Figure 1. Conditional and traditional observed survival rates of patients with head and neck cancer.
Figure 2. Conditional and traditional disease-specific survival rates of patients with head and neck cancer.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Thompson TL, Pagedar NA, Karnell LH, Funk GF. Factors Associated With Mortality in 2-Year Survivors of Head and Neck Cancer. Arch Otolaryngol Head Neck Surg. 2011;137(11):1100–1105. doi:10.1001/archoto.2011.179
Author Affiliations: University of Iowa Carver College of Medicine (Ms Thompson) and Department of Otolaryngology–Head and Neck Surgery, University of Iowa Hospital and Clinics (Drs Pagedar, Hynds Karnell, and Funk), Iowa City.
Objectives To determine conditional survival rates of 2-year survivors of head and neck cancer and to identify risk factors of increased mortality.
Design Prospective, observational study conducted from September 1, 2001, through September 31, 2008.
Setting Tertiary care institution.
Patients Two hundred seventy-six patients who survived 2 years after the diagnosis of their upper aerodigestive carcinoma.
Intervention Patients prospectively provided health-related information.
Main Outcome Measures The primary outcomes were observed (death from all causes) and disease-specific (cancer-related) survival for 2-year survivors.
Results Five-year observed (90.8%) and disease-specific (94.8%) survival rates were 29.7 and 25.0 percentage points higher, respectively, than rates calculated for all patients at diagnosis. Older age and advanced stage were associated with poorer survival, whether death was due to the cancer or from all causes. Patients with pain or poor overall quality of life at 2 years were more likely to die from all causes, whereas those still smoking 2 years after diagnosis were more likely to die from their cancer.
Conclusions In addition to older age and advanced stage, pain, poor overall quality of life, and tobacco use 2 years after diagnosis characterize patients who might need longer and more intense follow-up care to improve their observed and disease-specific survival. This information is useful in developing management plans for patients transitioning from a focus on cancer surveillance into survivorship.
A conditional survival rate is the probability of surviving after having already lived for a certain length of time. This concept is important in the care of patients with head and neck cancer (HNC) because it underlies the intuition of head and neck oncologists that mortality and recurrence rates are lower for patients further out from diagnosis. Quantification of conditional survival rates can therefore offer long-term survivors of HNC more accurate information about their prognosis. In addition, cancer is a primary consideration in understanding survival rates at the time of diagnosis, but other factors might be important 2 years after diagnosis. Information about factors associated with survival rates in patients who have already survived a number of years can also help physicians develop evidence-based plans for the follow-up management of this patient population.
To date, the few studies addressing conditional survival among patients with HNC have focused primarily on quantifying the extent to which survival rates improve after successful treatment. Fuller et al,1 in a 2007 study of 76 181 patients with HNC, reported a 5-year observed survival rate of 47.8%, whereas the group of patients who had already survived 3 years had a 64.4% likelihood of surviving an additional 5 years. This study suggested that, 6 years after diagnosis, patients' risk of dying from noncancer causes was equivalent to their risk of dying from their cancer. Yang et al2 also studied conditional survival in 339 patients with oral cancer. They described the same phenomenon with increased survival rates in the group who survived an initial period. Specifically, the benefit of having survived a certain length of time did not accrue until approximately 1 year after diagnosis.
Van der Schroeff et al3 studied conditional survival rates in a cohort of 7255 patients with HNC. They found that long-term survivors still had poorer 5-year conditional relative survival rates compared with age- and sex-matched equivalents in the general population. The authors theorized that the increased risk of death among long-term survivors of HNC, even 15 years after diagnosis, was due to comorbidity in this patient population, possibly related to smoking and alcohol abuse.
Age and cancer stage are examples of factors that predict survival in patients with cancer. The factors that predict conditional survival are not well described. The primary goal of this study was to identify factors predictive of observed and disease-specific conditional survival rates in patients with HNC. We chose to focus on 2-year survivorship because, at that point, recurrence risk is notably decreased and the follow-up care begins to encompass additional health issues beyond cancer surveillance. We sought to identify factors contributing to mortality in 2-year survivors of HNC.
The patient sample for this study was extracted from the University of Iowa's longitudinal Outcomes Assessment Project. All individuals with carcinomas of the upper aerodigestive tract are asked to participate by providing clinical and health-related quality of life (QOL) data before treatment, every 3 months during the first year, and then at 2, 3, 4, 5, 10, and 15 years after diagnosis.
This study included Outcomes Assessment Project participants who presented with a primary tumor and who survived without a recurrence for at least 2 years, requiring that their diagnosis occurred from September 1, 2001, through September 31, 2008. The project's accrual rate during this period was 73.0%, with 8.3% of eligible patients refusing to participate and 18.7% being missed.
