To evaluate the impact of comorbidities, symptoms, and patients' characteristics on the 5-year overall survival of patients who underwent surgery for cancer of the oral tongue or floor of the mouth and to improve the survival estimates by the creation of a new staging system.
Patients and Methods
A cohort of 110 patients with squamous cell carcinoma of the oral tongue or floor of the mouth, who were admitted to a tertiary cancer hospital from January 1, 1990, to December 31, 1994, and who underwent surgery was studied. Multivariate analysis distinguished that patients' characteristics, symptoms, and comorbidities have a significant impact on 5-year overall survival. This functional severity index combined with the TNM stage created the extended clinical severity staging system.
The 5-year overall survival was 33.4%. Survival by TNM cancer stage was 64.6% (stage I), 67.5% (stage II), 28.9% (stage III), and 13.1% (stage IV) (χ2 = 22.88, P<.001). When patients were categorized according to the extended clinical severity staging system, survival was as follows: 74.0% (stage 1), 47.1% (stage 2), 28.6% (stage 3), and 8.4% (stage 4) (χ2 = 38.67, P<.001).
Clinical variables have a prognostic impact on oral cancer that is surgically treated, and the consistency of results confirms that survival estimates can be improved by the addition of these elements to the TNM classification, creating a more powerful and precise system in the determination of a prognosis.
APPROXIMATELY 30,000 new cases of oral cavity and pharynx cancer are diagnosed annually in the United States, which represents almost 3% of all tumors in men.1 Fifty-eight percent of these neoplasms are located in the tongue and in other parts of the mouth. In Brazil, 7950 new cases of mouth cancer were estimated in 1999.2 Early diagnosis is the best chance for an effective treatment, with aesthetic and functional satisfactory results. Surgery and radiotherapy are the primary modalities of treatment, and the choice of therapy depends on factors related to the tumor, to the patient, and to the institutional experience.3 The most important prognostic factor is the anatomical extension of the disease, described through the TNM staging system.4 Clinical characteristics of the patients, such as severity of the symptoms related to the cancer and medical comorbidities (defined as concomitant diseases not related to the disease under study), are important for therapeutic planning and for determining the risk of complications and the prognosis of several types of cancer.5-12 The addition of these factors to the traditional TNM staging system permitted the creation of new staging systems, superior in the prediction of survival when compared with the TNM staging system alone.6-8,13
This study was designed to evaluate the impact of comorbid conditions, symptoms, and patients' characteristics on 5-year overall survival in patients who underwent surgery for cancer of the oral tongue or floor of the mouth.
Two hundred forty-seven medical records of patients with squamous cell carcinoma of the oral tongue or floor of the mouth, admitted to the Centro de Tratamento e Pesquisa Hospital do Câncer A. C. Camargo, São Paulo, Brazil, from January 1, 1990, to December 31, 1994, were reviewed. The following criteria were used for inclusion in the study: a histologically confirmed diagnosis; absence of previous oncological treatment for this primary tumor; and surgical treatment with a curative purpose, exclusive or as part of a multidisciplinary approach. A total of 110 patients met the criteria for inclusion in the study.
Data collection from the medical records was performed using a form specially designed for this purpose. These data included demographic information, symptoms and duration, smoking status, alcoholism status, associated diseases, TNM stage (Union Internationale Contre le Cancer or American Joint Committee on Cancer classification),4 tumor site, hematocrit, and details about treatment. Follow-up information contained development of recurrences or second primary tumors and patient status at the last objective evaluation. Outcome measures included 5-year overall survival, 5-year tumor-specific survival, recurrence, and disease-free survival rates. Patients were observed from the date of diagnosis to the date of the last objective examination or death. Only 4.5% of the patients were lost to follow-up.
To analyze the significance of the symptoms, we created a classification of severity of the symptoms, using a method described previously by Pugliano et al.13 Only those symptoms clearly attributed to the cancer were used in the classification of the severity. Using a χ2 test, we found that among the 8 evaluated symptoms at diagnosis (burning sensation in the mouth, neck lump, oral cavity pain, dysphagia, weight loss, odynophagia, earache, and oral cavity bleeding), 5 had predictive potential at the significance level of P =.25: oral cavity bleeding, earache, weight loss, dysphagia, and neck lump. Neck lump was registered only when patients reported having noted it by themselves. Among patients who described this symptom, 90% (18/20) were classified as having metastatic lymph nodes by the physician at the clinical examination, and all patients had cervical metastasis histologically confirmed.
