Equity in adjuvant radiotherapy utilization in locally advanced head and neck cancer: A SEER-data based study

Background: Not all patients with locally advanced head and neck cancer (HNC) who are eligible for adjuvant radiotherapy (RT) following upfront surgery appear to receive it. Methods: Data were obtained from the Surveillance, Epidemiology, and End Results (SEER) database. Selected patients from 2009 to 2018 had locally advanced HNC, underwent upfront surgery, and were eligible for adjuvant RT. Multivariable logistic regression and chi-squared test were used to analyze available patient and tumor characteristics. Results: Of 12 549 patients, 84.5% underwent adjuvant RT, 15.5% did not. Characteristics associated with lowest adjuvant RT utilization included cancers of the larynx ( p < 0.0001) and gingivae ( p < 0.0001), age 80 and above ( p < 0.0001), unpartnered status ( p < 0.0001), and residence within a nonmetropolitan area ( p < 0.0024). Conclusions: Tumor subsite, age, partnered status, and rural/urban residence correlate with omission of adjuvant RT in locally advanced HNC.


| INTRODUCTION
Head and neck cancers (HNC) comprise approximately 3.7% of annual new cancer diagnoses (53 000 cases) and approximately 2.4% of annual cancer deaths (10 860 deaths) in the United States. 1 Established risk factors include tobacco and alcohol consumption, infection with oncogenic forms of human papillomavirus (HPV), immunosuppression, and various environmental exposures. 2 In cases of locally advanced HNC (i.e., larger tumors and/or locoregional lymph node involvement, without sign of distant metastasis) a multidisciplinary assessment is conducted to determine the resectability of the tumor and the patient's candidacy for surgery. If the tumor is resectable and the patient is an operative candidate, then many patients undergo upfront surgical management followed by a riskadapted approach to adjuvant radiotherapy (RT), with or without concurrent chemotherapy, based on pathological findings. [3][4][5][6] This approach is supported by guidelines published by the National Comprehensive Cancer Network. 7 Indications for adjuvant RT typically include larger tumors, deeply invasive tumors, lymph node involvement, perineural or lymphovascular tumor invasion, close or positive surgical margins, or residual gross disease. 7 When indicated, adjuvant therapy improves locoregional control and overall survival compared to surgery alone in HNC. 6,[8][9][10][11][12][13] The prognosis of HNC has improved significantly since the early 1970s, and this has been attributed in large part to increased utilization of adjuvant RT and chemotherapy. 11 Note that there are also many patients, most notably those with cancer of the oropharynx, do not undergo surgery and instead undergo primary RT with or without concurrent chemotherapy. This is not the population in question for the present study.
Among those who undergo upfront surgery, there have been observed deviations from the recommended approach to adjuvant treatment, and not all patients who would benefit from adjuvant RT receive it. Adjuvant RT utilization is decreasing in locally advanced oral cancer. 14 Approximately 4% of patients undergoing surgical treatment for HNC refuse adjuvant RT, resulting in poorer survival outcomes. 15 Previous studies have shown that, in the United States, Black patients and patients with low incomes are more likely to experience treatment delays, lengthening the interval from surgery through the completion of adjuvant RT, which may decrease the benefit of RT and negatively impact survival. 16 Black patients are also less likely to receive definitive therapy in the form of upfront surgery. [17][18][19] Uninsured and underinsured patients are more likely to present with advanced disease and less likely to receive definitive therapy. 20,21 Socioeconomic factors appear to influence patterns of care and survival outcomes for HNC, but the way in which these specifically affect utilization of adjuvant RT following definitive surgery is not fully understood. Further study is therefore required in this clinical context to identify and rectify inequities in treatment delivery and outcomes.
Our aim was to gain insight into patterns of adjuvant RT utilization for locally advanced HNC through population-based analysis. Using data from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) database, 22 the rates of adjuvant RT use were assessed in patients in whom RT was indicated based on current evidence-based recommendations. Factors that may lead to omission of RT were studied, including age, race, partnered status, household income, and rural/urban residence influence.

