Matched Analysis of Survival in Patients With Squamous Cell Carcinoma of the Head and Neck Diagnosed Before and After 40 Years of Age

Objectives  To compare the survival rates of patients 40 years or younger and diagnosed with squamous cell carcinoma of the head and neck (SCCHN) with those of patients older than 40 years who underwent the same treatment. In 2 previous matched-pair analyses, the patients had been matched for tumor stage, site, sex, and date of presentation but not type of treatment.

Methods  Between 1995 and 2001, 46 patients 40 years or younger participated in a prospective epidemiologic study that included more than 500 patients newly diagnosed with SCCHN. We matched each of these patients by sex, race, tumor site, overall stage, and treatment modality with 2 patients older than 40 years. Ultimately, 31 of the younger patients were matched with 62 of the older patients. Survival analysis was performed using Cox proportional hazard models and accounting for the matched trios.

Results  There was no difference in overall, disease-specific, or recurrence-free survival rates between the patients who were 40 years or younger and those older than 40 years. Furthermore, matched survival analysis did not demonstrate a difference in overall survival rate (risk ratio [RR], 0.71; 95% confidence interval [CI], 0.22-2.29; P = .56), disease-free survival rate (RR, 0.83; 95% CI, 0.20-3.33; P = .79), or time to recurrence (RR, 1.46; 95% CI, 0.50-4.23; P = .49), and was not affected by adjustment for medical comorbidities or the severity of cancer-associated symptoms.

Conclusions  We found no evidence of a difference in the survival rates of patients with SCCHN who were 40 years or younger or older than 40 years and underwent similar treatment at our institution.

Most patients who develop squamous cell carcinoma of the head and neck (SCCHN) are older than 40 years. Only 3% of the patients who are diagnosed with SCCHN are younger than 40 years,¹ but it has been suggested that they develop a more aggressive type of the disease and that they may have a genetic predisposition that increases their risk for cancer.² However, studies about whether age at diagnosis affects prognosis have had conflicting results.^3-7 The major limitation of these studies has been the small number of patients younger than 40 years who develop SCCHN and the influence of confounding factors such as race, sex, smoking status, alcohol consumption, treatment bias, and the presence of comorbidities.

Matched-pair survival analysis allows to control for principal prognostic factors, and to adjust for confounding factors known to influence prognosis before a more precise comparison of survival rates between 2 groups is made. Two matched-pair analyses have analyzed age as a prognostic factor for SCCHN.^3,8 Both studies matched patients on the basis of sex, disease site, and date of presentation, and one study also matched them for overall disease stage. Neither study demonstrated a difference in overall or disease-free survival between younger and older patients. We present the first analysis of survival of patients younger than 40 years and diagnosed with SCCHN that matched patients for major prognostic indicators as well as the type of treatment received. We examined long-term survival to determine if age at time of diagnosis has any effect on disease-specific survival and overall survival in patients with SCCHN.

More than 500 patients with newly diagnosed, previously untreated SCCHN were enrolled in an epidemiologic study at The University of Texas M. D. Anderson Cancer Center between 1995 and 2001. Only incident cases of SCCHN confirmed by histopathologic studies were included in the present study. Patients with salivary gland, nasopharyngeal, or lip carcinoma, and those initially treated elsewhere, were not recruited. Forty-six of the 500 patients had been diagnosed with SCCHN at or before the age of 40 years and were included in this study. Two patients were subsequently excluded because they refused treatment, and 6 because it was discovered that they had received treatment elsewhere. The study comprised the remaining 38 patients.

All subjects participating in this study prospectively completed a self-administered epidemiologic questionnaire on presentation. The questionnaire quantified alcohol and tobacco consumption. "Former smokers" were defined as smokers who had quit smoking at least 1 year before presentation, and "never smokers" were defined as those who had smoked fewer than 100 cigarettes in their lifetime. "Drinkers" were defined as those who had had at least 1 alcoholic drink per week for at least 1 year, and "former drinkers" were defined as those who had quit such drinking at least 1 year prior to presentation.

Medical records were reviewed and data extracted for demographic information, primary site, clinical stage, treatment, histologic findings, outcome measures, and comorbidities, which were scored by using 2 previously described systems. One was the Cancer Associated Symptom Index, which scores the number of symptoms related to head and neck cancer prognosis.⁹ These include weight loss, dysphagia, neck mass, and otalgia. Symptoms are scored as none (0/4), mild (1/4), moderate (2/4), and severe (3/4 or 4/4). The other system was the Kaplan Feinstein Comorbidity Factor, which assesses related comorbidities as none to mild, moderate, and severe.¹⁰ Comorbid conditions included the following disorders: diabetes; hypertension; cardiac, renal, respiratory, cerebral/psychological, hepatic, locomotor, endocrine, and gastrointestinal peripheral vascular disease; alcoholism; other malignancies; and miscellaneous. These comorbidity scales have been demonstrated to be independent prognosticators in SCCHN.¹¹ Recurrent disease was defined as disease arising 3 or more months after completion of therapy; no patient had persistent disease at the end of treatment.