A 25% significance level was selected on the recommendation of Lemeshow and Hosmer14 for building multivariate models. This level of significance eliminates many insignificant variables from further analysis but ensures that all potentially explanatory variables are included in the multivariate analysis. The Cox proportional hazards model identified the following symptoms as independent predictors of survival: neck lump (P =.06), earache (P =.01), and oral cavity bleeding (P =.003). A symptom severity staging system was built based on the number of present symptoms (neck lump, earache, and oral cavity bleeding). The stage was defined as 0 for patients who did not have any of these symptoms. Stage 1 corresponded to the presence of 1 of these symptoms, and stage 2 corresponded to 2 or more of these symptoms.
In this study, Charlson and National Cancer Institute (NCI) indexes were used to classify comorbidity. The Charlson comorbidity index was created starting from a study of mortality rates among patients admitted to a unit of a university hospital in the period of 1 year. This index incorporates the number and the seriousness of the associated diseases. The system of punctuation of this instrument marks values of 1, 2, 3, and 6 for specific diseases present at hospital admission (Table 1), and later the comorbidity index score is determined by summing the weighted totals of all conditions in a given patient. The index score is then used to formulate the comorbidity stage, based on the medical record.15 Patients with a comorbidity index score of 0 belonged to comorbidity grade 0; those with a comorbidity index score of 1 or 2, comorbidity grade 1; patients with a comorbidity index score of 3 or 4, grade 2; and those with a comorbidity index score of 5 or greater, grade 3.
The NCI index was created in 1992 by a collaborative group of the National Institute on Aging (NIA) and the NCI to evaluate the prevalence of comorbidities in older patients (≥65 years) with cancer. This instrument was designed for the collection of information from hospital records. It included 24 comorbidities, such as a history of excessive alcohol intake, cardiovascular diseases, hip fracture, and urinary tract infection. Comorbid conditions were selected according to the leading causes of chronic diseases present in the community-dwelling population as reported by the National Center for Health Statistics, reports in the clinical literature for hospitalized patients, and reports on health status for selected populations.16 In this study, the NCI instrument was used and adapted for the NIA/NCI Colon Carcinoma Study Sample.17 The array of comorbidities and subcategories is listed in Table 2. Classifications recorded on the abstract form are grouped according to number of conditions present, level of current or historical impact of the condition, and no information available. The score was calculated by summing all conditions present. A comorbidity level of 1 was assigned to patients having 0 to 3 conditions, and a level of 2 was assigned to patients with 4 or more conditions.
The evaluation of staging systems can be done qualitatively or quantitatively. The qualitative comparison of different staging systems can be done based on face validity, clinical sensibility, or "common sense."18 A quantitative description of each development of the composite staging systems is provided by statistical analysis, and in our study the following quantitative techniques were used to compare the performance of the systems: the range of survival gradient, ie, the difference between the highest and lowest survival rates in each staging system (a wide range is obviously desirable); and −2 logarithm likelihood χ2, ie, the χ2 for covariates from logistic regression. For comparative purposes, the higher the value of χ2lt (lt denotes logarithm), the better.
The information contained in the forms was entered in a database (DBase for Windows; Borland International, Scotts Valley, Calif). Periodically, revisions were performed to verify the internal consistency of the data. For the statistical analysis, commercially available software (Statistical Product and Service Solutions for Windows, release 7.5; SPSS Inc, Chicago, Ill) was used. Descriptive statistics were used as a preliminary analysis of the relation between the baseline variables and outcome events. Continuous variables were categorized to facilitate data analysis and presentation. Survival analysis was performed using the Kaplan-Meier method (with the log-rank test value being used to compare groups), and the Cox proportional hazards model was chosen to identify independent prognostic factors.
The cohort of 110 patients included 93 men and 17 women; 93 were white, and 17 belonged to other ethnic groups. The patients' ages ranged from 31 to 80 years (mean, 57.5 years). Fifty-nine patients had tumors in the oral tongue, and 51 had tumors in the floor of the mouth. All patients underwent surgery as the primary treatment, and 69 underwent irradiation as adjuvant therapy.
