Functional Outcomes After Treatment of Squamous Cell Carcinoma of the Base of the Tongue

Objective  To compare functional outcome and quality of life after various treatments for squamous cell carcinoma of the base of the tongue.

Design  Retrospective survey using statistical comparison.

Setting  Academic medical center, institutional practice.

Participants  Patients treated for squamous cell carcinoma of the base of the tongue between 1976 and 2000. Living patients 3 or more months after treatment were eligible. Questionnaire packets including validated site-specific quality-of-life instruments were mailed to 105 qualifying patients. Sixty-one patients participated, forming a volunteer sample. Patient responses were grouped according to treatment modality, operative vs nonoperative.

Main Outcome Measures  The planned outcome was that nonoperative therapy would result in better function than operative treatment.

Results  Most comparisons indicated no statistical difference in outcome between operative and nonoperative groups. Significant differences (95% confidence interval) were calculated for age, interval since treatment, and T stage. Group comparisons of patient responses revealed significant differences only in xerostomia and days hospitalized.

Conclusions  The tongue remains dysfunctional after both surgical and nonoperative treatment. Nonoperative treatment might more adversely affect saliva. Surgery is associated with a longer hospital stay.

SQUAMOUS CELL carcinoma (SCC) of the base of the tongue (BOT) is an aggressive tumor associated with a poor prognosis. Generally, the stage of the tumor is advanced at presentation. Over the years, various treatments have been developed to improve survival rates and today, various therapeutic approaches for SCC of the BOT are available.¹ The optimal treatment may now be defined as that treatment that provides the best disease control with the least functional morbidity and greatest quality of life. Surgical resection for advanced tumors of the BOT provides significant functional disability and in some patients, may require total laryngectomy.¹ As a result, nonoperative therapies such as external beam radiation therapy (XRT), chemoradiation, and brachytherapy have replaced surgery in some centers.¹ For smaller lesions, there are various options. Some advocate that XRT alone or a combined treatment of surgery followed by radiation is the best course of action.² Others believe that for T1 or T2 tumors, surgery alone is the best treatment because XRT causes xerostomia, tissue fibrosis, and osteoradionecrosis, and may accelerate dental caries.^3,4 In recent years, however, combined treatment has been shown to be superior to any single therapy.^2,5-7 For example, primary XRT followed by brachytherapy resulted in an 85% 5-year survival in one study, while at the same time enabling patients to achieve excellent functional status and quality of life.⁸ Some centers have already compared primary surgery with primary XRT and concluded that nonsurgical treatment led to better performance.⁹ Recently, one study recommended that XRT and brachytherapy be considered only for T1 and T2 tumors.¹⁰ We evaluated patients treated at the University of Pittsburgh Medical Center, Pittsburgh, Pa, for SCC of the BOT to determine their functional outcome and quality of life. This study does not assess survival, but does give insight into functional capabilities of this patient group by therapeutic cohort.

This research project was approved by the biomedical institutional review board of the University of Pittsburgh Medical Center. The records of patients presenting to the University of Pittsburgh Medical Center with SCC of the BOT between 1976 and 2000 were reviewed. Patients with less than a 3-month interval since treatment, with insufficient data in their record, or who were deceased were excluded from the study. Three months was selected as the minimum interval since treatment because by this time the acute effects from treatment had likely stabilized.

