Abstract
Background:
Despite reports of socioeconomic disparities in rates of genetic testing and targeted therapy treatment for metastatic non-small cell lung cancer (NSCLC), little is known about whether such disparities are changing over time.
Methods:
We performed a retrospective analysis to identify disparities and trends in genetic testing and treatment with erlotinib. Using the Surveillance, Epidemiology, and End Results (SEER)-Medicare database, we identified 9,900 stage 4 NSCLC patients diagnosed in 2007-2011 at age 65 or older. We performed logistic regression analyses to identify patient factors associated with odds of receiving a genetic test and erlotinib treatment, and to assess trends in these differences with respect to diagnosis year.
Results:
Patients were more likely to receive genetic testing if they were under age 75 at diagnosis (odds ratio [OR]=1.55) independent of comorbidity level, and this age-based gap showed a decrease over time (OR=0.93). For untested patients, erlotinib treatment was associated with race (OR=0.58, black vs. white; OR=2.45, Asian vs. white), and was more likely among female patients (OR=1.45); for tested patients, erlotinib treatment was less likely among low-income patients (OR=0.32). Most of these associations persisted or increased in magnitude.
Conclusions:
Race and sex are associated with rates of erlotinib treatment for patients who did not receive genetic testing, and low-income status is associated with treatment rates for those who did receive testing. The racial disparity remained stable over time, while the income-based disparity grew larger.
Impact:
Attention to reducing disparities is needed as precision cancer treatments continue to be developed.
INTRODUCTION
Lung cancer, the most common cause of cancer death in the United States, is projected to account for 25% of cancer mortality in 2018 [1]. Fewer than 10% of those with stage 4 cancer survive longer than five years [2], and the majority of patients with non-small cell lung cancer (NSCLC), the most common lung cancer subtype, have stage 4 disease at time of diagnosis [3]. Since the late 1990s, the development of targeted cancer therapy has appreciably altered the landscape of lung cancer treatment by becoming a routine element of care for late-stage NSCLC [4].
Targeted therapy drugs inhibit specific molecular pathways associated with cancer growth, e.g., the pathway driven by the epidermal growth factor receptor (EGFR) tyrosine kinase. Approximately 3 in 10 NSCLC patients possess an EGFR mutation, with prevalence varying based on patient factors such as ethnicity [5, 6], and patients with certain types of EGFR mutation (i.e., exon 19 deletion or L858R mutation in exon 21) have better outcomes when treated with an EGFR tyrosine kinase inhibitor (TKI) than with standard chemotherapy [7]. Reports from as early as 2004 first indicated that EGFR mutations were associated with responsiveness to EGFR TKIs [8]. NCCN guidelines encouraged genetic testing in 2007, but it was not definitively recommended due to lack of consensus until 2011 [9, 10], at which point it was recommended for all advanced NSCLC patients considered for first-line EGFR targeted therapy regardless of patient characteristics such as age and sex [10]. EGFR TKIs are currently only one of several precision treatment options available for NSCLC. In recent years, lung cancer immunotherapies, including programmed death-1/programmed death ligand-1 (PD-1/PD-L1) inhibitors, have also shown promising results [11]. Biomarker testing more generally – including genetic testing for EGFR mutations and testing for elevated PD-L1 expression levels – is currently recommended for NSCLC patients to determine eligibility for lung cancer precision treatments [12, 13].
While precision treatments have yielded promising advancements in NSCLC treatment, utilization of some of these therapies is disproportionate across strata defined by race and socioeconomic status (SES). A reduced likelihood of receiving EGFR mutation testing is associated with factors suggesting lower socioeconomic status, including status as a Medicaid beneficiary [8] and patient residence in a relatively low-income area. The latter is also associated with lower rates of treatment with erlotinib, an EGFR TKI [14]. Hospitals located in areas with more high-income or more highly-educated residents are more likely to order EGFR testing for patients [9]. A recent study also showed that blacks and Hispanics were less likely and Asians more likely than whites to receive EGFR testing[8]. Racial disparities in cancer treatment have been well-documented [15-18]; with respect to targeted therapy treatments in particular, it has been found that black patients are less likely than white patients to receive human epidermal growth factors 2 (HER2)-targeted therapies for breast cancer [19], and black renal cancer patients show worse survival than whites even after the advent of targeted therapy based on vascular endothelial growth factor (VEGF) inhibition [20, 21]. Despite these studies on disparities in cancer treatment, little is known about whether such discrepancies are stable, growing, or shrinking over time within the context of the rapidly-evolving and relatively new field of precision cancer treatment.
We performed a retrospective study to investigate disparities in biomarker testing and precision treatment for NSCLC, and how the disparities changed over time. We analyzed genetic testing and erlotinib treatment among late-stage NSCLC patients diagnosed in 2007 to 2011 to discern patterns that may apply to current developments in targeted therapy and immunotherapy for cancer. By studying these disparities and trends, clinicians and other healthcare providers can be more aware of how different treatments are administered to different patient populations and aid in improving access to high quality care.
MATERIALS AND METHODS
Data source
We created our retrospective cohort using the Surveillance, Epidemiology, and End Results (SEER)-Medicare database (https://healthcaredelivery.cancer.gov/seermedicare/), which links records from the SEER cancer registries with Medicare claims data collected by the Center for Medicare and Medicaid Services (CMS). The National Cancer Institute’s SEER program (http://seer.cancer.gov/) collects information on cancer incidence, survival and patient demographics from the 17 SEER registries in the United States, comprising approximately 28% of the U.S. population. Medicare provides health insurance for approximately 97% of the United States population age 65 or older. Medicare coverage consists of several parts: Part A coverage is for hospital, skilled-nursing facility, hospice and home health care; Part B coverage is for physician and outpatient services; Part C refers to health maintenance organization (HMO) enrollment; and Part D coverage is for prescription drugs.
Study Population
Within the SEER-Medicare data, we identified patients diagnosed at age 65 or older with stage 4, NSCLC as their only cancer between 2007, the first year for which Part D data records were available, and 2011, the last year with diagnosis data available at the time of this study. Data was available for claims made through the end of 2013. Stage 4 lung cancer were identified as those with a Derived AJCC Stage Group (6th edition) of 70 through 74. Histology was captured through ICD-O-3 codes, with histological categories labeled according to the SEER Cancer Statistics Review [22]. Among the categories, NSCLC included those with a category of large cell, squamous cell, adenocarcinoma, or “other non-small cell type” within the other type category.
Patients were excluded if they were not enrolled continuously in Medicare Parts A, B and D from one year prior to diagnosis until death or end of the study period, or if they received Medicare benefits due to disability or end-stage renal disease. Patients were also excluded if they received benefits from an HMO during the study period, as HMOs do not submit detailed claims to Medicare.