| Data source
This study was conducted using data from the SEER database. The SEER database is an authoritative source of population-based cancer statistics collected from participating databases nationwide, representing approximately 34.6% of the US population. Specifically, we used the November 2020 submission of the SEER 18 Research Plus Database, linked to the SEER time-dependent county attributes database. 22 This included patients diagnosed with cancer in the following geographically-based registries: Alaska Native Tumor Registry, Connecticut, Detroit, Atlanta, Greater Georgia, Rural Georgia, San Francisco-Oakland, San Jose-Monterey, Greater California, Hawaii, Iowa, Kentucky, Los Angeles, Louisiana, New Mexico, New Jersey, Seattle-Puget Sound, and Utah. The data were void of any private identifiable health records. Patients are included regardless of insurance type, though data regarding insurance are not available in the current database. The SEER statistical software, SEER*Stat version 8.3.9, was used to conduct the database queries and extract the data. A research ethics board approval was not necessary as this data is publicly available and deidentified. 23

| Data collection
Data was collected by way of a database query, in which specific inclusion and exclusion criteria were selected and, based upon these, a dataset was generated in a single step. The goal was to select patients who underwent upfront surgery, had locally advanced HNC and were, based on available data, unambiguously eligible for F I G U R E 1 Flowchart description of treatment undergone by the final patient cohort. *All subsites were included except for nasopharynx. **Laryngeal cancer patients with T3 classification were not included.
T A B L E 1 Frequency of adjuvant radiotherapy utilization based on age, tumor subsite, T stage, N stage, race/origin, partnered status, median household income, and rural-urban classification. adjuvant RT ( Figure 1). The inclusion and exclusion criteria follow. Patients were identified who were diagnosed from 2009 to 2018 (inclusive) with squamous cell carcinoma of the lip, tongue, gingivae, floor of mouth, tonsil, nontonsil oropharynx, hypopharynx, and larynx as per International Classification of Disease for Oncology, 3rd Edition (ICD-O-3). 24 Data for patients with cancer of the nasopharynx were not extracted, as the multidisciplinary treatment paradigm differs significantly from the above subsites.
Only patients for whom this was their first cancer diagnosis were included, as previous cancer treatments, particularly with RT, might influence or prohibit further treatments due to unacceptable treatment-related toxicity. Only patients who underwent cancer-directed surgery were included, and patients who underwent RT prior to surgery were not included. This therefore left only patients who underwent upfront surgery ( Figure 1).
In terms of staging classification, the database query specified T3 or T4 (for all sites except larynx, for which only T4 was included) or N2 or N3 for all sites, and excluded any patients with metastatic disease (Figure 1). Patients with only clinical staging available, as opposed to the best available post-operative pathologic staging, or a classification that could not be interpreted (i.e., TX), were excluded from the query. Patients with metastatic disease or ambiguous M classification (i.e., MX) were also excluded, as these patients would typically not undergo treatment with curative intent. For analysis, stage III  tumors were defined broadly as T3N0, T3N1, T3N2,  T0N2, T1N2, and T2N2, and stage IV tumors were  defined broadly as T4N0, T4N1, T4N2, T4N3, T0N3,  T1N3, T2N3, T3N3, T4N3. Radiation recode data (i.e., whether the patient received RT, and the RT sequence with respect to surgery) were extracted for each patient. Sociodemographic data extracted for each patient included age, race/origin, partnered status at diagnosis, median household income of county of residence, and rural-urban classification of county of residence as determined by the US Department of Agriculture.

| Statistical analysis
Statistical analyses were performed with SAS software. Patient characteristics were summarized and compared between those receiving and those not receiving adjuvant RT using chi-square tests at the 0.05 level of significance (Table 1). Multivariable logistic regression was used to estimate odds ratios for receiving adjuvant RT associated with different patient characteristics ( Table 2). Variables included in the multivariable analysis included tumor stage, tumor subsite, age, race, partnered status, median household income, and rural-urban classification of the county of residence.