A total of 105 patients who underwent treatment for SCC of the BOT were eligible for study. The patients who qualified received questionnaire packets by mail to evaluate their performance status. Each packet contained an explanatory letter from a physician followed by 3 independent questionnaire instruments. The Performance Status Scale for Head and Neck Cancer Patients (List Questionnaire) assesses 3 areas of function: eating in public, understandability of speech, and normalcy of diet.¹¹ Eating in public and understandability of speech are measured using an ordinal scale of 25, 50, 75, and 100. Normalcy of diet is measured using an ordinal scale from 1 to 100 with increments of 10. This scale has been shown to be both reliable and sensitive to functional differences.¹¹ This scale was modified to instruct patients to self-report. The University of Washington Quality of Life Instrument¹² (UW-QOL) more thoroughly analyzes outcome and quality of life. Patients assessed their own pain, appearance, activity, recreation, swallowing, chewing, speech, shoulder, taste, saliva, and quality of life. Analysis of performance characteristics of the UW-QOL has shown this instrument to be highly consistent and reliable over time.¹³ Finally, we developed a third set of questions, the History of Treatment, to evaluate the hardships patients experienced directly due to their treatment (Figure 1). Self-addressed stamped envelopes were included in the packets to encourage patient participation. These questionnaires were administered once.

Of the 105 eligible patients, 61 agreed to participate in the study. Participating patients were characterized according to therapy. General cohort characteristics analyzed included age, interval since treatment, T stage, TNM stage, and presence of a comorbidity. Comorbidity was defined as a previous or current condition that could affect a patient's response or cohort assignment. Recurrences, second primary tumors, persistent disease, and other cancers were included in the comorbidity category. One patient was previously diagnosed as having pharyngeal stenosis and thus was considered to have a comorbidity.

Patients were separated into 2 groups: operative vs nonoperative. The operative group included all patients who had resection as part of their therapeutic plan. The nonoperative group consisted of patients treated with radiation therapy with or without chemotherapy or brachytherapy. The primary surgery cohort consisted of 43 patients. Each surgery performed was reflective of the stage of the primary tumor. All patients had a neck dissection appropriate to the preoperative staging. Transhyoid pharyngotomy was used to resect T1 and T2 tumors. These patients were reconstructed with primary closure. Patients with advanced primary disease underwent more extensive resections through a variety of approaches. Our data do not allow comparison of functional outcome among these surgical subgroups. The average age during treatment was 58.8 years. Following surgery, 87% (53/61) of the cohort underwent postoperative XRT. Eight patients did not receive adjuvant radiation due to refusal or prior XRT for positive neck nodes, carcinoma in the neck, or carcinoma of the larynx. Postsurgical brachytherapy of the BOT was used in 3 patients in addition to adjuvant XRT. Chemotherapy was given concurrently with radiation therapy in 8 patients. The primary surgery group had a mean interval since treatment of 66.8 months with a range of 4 months to 286 months.

The nonoperative cohort was composed of 18 patients. Their average age during treatment was 67.3 years. This cohort includes 1 patient (6%) who underwent chemoradiation; 1 (6%), brachytherapy; 2 (11%), XRT; 3 (17%), chemotherapy with brachytherapy; 4 (22%), XRT with brachytherapy; and 7 (39%), chemoradiation with brachytherapy. The chemotherapeutic approach varied according to protocol. Comparison of groups treated by different drugs was not possible. The nonoperative group had a mean interval since treatment of 29.5 months with a range of 3 months to 101 months.

The List Questionnaire already had a scale incorporated into it for analysis. A similar scale was necessary for the UW-QOL. The response to each question was translated into an ordinal scale—1 being the worst and 3, 4, 5, or 6 being the best, depending on the number of options for a particular question. No such scale was necessary for the History of Treatment instrument.

Initially, descriptive statistics (means, medians, and SDs) were used to examine the distributions of the variables. Exploratory data analytical techniques (boxplots) were used to determine any outliers. For continuous variables, the t test corrected for inequality of variances was used to compare group means (operative vs nonoperative). For ordinal variables and for skewed distributions, the Mann-Whitney test was used to compare groups. The χ² test was used for nominal variables. The Fisher exact test was used for tables with small cell sizes. The Spearman rank correlation test was used to examine the collinearity within each of the 3 instruments. To control for possible confounding variables, multivariate analysis (multiple linear regression and multiple logistic regression) was used to compare the groups after adjustment. Statistical significance was set at .05, assuming a 95% confidence interval. Statistical analysis was performed using SPSS software (SPSS Inc, Chicago, Ill).