Outcomes of interest
The two primary outcomes of interest were the receipt of a genetic test and treatment with erlotinib. In the years covered by our study, there was no CPT coding system for uniquely identifying a particular genetic test [23]. Genetic tests were identified in claims data using a set of Current Procedural Terminology (CPT) codes known as “stacking codes”, which corresponded to molecular pathology procedures performed in the genetic testing process [8, 24]. Use of stacking codes in claims data is highly accurate in identifying genetic tests in lung cancer patients [25]. We identified a patient as having received a genetic test if their records reflected a claim, occurring after the date of lung cancer diagnosis, corresponding to the following stacking codes: 83912 in the Carrier claims file; 83890-83892, 83894, 83896, 83898, 83901-83904, 83907, 83909, 83914, and 83912 in the Outpatient claims file. The code 83912 refers to the interpretation and report of a genetic test, while the other codes capture specific molecular pathology procedures typically performed in the course of such a test [24]. Treatment with erlotinib, which accounted for over 98% of lung cancer targeted therapy use in our patient cohort, was indicated for patients whose Part D event file records showed at least one prescription with a generic name of “Erlotinib HCL”.
Explanatory variables
The primary socioeconomic variables of interest were income level and residence in a high-poverty location. Eligibility for or receipt of a low-income subsidy for Part D prescription drug costs was used as a proxy for low income level. The Part D enrollment file contains data on the monthly subsidy status of each Medicare beneficiary. A patient was marked as having low income if, for at least one month in the study period, they received or were deemed eligible for the low-income subsidy (i.e., had the value of the Denominator Cost Share Group variable equal to 01 through 08). Residence in a high-poverty location was indicated for those patients whose Census Tract Poverty Indicator variable in the Patient Entitlement and Diagnosis Summary File (PEDSF), which measures the percentage of census tract residents living in poverty, had a value of at least 10% poverty.
Other explanatory variables included several demographic and clinical patient factors. Urban residence was captured through the Urban/Rural variable in the PEDSF file, which classifies a patient’s county of residence in terms of urban population size and adjacency to a metropolitan area according to a classification scheme created by the Department of Agriculture’s Economic Research Service (https://www.ers.usda.gov/data-products/rural-urban-continuum-codes/). This study designed urban locations as those coded “Big Metro”, “Metro”, or “Urban”, and less urban locations as those coded “Less Urban” or “Rural”. Other patient factors included in the analysis were sex, race (white, Asian, black, Hispanic, and other/unknown, with the last category collapsed to avoid small cell sizes), histology (adenocarcinoma vs. other non-small cell), and age at diagnosis (below 75 vs. above). Comorbidity scores were calculated using the adaptation by Deyo et al. of the Charlson comorbidity index, collapsed into categories of 0/1/2/3+ [26-29]. Trends in outcomes over time were measured with respect to diagnosis year, ranging from 2007 to 2011.
Statistical analysis
Descriptive statistics were calculated for the socioeconomic, clinical, and demographic variables. The associations between the explanatory and outcome variables were evaluated using Pearson’s chi-squared tests for categorical variables and Student’s T tests for continuous variables.
We used multivariate logistic regression models to compute odds ratios (ORs) and 95% confidence intervals (CIs) for receiving a genetic test and for using erlotinib, adjusting for demographic, clinical, and SES characteristics. We took a sequential modeling approach, showing results for models adjusting for demographic variables only and adjusting for both clinical and demographic variables. For analysis of erlotinib treatment, we divided the study cohort with regard to whether the patient had received a genetic test or not, and then applied the regression model to each of these two population subsets. Interaction terms for diagnosis year and each one of the other explanatory variables were used to examine trends over time in the adjusted prevalence of genetic testing and of erlotinib treatment. For each model, a Hosmer and Lemeshow chi-squared test was performed to confirm goodness of fit.
This study was approved by the Institutional Review Board at Massachusetts General Hospital. All statistical analyses were conducted in SAS 9.4 (SAS Institute, Cary, NC).
RESULTS
Our study population consisted of 9,900 patients, of whom 1,040 (10.5%) had a genetic test and 1,327 (13.5%) received at least one erlotinib prescription. Univariate analysis (Table 1) showed that the percentage of patients having a genetic test increased with diagnosis year from 1.6% in diagnosis year 2007 to 22.4% in diagnosis year 2011, while the percentage using erlotinib decreased from 15.4% to 11.7% over the same time period (Table 2). Low-income patients were less likely than non-low-income patients to have a genetic test (7.8% vs. 12.7%) or to receive erlotinib treatment (12.4% vs 14.1%). Patients residing in high-poverty areas were also less likely to have a genetic test (8.3% vs 13.2%) or to receive erlotinib treatment (12.2% vs 14.8%). Patients who were female, who have adenocarcinoma, or who reside in an urban area were more likely both to have a genetic test and to receive erlotinib treatment. Patients under age 75 at diagnosis were more likely to receive a genetic test and slightly less likely to be treated with erlotinib. Asians were most likely and blacks were least likely to receive either a genetic test or erlotinib. These patterns in erlotinib treatment occurred not only in the full patient cohort, but also within each subset based on testing status (Table 2).
Table 1:
Patient factors associated with genetic testing (univariate analysis).
| Had genetic test | No genetic test | ||
|---|---|---|---|
| N (%) | N (%) | P-value | |
| SES Variables | |||
| Income | <.0001 * | ||
| Low income | 319 (7.8) | 3893 (92.4) | |
| Not low income | 721 (12.7) | 4967 (87.3) | |
| High-poverty location | <.0001 * | ||
| High-poverty | 439 (8.3) | 4870 (91.7) | |
| Not high-poverty | 554 (13.2) | 3648 (86.8) | |
| Demographic and Clinical Characteristics | |||
| Sex | <.0001 * | ||
| Female | 590 (11.8) | 4432 (88.2) | |
| Male | 450 (9.2) | 4428 (90.8) | |
| Race | <.0001 * | ||
| White | 859 (10.8) | 7095 (89.2) | |
| Black | 53 (5.3) | 953 (94.7) | |
| Asian | 74 (14.8) | 425 (85.2) | |
| Hispanic | 15 (7.7) | 181 (92.3) | |
| Other or unknown | 39 (15.9) | 206 (84.1) | |
| Histology | <.0001 * | ||
| Adenocarcinoma | 787 (15.3) | 4345 (84.7) | |
| Other non-small cell | 253 (5.3) | 4515 (94.7) | |
| Age at diagnosis | 0.0038 * | ||
| Under 75 | 531 (11.5) | 4104 (88.5) | |
| 75 or older | 509 (9.7) | 4756 (90.3) | |
| Urban location | <.0001 * | ||
| Urban | 908 (11.0) | 7345 (89.0) | |
| Not urban | 86 (6.7) | 1208 (93.3) | |
| Diagnosis year | <.0001 * | ||
| 2007 | 31 (1.6) | 1900 (98.4) | |
| 2008 | 61 (3.2) | 1867 (96.8) | |
| 2009 | 182 (9.1) | 1810 (90.9) | |
| 2010 | 316 (15.5) | 1727 (84.5) | |
| 2011 | 450 (22.4) | 1556 (77.6) | |
| Charlson index | <.0001 * | ||
| 0 | 438 (14.4) | 2596 (85.6) | |
| 1 | 283 (9.6) | 2671 (90.4) | |
| 2 | 141 (8.7) | 1477 (91.3) | |
| 3+ | 165 (8.0) | 1909 (92.0) |
= Significant at the 0.05 level
Table 2:
Patient factors associated with erlotinib treatment (univariate analyses).
