registry Region
Variable CategoryHR(95% CI)
 Surgery + radiation1.28(1.24–1.32)
Tumor site (group)  
 Oral cavity2.83(2.67–3.00)
 Salivary gland1.97(1.83–2.12)
 Other mouth and pharynx3.08(2.66–3.56)
Time interval (year of diagnosis)  


The majority of reports concerning HNC have concluded that there has been no improvement in the 5-year survival rate in the last 2 or 3 decades despite improvements in treatment and diagnosis.17, 18, 19 Most of these reports have relied upon the SEER data set publication1, 2 to draw these conclusions. We hypothesized that prior methods of analysis have largely ignored site-specific trends in HNC, regarding incidence, clinical stage, treatment and prognosis.

Although the SEER database analyzes only about 10% of the US population, the cohort evaluated by SEER is representative of the general population with regard to most socioeconomic measures. Same criticisms to this claim are that the SEER population tends to reside in somewhat more urban areas as well as to have a higher rate of foreign-born persons.12 Despite this, the SEER Program is considered the gold standard among cancer registries around the world since it is the only comprehensive source of population-based information that includes clinical stage at cancer presentation and survival data within each clinical stage.12

Our initial observations were based upon the fact that there have been significant changes in site-specific incidence rates of head and neck cancer, as well as in clinical staging and treatment approaches. The clinical stage and the tumor site are the 2 most important predictors of outcome for HNC;4 therefore the differences in survival cannot be properly interpreted without stratifying by tumor site and clinical stage. The clinical stage classification adopted in the SEER summary guide staging14 is quite different from the AJCC classification.15 For example, the category “distant stage” in the SEER database is equivalent to all AJCC stage IV categories (clinical stages IVa, Ivb and IVc, e.g., T4 tumors and/or advanced nodal metastasis and/or distant metastasis). Although there are differences between the SEER and AJCC classifications, the SEER staging system has been applied in a uniform fashion to the SEER database since 197420 and has been used by other investigators to report changes in clinical stage in population-based statistics.7 Thus, use of this original clinical stage classification allows us to analyze trends regarding clinical stage at presentation since 1974.

In terms of clinical staging trends for specific sites, we expected that improvements in diagnostic techniques and increased awareness of cancer diagnosis would result in a bimodal shift in the clinical diagnoses away from moderate stages and toward local and distant stages. The proliferation of novel diagnostic techniques, fiberoptic laryngoscopy and awareness of patients and professionals regarding cancer diagnosis was expected to shift some diagnoses towards earlier stages. At the same time, novel diagnostic techniques, improved imaging techniques and fiberoptic laryngoscopy would also allow a more accurate diagnosis of disease extent, likely resulting in upstaging of some cases.21

Interestingly, we noted an increase in local clinical stage only for oral cavity cancers, favoring the “early detection” hypothesis. On the other hand, for larynx cancers we observed a decrease in local stage disease with a concomitant increase in distant stage disease. This observation supports favoring greater knowledge of the extension of disease thought more accurate diagnosis or “upstaging” hypothesis. However, for most sites, a decrease in patients diagnosed with disease limited to the primary site, as well as a decrease in patients with distant stage disease was seen, probably as a result of a mixture of “earlier detection” and “upstaging” hypothesis causing an increased diagnosis of regional disease.

In terms of treatment, there is a clear trend to incorporate new techniques and advances in treatment, particularly with the use of combined modality therapy, including surgery and radiotherapy or chemotherapy and radiation therapy.3, 4, 5, 6 However, the results regarding treatment should be interpreted with care, since changes in treatment also stem from individual and institutional biases influencing treatment selection.

The wide divergence in survival rates for lip cancer (5-year survival rate = 89.6%) and hypopharynx (5-year survival rate = 29.0%); demonstrates that the often cited 5-year survival rate of 56% for oral cavity/pharynx cancers2 is a less precise description of the survival rates for these malignancies. The latter figure also does not take into account the variance in trends of incidence, clinical stage and treatment within the specific sites of HNC over the past few decades.

If the assumption that more accurate staging has occurred during the later years within the examined time period were true, the under-diagnosis of the extension of the disease in the beginning of the studied period would be corrected, upstaging all groups in the latter period. This stage migration should result in an “artificial” better survival rates for every clinical stage group in the latter period, as originally under-staged tumors become correctly staged, due to an event known as “Will Rogers Phenomenon”.21, 22, 23 However, the improvement in survival for local stage disease was only observed for salivary gland and nasopharynx cancer patients. In addition, larynx cancer patients showed a significant decrease in the survival rates for local disease. Furthermore, for regional stage disease, the improvement was significant only for cancers of the pharynx (naso, oro and hypopharynx), while oral cavity cancer patients with regional stage disease experienced a significant decrease in their survival. Finally, only larynx cancer patients experienced an improvement in the prognosis for distant stage disease. The trend of worse prognosis of oral cavity regional stage and larynx local stage cancer is somewhat intriguing. There are 2 possible reasons for these trends. The first reason might be due to changes in the subsite distribution along the period. It is known that within the oral cavity, cancers from floor of the mouth have a worse prognosis than gingival cancer, considering the larynx cancer, the supra and subglottic cancers have worse prognosis than glottic cancers. Thus, increases in the rates of disease arising from those subsites, as opposed to others subsites, would result in worse prognoses. Second, changes in treatment for specific clinical stages could also play a role in these changes in prognoses. So far we have observed changes in treatments for a given specific site but not for a given clinical stage for that specific site. Thus, the increase in the use of combined treatment (surgery plus radiotherapy) for both, oral cavity and larynx cancer, as well as the decrease in the use of primary radiotherapy for oral cavity and of surgery alone for larynx cancer observed considering all cases for these specific sites would supposedly improve the prognosis. However a different trend would be observed considering treatment for each clinical stage within these tumor sites.

Using a stratified analysis combined with a multivariable model we have demonstrated significant changes in HNC presentation and survival during the period of collection of the SEER database. Overall, the prognosis for HNC has improved in the last decade compared to the 2 decades before, and this improvement is most significant in the last. The prognosis of disease also differs according to tumor site. Most notably, all pharynx cancer patients (naso, oro and hypo) demonstrated an improvement in prognosis, while larynx cancer patients showed a decrease in their survival rates.

The site-specific analysis provides a more accurate reflection of incidence rates,24 staging and treatment trends as well as survival rates for HNC. As such, site-specific investigation could also provide useful information to assist in the targeting of prevention and treatment approaches as well as for planning clinical trials in HNC. Prospective clinical trials for HNC are increasingly designed in a site-specific manner, placing a higher value on site-specific analysis of the historical survival data.5, 25, 26


We thank Dr. A.A. Forastiere and Dr. M.M. Kim for critical reading and valuable suggestions for the article and L. Kapp for editing assistance with the article.