Influences and Predictors of Long-term Quality of Life in Head and Neck Cancer Survivors

Objective  To examine the impact of clinical predictors (pretreatment variables) and other influences (treatment and posttreatment variables) on long-term quality of life (QOL) in patients treated for squamous cell carcinoma of the upper aerodigestive tract. We hypothesized that baseline QOL and comorbidity would be predictors of QOL 1 year after treatment.

Design  Retrospective cohort study.

Setting  Academic Medical Center in Seattle, Washington.

Patients  Patients (N = 173) with baseline (pretreatment) and 1-year posttreatment QOL data.

Main Outcome Measure  Head and neck–specific QOL scores at 1 year after treatment (as measured by the University of Washington Quality of Life [UW-QOL] scale).

Results  We identified strong relationships between 1-year UW-QOL scores and baseline UW-QOL scores (correlation coefficient [Pearson r] = 0.58; P < .001) and pretreatment comorbidity (as measured by the Adult Comorbidity Evaluation scale) (Spearman ρ = 0.23; P < .001). T stage and N stage were also predictive. Although not a predictive variable, the presence of a gastrostomy tube at 1 year also strongly influenced 1-year UW-QOL scores. Patients with gastrostomy tubes had UW-QOL scores 11.5 points worse than those without (P < .001), when a 7-point difference is considered clinically significant. In predictive multivariate regression models, pretreatment QOL scores, comorbidity, and T stage had the strongest prognostic impact on 1-year UW-QOL scores.

Conclusions  In bivariate analyses, the presence of a gastrostomy tube worsens UW-QOL scores at 1 year and requires further investigation. When considering predictive variables only, baseline QOL and comorbidity appear to have strong influences on posttreatment QOL and have greater impact than treatment modality. Greater attention to these baseline predictors should be given when counseling patients about long-term function after treatment.

Recent studies suggest that quality of life (QOL) outcomes are independent of treatment modality.^1-5 If not treatment, then what factors predict QOL after treatment for head and neck cancer? A variety of variables have been proposed,^6,7 but little attention has been paid to comorbidity and pretreatment function.

Although comorbidity has an important role in predicting survival outcomes after treatment of head and neck cancer,^8,9 its prognostic impact on QOL is less studied. Similarly, although baseline function such as performance status may be considered in treatment selection,^10-12 its influence on long-term function after treatment is less well characterized.

The University of Washington, Seattle, has a cohort of patients with head and neck cancer with detailed longitudinal data on head and neck function, as measured by the University of Washington–Quality of Life (UW-QOL) scale. The UW-QOL questionnaire is a widely used, psychometrically validated questionnaire.^13-15 The clinical interpretation of UW-QOL scores has been established, including understanding of what constitutes a minimal clinically important difference,¹⁶ which helps to determine whether observed differences between groups of patients or changes in scores over time have clinically significant meaning.

The goal of this study was to look for predictors and influences on posttreatment QOL. We define predictive variables as those that can be measured at baseline, before treatment. We hypothesized that baseline (pretreatment) comorbidity and QOL would have a strong impact on long-term QOL, even after adjustment of clinical variables. We also sought to explore other influences, such as the concomitant presence of long-term gastrostomy and tracheotomy tubes. These variables cannot be considered to be predictive, since they are not known preoperatively, but they may provide insight into the QOL of head and neck cancer survivors.

We have previously described a prospective cohort of over 500 patients with longitudinal QOL data.¹⁷ For this study, we included only patients with squamous cell tumors of the oral cavity, oropharynx, hypopharynx, and larynx. Only patients with pretreatment and 1-year UW-QOL scores were included, which left a subset of 173 patients. Human subject review board approval from the University of Washington was obtained prior to beginning the study.

The existing database did not contain detailed clinical data beyond TNM stage, tumor location, and treatment modality (surgery, radiation, or chemotherapy). We required data on comorbidity, as well as gastrostomy and tracheotomy status. We therefore performed a medical record review of this cohort of patients to collect a full set of clinical data:

• Pretreatment predictive variables (baseline) included demographic data, TNM stage and tumor location, disease status at presentation (recurrent or new primary), and comorbidity. We used the Adult Comorbidity Evaluation scale (ACE-27),⁹ a widely accepted and validated measure, to calculate comorbidity. American Society of Anesthesiology scores were also collected but were not available for patients who did not undergo surgery.