| Entire cohort | Had genetic test | No genetic test | ||||
|---|---|---|---|---|---|---|
| Treated with erlotinib (N (%)) |
P-value | Treated with erlotinib (N (%)) |
P-value | Treated with erlotinib (N (%)) |
P-value | |
| SES Variables | ||||||
| Income | 0.0131 * | 0.5622 | 0.1282 | |||
| Low income | 523 (12.4) | 73 (22.9) | 450 (11.6) | |||
| Not low income | 804 (14.1) | 177 (24.6) | 657 (12.6) | |||
| High-poverty location | .0002 * | 0.6300 | 0.0026 * | |||
| High-poverty | 646 (12.2) | 102 (23.2) | 544 (11.2) | |||
| Not high-poverty | 622 (14.8) | 136 (24.6) | 486 (13.3) | |||
| Demographic and Clinical Characteristics | ||||||
| Sex | <.0001 * | 0.0178 * | <.0001 * | |||
| Female | 807 (16.1) | 158 (26.8) | 649 (14.6) | |||
| Male | 520 (10.7) | 92 (20.4) | 428 (9.7) | |||
| Race | <.0001 * | 0.0004 * | <.0001 * | |||
| White | 966 (12.1) | 189 (22.0) | 777 (11.0) | |||
| Black** | 88 (8.8) | <11 (<20.8) | >77 (>8.1) | |||
| Asian | 169 (33.9) | 28 (37.8) | 141 (33.2) | |||
| Hispanic** | 33 (16.8) | <11 (<73.3) | >22 (>12.2) | |||
| Other or unknown | 71 (29.0) | 17 (43.6) | 54 (26.2) | |||
| Histology | <.0001 * | 0.1359 | <.0001 * | |||
| Adenocarcinoma | 868 (16.9) | 198 (25.2) | 670 (15.4) | |||
| Other non-small cell | 459 (9.6) | 52 (20.6) | 407 (9.0) | |||
| Age at diagnosis | 0.1351 | 0.1615 | 0.1736 | |||
| Under 75 | 596 (12.9) | 118 (22.2) | 478 (11.7) | |||
| 75 or older | 731 (13.9) | 132 (25.9) | 599 (12.6) | |||
| Urban location | <.0001 * | 0.0051 * | 0.0061 * | |||
| Urban | 1143 (13.9) | 228 (25.1) | 915 (12.5) | |||
| Not urban** | 127 (9.8) | <11 (<12.8) | >116 (>9.6) | |||
| Diagnosis year | 0.0002 * | <.0001 * | <.0001 * | |||
| 2007 | 298 (15.4) | 14 (45.2) | 284 (15.0) | |||
| 2008 | 270 (14.0) | 23 (37.7) | 247 (13.2) | |||
| 2009 | 291 (14.6) | 60 (33.0) | 231 (12.8) | |||
| 2010 | 233 (11.4) | 70 (22.5) | 163 (9.4) | |||
| 2011 | 235 (11.7) | 83 (18.4) | 152 (9.8) | |||
| Charlson index | 0.0088 * | 0.0088 * | <.0001* | |||
| 0 | 26 (15.8) | 126 (28.8) | 419 (16.1) | |||
| 1 | 32 (22.7) | 66 (23.3) | 332 (12.4) | |||
| 2 | 66 (23.3) | 32 (22.7) | 148 (10.0) | |||
| 3+ | 126 (28.8) | 26 (15.7) | 160 (8.4) | |||
= Significant at the 0.05 level
= Data masked to comply with SEER-Medicare policy regarding cells <11
After adjusting for demographic variables only, low-income status was significantly associated with a lower rate of genetic testing (OR=0.73, 95% CI 0.53-0.99) (Table 3). Residence in a high-poverty area, however, was not significant. When adjusting for clinical and demographic factors, only adenocarcinoma histology (OR=1.56, 95% CI 1.18-2.07), diagnosis age under 75 (OR=1.55, 95% CI 1.19-2.01), and Charlson index of 3 or higher (OR=0.53, 95% CI 0.29-0.97) had a significant association with odds of genetic testing. Of the differences in genetic testing rates, only the difference based on having age under 75 at diagnosis was found to be changing over time; the difference narrowed with increasing diagnosis year (OR=0.93, 95% CI 0.88-1.00).