Table 3 describes the 5-year overall survival rates according to 9 classification variables. The 5-year overall survival was 33.4%. Survival rates were almost equal for the 3 categories of the Charlson index and for both sexes. The largest differences in survival were observed for TNM clinical stage, the symptoms' staging system, alcohol consumption, comorbidities (NCI system), and hematocrit. The results show that TNM clinical stage, symptoms, comorbidities, age, alcohol use, and hematocrit produced the clearest distinctions in this sample. Figure 1, shows survival curves for all categories of the symptoms' staging system and comorbidities, respectively.
The first step in the organization of the clinical severity staging system was the conjunction of symptoms with comorbidities,8,9 producing a functional staging system with the following 5-year overall survival rates: asymptomatic with comorbidity level 1, 46% (20 of 44 patients alive); asymptomatic with comorbidity level 2, 43% (11 of 24 patients alive); symptom stage 1 with comorbidity level 1, 52% (7 of 12 patients alive); symptom stage 1 with comorbidity level 2, 0% (1 of 15 patients alive); symptom stage 2 with comorbidity level 1, 0% (1 of 8 patients alive); and symptom stage 2 with comorbidity level 2, 0% (0 of 7 patients alive).
The resulting survival rates allowed us to combine the patients into 2 categories of a functional severity staging system: α (including asymptomatic patients, despite the comorbidities' level, and patients with symptoms' stage 1 and comorbidities' level 1) and β (including patients with symptoms' stage 2 and comorbidities' level 1 and patients with comorbidities' level 2, despite the symptoms' stage). The survival for the 2 stages was as follows: α, 46% (38 of 80 patients alive); and β, 0% (2 of 30 patients alive). These rates demonstrate that survival decreases significantly with the increase of symptoms and comorbidities (P<.001). The survival curves for these functional stages are shown in Figure 2.
The conjunction of this functional system with the TNM staging system (second step) defined 8 different categories, and it was possible to show an important variation in the prognosis of patients with the same TNM stage and different grades of symptoms and comorbidities, especially in those with stages I, II, and IV (Table 4). The clinical severity staging system was then created through the consolidation of the categories, and it contains 4 stages: A, TNM stage I or II and functional severity stage α; B, TNM stage III and functional severity stage α; C, TNM stage IV and functional severity stage α; and D, any TNM stage and functional severity stage β. The 5-year survival for the clinical severity staging system differed significantly in the 4 groups: A, 74% (23 of 31 patients alive); B, 33% (9 of 26 patients alive); C, 25% (6 of 23 patients alive); and D, 0% (2 of 30 patients alive) (χ2 = 39.25, P<.001).
In addition to the symptoms and comorbidity, the following patient characteristics were added: age, alcohol use, and hematocrit. The Cox proportional hazards model identified, in a univariate analysis, 9 variables with an impact on the prognosis (P≤.10): daily alcohol consumption (hazards ratio [HR], 2.3; P =.008), neck lump (HR, 2.1; P =.008), dysphagia (HR, 2.1; P =.04), weight loss (HR, 1.7; P =.03), hematocrit of 0.35 or lower (HR, 1.9; P =.04), age older than 50 years (HR, 1.8; P =.06), NCI comorbidity index level 2 (HR, 1.6; P =.04), earache (HR, 2.0; P =.02), and oral cavity bleeding (HR, 2.6; P =.001). The clinical severity index was built through the multiplication of the values of the HR for each patient. When the condition was not present, we gave the value of 1 for that category. The score ranged from 1 to 258.23, and the results allowed grouping patients into 3 categories, based on the percentiles 30 and 60: high (score ≥8.7), intermediate (score >3.8 and <8.7), and low (score ≤3.8) grade of functional commitment. For example, if a 40-year-old patient presented with a neck lump, comorbidity level 2, and dysphagia, this corresponded to a score of 7.06 (2.1 × 1.6 × 2.1), which referred this patient to the category of intermediate grade of the functional severity index (FSI). Survival analysis also demonstrated a statistically significant difference between the 3 groups of the newly created FSI, described as follows: low grade, 63.6%; intermediate grade, 35.2%; and high grade, 7.9% (χ2 = 27.91, P<.001). Therefore, according to the method previously described, the next step was the conjunction of this other FSI with the TNM staging system. Once again, the categories of the conjunction of the 2 classifications were consolidated to create an extended clinical severity staging system, also composed of 4 stages (Figure 3). Five-year overall survival for this extended clinical severity staging system was as follows: stage 1, 74% (22 of 30 patients alive); stage 2, 47% (8 of 16 patients alive); stage 3, 29% (4 of 14 patients alive); and stage 4, 8% (6 of 50 patients alive). These rates were similar to those previously obtained for stages A, B, C, and D, and with important statistical significance (χ2 = 38.67, P<.001). The survival curves are shown in Figure 4.