• Treatment information included details on antineoplastic (surgical extirpation of the primary tumor, neck dissections, reconstructive surgery, radiation therapy, and chemotherapy) and ancillary (gastrostomy and tracheotomy) therapy.

• Posttreatment data included the primary outcome of UW-QOL scores at 1 year after the initiation of therapy. The UW-QOL is a 12-item scale with 3 additional global questions and is scored from 0 to 100. A 7-point difference is considered to represent a clinically important difference (B.Y., unpublished data, 2007). Several iterations of the UWQOL inventory were used, however scaled scores remained consistent throughout the years. We used the overall composite UW-QOL score for these analyses. We also documented the presence of gastrostomy and tracheostomy tubes, as well as disease status. Vital status was not relevant because only cancer survivors were included.

Data from paper and electronic medical records were extracted onto a standardized form in a Microsoft Access database (Microsoft Corp, Redmond, Washington). Well-accepted methods for extraction and classification of archival data for patients with cancer were used.¹⁸ In addition, study variables and potential values were explicitly defined in a handbook to ensure reproducibility. A separate coding guide was used to convert each data point into a numerical value for data analyses. Discrepancies about clinical variables in the medical record were addressed with the following conventions: (1) Symptoms or findings were recorded as present if noted by at least 1 member of the medical or nursing staff. (2) When dimensional data (eg, the amount of weight loss, tumor diameter) were discrepant, the largest number was recorded. (3) To assess interabstractor agreement, 5% of medical records were abstracted by two of us (M.E.-D. and B.Y.). (4) Analysis of these abstractions demonstrated good interobserver agreement, so the remainder of the records were abstracted independently by the first author (M.E.-D.).

Bivariate statistical analyses, including the unpaired t tests for dimensional data, χ² tests for categorical data, Spearman ρ and Kendall τ tests for ordinal data, and analysis of variance, were performed with the SAS version 8.0 (SAS Institute, Inc, Cary, North Carolina). Variables that were hypothesized to have predictive value, as well as those that were significant in bivariate analyses, were entered in a multivariate linear regression model. Two regression models were used—one including only predictive variables (baseline only) and a second using all relevant variables. P < .05 was considered statistically significant.

The overall 1-year UW-QOL score for the cohort of 173 patients was 68.9 (Table 1). Among predictive variables, strong bivariate relationships were observed between T stage, N stage, comorbidity, and baseline UW-QOL scores. The correlation coefficient (Pearson r) between baseline and 1-year UW-QOL scores was 0.58. The relationships were both statistically (Table 1) and clinically significant (Figure 1), since the magnitude of the observed differences were larger than the minimal clinically significant difference of 7 points.

Strong bivariate relationships were also observed with treatment variables (patients who underwent free tissue transfer and radiation therapy had worse UW-QOL scores at 1 year). Patients with gastrostomy tubes had UW-QOL scores 11.5 points (P < .001) worse than those without gastrostomy tubes. The presence of recurrent or persistent disease at the 1-year follow-up assessment had UW-QOL scores only 7.5 points worse than those without disease (Figure 2).

Our multivariable analyses focused on predictive variables (Table 2). Baseline UW-QOL scores, T stage, and comorbidity had a statistically significant impact on 1-year scores. Demographic, tumor site, and treatment modality including type of reconstruction had little predictive value for long-term UW-QOL scores. The coefficients listed in Table 2 provide insight into the magnitude of the anticipated impact. For each unit increase in the predictive variable, the 1-year UW-QOL changes by the listed coefficient. For example, since the coefficient for baseline UW-QOL scores is 0.63, a 10-point improvement in baseline UW-QOL scores would predict a higher final UW-QOL score of 6.3 (6.3 × 10) points. With a negative coefficient of −3.38 for T stage, a T4 tumor (compared with a T2 tumor) would be expected to have a lower 1-year UW-QOL score by 6.76 (3.38 × 2). Our findings were robust because all models confirmed the statistically significant impact of comorbidity and baseline QOL.

Although organ preservation rates have improved, survival rates for squamous cell carcinoma of the upper aerodigestive tract have remained stable in numerous recent randomized trials of new treatment paradigms.^19,20 As a result, increasing attention is focused on QOL and function, and we therefore need to improve understanding of the prognostic factors for these alternative outcomes. Although a number of clinical variables have been examined,^6,7 little attention has been given to baseline QOL and comorbidity.