Table 3:
Patient factors associated with genetic testing (multivariate analysis)
| Model 1 – Without clinical variables |
Model 2 – With clinical variables included |
|||
|---|---|---|---|---|
| OR (CI) | P-value | OR (CI) | P-value | |
| SES Variables | ||||
| Low income | 0.73 (0.53, 0.99) | 0.0425 * | 0.78 (0.57, 1.08) | 0.1335 |
| High-poverty location | 0.84 (0.64, 1.1) | 0.2146 | 0.82 (0.62, 1.08) | 0.1553 |
| Demographic and Clinical Characteristics | ||||
| Female | 1.13 (0.88, 1.46) | 0.3408 | 1.14 (0.88, 1.48) | 0.3275 |
| Black (vs. white) | 1.20 (0.47, 3.1) | 0.7012 | 1.16 (0.43, 3.14) | 0.7639 |
| Asian (vs. white) | 0.96 (0.35, 2.64) | 0.9422 | 0.95 (0.34, 2.62) | 0.9161 |
| Hispanic (vs. white) | 1.77 (0.36, 8.67) | 0.4795 | 1.95 (0.4, 9.43) | 0.407 |
| Other/unknown (vs. white) | 0.53 (0.14, 2.04) | 0.3534 | 0.50 (0.13, 1.96) | 0.318 |
| Urban location | 1.59 (0.96, 2.64) | 0.0706 | 1.60 (0.96, 2.66) | 0.0725 |
| Diagnosis year | 2.17 (1.82, 2.6) | <.0001 * | 2.18 (1.81, 2.63) | <.0001 * |
| Adenocarcinoma | - | - | 1.56 (1.18, 2.07) | 0.0021 * |
| Under 75 at diagnosis | - | - | 1.55 (1.19, 2.01) | 0.001 * |
| Charlson index 3+ (vs. 0) | - | - | 0.53 (0.29, 0.97) | 0.0393 * |
| Charlson index 2 (vs. 0) | - | - | 1.41 (0.82, 2.41) | 0.214 |
| Charlson index 1 (vs. 0) | - | - | 0.81 (0.51, 1.29) | 0.3857 |
| Diagnosis Year Interaction Variables | ||||
| Low income | 1.01 (0.94, 1.09) | 0.7793 | 1.01 (0.93, 1.09) | 0.8961 |
| High-poverty location | 1.00 (0.94, 1.07) | 0.8947 | 1.02 (0.95, 1.09) | 0.5733 |
| Female | 1.00 (0.95, 1.07) | 0.8785 | 0.99 (0.93, 1.05) | 0.713 |
| Black (vs. white) | 0.82 (0.65, 1.04) | 0.0989 | 0.83 (0.65, 1.07) | 0.1464 |
| Asian (vs. white) | 1.14 (0.9, 1.46) | 0.2786 | 1.14 (0.89, 1.45) | 0.3126 |
| Hispanic (vs. white) | 0.82 (0.55, 1.23) | 0.3386 | 0.81 (0.55, 1.21) | 0.2998 |
| Other/unknown (vs. white) | 1.31 (0.95, 1.82) | 0.0993 | 1.31 (0.95, 1.82) | 0.1038 |
| Urban location | 0.94 (0.83, 1.05) | 0.2852 | 0.94 (0.83, 1.05) | 0.2743 |
| Adenocarcinoma | - | - | 1.01 (0.94, 1.08) | 0.8285 |
| Under 75 at diagnosis | - | - | 0.93 (0.88, 1) | 0.0353 * |
| Charlson index 3+ (vs. 0) | - | - | 1.10 (0.96, 1.27) | 0.1648 |
| Charlson index 2 (vs. 0) | - | - | 0.90 (0.79, 1.02) | 0.0972 |
| Charlson index 1 (vs. 0) | - | - | 1.04 (0.93, 1.16) | 0.4995 |
= Significant at the 0.05 level
Among untested patients, Asian race was significantly associated with higher rate of erlotinib treatment (OR=2.45, 95% CI 1.49-4.02) and black race with a lower rate (OR=0.58, 95% CI 0.35-0.97) as compared to non-Hispanic whites (Table 4). Female sex (OR=1.45, 95% CI 1.24-1.70) was also associated with erlotinib use. The difference in erlotinib treatment rate associated with female sex changed (narrowed) significantly as diagnosis year increased. Among patients who did receive a genetic test, low income status was associated with lower likelihood of erlotinib use (OR=0.32, 95% CI 0.13-0.79 for tested patients) (Table 5), a difference that grew in magnitude with diagnosis year (OR=1.26, 95% CI 1.02-1.56). Results of multivariate analysis of erlotinib treatment based on the full study cohort (without partitioning by genetic testing status) are available in Supplementary Table 1.
Table 4:
Patient factors associated with erlotinib treatment (multivariate analysis) – Patients who did not have genetic test
| Model 1 – Without clinical variables |
Model 2 – With clinical variables included |
|||
|---|---|---|---|---|
| OR (CI) | P-value | OR (CI) | P-value | |
| SES Variables | ||||
| Low income | 0.82 (0.69, 0.97) | 0.024 * | 0.86 (0.72, 1.02) | 0.0896 |
| High-poverty location | 1.05 (0.9, 1.24) | 0.5348 | 1.06 (0.9, 1.25) | 0.5066 |
| Demographic and Clinical Characteristics | ||||
| Female | 1.50 (1.29, 1.75) | <.0001 * | 1.45 (1.24, 1.7) | <.0001 * |
| Black (vs. white) | 0.52 (0.31, 0.85) | 0.0093 * | 0.58 (0.35, 0.97) | 0.0382 * |
| Asian (vs. white) | 2.34 (1.45, 3.78) | 0.0005 * | 2.45 (1.49, 4.02) | 0.0004 * |
| Hispanic (vs. white) | 0.83 (0.36, 1.93) | 0.6679 | 0.64 (0.26, 1.58) | 0.3304 |
| Other/unknown (vs. white) | 1.80 (0.97, 3.35) | 0.0636 | 1.85 (0.97, 3.53) | 0.0602 |
| Urban location | 1.01 (0.8, 1.28) | 0.943 | 1.02 (0.81, 1.3) | 0.8552 |
| Diagnosis year | 0.90 (0.8, 1.01) | 0.0758 | 0.89 (0.79, 1.01) | 0.0735 |
| Adenocarcinoma | - | - | 1.14 (0.98, 1.33) | 0.0989 |
| Under 75 at diagnosis | - | - | 0.91 (0.78, 1.06) | 0.2072 |
| Charlson index 3+ (vs. 0) | - | - | 0.79 (0.58, 1.09) | 0.1489 |
| Charlson index 2 (vs. 0) | - | - | 0.97 (0.7, 1.34) | 0.8541 |
| Charlson index 1 (vs. 0) | - | - | 1.05 (0.82, 1.36) | 0.6808 |
| Diagnosis Year Interaction Variables | ||||
| Low income | 1.02 (0.92, 1.02) | 0.5302 | 1.03 (0.97, 1.09) | 0.4011 |
| High-poverty location | 0.97 (0.96, 1.08) | 0.2537 | 0.97 (0.92, 1.03) | 0.3238 |
| Female | 0.95 (0.91, 1) | 0.8955 | 0.95 (0.9, 1) | 0.0439 * |
| Black (vs. white) | 0.96 (0.81, 1.13) | 0.5901 | 0.94 (0.8, 1.12) | 0.4964 |
| Asian (vs. white) | 1.04 (0.9, 1.22) | 0.5723 | 1.02 (0.87, 1.19) | 0.835 |
| Hispanic (vs. white) | 1.04 (0.8, 1.35) | 0.7945 | 1.11 (0.84, 1.46) | 0.4608 |
| Other/unknown (vs. white) | 0.99 (0.8, 1.21) | 0.8963 | 0.96 (0.78, 1.19) | 0.7338 |
| Urban location | 1.01 (0.93, 1.09) | 0.8955 | 1.00 (0.92, 1.08) | 0.9808 |