| RESULTS
A total of 12 549 patients were identified by the database query, formed the final patient cohort, and underwent statistical analysis. Of these, 84.50% underwent adjuvant RT, and 15.50% did not. The most common subsites were tongue (3594 patients), tonsil (3384 patients), and oral cavity excluding floor or mouth (2168 patients). Table 1 summarizes demographic and tumor characteristics of patients and the utilization of adjuvant RT.  (Table 2). There was no difference in adjuvant RT use between stage III and stage IV tumors (Table 2).
There was a general decrease in RT uptake with age. The highest rates of adjuvant RT utilization were in those younger than 50 and those age 50-59 (89.07% and 89.05% of these patients underwent RT, respectively) ( Table 1). For those age 60-69, 85.39% of patients underwent RT, and for those 70-79, 77.73% underwent RT (Table 1). Patients age 80 and above were less likely to undergo adjuvant RT than any of the younger age groups, and these differences were significant (Table 2). Partnered patients (i.e., married or in common-law relationships) were 1.47 times more likely to undergo RT than those who were not partnered (OR 1.470, CL 1.332-1.623, p < 0.0001) ( Table 2). Patients living in nonmetropolitan areas that were not adjacent to metropolitan areas were less likely to undergo adjuvant RT than those in metropolitan areas of all sizes, and those in nonmetropolitan areas adjacent to metropolitan areas ( Table 2). This was found to be statistically significant for those in metropolitan areas adjacent to metropolitan areas (OR 1.434, CI 1.109-1.854, p = 0.0059), those in metropolitan areas of less than 250 000 (OR 1.427, CI 1.092-1.864, p = 0.0091), and those in metropolitan areas of 250 000 to 1 million (OR 1.462, CI 1.145-1.867, p < 0.0024) ( Table 2).
In terms of race/origin, all categories except for Non-Hispanic American Indian/Alaska Native were less likely to undergo adjuvant RT than Non-Hispanic White patients, with Non-Hispanic Black patients the least likely (Table 1). However, these differences were not found to be statistically significant ( Table 2). There were no statistically significant differences observed based on median household income in patients' counties of origin ( Table 2).

| DISCUSSION
All patients included in this analysis underwent definitive surgery and, based on available tumor and patient characteristics in the SEER database, were eligible for adjuvant RT. However, approximately 15% of these patients did not receive it. Our results demonstrate that age, partnered status, rural/urban residence, and tumor subsite significantly correlate with the utilization of adjuvant RT (Table 2). Differences based on race and average income of patients' counties of residence were not found to be statistically significant, though trends based on race were observed.

| Tumor factors
We did not find statistically significant difference in utilization of adjuvant RT between patients with stage III and stage IV disease ( Table 2). Our results demonstrate that the patients most likely to receive adjuvant RT are those with primary cancer of the tonsil, and the least likely are those with cancers of the gingivae and larynx. There is little published data on the variation in adjuvant RT by subsite for locally advanced HNC. Of note, there is a movement toward deintensification of adjuvant treatment for patients with HPV-associated cancers, but this primarily applies to those with cancers of the oropharynx, including the tonsil. 11 oropharyngeal cancer undergo definitive RT with or without chemotherapy, so it is possible that those who instead underwent upfront surgery, and thus were included in this investigation, had contraindications to radiotherapy that could not be captured in a populationbased study. Treatment decisions may also be individualized based on the anticipated treatment volumes and subsequent toxicities and expectations for functionality. Primary surgical management of locally advanced laryngeal cancer, for example, often requires laryngectomy, a procedure with complex postoperative recovery and longterm care needs due to changes in speech and swallowing function, including adaptation to a tracheostoma. 26,27 Adjuvant RT following laryngeal surgery is known to worsen functional outcomes and lead to lower patientreported quality of life. 28,29 It therefore stands to reason that patients recovering from laryngeal surgery and adapting to these major lifestyle changes may forego recommended adjuvant RT.