From this cohort, younger patients (≤40 years old at the time of diagnosis) were matched with older patients (>40 years old at the time of diagnosis). Matching variables were sex, race, site of primary tumor (oral cavity, oropharynx, hypopharynx, larynx, or cervical metastasis of unknown primary origin), disease stage (stage I/II vs stage III/IV), and treatment received (surgery; radiation therapy; surgery and radiation therapy; chemotherapy and radiation therapy; or chemotherapy, radiation therapy, and surgery). Each younger patient was matched to 2 older controls. Matching each case to 2 controls reduces the risk of matching cases to unrepresentative controls. Of the 38 younger patients available for matching, 31 could be matched to 2 controls of the same sex, disease site, stage, and treatment who had been diagnosed after the age of 40 years. Four of the 7 younger patients who could not be matched were treated with induction chemotherapy, but no control also treated with induction chemotherapy could be identified. The remaining 3 younger patients were not matched because of the small number of Hispanic, Asian, and African American women in the study. Therefore, 31 trios of younger and older patients were available for survival analysis.

Survival rates of the younger and older groups were compared using Kaplan-Meier estimates and the log-rank test for equality of survival curves. Matched survival analysis was completed using Cox proportional hazard models. Survival was measured from the time of the first appointment by using overall death, death from disease, and recurrence as censoring variables. Matching was accounted for in the Cox proportional hazard models by including a matching variable that accounted for the matching based on sex, disease site, overall stage, and treatment. The covariates smoking status, alcohol consumption status, Kaplan Feinstein Comorbidity Factor, and Cancer Associated Symptom Index score were also included in the survival model. We evaluated factors that were not matched (smoking status, alcohol consumption status, Kaplan Feinstein Comorbidity Factor, and Cancer Associated Symptom Index score, as well as T stage and N stage) by using the Pearson χ² test and the 2-tailed Fisher exact test to determine whether there was any significant difference between the groups of younger and older patients. This study was powered at 80% to detect a 2-fold difference in survival risk ratio between the younger and the older patients. The power of the study to detect the relative risk obtained was based on the Dupont model.¹²

The characteristics of the matched groups are listed in Table 1 to demonstrate appropriate matching. The patients were matched exactly by sex, race, disease site, overall stage (I/II vs III/IV), and treatment. As expected, no significant difference existed between the 2 groups with respect to the matching variables.

The age of the younger patients ranged from 24.2 to 40.9 years (mean, 34.5 years; median, 35.7 years). The age of matched older patients ranged from 41.0 to 83.0 years (mean, 64.1 years; median, 64.9 years). The specific staging and comorbidity characteristics of the 2 groups are listed in Table 2. The 2 groups were similar except for 2 variables: there were significantly more nonsmokers among the younger group (P<.001), and the Kaplan Feinstein Comorbidity score was lower in the younger group (P<.001).

Follow-up time ranged from 4.8 to 55.7 months (mean ± SD, 24.6 ± 12.0 months; median, 26.5 months) in the younger group and 1.5 to 65.2 months (mean ± SD, 23.5 ± 16.0 months; median, 19.0 months) in the older group. There were 13 events: 8 deaths due to disease and 5 deaths due to other causes. Mean survival duration until death from any cause (overall death) was 50.2 ± 3.6 months in the younger group and 53.8 ± 3.4 months in the older group (P = .73). Figure 1 shows the Kaplan-Meier overall survival curve for the younger and older patients. Mean survival until death from disease was 50.2 ± 3.0 months in the younger group and 58.4 ± 3.0 months in the older group (P = .88). Figure 2 compares the Kaplan-Meier disease-specific survival curve for the younger and older patients. Six (19.4%) of the younger patients and 10 (16.1%) of the older patients experienced disease recurrence. Mean survival until recurrence was 32.6 ± 2.6 months in the younger group and 51.8 ± 3.0 months in the older group (P = .40). Figure 3 compares the Kaplan-Meier recurrence-free survival curve for the younger and older patients.

There were no significant differences in risk of overall death, death due to disease, or time to recurrence between the young and older patients (Table 3). The risk was unchanged after adjusting for medical and symptom comorbidities in the survival model (Table 3). Furthermore, there was no difference in overall death, death due to disease, or time to recurrence after adjustment for T stage, N stage, or tobacco/alcohol consumption (data not shown).