| Adenocarcinoma | - | - | 1.05 (0.99, 1.1) | 0.09 |
| Under 75 at diagnosis | - | - | 1.05 (0.99, 1.1) | 0.0824 |
| Charlson index 3+ (vs. 0) | - | - | 0.99 (0.89, 1.09) | 0.7842 |
| Charlson index 2 (vs. 0) | - | - | 0.97 (0.87, 1.07) | 0.5177 |
| Charlson index 1 (vs. 0) | - | - | 1.00 (0.92, 1.08) | 0.9733 |
= Significant at the 0.05 level
Table 5:
Patient factors associated with erlotinib treatment (multivariate analysis) – Patients who had a genetic test
| Model 1 – Without clinical variables |
Model 2 – With clinical variables included |
|||
|---|---|---|---|---|
| OR (CI) | P-value | OR (CI) | P-value | |
| SES Variables | ||||
| Low income | 0.31 (0.14, 0.72) | 0.0064 * | 0.32 (0.13, 0.79) | 0.0131 * |
| High-poverty location | 1.16 (0.61, 2.19) | 0.6555 | 1.12 (0.58, 2.17) | 0.7304 |
| Demographic and Clinical Characteristics | ||||
| Female | 0.76 (0.42, 1.38) | 0.3696 | 0.68 (0.36, 1.3) | 0.2422 |
| Black (vs. white) | 1.72 (0.13, 23.64) | 0.684 | 2.14 (0.14, 32.75) | 0.5843 |
| Asian (vs. white) | 2.64 (0.27, 25.97) | 0.4048 | 2.64 (0.26, 26.73) | 0.4108 |
| Hispanic (vs. white) | 0.06 (0, 7.91) | 0.2529 | 0.03 (0, 3.77) | 0.1547 |
| Other/unknown (vs. white) | 4.04 (0.1, 156.89) | 0.454 | 6.45 (0.16, 260.1) | 0.3229 |
| Urban location | 2.13 (0.49, 9.24) | 0.3115 | 2.16 (0.48, 9.71) | 0.3159 |
| Diagnosis year | 0.93 (0.55, 1.56) | 0.7763 | 0.87 (0.5, 1.51) | 0.6181 |
| Adenocarcinoma | - | - | 1.26 (0.59, 2.66) | 0.5506 |
| Under 75 at diagnosis | - | - | 1.13 (0.6, 2.12) | 0.7003 |
| Charlson index 3+ (vs. 0) | - | - | 0.78 (0.16, 3.93) | 0.7659 |
| Charlson index 2 (vs. 0) | - | - | 3.40 (0.83, 13.89) | 0.0879 |
| Charlson index 1 (vs. 0) | - | - | 0.49 (0.15, 1.59) | 0.2347 |
| Diagnosis Year Interaction Variables | ||||
| Low income | 1.26 (1.03, 1.53) | 0.0248 * | 1.26 (1.02, 1.56) | 0.034 * |
| High-poverty location | 0.98 (0.83, 1.14) | 0.771 | 0.99 (0.84, 1.16) | 0.8966 |
| Female | 1.14 (0.98, 1.32) | 0.0816 | 1.16 (0.99, 1.35) | 0.0745 |
| Black (vs. white) | 0.72 (0.37, 1.4) | 0.3303 | 0.72 (0.36, 1.43) | 0.3462 |
| Asian (vs. white) | 0.88 (0.52, 1.51) | 0.6526 | 0.87 (0.51, 1.51) | 0.6246 |
| Hispanic (vs. white) | 2.15 (0.65, 7.07) | 0.2074 | 2.42 (0.75, 7.78) | 0.1385 |
| Other/unknown (vs. white) | 0.85 (0.36, 1.99) | 0.7086 | 0.76 (0.32, 1.79) | 0.5247 |
| Urban location | 0.90 (0.64, 1.27) | 0.552 | 0.89 (0.63, 1.27) | 0.5349 |
| Adenocarcinoma | - | - | 0.97 (0.8, 1.17) | 0.753 |
| Under 75 at diagnosis | - | - | 0.93 (0.8, 1.09) | 0.3778 |
| Charlson index 3+ (vs. 0) | - | - | 0.96 (0.65, 1.41) | 0.8201 |
| Charlson index 2 (vs. 0) | - | - | 0.74 (0.52, 1.06) | 0.0966 |
| Charlson index 1 (vs. 0) | - | - | 1.23 (0.93, 1.64) | 0.1532 |
= Significant at the 0.05 level
DISCUSSION
This retrospective study used SEER-Medicare data to examine whether disparities associated with socioeconomic, demographic, or clinical patient factors existed in genetic testing and erlotinib treatment among late-stage NSCLC patients with Medicare diagnosed in 2007 through 2011, and whether any such disparities changed over time. Our results show that low income status is associated with lower rates of genetic testing after accounting for demographic factors, and lower rates of erlotinib treatment among tested patients after adjusting for all factors. Race and sex were associated with erlotinib treatment for untested patients. Additionally, adenocarcinoma histology was associated with genetic testing, as was younger age at diagnosis. Of all discrepancies found, only the ones due to age at diagnosis in genetic testing and female sex in erlotinib treatment among untested patients were found to be decreasing over the study time period, while income-based disparities in erlotinib treatment among tested patients were increasing.
Little is known about disparity trends in genetic testing or targeted therapy usage. (Racial disparity trends in receipt of surgery or chemotherapy as cancer treatment have been investigated and shown to be relatively stable over time [16, 30].) Given that knowledge about the role of genetic testing and the proper indication for erlotinib in NSCLC was continually developing over our study time period, we sought to investigate whether increasing knowledge benefitted all patients equally.
Erlotinib was first approved by the Food and Drug Administration in 2004 for use in locally advanced or metastatic NSCLC. Although it was not until 2016 that erlotinib’s FDA approval was limited to patients who have certain EGFR mutations, there is evidence that the rate of genetic testing among lung cancer patients had been increasing prior to this [8, 23] as awareness grew of its potential benefits (as reflected by, e.g., a provisional clinical opinion from the American Society of Clinical Oncology, published in 2011, recommending a genetic test prior to first-line EGFR TKI treatment [10]). Our results showing that later diagnosis years in the 2007-2011 time period correspond to higher genetic testing rates is in line with this trend. Increasing awareness of biomarkers’ relevance to treatment choice may also explain our results showing erlotinib treatment not similarly increasing with time, given that only a minority of NSCLC patients harbor an EGFR mutation [31].