| Patient factors
Radiotherapy of the head and neck causes substantial acute and late toxicities that can have major quality of life consequences and must be weighed against the risk of recurrent disease and the substantial morbidity therein. 30,31 For this reason, decisions are based on factors such as life expectancy, performance status, comorbidities, and available support. [31][32][33] Our results show a trend toward decreased adjuvant RT in older patients, particularly in those older than 80, with patients younger than 50 having almost six times the odds of undergoing adjuvant RT as those older than 80 (OR 5.579, CL 4.689-7.167, p < 0.0001) ( Table 2). This is consistent with literature exploring the use of RT in elderly populations. Schwam et al. found that HNC patients 65 years and older were more likely to refuse RT than those younger than 65, and Amini et al. found that HNC patients 90 years and older were most likely to refuse RT. 15,34 There is a paucity of high-level evidence for the use of RT in elderly populations, as patients older than 70 are underrepresented in clinical trials. 35,36 However, studies of HNC patients have found that those older than 75 had no objective differences in treatment tolerability, interruptions, completion, or treatment-related toxicity or mortality than those younger than 75, and concluded that chronological age alone should not be a major factor in treatment decisions. 37,38 Patients in the present study population had undergone primary surgery, and it has been documented that patients age 85 and older face higher incidence of postoperative complications following surgery for HNC. 39 These patients and their physicians may therefore seek to avoid lengthy RT regimens and associated toxicities. Studies of breast, prostate, and rectal cancer treatment have investigated alternative fractionation regimens for elderly patients in effort to maintain efficacy and improve tolerability. [40][41][42] This may be a direction for future research in HNC, as highly conformal treatments, hypofractionation, and accessible supportive services may reduce toxicity and make RT a more tolerable and therefore more commonly selected treatment option for elderly patients who would benefit from it. The impact of partnered status on treatment patterns and outcomes has been well-documented for many cancer sites. 14,[43][44][45][46][47][48][49] Our results confirmed that partnered patients were more likely to undergo adjuvant RT than unpartnered patients (OR 1.470, CL 1.332, 1.623, p < 0.0001) ( Table 2). This is in concordance with previous findings that married HNC patients are more likely to receive definitive treatment in the form of surgery and/or RT. 14,43 Similar findings have been reported in breast, esophageal, and renal cancer, with the discrepancies associated with partnered status in some cases even superseding those associated with socioeconomic status and comorbidities. 44,45,49 Our findings further establish the significant role of partner support in receipt of complete definitive therapy in HNC, likely due to the particularly challenging nature of treatment. The availability of more supportive resources for unpartnered patients may help to close this gap.
For HNC, in particular, RT is often delivered daily over several weeks. For this reason, longer travel distance and time to RT centers can pose challenges, and potentially influence treatment decisions. We found that patients residing in nonmetropolitan areas not adjacent to metropolitan areas were significantly less likely than those in metropolitan or adjacent areas to receive adjuvant RT (Table 2). This is unsurprising, as geographic access has been established as a barrier to RT across various health care systems. [50][51][52][53][54][55][56][57][58] Increasing travel distance to RT center is associated with treatment refusal, inappropriate treatment regimens and worse quality of life. 15,50 This might explain why patients facing high travel burdens might proceed with upfront surgery but not adjuvant RT. Patients with breast cancer and prostate cancer have also been found to make treatment decisions that omit or minimize RT due to travel distances. 52,[54][55][56][57]59 Further research at local and regional levels should seek to understand the challenges associated with longer travel distances, and how mindful treatment decisions, support service availability, and RT center location might alleviate these barriers.
Race has been a thoroughly investigated driver of inequities in cancer treatment patterns and outcomes.
Black patients with HNC are more likely to present at advanced stages, less likely to undergo definitive treatment, and more likely to die of cancer than White patients. 17,18,60,61 Considering these well-documented disparities, it is unsurprising that our analysis identified trends indicating that non-white patients were less likely to receive adjuvant RT than white patients in most cases. However, these differences were not statistically significant. Similarly, our study found no significant differences in adjuvant RT uptake based on county-level average household income, a surrogate variable for socioeconomic status (SES) that, in many cases, also correlates with insurance status. Unfortunately, individual insurance status is currently not available in the SEER database.
It should be noted again that all patients included in our analysis underwent definitive surgery. Literature exploring the racial disparities in treatment patterns suggests that the barriers to definitive treatment for Black patients lie in barriers to surgery. 18,19 Previous investigations have shown that HNC patients of low SES and with no insurance or Medicaid were more likely to present with metastatic disease and less likely to undergo definitive treatment than those with private insurance. 62 Uninsured and Medicaid-insured patients with oral cavity cancer have been found to undergo definitive surgery at lower rates than privately insured patients, but undergo RT and chemotherapy at higher rates. 20 The lack of statistical significance of our results supports the impression that, among those who have overcome potential socioeconomic barriers and undergone definitive surgery, race and SES may not play large roles in the inclusion or omission of adjuvant RT.

| Study limitations
By nature of population-based analysis, we are not privy to all factors that play a role in clinical decisions, such as detailed pathology reports indicating margin status, perineural or lymphatic space invasion, or completeness of resection. HPV status of these tumors is an important factor, particularly for cancers of the oropharynx, which may inform treatment decisions and was not available in the SEER database. We do not have information on comorbidities or risk factors of particular patients, nor do we have information regarding an individual patient's socioeconomic status, as we only have access to countylevel income averages. The most recent SEER submission does not contain patient insurance information. We do not know the intent (i.e., palliative or curative) or dose of RT. The use of concurrent chemotherapy was outside the scope of this investigation, and this is a potential direction for future research. SEER is also limited in that it only covers a limited geographic selection of the United States. It should be noted that, if patients were to receive treatment at multiple centers (e.g., surgery at one facility and RT at a different facility), then both forms of treatment may not be reported to SEER if these facilities are in different jurisdictions or differ in their data collection and reporting capabilities. It has been suggested that the SEER database may underestimate rates of radiotherapy utilization, and may be limited in its ability to directly compare geographical variation. 63 Finally, the p-values in Table 2 come from a single multivariable logistic regression model but were not adjusted for multiple hypothesis testing.
Despite these limitations, SEER remains a powerful database to assess treatment patterns at the population level and generate hypotheses and opportunities for further investigation, which our study has accomplished by identifying several factors that significantly correlate with utilization of adjuvant RT in locally advanced HNC.

| CONCLUSIONS
In conclusion, our analysis found that approximately 15% of patients with locally advanced HNC do not undergo adjuvant RT following definitive surgery as recommended by current evidence-based guidelines. Those least likely to receive adjuvant RT are patients with cancers of certain subsites including larynx and gingivae, patients of advanced age, unpartnered patients, and patients living in nonmetropolitan areas not adjacent to metropolitan areas. Race and county-level average income do not have a statistically significant correlate with adjuvant RT uptake, though trends suggest lower rates of adjuvant RT in non-White populations. This suggests that the barriers faced by patients of racial minorities and patients of low SES are primarily barriers to receipt of definitive treatment in the first place, and therefore those who do undergo definitive surgery may face less significant barriers to adjuvant treatment based on these race and income. All treatment decisions are individualized, and vital patient-specific information is not available at the population level. Further research should assess each of these potential barriers within health systems at local and regional levels in order to develop properly contextualized solutions.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in the SEER 18 Research Plus Database, November 2020 submission, linked to the SEER timedependent county attributes database, publicly available as detailed at https://www.seer.cancer.gov. ORCID Matthew Beckett https://orcid.org/0000-0002-4711-1002