Kaplan-Meier plot of survival with 95% confidence interval.
Mehanna HM, Morton RP. Does Quality of Life Predict Long-term Survival in Patients With Head and Neck Cancer?. Arch Otolaryngol Head Neck Surg. 2006;132(1):27-31. doi:10.1001/archotol.132.1.27
To assess whether pretreatment and posttreatment quality of life (QOL) is associated with long-term survival in patients with head and neck cancer.
Ten-year follow-up of an inception cohort.
Regional tertiary referral center.
The study included 200 consecutive patients with primary epithelial head and neck cancer.
Quality of life and several recognized risk factors for death were assessed prospectively using the Auckland QOL questionnaire before treatment and 12 months after treatment; survival was determined at 10 years.
Main Outcome Measures
Survival and odds of death (hazards ratio) were measured.
At 10 years, 136 patients (68%) were deceased, 48 patients (24%) were alive, and the status of 16 patients (8%) was unknown. Median survival was 6 years (interquartile range, 4.4-7.7). Before treatment, patients with low QOL had no significantly increased odds of death (hazard ratio, 1.4; 95% confidence interval, 0.8-2.4). In contrast, after treatment, patients with low QOL at 1 year had significantly increased odds of death (2.5; 95% confidence interval, 1.4-4.3; P = .001) even after adjustment for covariates.
Findings suggest potential survival benefits from improvements in QOL. However, the observed associations between survival benefit and QOL at 1 year may be confounded by comorbidity, which was not measured and deserves further investigation.
During the past 20 years, there has been increased awareness of quality of life (QOL) as an outcome measure of cancer management.1 More recently, there has been interest in QOL as a predictor of survival. Strong associations between overall QOL, a variety of QOL variables, and short-term survival have been reported in patients with a variety of advanced cancers treated with palliative intent, including breast cancer,2 lung cancer,3- 5 and melanoma.6 However, QOL variables predicting survival in these patients may not be relevant when considering long-term survival in patients with cancers treated with curative intent.
Therefore, studies evaluating the relationship of QOL with long-term survival have been published. They are, however, much fewer because of the difficulties associated with conducting long-term longitudinal studies.7 These studies have shown variable and sometimes contradictory results, with some demonstrating a strong, clear association between overall QOL and long-term survival8- 10 and others demonstrating no relationship3 or a relationship with only one component QOL variable.11- 13
For head and neck cancer (HNC), the limited evidence available suggests that pretreatment overall QOL and long-term survival are not associated.14 However, studies have found associations between long-term survival and individual variables, such as cognitive functioning,14 fatigue,15 or perceived self-efficacy.7 To date, no studies have examined the role of QOL after treatment, when patients have adapted to the diagnosis and treatment effects, in predicting long-term survival.
These apparent inconsistencies in QOL research in patients with HNC raise important questions. Does poor QOL increase the risk for death in the long term or is it a surrogate for other risk factors, such as tumor site and stage? And since QOL changes significantly after treatment, would QOL after treatment be a more accurate prognosticator of long-term survival?
We analyzed the association between QOL before and 1 year after treatment, and long-term survival in a cohort of patients with HNC. Several other factors that could affect this relationship were also analyzed, including sociodemographic data, alcohol intake and smoking history, disease characteristics, treatment, and psychological and functional disability.
A cohort of 200 patients with primary epithelial HNC were recruited for a prospective study of QOL after written informed consent was obtained by a trained nurse, as approved by the local ethics committee. The exclusion criteria were the inability to understand or read English, blindness, and having learning disabilities. The cohort was followed up for 10 years.
Patients completed a QOL questionnaire at diagnosis. Twelve months later, those patients free of recurrent disease completed the same QOL questionnaire again. At diagnosis, the following risk factor data were also collected: age, gender, alcohol intake and smoking status, disease type and stage (according to the American Joint Committee on Cancer Classification16), and subsequent treatment.
The Auckland Quality of Life Questionnaire was used in this study. This questionnaire is self-reported, patient oriented, and validated (see previous publications for further details17- 19) and is a composite of 3 questionnaires. The first, the Life Satisfaction Score, is a measure of general well-being and is considered a reliable measure of overall QOL.17 It consists of 10 questions, each scored on a Likert scale of 1 (poor) to 7 (best). The 10 scores are summed to give an aggregated score, with a range of 10 (poorest QOL) to 70 (best QOL).