The comparison among the systems demonstrated that both clinical severity staging systems (the "classic" and the extended elaborated systems starting from our new FSI) exhibited similar survival gradients in 5 years (74.2% and 65.6%, respectively). Also, they overcame the TNM stage that presented the lowest value of the χ2 test (22.88) and of the survival gradient (54.4%).
The evident advantage of this extended clinical severity staging system (addition of the variables age, alcoholism, hematocrit, symptoms, comorbidities, and TNM stage) over the other classifications in the prediction of tumor-specific survival could also be recognized. The extended clinical severity staging system also predicts recurrence rates and disease-free survival (Table 5).
For more than 30 years, the TNM staging system4 has been universally accepted and widely used as the basis of cancer staging. The system's macroscopic and microscopic classifications provide a reasonably precise description of the extent of disease. The system fails, however, for not including information about the clinical biological features of the cancer, which is expressed by structural format and physiological function in the patient. The gross anatomical features (extent of the disease), the microscopic appearance (cell type and degree of differentiation), and the biomolecular characteristics (tumor markers and ploidy) are different ways to describe tumor morphologic structure.19 Cancer symptoms (type, duration, and severity)20 and the performance status of the host21 are clinical elements that represent the severity of illness in a patient. Comorbidity is another important aspect of the clinical biological features for being able to affect the choice of treatment and prognosis, despite the fact that it is not related to the cancer itself.5,7,8,11
Symptoms are the result of the interaction between the host patient and the malignant neoplasm, providing important prognostic information already described in previous studies7,8 for different types of cancer. Previous studies9,13 on laryngeal and oropharyngeal cancer prove that symptom severity contributes, with additional prognostic data not available from anatomic staging alone. Pugliano et al22 demonstrated in patients with head and neck cancer that 4 symptoms—dysphagia, earache, neck lump, and weight loss—were found to be independent predictors of survival duration. A composite staging system was created based on the 4 symptoms, and when symptom severity stage was entered in a proportional hazards model along with TNM stage, comorbidity, age, and alcohol use, all 5 variables were independently predictive of survival duration.22 In fact, initially, we used the same method of Pugliano et al13 to build the symptom staging system. Our significant symptoms were different probably because we studied only patients with carcinomas of the oral tongue or floor of the mouth, but our findings that symptoms such as neck lump, earache, dysphagia, weight loss, and oral cavity bleeding are prognostic determinants substantiate similar results published in the literature13,22,23 for patients with head and neck cancer.