The presence of SCCHN in younger patients was first reported by Byers² in 1975, when he described 11 patients with squamous cell carcinoma of the oral tongue, all of whom were younger than 30 years; their 2-year survival rate was 45%, whereas it was 65% in a historical cohort of older patients. While Byers suggested that squamous cell carcinoma of the oral tongue seemed to be more aggressive in patients younger than 30 years, the number of patients was too small to draw a valid statistical conclusion. Since this landmark article, conflicting results have been reported regarding whether age at diagnosis influences prognosis.^2-8

Age at diagnosis as a prognostic factor in squamous cell carcinoma has been most extensively studied in patients with squamous cell carcinoma of the oral tongue. Myers et al¹³ reviewed 37 patients with squamous cell carcinoma of the oral tongue treated at our institution between 1973 and 1995, and found that the 5-year survival rate of younger patients was similar to published findings for more typical, older patients. The authors also noted that the incidence of squamous cell carcinoma of the oral tongue in young adults appears to have increased since 1973. Friedlander et al¹⁴ performed a matched-pair analysis of younger patients with squamous cell carcinoma of the oral tongue, matching for sex, year of diagnosis, site of disease (oral tongue) and stage of disease, and did not find any difference in the 5-year survival rate of younger and older patients (62% vs 69%). There was a statistically higher rate of locoregional recurrence in the younger patients. However, the groups were not matched for treatment type, and there may have been a selection bias against postoperative adjuvant radiation therapy in the younger group. Pitman et al¹⁵ conducted a meta-analysis of patients diagnosed with squamous cell carcinoma of the tongue before the age of 40 years. They matched a group of 128 younger patients by sex, site (all oral tongue), and disease stage to patients diagnosed with squamous cell carcinoma of the tongue after the age of 40 years. The disease-free survival rate was similar for the 2 groups. Again, no attempt was made to match for type of treatment or race. Lacy et al⁴ found that patients younger than 41 years who had squamous cell carcinoma at any site appeared to have an overall survival advantage, but the authors did not examine disease-specific survival or attempt to match for prognostic factors.

Two previous matched-pair analyses examined the effect of age at time of diagnosis on survival in patients with SCCHN at all sites. Neither study found any difference in survival rates based on age at diagnosis. Verschuur et al⁸ examined 185 patients who developed SCCHN before 41 years of age. Matching was based on site, sex, and date of presentation. Race, disease stage, and treatment were not matching variables; however, most (86%) of the patients were treated with radiation therapy alone. The authors did not find any difference in rates of death due to disease within 5 years between the younger and older patients (72% vs 68%). Excluding disease stage or tumor stage as matching variables raises concern that the groups could have been dissimilar with respect to stage. Schantz et al³ compared 83 younger patients with SCCHN with patients older than 40 years who were matched for sex, site, disease stage, and year of presentation, and found no difference in survival rates, but again the groups were not matched for type of treatment.

Our study confirms previous reports that found no detectable difference in survival rates between younger and older SCCHN patients matched for tumor stage and site. Furthermore, we extended the matching criteria of the previous studies to include race and treatment type, and included comorbid conditions and symptoms in our multivariate analysis. The inclusion of treatment type as a matching variable controls for potential variability caused by selection bias, as certain patients may receive more aggressive treatment. Furthermore, medical and symptom comorbidities have been demonstrated to be independent prognosticators in SCCHN.¹¹ Because such comorbidities may be more common in some age groups than in others, we included these potential confounders in our multivariate adjustment, unlike previous studies assessing survival of the younger patient with SCCHN. Finally, every patient in this study participated in a prospective epidemiologic study in which clear documentation of tobacco and alcohol consumption were obtained, which allowed assessment of their impact. Tobacco and alcohol consumption did not appear to influence outcome risk estimates.

The major limitation of this study is the small number of patients; however, we tried to limit the effect of potential confounders by matching and included a larger control group of older individuals. This study was powered to detect a 2-fold relative risk between the younger and older patients, but this power calculation did not include the additional power afforded by the extensive matching process. The matching for 5 variables known to affect prognosis (sex, race, site, tumor stage, and treatment) make it more likely that any detectable difference was due to age. It is likely that the actual minimal detectable relative risk is smaller than calculated. The number of potential matches was limited by the ability to tightly match for several variables. A large study with strict matching criteria and the ability to detect a very small survival difference is not feasible because of the low incidence of disease among individuals younger than 40 years. Histologic factors such as extracapsular spread and perineural invasion could not be analyzed: because few patients underwent surgical procedures, few pathological specimens were available.

In conclusion, we found no evidence of a difference between the survival rates of patients diagnosed with SCCHN at or before the age of 40 years and those diagnosed at a later age who undewent the same treatment—even after controlling for major prognostic, demographic, and comorbidity indicators.

Correspondence: Erich M. Sturgis, MD, Department of Head and Neck Surgery, Unit 441, The University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Blvd, Houston, TX 77030 (esturgis@mail.mdanderson.org).

Submitted for publication June 10, 2003; final revision received October 28, 2003; accepted November 17, 2003.

This study was presented at the annual meeting of the American Head and Neck Society; May 4, 2003; Nashville, Tenn.

We thank Maureen Goode, PhD, for the scientific editing and Deanna Thomas, BS, for manuscript preparation.