Abstract
Background: Relative survival (RS) for patients with laryngeal cancer in the US population has yet to be described. Additionally, survival of patients with laryngeal cancer has demonstrated socioeconomic and racial disparities which have not been previously examined through the lens of RS. Materials and Methods: Data on 58,814 patients with laryngeal cancer were obtained from the National Cancer Database. Patients were diagnosed between 1998 and 2006, and had at least 5 years of follow-up. Birth-, year-, age-, sex- and race-specific matched life expectancies were used to estimate expected survival. Stage-stratified RS rates were calculated with multiple factors assessed for significance. Excess mortality ratios were estimated in multivariate analysis utilizing Poisson regression. Results: Younger age, African-American race, uninsured or Medicaid insurance, and treatment at an academic center were all significantly associated with stage IV disease. Uninsured and Medicaid patients demonstrated decreased RS when compared to privately insured individuals. Excess mortality was markedly pronounced in the first year for patients over 50 years old with stage II-IV disease, especially for the over 75-year-old cohort. Conclusion: Survival disparities for uninsured and Medicaid populations were found, with these patients exhibiting increased excess mortality. Additionally, RS calculations highlight the negative effects of increasing age on survival of patients with laryngeal cancer.
Laryngeal squamous cell carcinoma (LSCC) accounted for approximately 13,560 new cancer diagnoses and 3,640 deaths in 2015 within the US (1, 2). Due to its association with long-term tobacco and alcohol abuse, LSCC presents at a median age of 65 years and much more commonly in males, with an approximate 8:1 ratio to females (1). Additionally, these patients often have multiple comorbidities, that significantly alters the LSCC population's overall survival (3). Given these statistics, this cohort of patients would likely demonstrate a high mortality rate regardless of their stage of laryngeal cancer.
Relative survival (RS) has emerged as an alternative for estimating cause-specific or net survival in population-based studies (4). Cause-specific survival is estimated by using the cause of death due to cancer. However, the use of cause-specific survival methods can be problematic and inaccurate since information on cause of death is often unavailable. In order to overcome these obstacles, RS is an estimate of the difference or the ratio of the observed survival of a patient population to the expected survival of a comparable cohort within the general population. Therefore, RS can be used as an alternative way of estimating cause-specific survival. RS can act as a measure of the excess mortality experienced by patients with cancer, regardless of whether the excess mortality can be directly or indirectly attributed to cancer.
Previous studies have demonstrated the existence of racial and socioeconomic survival disparities for head and neck cancer, and more specifically, for laryngeal cancer (3, 5-8). These disparities seemingly influence all aspects of LSCC, including advanced-stage presentation, delay in diagnosis and treatment, higher rates of comorbidities, and difference in treatments received. Additional investigations have shown that the disparate survival outcomes for African-American and uninsured or publicly insured patients appear to be largely unrelated to actual tumor biology (3), and can be overcome with equal access to care (9).
The current research investigated the potential prognostic variables for LSCC by using relative survival from the National Cancer Database (NCDB) and the extent of racial and socioeconomic disparities for relative survival of patients with LSCC on a population level.
Stage distribution of laryngeal cancer.
Materials and Methods
Study population. The patient data set was obtained from the NCDB with the primary site of C320,C321,C322,C323, C328, and C329 based on the International Classification of Diseases for Oncology, third edition code (ICD-O-3) (10). These patients were diagnosed between 1998 and 2006 and followed-up until the end of 2011. The final analysis included only patients with stage I to stage IV disease, aged 18 to 90 years old, of White, African-American and other racial (Asian and Hispanic) ethnicity. Insurance status was classified as uninsured, private, Medicaid, Medicare (or other government insurance plan), or unknown status. Treatment facility was classified as Community Cancer Program, Comprehensive Cancer Program, Academic and Research Cancer Program, or Other Services and Clinics as defined by the NCDB. The following patients were excluded from the analysis: patients with missing survival time, unknown vital status, unknown facility identification, and non-primary cancer, carcinoma in situ, non-carcinoma or borderline status.
Expected survival and RS. Several methods are available to estimate the expected survival and RS (11-13). US calendar year of diagnosis, age at diagnosis, sex, and race specific life tables were used to provide expected survival rates for the general population (2). The calendar year of diagnosis-, age at diagnosis-, sex-, and race-matched survival rates were used to estimate the expected survival for the matched patient cohort. In the current study, the Ederer II method was used to estimate the cumulative RS (11) and to create RS data for estimation and modeling.
Smoothed hazard estimate plots are shown by stage. Mortality is highest in the first two years following diagnosis and tapers off after 5 years for all stages. As expected, mortality is highest for those with stage IV disease.