Patient-reported outcomes were captured using standardized surveys and single-item queries. The Medical Outcomes Survey Short-Form 36 was used to measure patients' perception of their general health.4,5 This 36-item survey assesses multiple areas of physical and mental health and well-being that are aggregated into 2 weighted and standardized composite health scores based on the means and SDs of scores from the 1998 general US population. Each patient has a normative score representing the mean Medical Outcomes Survey Short-Form 36 physical health score from the age-matched individuals in the general population, which represents the average for a single year of age vs published 10-year year aggregates. Cutoff scores representing the quartiles of these patients' normative scores were then used to create groupings of low (first quartile), intermediate (second and third quartiles), and high (fourth quartile) levels of physical functioning.
The Michigan Alcoholism Screening Test, a 13-item survey that screens for problems that might arise from alcohol use, was used to identify individuals who had a history of abusing alcohol.6,7 Patients marked their level of pain on a visual analog scale ranging from 0 (no pain) to 10 (worst possible pain). For statistical analysis, pain was analyzed as being either present (level 1-10) or absent (level 0). Overall QOL, reported by patients using a Likert-type scale ranging from 1 (very poor) to 5 (excellent), was analyzed as being high (level 4-5), intermediate (level 3), or low (level 1-2).
Survival analyses, performed using SPSS, version 19 (SPSS, Inc, Chicago, Illinois), included the calculation of 3-, 4-, and 5-year observed (death from all causes) and disease-specific (death with cancer present) rates for individuals who had survived 2 or more years. The rates calculated in this study reflect those at 3, 4, and 5 years from the date of diagnosis on the condition that the patient survived for 2 years. Life-table survival analyses were then performed to determine differences in conditional survival rates based on patient, disease, and health-related QOL variables assessed using the Wilcoxon statistic. Although the analysis of covariate effects on survival was based on patients' longest follow-up time, the rates for the third through fifth year after diagnosis are provided in the tables to illustrate the magnitude and trend of survival patterns. Stepwise Cox proportional hazards regression analyses were then performed to determine which of the variables that were statistically significant in the univariate analysis were included in the multivariate analyses, using the P ≤ .10 level to increase the likelihood of inclusion.
The characteristics of the 276 participants included in this study are outlined in Table 1. Their mean age was 58.1 years, and 66.7% were male. Most of these primary tumors originated in the oral cavity (30.8%), the oropharynx (27.5%), or the larynx (20.3%), and 54.0% were advanced stage. Tobacco use was common, with 59.9% of the patients being previous users and 11.4% being current users 2 years after diagnosis. Of the 237 patients for whom alcohol abuse status was known, 20.3% had a history of being a problem drinker. Two years after diagnosis, 86.0% of these survivors were eating a full diet and 80.5% reported having no pain.
Almost 28% of these 2-year survivors were in the low physical health function group, which represents the lowest 25% of scores from age- and sex-matched individuals in the general population. The 60.5% survivor rate in the high physical function group, however, is much higher than the expected 25% on the matched general population with scores in this range.
Figure 1 and Figure 2 illustrate the differences between traditional survival, representing observed and disease-specific rates for all patients, with the starting point being time of diagnosis, and conditional survival representing rates for 2-year survivors, with the starting point being time of diagnosis. The 5-year traditional observed survival rate was 61.1% for all patients compared with the conditional rate of 90.8% for 2-year survivors. The 5-year traditional disease-specific survival rate was 69.8% for all patients compared with the conditional rate of 94.8% for 2-year survivors.
Numerous variables demonstrated significant association (at the P ≤ .10 level) with observed conditional survival: stage, alcohol and tobacco abuse status at 2 years, pain, physical health, diet, and overall QOL (Table 2). Although the P value for age was greater than .10, this variable was included in the multivariate analysis because of the natural association of decreasing survival rates with increasing age, and because of the results of the post hoc pairwise comparisons that indicated a P value of .08 for the difference between the survival rates of those younger than 55 and those aged 55 to 69. In the multivariate analysis, 4 of these variables emerged as significant independent predictors of observed conditional survival: age (P = .01), stage (P = .03), pain (P = .03), and overall QOL (P = .02) (Table 2).
The variables that were significantly associated with disease-specific conditional survival included stage, alcohol abuse, tobacco use, physical health, and diet (Table 3). As in the analysis of observed conditional survival, despite having a univariate P value of .18, age was included in the multivariate analysis of disease-specific conditional survival because of the natural association of decreasing survival rates with increasing age and because of a P value of .06 in the post hoc pairwise comparison of the survival rates of those younger than 55 and those aged 55 to 69. In the multivariate analysis that included all variables with a value of P ≤ .10 in the model, the 3 independent predictors of disease-specific conditional survival were age (P = .03), stage (P = .008), and tobacco use at 2 years (P = .003) (Table 3).
Table 4 describes the results of the univariate and multivariate analyses along with hazard ratios for observed and disease-specific survival. Of particular interest were the hazard ratios for pain and overall QOL for observed conditional survival and continued tobacco use with disease-specific conditional survival. The likelihood of death was 4 times lower for patients reporting high overall QOL than for those reporting low QOL and 2 times higher for those who reported the presence of pain. Those who continued to use tobacco had a likelihood of death from cancer 4 times higher than those who had quit or had never used tobacco.