Low-income patients in our study were identified through the proxy of eligibility for a low-income subsidy for Part D prescription costs. According to CMS, in 2018, a single beneficiary may qualify for the low-income subsidy with up to $18,210 in yearly income and up to $14,000 in resources, and a beneficiary may be deemed eligible automatically due to having full Medicaid coverage, enrollment in the Medicare Savings Program, or receipt of Supplemental Security Income benefits (https://www.medicare.gov/your-medicare-costs/help-paying-costs/get-help-paying-costs.html/). We found that after adjusting for demographic variables, low-income status was associated with significantly lower rates of both genetic testing and erlotinib treatment, but the statistical significance of these associations was weakened after adjusting for three clinical factors (histology, age at diagnosis, and comorbidity index). This was particularly pronounced in the case of genetic testing: a Charlson comorbidity index of 3 or higher was the explanatory measure most negatively associated with testing. EGFR testing usually requires biopsy [32], a surgical procedure that may present risks for patients with more comorbidities, and there is a well-established relationship between lower socioeconomic status and higher comorbidity level [33, 34]. Thus, while low socioeconomic status per se may not impede access to genetic testing, the higher comorbidity rate in low-SES populations appears to create a de facto socioeconomic disparity.
We found that low patient income was also associated with lower erlotinib treatment rates, particularly for patients with known mutation status, and that this disparity was widening over time for these patients. Erlotinib, like other targeted therapies, is known to be a high-cost medication [35], and its cost increased substantially over the study period[36]. It is noteworthy that we found the low-income patients in our study paid much less out-of-pocket for erlotinib (as measured by the patient pay amount corresponding to each prescription in the Medicare Part D Prescription Drug Event file): $42 versus $1,247 per patient per month for eligible patients versus non-eligible. Therefore, it is not clear that unaffordable out-of-pocket expense serves as the mechanism for reduced erlotinib treatment in our low-income cohort. More research is needed to elucidate this mechanism. It has been suggested that providers may “implicitly ration” treatment based on patient income level [15].
Previous work has shown that a high-poverty residence is associated with lower rates of genetic testing and erlotinib treatment [14], as well as of cancer treatment generally [37, 38]. However, we found that after correcting for an individual’s income level, a significant association with community-level poverty did not remain for either outcome. This is notable given that community poverty level has been used as a proxy for individual SES in a number of studies examining relationships between SES and cancer outcomes. In our data, the association between low-income status and high-poverty residence was found to be weak (Cramer’s V = 0.2791), though highly significant (p < 0.0001).
Patients who are female, Asian, or have adenocarcinoma are known to have higher prevalence of EGFR mutation [39, 40]. Previous studies have indicated increased genetic testing rates for female and Asian patients [8, 23], whereas our study does not; this may be because we have adjusted for adenocarcinoma histology in our genetic testing analysis, while the other studies did not. However, the lack of racial or sex-based discrepancies in genetic testing is a positive sign, suggesting that perceived relative likelihood of having an EGFR mutation based on these factors is not affecting genetic testing rates.
The erlotinib treatment patterns among untested patients, on the other hand, suggest the existence of race- and sex-based disparities. While it is not known from our study what causes the association of black race with erlotinib treatment rates, one possibility is that this disparity reflects whatever underlying causes drive other disparities for black patients observed in the context of cancer treatments, including lower rates of chemotherapy and surgery for lung cancer [16, 30]. Another mechanism that may be at play, particularly with respect to the associations with female sex and Asian race, is that these characteristics are taken into account when evaluating how likely the patient is to possess an EGFR mutation, and consequently to be eligible for erlotinib treatment. In the case of tested patients, presumably the decision to treat with erlotinib would be driven more directly by the patient’s known actual EGFR mutation status. It is also possible that patients who do not receive a genetic test from their healthcare provider represent a subset who are also more likely to experience racial treatment disparities due to latent variables pertaining to quality of care for these patients.
Our finding that genetic testing rate is associated negatively with older age at diagnosis (as shown previously [8, 14, 23]) may reflect a relationship that has been demonstrated between advanced age and lung cancer treatment more broadly. Patients of advanced age are generally more likely to refuse cancer treatment, due to, for example, comorbid conditions [41]. The age disparity that remains even after correcting for comorbidity level, however, does raise the question of whether older patients have lower access to genetic testing not because they are sicker, but simply because they are older. Comorbidity-independent age-based disparity has already been documented in the more general context of active treatment for NSCLC [42-44], and our result implies that this disparity may also exist in targeted therapy treatment specifically. Further investigation is warranted to determine whether this disparity could be due to patient preferences or to inappropriate undertreatment. Nevertheless, the decreasing trend in the genetic testing disparity based on age may indicate that, over time, more older patients are at least considering the possibility of targeted therapy treatment.
Our study demonstrates that even as precision therapy has yielded increasingly effective treatments for NSCLC patients, some disparities in treatment rates have persisted. During the study period, the field of targeted therapy was rapidly evolving, and this remains true for precision therapy in NSCLC even in the present day. For example, it has been established in recent years that a particular EGFR mutation, known as T790M, is associated with acquired resistance to standard EGFR TKIs including erlotinib [45], and subsequently a newer EGFR TKI, osimertinib, has been shown to be more effective than erlotinib in patients with T790M [46]. This discovery has raised new questions about optimal targeted therapy regimens. Cancer immunotherapy is also a very active field of research. PD-L1 expression testing, in particular, has much in common with EGFR testing in our study: it is currently recommended but not required for NSCLC patients [12]; PD-L1 expression predicts outcomes for immunotherapy patients, although only a minority of NSCLCs are associated with high PD-L1 expression [47]; and patient factors are associated with high expression levels [47]. As precision treatments continue to be refined, healthcare providers must be aware that disparities can persist even as the field advances.
This study has several limitations. Given that this is a retrospective study, it was not possible to fully control for potential confounding variables. There are elements of socioeconomic status beyond income and residence – education level, for example – that we were not able to include in our analysis. Furthermore, due to the nature of genetic test coding during the study period, it was not possible to distinguish an EGFR test from a genetic test more generally, and the results of patients’ genetic tests were not available. In addition, the SEER-Medicare database, which represents patients from 17 registries, may contain geographic bias and not be representative of the United States as a whole. Another limitation common to SEER-Medicare analyses is that the restriction to patients continuously enrolled over the study period and patients without HMOs, done to ensure a complete longitudinal analysis, may have excluded patients who were disenrolled from Medicare (for example, due to non-payment of premiums) and thus had an effect on the socio-demographic makeup of the study cohort.