The patients' general health status directly influences treatment planning and estimates of prognosis. A less aggressive or even palliative treatment may be proposed to a patient who is considered "too sick" to tolerate preferred treatment. Therefore, an analysis of comorbidity should be included in any interpretation of outcome.5,9 The NCI recommends that future multi-institutional studies should be stratified according to variables regarding medical comorbidities, performance status, and a measure of alcohol and tobacco use, considered as definitely important and easy to obtain.24
The impact of comorbidity is most clearly evident in cancers that are not rapidly fatal. Thus, when comorbidity is included in a staging system, patients with higher survival rates show the best improvements in prognostic stratification. On the other hand, the effects of the comorbid conditions have been found to be more important in older patients and need to be assessed independently from functional status.25
We cannot apply the Kaplan-Feinstein index to measure comorbidity because it is significantly more difficult to use and apply than the Charlson index. This can be related to the multiple criteria required for the application of the Kaplan-Feinstein index that divide comorbid conditions into 13 categories, each having 3 severity grades. Severity grading within each category requires specific documentation of many tests and evaluations to establish the comorbid stage.26 In the retrospective study by Singh et al,27 the Kaplan-Feinstein index was significantly more difficult to use and apply than the Charlson index, suggesting that the Charlson index may be better suited for use in retrospective studies of comorbidity because it is easier to use and comparable to the Kaplan-Feinstein index in the prediction of survival. In 1992, the NIA and the NCI initiated a study to assess the prevalence of comorbid conditions in elderly patients with cancer. Seven cancer sites were selected for the study: breast, cervix, ovary, prostate, colon, stomach, and urinary bladder.28 In 1996, a report on approximately 7600 patients in the study sample described the NIA/NCI approach to developing information on comorbidity in elderly patients and addressed the chronic disease burden (comorbidity) and severity for 6 particular conditions: arthritis, chronic obstructive pulmonary disease, diabetes, gastrointestinal tract problems, heart-related conditions, and hypertension. Comorbidity data were matched with data from the conventional Surveillance, Epidemiology, and End Results program monitoring system. Analyses showed that hypertension is the most prevalent condition, as in our patients, and is also much more common as a management problem rather than as a history for the NIA/NCI Surveillance, Epidemiology, and End Results program study patients.16
As described by Guralnik29 in 1996, various assessment techniques have been used for the measurement of comorbidity and have demonstrated the association of increased level of comorbidity with various adverse health outcomes. The most basic measure of comorbidity is a sum of the number of conditions present. Although there is probably no single best way to assess comorbidity in all circumstances, many different approaches have proved valuable in demonstrating the presence of substantial comorbidity and in studying its impact. When evaluating patients' conditions, whether in observational epidemiological studies or in clinical trials, it is clearly important to consider the potential effect that comorbidity may exert on the outcomes of interest. Therefore, continued research is needed to evaluate techniques for assessing comorbidity and to develop new approaches to measuring this important concept.29
Several studies have applied multivariate analysis to large patient populations in an attempt to identify significant prognostic factors in head and neck cancer. Age30 and alcohol use31 have been reported as correlated with survival. Moderate anemia appears to be an independent prognostic factor in squamous cell carcinoma treated with radiation therapy alone.32 An alcoholic severity staging system developed by Deleyiannis et al31 demonstrated a distinct prognostic gradient across stage for all sites of head and neck cancer. The inclusion of alcohol use in our FSI is based on the high prevalence of alcohol use in patients with head and neck cancer that requires a comorbidity measure containing alcohol-specific information. Our results confirm that the incorporation of alcoholism, and age and hematocrit, in this index provides even greater prognostic information than previous comorbidity instruments. Furthermore, it complies with 1 of the 5 major purposes of the multivariable analysis: to assign simple rating scores to important variables and combine them into a single risk score to predict outcomes of individual patients.33
This study revealed an improved capacity of this new system to predict rates of recurrence and disease-free survival, when compared with other systems. The disease-free and tumor-specific survival were lower for patients with advanced stages of the new clinical severity system, independent of the treatment (surgery only or surgery combined with radiotherapy). This is similar to the results of a previous study27 in the literature in young patients with head and neck cancer and advanced comorbidity. The reasons can be a lower level of antitumor activity in patients with advanced comorbidity or simply biases of the physician in the treatment planning. These findings suggest that aggressive follow-up is especially important for patients with advanced comorbidity to warrant earlier detection of cancer recurrence.27 However, other studies13,23 for oropharyngeal and oral cavity cancer did not prove that the clinical severity staging system can do well at predicting recurrence rates.
This study demonstrates that clinical variables have prognostic impact on cancer of the oral tongue and the floor of the mouth that is surgically treated, and the consistency of results confirms that survival estimates can be improved by the addition of these elements to the TNM staging system, creating a more powerful and precise system in the determination of prognosis.
Accepted for publication March 22, 2000.
Presented at the Annual Meeting of the American Head and Neck Society, Palm Desert, Calif, April 24-27, 1999.
Reprints: Luiz Paulo Kowalski, MD, PhD, Centro de Tratamento e Pesquisa Hospital do Câncer A. C. Camargo, R. Professor Antônio Pudente, 211, CEP 01509-010, São Paulo-SP, Brazil (e-mail: email@example.com).
Ministério da Saúde, Estimativa da Incidência e Mortalidade por Câncer no Brasil para 1999. Rio de Janeiro, Brazil INCA1999;
RM Factors affecting the choice of initial therapy in oral cancer. Semin Surg Oncol.