Our findings suggest that pretreatment function as well as comorbidity play an important role not only in survival but also in long-term QOL. It is important to note that pretreatment UW-QOL scores are not just another marker for disease burden; pretreatment scores, comorbidity, and T stage were each a statistically significant, independent predictor of long-term QOL. Although treatment has been shown to have little influence on long-term QOL, clinicians continue to recommend treatment based on the projected impact of treatment on QOL.^1-5,21 Given that several studies have shown that QOL affects long-term survival,^22-27 it seems prudent to take pretreatment function and comorbidity into context when counseling patients regarding treatment outcomes beyond survival.

More important, however, is the notion that pretreatment function may have a greater impact on posttreatment function than treatment modality. This concept suggests that either posttreatment function is to some degree anchored to pretreatment function, or that innate personality characteristics of patients allow them to do well (or poorly) when others cannot. Initially optimistic patients may be more likely to maintain their optimism after treatment, thus influencing their subjective interpretation of their posttreatment results. Conversely, the opposite may be true for patients with an inherently pessimistic approach to their disease. This may support the rationale for more individualized treatment, with the primary goal of treatment selection (when survival is equivalent) being to restore the patient's perceived pretreatment functional deficits. Certainly, this data suggests that “cookie-cutter” approaches play little role in the treatment of squamous cell carcinoma of the upper aerodigestive tract.

Prognostic models are ideally based on pretreatment factors alone, as was the case in this study. However, our exploratory analyses suggested intriguing relationships between persistent gastrostomy tubes and worsened long-term UW-QOL scores. The presence of a gastrostomy tube had a greater influence on patient posttreatment scores (11.5-point deficit) than even the persistence of disease (7.5-point deficit), based on nonadjusted analyses. Both effects were clinically significant, since they were both greater than the minimal clinically important difference of 7 points. We caution that 3 of the 12 UW-QOL questions (swallowing, chewing, and taste) could be strongly influenced by a gastrostomy tube, and therefore its impact on UW-QOL scores may be exaggerated when compared with other head and neck–specific scales with more items.

The presence of a long-term tracheotomy tube also had a negative impact on 1-year QOL (5.8-point deficit), but surprisingly, this impact was neither statistically nor clinically significant. We note that these relationships are merely exploratory because they were not hypothesized a priori and therefore must be interpreted with caution. Furthermore, these results are unadjusted for confounding variables. For example, the bivariate analyses also suggest that patients undergoing free tissue transfer have worse long-term QOL, but this is likely to be confounded by more severe baseline disease. Nonetheless, it is consistent with evidence that swallowing function in patients with head and neck cancer is paramount.^6,28-30 Consideration of posttreatment swallowing function should be a primary consideration when selecting treatment and ancillary care.³¹

A strength of this study is the careful prospective collection of detailed QOL data. Furthermore, the UW-QOL scale is well understood and widely used, and we are able to provide interpretation of the clinical significance of the scores.³² The Head and Neck Oncology Team at the University of Washington has used both distribution-based techniques and anchor-based methods to estimate the minimal clinically important differences (7 points).^16,32 A detailed description of these methods is beyond the scope of this article. Our findings suggest that not only do T stage, comorbidity, and baseline QOL scores have statistically significant effects 1-year QOL scores, but that these effects are clinically important as well.

Correspondence: Mark W. El-Deiry, MD, Department of Otolaryngology–Head & Neck Surgery, University of South Florida, 12901 Magnolia Ave, Tampa, FL 33612 (mark.eldeiry@moffitt.org).

Submitted for Publication: July 16, 2008; final revision received December 5, 2008; accepted July 18, 2009.

Author Contributions: Dr Yueh had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: El-Deiry, Futran, and Yueh. Acquisition of data: El-Deiry, Weymuller, and Yueh. Analysis and interpretation of data: El-Deiry, Futran, McDowell, and Yueh. Drafting of the manuscript: El-Deiry and Yueh. Critical revision of the manuscript for important intellectual content: El-Deiry, Futran, McDowell, Weymuller, and Yueh. Statistical analysis: Yueh. Obtained funding: Yueh. Administrative, technical, and material support: Futran and Yueh. Study supervision: Futran and Yueh.

Financial Disclosure: None reported.

Disclaimer: The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.

Previous Presentation: This study was presented at the Seventh International Conference on Head and Neck Cancer of the American Head and Neck Society; July 20, 2008; San Francisco, California.