In conclusion, our results demonstrate that discrepancies, including socioeconomic and racial gaps, have persisted in the complex and continually evolving field of precision medicine for lung cancer, and that renewed attention to narrowing disparities is needed so that all patients may benefit equally from paradigm-changing advances in cancer treatment.
Supplementary Material
Acknowledgments
L.L.P., D.F.S., A.C.T., and C.Y.K. were supported by funding from the National Cancer Institute (https://www.cancer.gov/) in the form of grants #U01CA199284 and #R01CA202956. The statements contained herein are solely those of the authors and do not represent or imply concurrence or endorsement by NCI.
Footnotes
The authors declare no potential conflicts of interest.
REFERENCES
- 1.Siegel RL, Miller KD, and Jemal A, Cancer statistics, 2018. CA Cancer J Clin, 2018. 68(1): p. 7–30. [DOI] [PubMed] [Google Scholar]
- 2.Goldstraw P, Chansky K, Crowley J, Rami-Porta R, Asamura H, Eberhardt WE, et al. , The IASLC Lung Cancer Staging Project: Proposals for Revision of the TNM Stage Groupings in the Forthcoming (Eighth) Edition of the TNM Classification for Lung Cancer. J Thorac Oncol, 2016. 11(1): p. 39–51. [DOI] [PubMed] [Google Scholar]
- 3.Crino L, Weder W, van Meerbeeck J, Felip E, and E.G.W. Group, Early stage and locally advanced (non-metastatic) non-small-cell lung cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol, 2010. 21 Suppl 5: p. v103–15. [DOI] [PubMed] [Google Scholar]
- 4.Li T, Kung HJ, Mack PC, and Gandara DR, Genotyping and genomic profiling of non-small-cell lung cancer: implications for current and future therapies. J Clin Oncol, 2013. 31(8): p. 1039–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Zhang YL, Yuan JQ, Wang KF, Fu XH, Han XR, Threapleton D, et al. , The prevalence of EGFR mutation in patients with non-small cell lung cancer: a systematic review and meta-analysis. Oncotarget, 2016. 7(48): p. 78985–78993. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Midha A, Dearden S, and McCormack R, EGFR mutation incidence in non-small-cell lung cancer of adenocarcinoma histology: a systematic review and global map by ethnicity (mutMapII). Am J Cancer Res, 2015. 5(9): p. 2892–911. [PMC free article] [PubMed] [Google Scholar]
- 7.Rosell R, Carcereny E, Gervais R, Vergnenegre A, Massuti B, Felip E, et al. , Erlotinib versus standard chemotherapy as first-line treatment for European patients with advanced EGFR mutation-positive non-small-cell lung cancer (EURTAC): a multicentre, open-label, randomised phase 3 trial. The Lancet Oncology, 2012. 13(3): p. 239–246. [DOI] [PubMed] [Google Scholar]
- 8.Lynch JA, Berse B, Rabb M, Mosquin P, Chew R, West SL, et al. , Underutilization and disparities in access to EGFR testing among Medicare patients with lung cancer from 2010 - 2013. BMC Cancer, 2018. 18(1): p. 306. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Lynch JA, Khoury MJ, Borzecki A, Cromwell J, Hayman LL, Ponte PR, et al. , Utilization of epidermal growth factor receptor (EGFR) testing in the United States: a case study of T3 translational research. Genet Med, 2013. 15(8): p. 630–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Keedy VL, Temin S, Somerfield MR, Beasley MB, Johnson DH, McShane LM, et al. , American Society of Clinical Oncology provisional clinical opinion: epidermal growth factor receptor (EGFR) Mutation testing for patients with advanced non-small-cell lung cancer considering first-line EGFR tyrosine kinase inhibitor therapy. J Clin Oncol, 2011. 29(15): p. 2121–7. [DOI] [PubMed] [Google Scholar]
- 11.Yu H, Boyle TA, Zhou C, Rimm DL, and Hirsch FR, PD-L1 Expression in Lung Cancer. J Thorac Oncol, 2016. 11(7): p. 964–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Riely GL, What, When, and How of Biomarker Testing in Non-Small Cell Lung Cancer. J Natl Compr Canc Netw, 2017. 15(5S): p. 686–688. [DOI] [PubMed] [Google Scholar]
- 13.Ettinger S, W.D., Aisner DL, et al. , Non–Small Cell Lung Cancer, Version 5.2017. J Natl Compr Canc Netw 2017(15): p. 504–535. [DOI] [PubMed] [Google Scholar]
- 14.Enewold L and Thomas A, Real-World Patterns of EGFR Testing and Treatment with Erlotinib for Non-Small Cell Lung Cancer in the United States. PLoS One, 2016. 11(6): p. e0156728. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Woods LM, Rachet B, and Coleman MP, Origins of socio-economic inequalities in cancer survival: a review. Ann Oncol, 2006. 17(1): p. 5–19. [DOI] [PubMed] [Google Scholar]
- 16.Gross CP, Smith BD, Wolf E, and Andersen M, Racial disparities in cancer therapy: did the gap narrow between 1992 and 2002? Cancer, 2008. 112(4): p. 900–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Singh GK, Williams SD, Siahpush M, and Mulhollen A, Socioeconomic, Rural-Urban, and Racial Inequalities in US Cancer Mortality: Part I-All Cancers and Lung Cancer and Part II-Colorectal, Prostate, Breast, and Cervical Cancers. J Cancer Epidemiol, 2011. 2011: p. 107497. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Gross C, Wong N, Dubin JA, Mayne ST, Krumholz HM., Enrollment of Older Persons in Cancer Trials After the Medicare Reimbursement Policy Change.. Arch Intern Med, 2005. 165(13): p. 1514–1520. [DOI] [PubMed] [Google Scholar]
- 19.Reeder-Hayes K, Peacock Hinton S, Meng K, Carey LA, and Dusetzina SB, Disparities in Use of Human Epidermal Growth Hormone Receptor 2-Targeted Therapy for Early-Stage Breast Cancer. J Clin Oncol, 2016. 34(17): p. 2003–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Vaishampayan U, Vankayala H, Vigneau FD, Quarshie W, Dickow B, Chalasani S, et al. , The effect of targeted therapy on overall survival in advanced renal cancer: a study of the national surveillance epidemiology and end results registry database. Clin Genitourin Cancer, 2014. 12(2): p. 124–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Rose TL, Deal AM, Krishnan B, Nielsen ME, Smith AB, Kim WY, et al. , Racial disparities in survival among patients with advanced renal cell carcinoma in the targeted therapy era. Cancer, 2016. 122(19): p. 2988–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.SEER Cancer Statistics Review, 1975-2015, H.N. Noone AM, Krapcho M, Miller D, Brest A, Yu M, Ruhl J, Tatalovich Z, Mariotto A, Lewis DR, Chen HS, Feuer EJ, Cronin KA Editor. 2018, National Cancer Institute: Bethesda, MD. [Google Scholar]