Table 1 provides general characteristics of our sample population. The mean age and interval since treatment were 61.2 years and 56.2 months, respectively. The T stage was T3 or T4 in 26 patients (43%). The TNM stage was III or IV in 49 patients (80%). Twenty-eight percent (17/61) of the sample population had a significant comorbidity. In addition to these general characteristics, Table 1 also provides insight into how this sample population as a whole responded on the 3 quality-of-life instruments.

All cohort comparisons are listed in Table 2. The primary surgery cohort compared with the primary nonoperative cohort was younger, had a longer interval since treatment, and a lower T stage (P = .003, .003, and .04, respectively). The mean age was 58.8 years for the surgical cohort and 67.3 years for the nonoperative cohort. The mean interval since treatment was 66.8 months for the surgical cohort and 29.5 months for the nonoperative cohort. The primary tumor stage was T3/T4 in 37% (16/43) of the surgical cohort and 72% (13/18) for the nonoperative cohort.

There were no statistically significant differences between the surgical and nonoperative cohorts according the List Questionnaire. The UW-QOL results revealed that patients treated with surgery were left with significantly better saliva production (P = .02) when compared with patients treated nonoperatively. Nine patients (21%) in the surgical cohort had a saliva score of 4. Of the nonoperative cohort, 1 patient (6%) had a score of 4 or greater on the saliva scale. The History of Treatment questionnaire revealed a significant difference in days hospitalized between the surgical and nonoperative cohorts. The surgical cohort's mean days hospitalized was 12.3 days vs the nonoperative cohort's mean of 6.3 days. The corresponding P value was .002.

After these initial cohort comparisons, the variables in Table 2 were reanalyzed multivariately, adjusting for age, interval since treatment, and T stage. However, due to the small sample sizes in the surgical (n = 43) and nonoperative (n = 18) groups, it was impossible to incorporate all 3 variables in the same equation. Adjusting only for age was possible due to the fairly symmetric distribution. The adjustment did not change the direction or significance of any of the dependent variables. Adjusting only T stage resulted in cells with zero frequency. Adjusting only for interval since treatment was too skewed and would require further transformations.

The treatment of SCC of the BOT continues to evolve. While patient survival remains our primary concern, the development of new therapeutic modalities and the growing recognition of the importance of quality of life and functional outcome have stimulated new introspection. If survival outcomes for certain treatment modalities are similar, then cost to the individual (quality of life) and to society (expenses) assume increasing importance.

There are now numerous acceptable treatments for SCC of the BOT. While these therapies can be used in almost any combination to create a complete multimodality treatment, they can also be reduced to 2 basic categories—operative and nonoperative. We used this basic distinction between the various therapeutic options to assess differences in quality of life or function. We were surprised at the overall similarity in perceived morbidity and quality of life between the surgical and nonoperative cohorts. Even so, some differences did appear.

To determine if the significant differences calculated could be attributed to general cohort characteristics, patient responses underwent multivariate analysis according to age, interval since treatment, and T stage. TNM stage was not significantly different between the 2 cohorts. When reanalyzed according to age, there were no significant changes in the data. Therefore, patients' responses could not have been affected by age. This result confirms that the 8-year difference in age between the 2 primary cohorts does not affect function or quality of life.

Due to the small sample sizes, the data could not be reanalyzed after accounting for interval since treatment. While this study cannot determine if any patient responses were affected by interval since treatment, our experience is that pain usually diminishes as the interval since treatment increases. Pain is subjective and can be affected by patient outlook. It seems that long-term survivors place less emphasis on the details of their daily functioning and are simply happy to have been cured. Furthermore, we expected that the number of days hospitalized would have been affected by interval since treatment; health care economics has caused a shift toward briefer hospital stays in the last few years.

The data also could not be reanalyzed accounting for T stage. While this study could not determine if any patient responses were affected by T stage, we expected that patients treated surgically for large tumors to have more extensive dysfunction than for small tumors. While TNM stage predicts survival, T stage may predict the extent of tissue damage and consequently function.