RS was calculated by dividing the observed survival rate of patients (So) in a particular period by the expected survival rate (Sp) for members of the general population matched for age, sex, and race and year of birth.
RS modeling and modeling excess mortality. The RS is modeled using the concept of excess hazards as briefly described as follows.
The hazard λo (t) at time t since diagnosis for patients diagnosed with cancer was modelled as the sum of the known baseline hazard, λp (t), and the excess hazard due to a diagnosis of cancer, λe (t). λo (t, x) = λp (t, x) + λe (t, x)
Methods developed by Dickman et al. in Stata (14) were used to perform RS analysis. Chi-square test was used to compare the distribution of stage by various factors. Poisson modeling was used to estimate the stage-specific excess mortality ratio. Statistical software version 9.4 of SAS system for windows (SAS Inc., Gary, NC, USA) and Stata Statistical Software STATA 13.1 (StataCorp LP. College Station, TX, USA) were used for data management and statistical analysis and modeling.
Results
A total of 58,814 patients met the inclusion criteria during the time period of 1998-2006 with 5-year follow-up ending in 2011. Diagnosis of stage IV laryngeal cancer was statistically significantly associated (p<0.001) with younger age, African-American race, being uninsured or having Medicaid insurance, and treatment at an academic center (Table I). In terms of age, 34% of the 18- to 49-year-old age group presented with stage IV disease at diagnosis, which incrementally decreased to 19.7% in the group aged 75 years or older. Lack of insurance or Medicaid coverage was highly associated with stage IV disease at diagnosis (42.5% and 46.9%, respectively), especially when compared to private or Medicare insurance status (24% and 24.9%, respectively).
Smoothed hazard estimate plots are shown by age group. This plot demonstrates that mortality among patients aged more than 75 years is highest among all age groups; the higher the age group, the higher the mortality.
In univariate analysis, the smoothed hazard (mortality) estimate plots demonstrate stage IV patients died at an increased rate, especially during the first two years after diagnosis (Figure 1). The plot based on age group also demonstrates that age has a large effect on mortality. Patients aged 75 years or older were found to be at increased risk for mortality as follow-up time increased and their risk of death remained higher than that for other age groups at all time points after diagnosis (Figure 2).
In Table II, for patients diagnosed in 1998 to 2002, compared to the matched, general population, the 5-year RS rate was 0.62 [95% confidence interval (CI)=0.61-0.63] and those patients treated from 2003 to 2006 had an RS of 0.89 (95% CI=0.88-0.90). Uninsured patients and Medicaid patients had a 5-year RS of 0.59 (95% CI=0.57-0.61) and 0.5 (95% CI=0.48-0.52), respectively. Medicare and private insurance patients had a 5-year RS of 0.73, and 0.79, respectively.
Multivariate generalized linear modeling was used to model the effect of factors on relative survival of patients with LSCC. As seen in Table III, stage had a significant effect on relative survival. Excess mortality ratios (eMR) were 3.63, 6.03, and 10.66 for patients with stage II, III, and IV disease compared to those with stage I. Adjusted for age, patients aged 50 years or older presented a 23% excess mortality, which varied for stage II, III and IV disease, with excess mortality of 39%, 29% and 20%, respectively, when compared to individuals younger than 50 years. Overall, privately insured patients demonstrated a reduced risk of mortality by 23% compared to uninsured patients. Additionally, at all stages of diagnosis, insurance status consistently demonstrated a significant effect on mortality rate. Privately insured patients had eMRs of 0.48, 0.75, 0.69, and 0.85 for those with stages I to IV, respectively, compared to uninsured patients. However, Medicaid patients demonstrated increased eMR when compared to uninsured patients at every stage (Table III). Treatment at an academic center compared to a community hospital was associated with a reduced risk of excess mortality by 16% for disease at all stages; in particular, there was a 28%, 20%, and 11% decrease in excess mortality for stage II, III and IV disease, respectively. In addition, higher zip code income, longer distance travelled to care facility, treatment and diagnosis in the different facilities were associated with a reduced mortality across all stages. Treatment delays were associated with increased eMR for stage I and II.
Five-year cumulative relative survival ratio by payer status and year of diagnosis in patients with laryngeal cancer.
To further illustrate the stage-specific interaction effect of age and follow-up interval, the eMR results are displayed in Table IV. The general trends suggest that excess mortality is markedly pronounced in the first year for patients age over 50 years with stage II-IV disease when compared to those younger than 50 years, especially for the cohort of those aged older than 75 years. Stage I disease demonstrates lower eMR for patients compared to the general population for the first three years of follow-up after diagnosis of laryngeal cancer for those aged 65-74 years.