1995;11183- 189Google ScholarCrossref
LHed UICC TNM Classification of Malignant Tumors. 4th ed. 2nd rev. Berlin, Germany Springer-Verlag1992;
AR The pre-therapeutic classification of comorbidity in chronic disease. J Chronic Dis.
1970;23445- 468Google ScholarCrossref
CK Cancer of the larynx: a new staging system and a re-appraisal of prognosis and treatment. J Chronic Dis.
1977;30277- 305Google ScholarCrossref
RI Comorbid and clinical determinants of prognosis in endometrial cancer. Arch Intern Med.
1984;1442004- 2009Google ScholarCrossref
S A new clinical-anatomic staging system for evaluating prognosis and treatment of prostatic cancer. J Chronic Dis.
1986;39913- 928Google ScholarCrossref
AR New clinical severity staging system for cancer of the larynx: five-year survival rates. Ann Otol Rhinol Laryngol.
1994;10383- 92Google Scholar
CK A clinical-severity staging system for patients with lung cancer. Medicine (Baltimore).
1990;691- 33Google ScholarCrossref
DR The effect of comorbidity on 3-year survival of women with primary breast cancer. Ann Intern Med.
1994;120104- 110Google ScholarCrossref
EW A reappraisal of staging and therapy for patients with cancer of the rectum. Arch Intern Med.
1975;1351441- 1453Google ScholarCrossref
et al. Clinical-severity staging system for oropharyngeal cancer. Arch Otolaryngol Head Neck Surg.
1997;1231118- 1124Google ScholarCrossref
D Applied Logistic Regression Analysis. New York, NY John Wiley & Sons Inc1989;
R A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis.
1987;40373- 383Google ScholarCrossref
et al. Cancer and comorbidity in older patients: a descriptive profile. Ann Epidemiol.
1996;6399- 412Google ScholarCrossref
et al. Comorbidity and age as predictors of risk for early mortality of male and female colon carcinoma patients: a population-based study. Cancer.
1998;822123- 2134Google ScholarCrossref
AR Clinimetrics. New Haven, Conn Yale University Press1987;
AR Clinical symptoms and comorbidity: significance for the prognostic classification of cancer. Cancer.
1996;77834- 842Google ScholarCrossref
WA Comorbidity and functional status in older women with breast cancer: implications for screening, treatment, and prognosis. J Gerontol.
24- 31Google Scholar
JR Symptoms as an index of biologic behavior in head and neck cancer. Otolaryngol Head Neck Surg.
1999;120380- 386Google ScholarCrossref
JR Clinical-severity staging system for oral cavity cancer: five-year survival rates. Otolaryngol Head Neck Surg.
1999;12038- 45Google ScholarCrossref
Jr Clinical staging and operative reporting for multi-institutional trials in head and neck squamous cell carcinoma. Head Neck.
1997;19650- 658Google ScholarCrossref
L Comorbidity and functional status are independent in older cancer patients. J Clin Oncol.
1998;161582- 1587Google Scholar
AR The importance of classifying initial comorbidity in evaluating the outcome of diabetes mellitus. J Chronic Dis.
1974;27387- 404Google ScholarCrossref
et al. Impact of comorbidity on outcome of young patients with head and neck squamous cell carcinoma. Head Neck.
1998;201- 7Google ScholarCrossref
et al. The National Institute on Aging and the National Cancer Institute SEER collaborative study on comorbidity and early diagnosis of cancer in the elderly. Cancer.
1994;742101- 2106Google ScholarCrossref
JM Assessing the impact of comorbidity in the older population. Ann Epidemiol.
1996;6376- 380Google ScholarCrossref
Jr A case-control study on complications and survival in elderly patients undergoing major head and neck surgery. Am J Surg.
1994;168485- 490Google ScholarCrossref
S Alcoholism: independent predictor of survival in patients with head and neck cancer. J Natl Cancer Inst.
1996;88542- 549Google ScholarCrossref
et al. Anemia is associated with lower local-regional control and survival after radiation therapy for head and neck cancer: a prospective study. Radiology.
1996;201553- 558Google Scholar
TR The risk of determining risk with multivariable models. Ann Intern Med.
1993;118201- 210Google ScholarCrossref