The UW-QOL revealed a significant difference in saliva production. The general cohort characteristics have been shown not to affect these categories. The surgical cohort fared better in saliva production than the nonoperative cohort. These results were surprising since surgical patients usually receive postoperative XRT. However, almost 20% of the surgical patients in this study did not receive XRT. These patients might have skewed the results since postoperative XRT is the mainstay of surgical treatment. The results may also be due to differences in tumor stage and location, which affect the radiation field and dosage. Nonetheless, these results emphasize that nonoperative treatment affects saliva, potentially more severely than surgical therapy. A final important result from the UW-QOL is the absence of any differences between the 3 quality-of-life assessment questions or in the summation of these 3 questions into 1 total score. These questions try to unmask differences in the populations other than for what is being tested. These results indicate that the populations are likely similar in other ways.

The History of Treatment questions that we developed also provided important insights. There was a significant difference between the cohorts calculated for days hospitalized. The nonoperative cohort had a more favorable outcome than the surgical cohort. This implies that surgery could impede more on the patient's life. This extra impedance could be costly to the patient; however, the measurement ability of this instrument has not been examined. More accurate assessment of cost requires direct investigation of these factors.

Our results reveal that patients treated with cytoreductive therapy have similar dysfunction as patients treated surgically. The similarity between the surgical and nonoperative cohorts is the most striking finding of this study. We expected the nonoperative cohort would report significantly better function as measured by chewing, swallowing, and speech. These results indicate that after radiation therapy, significant morbidity results from fibrosis, xerostomia, and other side effects.

Additionally, this study uses 2 different instruments that attempt to measure the outcome of treatment for head and neck cancer patients. The List Questionnaire did not reveal any significant differences between cohorts while the UW-QOL revealed 1 difference. This result could indicate that the UW-QOL might be more sensitive to the outcomes of this particular patient population; however, a more thorough comparison is needed to determine if one instrument has greater utility over the other.

There are a number of limitations to our study. While we were pleased with the percentage of eligible patients that responded, there were still 44 patients who did not respond. This is a potential source of sampling bias. Nevertheless, our study was not institutional review board approved for additional attempts at patient participation. Unfortunately, the 2 cohorts consisted of relatively small populations. Even if every category initially showed differences between the surgical and nonoperative cohorts, the small populations may have prevented these differences from being statistically significant due to a reduced statistical power. A multicenter study could afford larger populations, allowing for statistical tools to be more sensitive to differences in patient responses.

In addition, for 2 of the 3 instruments used, patients were forced to choose from limited scales. The patients might have been forced into replies that were either too constricting or too broad. The first 2 questionnaires asked about current functioning and thus the large range in interval since treatment did not affect the accuracy of responses. The third instrument, however, asked patients about the details of their treatment. Accurate recall about the treatment decreases as interval since treatment increases. Finally, as with many quality-of-life assessments, the study is subjective in nature. It relies on patients selecting what they considered their level of functioning to be in the various categories. There is potential for skewed results in any subjective analysis. A long-term survivor who is relieved to be cured might believe his or her functional outcome to be better than that witnessed by an objective grader. However, this raises the question of which is more important: the actual outcome or the patient's perception of the outcome. Nevertheless, one recent self-assessment study that measures the impact of dysphagia on quality of life of patients with head and neck cancer has been shown to be valid and reliable.¹⁴ This study could prove valuable for future assessments.

While the results of this study are both surprising and interesting, it calls for further investigation. Larger populations could better account for age, interval since treatment, T stage, and subjectivity in patient responses, further elucidating how patients function after the various treatments currently offered for SCC for the BOT.

Accepted for publication February 13, 2002.

Corresponding author: Jonas T. Johnson, MD, Department of Otolaryngology, The University of Pittsburgh School of Medicine, The Eye and Ear Institute, Suite 500, 200 Lothrop St, Pittsburgh, PA 15213.