Discussion
The results of the NCDB RS analysis for patients with laryngeal cancer diagnosed from 1998-2006 demonstrate that stage presentation for patients significantly varied depending on age, race and insurance status. Relative survival outcomes for patients also significantly varied depending on stage, age, and insurance status. Through RS analysis, excess mortality attributable to laryngeal cancer was estimated in a large cohort of patients to further address the potential confounding effects of age on survival of patients with laryngeal cancer.
Age demonstrated significant effects on laryngeal cancer progression in our study cohort. As patients aged, they were much less likely to present with advanced-stage (III, IV) disease. Younger patients were almost twice as likely to present with stage IV disease than patients over 75 years old. However, older patients were more likely to die during the study period; for those with stage II-IV disease, the relative survival of those aged 65-74 years was 31% more affected by laryngeal cancer when compared to the cohort aged 18-49 years (Table III). These results are similar to those seen in the German population-based study by Ramroth et al. from 2011 (15). The univariate analysis in the Ramroth et al. study demonstrated a higher mortality by 5-year overall survival for patients over 70 years old compared to peers under 60 years old. Jean et al. showed similar results utilizing the Surveillance, Epidemiology, and End Results Program Medicare database; the adjusted hazard ratio significantly increased as the population aged (16).
The RS rate for the 2003-2006 cohort demonstrated survival improved by 16% and ranged from 41% to 11% for those with stage I to IV disease, respectively, compared to patients diagnosed for the period 1998-2002.
Several studies have noted significant racial disparities in head and neck cancer, and specifically, in laryngeal cancer. Gourin et al. reported in 2006 that Black patients with head and neck cancer had a substantially lower disease-specific survival compared to White patients (29.3% vs. 54.7% at 5 years), with the cause of the discrepancy being multifactorial (17). Our results reflect this disparity with regards to the disproportionately high presentation of advanced-stage cancer in African-American patients (stage IV=39% vs. 26% for White patients). However, we found eMRs by stage to be similar between the White and African-American patients (Table III) except for stage IV in our study, possibly suggesting lower survival rates for Black patients seen in previous studies is due to increased basal mortality rates among the Black U.S. population. A similar finding was seen in the study by Zakeri et al. (3), which showed that the univariate prognostic effect of race on disease-free survival for patients with advanced head and neck cancer was not significant in multivariate analysis. However, the study demonstrated that Black patients were twice as likely to die from comorbidities during the study time period as White patients (3).
Generalized linear modeling of relative survival by stage of patients with laryngeal cancer diagnosed in the US during 1998-2006 (National Cancer Database data).
Generalized linear modeling of relative survival (age and follow-up interaction) by stage of patients with laryngeal cancer diagnosed in the US during 1998-2006 (National Cancer Database data).
The current analysis highlighted the disparities in stage at diagnosis seen with insurance status which may reflect access to care: 42.5% and 46.9% of Medicaid and uninsured patients, respectively, presented with stage IV disease. When controlling for stage, uninsured and Medicaid patients still had significantly higher excess mortality relative to those with private insurance. Several other studies have demonstrated similar disparate outcomes for head and neck cancer in the US (18, 19) and that equal access to care can possibly overcome these disparities (9, 20). In a University of Pittsburgh study, patients with Medicaid/uninsured patients had a hazard ratio of 1.50 compared to patients with private insurance (18); our findings agree with this.
There were several limitations to our study. The NCDB did not collect comorbidity information consistently until after 2003; thus, we were not able to specifically include it in our analysis. However, by utilizing RS analysis, the confounding effect of comorbidity can be partially tempered. Additionally, the NCDB does not maintain data on tobacco and alcohol abuse, which are not only well-known causal factors of laryngeal cancer, but can also negatively impact treatment outcomes in a dose-dependent manner. The potential confounding effect of these substance abuse variables remains unaccounted for in our analysis.
This study examined RS for patients with laryngeal cancer on a U.S. population level. These data demonstrate that age significantly impacts stage presentation and RS outcomes of patients with laryngeal cancer. Relative survival analysis demonstrated that race was not a significant factor in survival disparity. Finally, our RS analysis highlighted the disparities seen with payer status.
Acknowledgements
The Authors wish to acknowledge the Commission on Cancer of the American College of Surgeons and the American Cancer Society for making public data available through the National Cancer Database (NCDB). The data used in this study were derived from a de-identified NCDB file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology employed or the conclusions drawn from these data by the investigator. The Auhtors would also like to acknowledge Matthew Clavenna for his assistance of manuscript preparation.
- Received October 6, 2015.
- Revision received November 3, 2015.
- Accepted November 6, 2015.
- Copyright© 2016 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.