- 23.Shen C, Kehl KL, Zhao B, Simon GR, Zhou S, and Giordano SH, Utilization Patterns and Trends in Epidermal Growth Factor Receptor (EGFR) Mutation Testing Among Patients With Newly Diagnosed Metastatic Lung Cancer. Clin Lung Cancer, 2017. 18(4): p. e233–e241. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Romanus D, The Value of Targeted Therapies in Lung Cancer. 2014, Harvard University. [Google Scholar]
- 25.Vachani A, et al. , Validation of Molecular Pathology Codes for the Identification of Mutational Testing in Lung and Colon Cancer. Med Care, 2017. 55: p. e131–e136. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987; 40: 373–383. [DOI] [PubMed] [Google Scholar]
- 27.Deyo RA, C.D., Ciol M, Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases.. J Clin Epidemiol., 1992(45): p. 613–619. [DOI] [PubMed] [Google Scholar]
- 28.Klabunde CN, P.A., Legler JM, Warren JL., Development of a comorbidity index using physician claims data.. J Clin Epidemiol., 2000(53): p. 1258–67. [DOI] [PubMed] [Google Scholar]
- 29.Romano PS, R.L., Jollis JG., Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives.. J Clin Epidemiol., 1993(46): p. 1075–79. [DOI] [PubMed] [Google Scholar]
- 30.Hardy D, Liu CC, Xia R, Cormier JN, Chan W, White A, et al. , Racial disparities and treatment trends in a large cohort of elderly black and white patients with nonsmall cell lung cancer. Cancer, 2009. 115(10): p. 2199–211. [DOI] [PubMed] [Google Scholar]
- 31.Rosell R, Moran T, Queralt C, Porta R, Cardenal F, Camps C, et al. , Screening for epidermal growth factor receptor mutations in lung cancer. N Engl J Med, 2009. 361(10): p. 958–67. [DOI] [PubMed] [Google Scholar]
- 32.Goldman JW, Noor ZS, Remon J, Besse B, and Rosenfeld N, Are liquid biopsies a surrogate for tissue EGFR testing? Annals of Oncology, 2018. 29(suppl_1): p. i38–i46. [DOI] [PubMed] [Google Scholar]
- 33.Louwman WJ, Aarts MJ, Houterman S, van Lenthe FJ, Coebergh JW, and Janssen-Heijnen ML, A 50% higher prevalence of life-shortening chronic conditions among cancer patients with low socioeconomic status. Br J Cancer, 2010. 103(11): p. 1742–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Schrijvers C, et al. , Socioeconomic Status and Comorbidity among Newly Diagnosed Cancer Patients. Cancer, 1997. 80: p. 1482–8. [PubMed] [Google Scholar]
- 35.Faden RR, Chalkidou K, Appleby J, Waters HR, and Leider JP, Expensive cancer drugs: a comparison between the United States and the United Kingdom. Milbank Q, 2009. 87(4): p. 789–819. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Shih Y, et al. , Rising Prices of Targeted Oral Anticancer Medications and Associated Financial Burden on Medicare Beneficiaries. J Clin Oncol, 2017. 35: p. 2482–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Johnson AM, Hines RB, Johnson JA 3rd, and Bayakly AR, Treatment and survival disparities in lung cancer: the effect of social environment and place of residence. Lung Cancer, 2014. 83(3): p. 401–7. [DOI] [PubMed] [Google Scholar]
- 38.Ou SH, Zell JA, Ziogas A, and Anton-Culver H, Low socioeconomic status is a poor prognostic factor for survival in stage I nonsmall cell lung cancer and is independent of surgical treatment, race, and marital status. Cancer, 2008. 112(9): p. 2011–20. [DOI] [PubMed] [Google Scholar]
- 39.Shigematsu H, Lin L, Takahashi T, Nomura M, Suzuki M, Wistuba II, et al. , Clinical and biological features associated with epidermal growth factor receptor gene mutations in lung cancers. J Natl Cancer Inst, 2005. 97(5): p. 339–46. [DOI] [PubMed] [Google Scholar]
- 40.Gazdar AF, Tyrosine kinase inhibitors and epidermal growth factor receptor (EGFR) mutations in non-small cell lung cancer: to test or not to test? Medicine (Baltimore), 2011. 90(3): p. 168–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Puts MT, Tapscott B, Fitch M, Howell D, Monette J, Wan-Chow-Wah D, et al. , A systematic review of factors influencing older adults’ decision to accept or decline cancer treatment. Cancer Treat Rev, 2015. 41(2): p. 197–215. [DOI] [PubMed] [Google Scholar]
- 42.Wang S, Wong ML, Hamilton N, Davoren JB, Jahan TM, and Walter LC, Impact of age and comorbidity on non-small-cell lung cancer treatment in older veterans. J Clin Oncol, 2012. 30(13): p. 1447–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Wong ML, McMurry TL, Stukenborg GJ, Francescatti AB, Amato-Martz C, Schumacher JR, et al. , Impact of age and comorbidity on treatment of non-small cell lung cancer recurrence following complete resection: A nationally representative cohort study. Lung Cancer, 2016. 102: p. 108–117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Brown J, et al. , Age and the treatment of lung cancer. Thorax, 1996. 51: p. 564–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Oxnard GR, Arcila ME, Sima CS, Riely GJ, Chmielecki J, Kris MG, et al. , Acquired resistance to EGFR tyrosine kinase inhibitors in EGFR-mutant lung cancer: distinct natural history of patients with tumors harboring the T790M mutation. Clin Cancer Res, 2011. 17(6): p. 1616–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Soria JC, Ohe Y, Vansteenkiste J, Reungwetwattana T, Chewaskulyong B, Lee KH, et al. , Osimertinib in Untreated EGFR-Mutated Advanced Non-Small-Cell Lung Cancer. N Engl J Med, 2018. 378(2): p. 113–125. [DOI] [PubMed] [Google Scholar]
- 47.Petrelli F, Maltese M, Tomasello G, Conti B, Borgonovo K, Cabiddu M, et al. , Clinical and Molecular Predictors of PD-L1 Expression in Non-Small-Cell Lung Cancer: Systematic Review and Meta-analysis. Clin Lung Cancer, 2018. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.