The second component is the General Health Questionnaire. This well-validated instrument measures psychological distress.17,18 We used the 12-item version, with an aggregated score ranging from 0 (no distress) to 36 (maximum distress). The final component of the Auckland Quality of Life Questionnaire is the Functional Ability Questionnaire, which inquires about the severity of several important physical symptoms of HNC, including cough, speech and swallowing difficulties, head and neck pain, and shoulder pain. For clarity during the reporting of this study, some Functional Ability Questionnaire scores have been reversed where appropriate so that the direction of all scores measuring physical symptoms coincide; that is, higher scores denote better function and less severe symptoms.
The primary outcome measure for this study was death from all causes. This was determined from patient medical records, a centralized national clinic booking system, the patient's last recorded address, and the family physician or the next of kin.
Associations between QOL and other baseline characteristics were examined with logistic regression analysis, with low or not low QOL as the outcome measure. Cox proportional hazards regression analysis with time-constant covariate was used to examine associations between baseline variables and death from all causes, as well as QOL 1 year after diagnosis and death from all causes. The proportionality of hazards was assessed by plotting Schoenfeld residuals20 against each covariate, and no significant departures from the base model were discovered. The results are reported as hazard ratios and 95% confidence intervals. The nature of the associations between low QOL (at both baseline and 1 year) and death were examined using a linear term, quintiles, and a binary measure in the models. For ease of interpretation, continuous risk factors in the final models were expressed using categorical classifications because it had been established that the results were not sensitive to continuous or categorical expression. When analyzing the association between 1-year QOL and death (Table 3), patients who died within 12 months were excluded.
At 10 years, 136 (68%) of 200 patients were deceased. Of the remaining 64 patients, 48 were alive and the status of 16 could not be determined. The overall median survival was 6 years (95% confidence interval, 4.4-9.1; Figure). At the 1-year assessment, 3 patients had recurrent disease and were given palliative treatment. Data were available for 137 of 140 patients who were alive at 1 year without recurrence of disease. Patients' baseline characteristics are detailed in Table 1.
The associations between baseline characteristics and both low QOL (Life Satisfaction Score <55) before treatment and death are given in Table 2. Participants who smoked, had nodal involvement, underwent neck dissection, or reported swallowing difficulties were more likely to have low QOL before treatment.
Age, disease severity, pretreatment low QOL, and high psychological disability were all strongly associated with death, as were reported shoulder and arm pain and swallowing difficulty (Table 2). However, the association between pretreatment QOL and death decreased substantially after adjustment for covariates and became statistically not significant (Table 3).
In contrast, the association between low QOL after treatment and death remained strong even after adjustment for demographic data and disease severity (Table 3). The only other QOL measure to remain significant after adjustment was head and neck pain.
The results of this prospective study demonstrate a strong independent association between long-term survival and QOL after treatment of HNC. To our knowledge, this is the first time this has been reported in the HNC literature.
Before treatment, overall or global QOL was not strongly associated with long-term survival after adjustment for disease severity, supporting the findings of the only other substantive studies in this area in the HNC literature.7,14,15 During the pretreatment phase, classic predictors of survival (age, tumor stage, and site) were found to be more important in determining long-term survival. This is consistent with existing literature,7,14 further supporting the validity of our findings.
In contrast to previous studies that only examined QOL before or just at the start of treatment,2- 15,21 we also examined QOL 1 year after diagnosis. We hypothesized that, intuitively, it is more likely that the steady-state QOL, after patients had adjusted to the effects of the diagnosis and of treatment and had mobilized their coping strategies accordingly, would be the determinant of long-term survival,22,23 rather than pretreatment QOL. This was based on the observations that QOL status usually decreases noticeably during and in the period immediately after treatment and that patients return to a steady-state QOL at about 1 year after diagnosis.24,25 The findings of this study seem to corroborate this premise, and, to our knowledge, this is the first time this relationship between posttreatment QOL and long-term survival has been demonstrated in the HNC literature. It would also explain why a strong association between pretreatment overall QOL and long-term survival could not be demonstrated in our study and those of others,7,14,15 because the pretreatment QOL may be strongly confounded by disease severity and uncertainty about treatment.