These data confirm the differences between survival expected at the time of diagnosis and that expected after 2 years. The findings, with an observed survival rate of 91% at 5 years from diagnosis, are similar to those presented by Fuller et al1 and Van der Schroeff et al.3 The demographic and clinical characteristics of the 276 patients included in this study, such as the preponderance of men and advanced-stage cases, indicated that this patient sample was representative of patients with HNC in general.
Age and stage continue to be important predictors of length of survival several years after diagnosis. Older patients would be expected to have an increased risk of dying from all causes, just as patients with advanced-stage disease would be expected to have an increased risk of experiencing a cancer-related death. These results show, however, that older patients are also at increased risk of dying because of their HNC, whereas patients with advanced-stage disease were also at increased risk of dying from causes other than their HNC. It is likely that older patients have a more difficult time recovering from the debilitating effects of their cancer and its treatment, even after having survived at least 2 years. In like fashion, patients who present with advanced disease may have comorbidities related to smoking and alcohol, or even to aggressive cancer treatment, that put them at higher risk of dying from any and all causes. These findings indicate that older age and advanced stage are risk factors for mortality 2 or more years after diagnosis.
Patients who were still smoking 2 years after the diagnosis of their HNC were 4 times more likely to die of their cancer than previous smokers or those who never smoked. Although the detrimental effects of continued use of tobacco are no surprise, this finding reiterates the need to educate patients and help them in an effort to quit smoking. The association between continued tobacco use and disease-specific mortality might push clinicians toward efforts to bring about smoking cessation.
The observed survival rates (death from all causes) were significantly lower for patients who reported higher levels of pain and who reported lower levels of overall QOL at 2 years. It is likely that pain and global QOL are associated with death from all causes because both would be expected to increase as individuals' health deteriorates due to life-threatening comorbidities. A previous study8 has shown that pain reported during the first year following diagnosis is highly correlated with the presence of a recurrence. We can hypothesize only that the lack of such a relationship between these 2 variables and disease-specific survival is due to other variables, such as stage, being stronger, and independent reflections of the increased likelihood of dying from the cancer itself.
The factors identified in this study that are predictive of observed and disease-specific survival may allow clinicians to select more or less intensive surveillance regimens on the basis of patients' risk profile. For example, young nonsmoking patients who have survived 2 years, being at lower risk of death from cancer, may benefit less from subsequent follow-up imaging. Conversely, older patients or those who report persistent pain and poor overall QOL are at higher risk of all-cause mortality, and clinicians may choose to target both cancer-related and general health-related interventions for these patients.
Studies9 have shown that patients with human papillomavirus (HPV)–positive tumors tend to have better survival rates despite their propensity to present with more advanced disease compared with their HPV-negative counterparts. Furthermore, patients with HPV-positive tumors are less likely to have tobacco and alcohol as risk factors.10 For patients in our study group, HPV status was not routinely assessed. It would be valuable to understand how HPV status affects conditional survival and the predictive factors we have identified. Additional studies are required in this regard.
In conclusion, the analysis of risk factors associated with conditional survival in this study of patients with HNC can help guide clinicians in the appropriate management of these survivors. In addition to older age and advanced stage, which are known to have a negative effect on survival, the presence of pain and continued tobacco use should flag patients who might need longer and more intense follow-up care to improve their observed and disease-specific survival rates. This informa-tion is useful for clinicians in the development of management plans for patients who are transitioning from treatment into survivorship.
Correspondence: Nitin A. Pagedar, MD, Department of Otolaryngology–Head and Neck Surgery, University of Iowa Hospitals and Clinics, 200 Hawkins Dr, Room 21261 PFP, Iowa City, IA 52242 (email@example.com).
Submitted for Publication: March 15, 2011; final revision received May 27, 2011; accepted September 4, 2011.
Author Contributions: All authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Thompson, Pagedar, Karnell, and Funk. Acquisition of data: Karnell and Funk. Analysis and interpretation of data: Thompson, Pagedar, and Karnell. Drafting of the manuscript: Thompson and Karnell. Critical revision of the manuscript for important intellectual content: Thompson, Pagedar, Karnell, and Funk. Statistical analysis: Pagedar and Karnell. Obtained funding: Funk. Administrative, technical, and material support: Karnell. Study supervision: Pagedar and Karnell.
Financial Disclosure: None reported.
Funding/Support: This work was supported by National Institutes of Health American Recovery and Reinvestment Act grant 3R01CA106908-04S1 through the Office of Cancer Survivorship.
Previous Presentation: This study was presented at the American Head and Neck Society 2011 Annual Meeting; April 28, 2011; Chicago, Illinois.
Additional Contributions: We acknowledge Amy Trullinger, who enrolled eligible patients and collected their data for this study.
Create a personal account or sign in to: