Real-World Patterns of EGFR Testing and Treatment with Erlotinib for Non-Small Cell Lung Cancer in the United States

Despite being the basis of one of the most effective interventions in lung cancer, little is known about the patterns of epidermal growth factor receptor (EGFR) mutation testing in the general population. We assessed the frequencies and determinants of EGFR testing and erlotinib treatment among a population-based sample. A random sample (n = 1,358) of patients diagnosed in 2010 with histologically-confirmed NSCLC, as reported to the Surveillance Epidemiology and End Results (SEER) program, had their medical records abstracted and treating physicians queried. Logistic regression was used to identify factors associated with EGFR testing and erlotinib treatment. Survival was examined using Cox proportional hazards regression. The frequency of EGFR testing was 16.8% overall and 22.6% for stage IV adenocarcinoma patients. Given an EGFR mutation, 33.6% of all patients and 48.3% of stage IV patients received erlotinib. Among stage IV patients, increased age, Medicaid/no/unknown insurance status, death within 2 months of diagnosis and comorbidity were inversely associated with EGFR testing; erlotinib treatment was less likely among smokers and patients with non-adenocarcinomas. EGFR-mutation was associated with improved survival, albeit only among stage IV adenocarcinomas. Less than a quarter of NSCLC patients diagnosed in 2010 received EGFR testing and less than half of the patients with EGFR-mutant stage IV tumors received erlotinib. Significant disparities were observed in EGFR mutation testing by health insurance status, comorbidity and age. A national strategy is imperative to ensure that resources and processes are in place to efficiently implement molecular testing of cancer.


Introduction
Lung cancer is the leading cause of cancer-related mortality in the United States [1]. Non-small cell lung cancer (NSCLC), which constitutes over 80% of all lung cancer, is most often diagnosed at advanced stages and, as such, is largely treated with systemic therapy. Although distinct histological subtypes of NSCLC have been recognized since the 1950s, NSCLC was considered a single disease entity in terms of selection of treatment options until the mid-2000s [2].
It has since been recognized that a subgroup (10-28%) of NSCLCs have mutations in the epidermal growth factor receptor (EGFR) gene that predict sensitivity to EGFR tyrosine kinase inhibitors (TKIs), such as gefitinib, erlotinib or afatinib [3][4][5][6]. Clinical trials have consistently demonstrated response rates of over 60% with EGFR TKIs in patients with EGFR-mutant NSCLC [7][8][9][10]. Marked improvements in progression-free survival have also been demonstrated with EGFR TKIs over first-line chemotherapy, albeit with no overall survival benefit potentially due to patient crossover from chemotherapy to EGFR TKI [7][8][9][10]. Erlotinib, afatinib and gefitinib are currently approved in the United States for treatment of patients with advanced, EGFR-mutant NSCLC tumors.
The observation that tumor genotype affects treatment response has transformed the recommended care of patients with NSCLC; practice guidelines recommend molecular testing of tumor specimens to inform treatment decisions [11][12][13]. Both the National Comprehensive Cancer Network and American Society of Clinical Oncology guidelines recommended EGFR mutation testing for NSCLCs, particularly for advanced, non-squamous tumors [14][15]. However, barriers to molecular testing have been identified [16] and the patterns of EGFR mutation testing and erlotinib treatment in the general population have not previously been described.
The primary aims of this study were to estimate the frequency and determinants of EGFR mutation testing and erlotinib treatment among a population-based sample of NSCLC patients. All stages of NSCLC were studied, in order to gain broad insights into community practice. A secondary aim was to determine if EGFR mutation status and/or receipt of erlotinib were associated with survival. In order to carry out these research aims, we analyzed data from the most recent National Cancer Institute (NCI) Patterns of Care (POC) lung cancer study. The included participants were diagnosed in 2010 with histologically confirmed NSCLC and were ascertained through the Surveillance, Epidemiology and End Results (SEER) program. When these patients were diagnosed, based on the FDA labeling, erlotinib was the preferred EGFR TKI. Although it was not until 2011 that professional guidelines recommended EGFR testing, the utility of EGFR testing to predict EGFR TKI sensitivity and the benefits of EGFR TKIs in EGFR-mutant tumors were well recognized in 2010 [6,7,[17][18][19]. We believe that the findings described herein provide insights into the early dissemination of NSCLC management practices and could inform ongoing efforts to improve uptake of molecular testing in the general population.

Data source
A NCI POC study was conducted among NSCLC patients who were ascertained through the SEER program. The SEER program consists of multiple population-based registries that collect data, mainly from hospital records, on incident cancer arising within specified geographic regions across the nation. For this POC study, a random sample of eligible NSCLC SEER patients was selected after stratification by registry, sex and race/ethnicity. To increase precision among minority groups, African Americans, Hispanics, Asian Pacific Islanders (APIs) and American Indians/Native Alaskans (AI/NA) were oversampled. Using survey instruments developed for the POC study, trained abstractors at each registry reviewed medical records and treating physicians were queried. Data were collected at least one year post-diagnosis and included demographics, diagnostics, staging, tumor characteristics and treatment. For more details see S1 Methods. Each SEER registry obtained institutional review board approval prior to initiating the POC study.

Study population
Patients were eligible for the POC study if they were diagnosed in 2010 with invasive, histologically-confirmed, primary NSCLC [International Classification of Diseases for Oncology, 3rd Edition (ICD-O3): C34.X and 8000-8040, 8046-8671 and 8940-8941]. Patients were not eligible if they had a history of cancer, except non-melanoma skin cancer, were simultaneously diagnosed with more than one cancer within 60 days, were diagnosed at autopsy or via death certificate only or were younger than 20 years at diagnosis. For the current analyses, patients with neuroendocrine carcinoma, not otherwise specified (ICD-O3: 8246: n = 25) and patients with unknown tumor stage (n = 22) were also excluded.

Statistical analysis
To obtain estimates that reflected all eligible NSCLCs diagnosed within the SEER program in 2010, sample weights, defined as the inverse of the sampling proportion for each sampling stratum, were applied. To account for the sample design, all analyses were performed using SAS (version 9.3; SAS Institute Inc., Cary, NC) and SAS-callable SUDAAN (version 10.0.1; Research Triangle Institute, Research Triangle Park, NC).
The weighted percentages of patients who had EGFR testing or received erlotinib were calculated among all patients combined and stratified by tumor characteristics (stage, histology and EGFR status). Due to the infrequency of EGFR testing and administration of erlotinib among patients with non-metastatic disease (stage I-III), subsequent analyses excluded these patients. Among stage IV patients, factors associated with EGFR testing and erlotinib treatment were assessed separately. Variables that were associated with either outcome (p<0.10) based on bivariate Chi-square tests were included in a multivariate logistic model and were retained if they remained significant. The weighted percentage of stage IV patients with EGFR mutations was also calculated overall and stratified by race/ethnicity and histology. Finally, whether or not EGFR mutation status and/or receipt of erlotinib were associated with all-cause mortality was assessed by constructing a multivariate Cox proportional hazards regression model. Follow-up began on the first day of the month of cancer diagnosis; exact diagnosis date was not available. Survival was calculated through the date of death, date of last contact or December 31, 2011, whichever came first. EGFR mutations are known to be more common in adenocarcinomas; therefore, sensitivity analyses were also conducted excluding all non-adenocarcinoma patients. All tests were two sided and statistical significance was assessed using an alpha of 0.05.

Patient demographics and characteristics
This study included 1,358 NSCLC patients. The mean age was 67.7 years and 74.6% of patients were over 60 years (Table 1). 54.6% were male and 73.8% were non-Hispanic white. Most patients (82.4%) were smokers and had adenocarcinomas (50.5%). According to the American Joint Committee on Cancer 7th edition (AJCC-7), 26.5% were stage I-II, 18.2% were stage III and 55.3% were stage IV.

Frequency of EGFR testing
Overall 16.8% of NSCLC patients underwent EGFR testing (Table 2). When stratified by histology, the frequency of EGFR testing ranged from 2.7% among large cell tumors to 20.8% among

Frequency of EGFR mutations
Overall 30.4% of the patients with stage IV tumors who underwent EGFR testing were found to have an EGFR mutation (data not shown). When stratified by race/ethnicity, EGFR mutations were least common among non-Hispanic whites (21.2%) compared to Non-Hispanic blacks (42.5%), APIs (49.0%) and Hispanic patients (50.1%). EGFR mutations were also more common in adenocarcinomas (35.7%) than squamous cell tumors (9.9%) and tumors of other histology (22.4%).

Determinants of erlotinib treatment
Among patients with stage IV disease, treatment with erlotinib was associated with Hispanic and API heritages, not smoking, having an adenocarcinoma, having an EGFR mutation, living at least two months after cancer diagnosis and being treated at a larger hospital (Table 4). Multivariate analysis indicated that erlotinib treatment was significantly less likely among smokers compared to non-smokers (OR: 0.27; 95% CI: 0.12-0.59) and patients with other/not specified NSCLC histologies compared to patients with adenocarcinoma (OR: 0.14; 95% CI: 0.04-0.54) and more likely among patients with EGFR mutations (OR: 9.90; 95% CI: 3.04-32.24). Among patients with stage IV adenocarcinomas, receipt of erlotinib remained significantly lower among smokers and higher among patients with EGFR mutations. Residence in a higher median income area was also significantly associated with receipt of erlotinib. Of patients who received erlotinib, 87.0% of patients with EGFR-mutant tumors started erlotinib as their firstline therapy compared to 36.2% of patients with EGFR wild-type tumors (data not shown).

Survival
Among patients with stage IV disease, EGFR-mutant tumors and treatment with erlotinib were both associated with better survival during bivariate analyses (

Discussion
This study establishes the general population-based patterns of EGFR mutation testing and treatment with erlotinib for NSCLC in the United States in 2010. An estimated 16.8% of all newly diagnosed NSCLC patients underwent EGFR mutation testing. Among patients with stage IV tumors, EGFR testing varied significantly by age, insurance and comorbidity level. Furthermore, an estimated 33.6% of NSCLC patients with EGFR-mutant tumors received erlotinib, which was also administered to 5.9% of NSCLC patients with EGFR-wild type tumors. Among patients with stage IV tumors, erlotinib treatment was less likely among smokers and patients with non-adenocarcinomas. EGFR-mutation was associated with improved survival, albeit only among stage IV adenocarcinomas. The reason why a large proportion of patients with stage IV disease in the current study were not assessed for EGFR mutations is likely multifactorial. For example, assay costs and issues related to tissue acquisition and turnaround time may have contributed to the low testing rate [20,21]. Additionally, and maybe more importantly, professional guidelines did not recommend routine testing for EGFR mutations until 2011 [11][12][13]. However, given that the benefit of EGFR-directed therapy in selected patients was recognized well before professional societies formally recommended testing, it was still surprising that less than a quarter of the patients underwent EGFR testing in 2010.
More concerning is our finding of significant disparities in EGFR mutation testing. In addition to observing variations by health insurance status, comorbidity and older age were associated with significantly lower EGFR mutation testing rates. Although aggregate residential income level was not associated with EGFR testing rates, it cannot be ruled out that the association with insurance status may at least partially be due to confounded by unmeasured variations in patient-level variables (e.g., income and education). The observed variations in EGFR testing rates by insurance status may also reflect the fact that professional guidelines, which can impact insurance coverage policy, had yet to recommend routine EGFR testing by 2010. It is worth noting however that previous POC analyses have indicated that patients with  Medicaid or Medicare-only are often under treated [22]; thus, our results may extend these findings to the realm of molecular testing. Given the manageable toxicity profile and higher efficacy compared with chemotherapy, EGFR TKIs are recommended in patients with tumors harboring EGFR sensitizing mutations regardless of their performance status [23]. Sustained, clinically relevant improvements in quality of life have been observed in patients with EGFRmutant tumors after EGFR TKIs treatment compared to chemotherapy [7]. Our data indicates that in the general practice, comorbid conditions and limited life expectancy, both surrogates for poor performance status, were significant negative determinants of EGFR mutation testing.
Although the progression-free survival benefits associated with erlotinib have not been found to vary by age, greater toxicities have been observed among older patients [24]. Therefore, the anticipation of greater toxicities with erlotinib among older patients may explain why older age was associated with a lower likelihood of EGFR testing. Disparities were also observed for the receipt of erlotinib related to smoking status and possibly median residential income level. Although further studies are needed to confirm these findings, it is possible that clinicians are less inclined to administer erlotinib to smokers because smoking increases the metabolic clearance of erlotinib and, thereby, diminishes the effectiveness [25]. The finding that lower income patients were less likely to receive erlotinib may be a true indication of a cost barrier. However, this finding should be interpreted cautiously because individual income level was not available.
The incidence of EGFR mutation varies by race/ethnicity. Previous studies have estimated that 15-20% of white and 50-55% of API NSCLC patients have EGFR-mutant tumors [26][27][28][29]. The frequency of EGFR mutations among African American and Hispanic NSCLC patients is less clear. Possibly due to small sample sizes and heterogeneity across studies, the frequency of EGFR-mutant tumors among African American NSCLC patients has ranged between 2-20% [26][27][28][29][30]. Only one study has assessed the frequency of EGFR-mutant NSCLC in Hispanic patients (15%) [31]. In contrast to these previous studies, the frequency of EGFR-mutant tumors in this study was assessed among patients selected on the basis of clinico-pathologic characteristics (e.g., patients with adenocarcinomas and/or non-smokers were more likely to have an EGFR test). Thus, the observed frequencies in the current study demonstrate selective testing tends to enrich the frequency of EGFR-mutant tumors, particularly among non-Hispanic black and Hispanic patients. Higher than expected frequencies of EGFR-mutant tumors among non-Hispanic black and Hispanic patients, may be largely due to racial/ethnic variation in smoking and histology. For example, Hispanic patients were less likely to be smokers than NHW (71.1% vs. 85.4%, respectively, p<0.01) and Hispanic and non-Hispanic black patients were more likely to have had adenocarcinomas than NHW patients (56.7%, 57.5% and 48.5%, respectively p0.14; data not shown).
Albeit only among patients with stage IV adenocarcinomas, we found that EGFR-mutant tumors were associated with better survival. This finding is consistent with previous reports in patients with advanced NSCLC, which suggest that EGFR mutation status by itself is a favorable prognostic marker [32,33]. Erlotinib was not independently associated with survival. However, due to the high frequency of patients in whom EGFR testing was not done, the small number of patients with EGFR mutations and the observational nature of this study, which makes it prone to confounding by indication, these results should be interpreted with caution.
To our knowledge only a limited number of prior studies have addressed the question of adoption of EGFR mutation testing in general practice. Based on retrospective data from the US Oncology Network data, Pan et al. estimated that 15.2% of patients with stage IV NSCLC underwent EGFR testing and that 50.0% of patients with stage IV EGFR-mutant tumors received erlotinib [34]. Our results which are based on a larger, population-based sample are consistent with these findings.
The current study also provides data on EGFR mutation testing and erlotinib use among patients with early-stage NSCLC, a group for which current guidelines do not recommend testing or treatment [13,26]. Interestingly, 13.9% of NSCLC patients with stage I-II disease received EGFR mutation testing. Notwithstanding the relatively high frequency of EGFR mutation testing in this population, a very low number of the patients received erlotinib. The role of adjuvant erlotinib in NSCLC remains under investigation (NCT02194738).
The strengths of this study include the population-based data, oversampled minority groups, and physician verified treatment. This study had several limitations. We were not able to assess all factors that may have influenced the decisions to have EGFR testing or treatment with erlotinib. For example, tumors that were classified as non-adenocarcinoma may have had an adenocarcinoma component, which may partially explain the higher than expected frequency of EGFR mutations among the non-adenocarcinoma categories. Additionally, we were not able to assess how testing and subsequent treatment were impacted by inadequate tissue samples and variable laboratory turnaround times. Variations by specific mutation were also not assessable because this information was not recorded. Additionally, small sample size precluded the ability to identify factors associated with EGFR testing and receipt of erlotinib among patients with earlier stage tumors and non-adenocarcinomas. Small sample size may also have impacted the observed mutation rate among racial subgroups. Finally, because NSCLC has not been selected as a POC study cancer site since 2010, it was not possible to assess more recent clinical practices. It is likely that EGFR testing frequency has increased in the recent years with it being recommended by professional societies [11,12]. Despite the limitations, we were able to examine EGFR testing and erlotinib use among NSCLC patients that were representative of those seen in the general clinical practice.
In conclusion, targeted therapy in molecularly selected patients is transforming lung cancer treatment. Although the current testing rates are likely substantially higher than rates reported here, the results from the current study indicate patterns of early dissemination of EGFR mutation testing and erlotinib treatment. The complexity of testing and treatment for lung cancer patients will likely increase as additional targets and therapies are identified. A national strategy is imperative to ensure that resources and processes are in place to more widely implement molecular testing.
Supporting Information S1 Methods. Description of analysis of individual variables. (DOC)