Untreated Head and Neck Cancer: Natural history and associated factors

[ginfreely]

 

[ginfreely]

Abstract​

Objectives:​

This study aims to provide modern estimates of survival for untreated HNC and to identify patient- and tumor-specific factors associated with not receiving definitive treatment.

Methods:​

Adults with mucosal HNC were identified from the SEER database. Median overall survival of untreated patients was calculated and stratified by site and stage. Cause-specific survival and factors associated with no treatment were investigated with advanced statistics.

Results:​

The study cohort included 6,477 patients who received no treatment. Overall, untreated patients had a median survival of 12 months, with significant variability by site and stage. Multivariable analysis identified advanced age, black race, unmarried status, and lack of private insurance as associated with no treatment.

Conclusion:​

Survival of HNC without treatment is dismal, yet substantial variation exists by tumor site and stage. Higher rates of no treatment among the poor and racial minorities are potentially driven by disparities in care availability and provision.

 

[ginfreely]

INTRODUCTION​

Head and neck cancer (HNC) is one of the ten most common cancers in the United States with roughly 53,000 new cases each year[1]. Despite advances in surgical techniques, radiation, and chemotherapy, the median five-year survival remains 60 percent[2]. Recent literature has demonstrated that definitive surgery for patients with advanced age can still be performed safely[3]. However, in patients with advanced disease, treatment is often morbid, requiring disfiguring surgeries, increased risk of dysphagia, dysphonia, and worsened quality of life. In a recent study of 1,729 patients who underwent treatment for HNC, 61% felt some degree of regret over their treatment decision due to symptoms[4]. The challenges facing practitioners counseling elderly or comorbid patients are two-fold: 1) how to properly prognosticate outcomes for patients interested in palliative non-oncologic treatment only and 2) how to identify patients likely to forgo treatment to provide additional support and resources to ensure they are being appropriately managed.

The literature on the natural history of untreated head and neck cancer to date is mostly comprised of single institution studies from more than a decade ago. Several single institution studies have been performed with median survival ranging from 11.5 weeks to 9 months but were all limited by variable patient selection, inclusion criteria, and poor generalizability[5–8]. One database study performed in 2017 identified a median survival of 4 months and found African-American race, pharynx subsites and stage II disease associated with lack of treatment[6]. However, this study only focused on relative survival, which has limited reliability in a population with higher than average comorbidities, such as HNC patients and those most likely to forgo treatment.

Due to the discrepancies in the literature and lack of recent relevant publications, the objectives of this study are to identify the patient and oncologic factors associated with not receiving definitive treatment and provide prognostic data for patients with untreated HNC based on nationally representative data.

 

[ginfreely]

METHODS​

Adults (age ≥ 18 years) with mucosal squamous cell carcinoma of the head and neck were identified from the Surveillance, Epidemiology, and End Results (SEER) 18 database. Patients were included from 2004 to 2014 to allow use of the TNM staging system throughout. Patients were excluded for other malignant diagnoses, lack of histologic confirmation, unknown survival or missing treatment data (Fig. 1).

 

[ginfreely]

In order to determine factors associated with no treatment, the primary outcome was untreated status. Patients with no treatment recorded for any available treatment variables, including surgery, radiation, chemotherapy, and immunotherapy, were classified as untreated. Surgery limited to biopsies was not considered oncologic treatment. Tumor stage, site, grade, size, treatments, and sociodemographic variables were collected. The 6th edition of the American Joint Committee on Cancer staging system was used since this was available for throughout the study timeframe. Poverty, education, and smoking rate were assessed at the county level.

The cohort of treated and untreated patients was characterized with descriptive statistics. Associations with no treatment were assessed by univariable and multivariable logistic regression models. Variables whose odds ratio (OR) excluded 1 with 99% confidence were included in the multivariable model. Additionally stage was included in all models based on its clinical importance. A planned subset analysis was performed for Stages I-IVa only. This eliminates stages IVb and IVc, which are not generally considered curable by established treatments, and therefore the decision whether to pursue treatment may differ. Effect sizes are reported as odds ratios (OR) for univariable analysis and adjusted odds ratios (aOR) for multivariable analysis with precision reported as 99% confidence intervals (CI). A sensitivity analysis was performed by including chemotherapy only in the untreated group since chemotherapy alone can prolong life but is not recommended for curative intent[9, 10].

Overall survival of untreated versus treated patients was calculated with Kaplan-Meier methods and stratified by site and stage. Median survival was reported for each site and stage combination and reported with 95% CI. To account for substantial mortality from causes other than the index cancer among HNC patients, especially those declining treatment, survival differences were further investigated using competing risk models. Cumulative mortality functions were created using two endpoints: death attributed to the index HNC cancer and death from any other cause. Cause of death was classified based on the SEER methodology which uses additional oncologic and patient data to improve classification accuracy compared to death certificates alone[11]. A multivariable Fine and Gray competing risk regression model was used to assess cancer-specific survival of untreated HNC by site with stage included as a covariate. Results were reported as sub-distribution hazard ratios (sHR) with 99% CI.

Data was extracted using SEER*Stat version 8.3.5 and exported to R version 3.5.1 for analysis using the survival and cmprsk packages.

 

[ginfreely]