We are aware of the potential limitations of this kind of study. Death from all causes was used as the end point, inasmuch as we were unable to always accurately define the cause of death because we frequently used death certificates for information, with the accompanying unreliability of this source as an indicator of true cause of death in patients with cancer. It is our experience that patients are interested in their overall survival prognosis rather than disease-specific survival.
As with most studies of HNC, as a result of the relatively low incidence of the disease and the heterogeneity of the tumor sites, there are few patients in each tumor site subgroup. This may lead to analysis bias because significant associations may not be detected owing to low sample numbers. Furthermore, there was incomplete data for all included patients, leading to attrition of patients available for analysis. Confidence intervals were provided to adjust for this.
We did not collect data on comorbidities except for smoking and alcohol consumption because at the time of inception of the cohort and at 1 year after diagnosis the idea of 10-year QOL follow-up and the role of comorbidities as a determinant of long-term QOL were not widely contemplated or considered by us or the literature available at that time. Therefore, we have not adjusted for comorbidities in our study. This raises the possibility that the association between QOL and long-term survival may be confounded by comorbidity and may be a reflection of it, as some authors have found comorbidity to be a significant prognostic indicator of survival. However, there are others, including some examining HNC,15 who have found that comorbidity does not have a major role.
We excluded patients with recurrent disease undergoing palliative treatment when reassessing QOL at 1 year. These patients are known to have both poor QOL and short-term survival.2- 4 Inasmuch as this study examined long-term survival, the authors thought that including these patients in the analysis was inappropriate because it would bias the results by strengthening the association between QOL and survival.
This line of research in other cancers, such as breast and lung cancers,8- 10 has demonstrated some strong associations between pretreatment overall QOL and survival, which is at variance with our findings in HNC. However, other long-term studies have not found this relationship or found only limited associations with single-component variables.11- 13 The reasons for these inconsistencies are unclear and are likely to be multiple, including methodological differences, small study numbers, differences in tumor types and behavior, and confounding by other factors.
There are also conflicting findings in the literature about the role of psychological factors, such as depression, in determining survival after cancer. Some report that psychosocial complaints are independent prognostic factors of survival after cancer,2 including HNC.7 Others,3,11,14 however, have not demonstrated such a relationship. This study did not find a strong association between psychological distress, as measured with the General Health Questionnaire, and long-term survival. Hence, it would seem that the relationship between QOL and survival may be in some, possibly large, part independent of psychological state.
Notwithstanding the difficulties with data and analysis, we believe we have produced sufficient evidence to warrant further investigation of the association between global QOL and survival in patients with HNC. While no causative relationship can be proved from this study, our findings may mean that interventions to improve QOL can potentially improve survival. Elucidating and understanding these associations and the prognostic role of comorbidities more clearly would be important in establishing whether QOL can be manipulated to enhance survival. Ultimately, the relationship between QOL and long-term survival will only be considered useful if interventions can effect any therapeutic or survival advantage.
To our knowledge, this study has demonstrated for the first time an independent and strong association between QOL after treatment and long-term survival in patients with HNC. While a weak association between pretreatment QOL and survival was demonstrated, recognized risk factors such as age and disease site and extent were found to be the main prognostic factors in the pretreatment period. The findings suggest potential survival benefits from improvements in QOL. However, the observed associations between survival benefit and 1-year QOL may be confounded by comorbidity, which was not measured and deserves further investigation.
Correspondence: Randall P. Morton, MB, BS, MSc, FRACS, Department of Otolaryngology and Head and Neck Surgery, Manukau Surgery Centre, Auckland City Hospital, PO Box 98743, South Auckland Mail Centre, Auckland 6, New Zealand.
Submitted for Publication: January 19, 2005; final revision received August 15, 2005; accepted August 22, 2005.
Author Contributions: Dr Morton 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.
Financial Disclosure: None.
Funding/Support: This study was supported by the Green Lane Research and Education Trust Fund, Auckland, the New Zealand Lotteries Commission, Wellington, and the Head and Neck Trust, Auckland (Dr Morton).
Previous Presentation: This study was presented at the Sixth International Conference on Head and Neck Cancer sponsored by the American Head and Neck Society; August 8, 2004; Washington, DC.
Acknowledgment: We thank Teena West, MSc, biostatistician, for her work and patience with the statistics.