Epidemiology of acute myeloid leukemia in Virginia: Excellent survival outcomes for patients in rural Appalachia

Abstract Background Acute myeloid leukemia, the most common acute leukemia in adults, has a poor overall survival. Studies have suggested that certain socioeconomic factors such as living in a rural or farming area are associated with worse outcomes. Since 42% of acute myeloid leukemia patients seen in our academic center reside in a rural area, we have a unique opportunity to study outcomes of patients in rural versus urban settings. Aim This analysis evaluates the effect of geography and socioeconomic factors on the biology, treatment, and overall survival of patients with acute myeloid leukemia, with the goal of understanding health care disparities. Methods and results Patient characteristics, cytogenetic data, treatment history, and overall survival were collected and analyzed to identify differences between urban and rural residency. This cohort included 42% of patients who resided in a rural area at the time of acute myeloid leukemia diagnosis. There was no difference in overall survival between the cohorts. The 1 year overall survival for the entire cohort was 47.9%. There was no difference detected in rates of adverse cytogenetics between the rural and urban cohorts. Similar numbers of patients received induction chemotherapy or proceeded to allogeneic stem cell transplant between the cohorts. Conclusions This study highlights that similar outcomes can be achieved in rural and urban patients, suggesting that intensive efforts at telehealth, education, and collaboration with local oncology practices may be beneficial.

area of residency, specifically living in a rural area as compared to an urban area, on the mutational frequency and OS in AML is unknown.
According to the 2010 census, 24.5% of Virginia's population currently lives in a rural area. 2 Our cancer center provides care to a significant number of patients residing in rural communities, including many from rural Appalachia. This provides a unique opportunity to study health care disparities in our state, and the effects of geography and socioeconomic factors on the genetics and outcomes of patients with AML.
Cytogenetic analyses and molecular genetic testing have prognostic importance in AML. The 2017 European Leukemia Net (ELN) criteria are the most widely used risk model and stratifies patients into three prognostic groups: favorable, intermediate, and adverse. 3 Unfortunately, there are many gene mutations which are not currently characterized in this model that may have a significant impact on prognosis such as NRAS/KRAS or DNMT3a mutations. [4][5][6] Recent advances using targeted therapy in patients who are not candidates for intensive induction chemotherapy and who have relapsed or refractory disease have shown response rates of 20-60%. 1,[7][8][9][10][11] Despite the progress made in cancer detection, treatment, and survival, patients from rural Appalachia continue to have higher incidence of cancer diagnosis and cancer-related mortality compared to patients from non-Appalachia. [12][13][14] Although the gap in incidence of cancer has narrowed, patients from Appalachia continue to have increased rates of tobacco-related cancers, specifically lung cancer, oral cavity cancer, pharyngeal cancer and cancer of larynx. 15 In Virginia, the cancer mortality rate in Appalachia is 10% higher than the national rate and 11% higher than non-Appalachian Virginia. 16 Our Cancer Center provides strategic services to rural residents including those from Appalachia and throughout Virginia to address disparities in cancer mortality. This rural cancer program benefits from extensive local, regional, and state-wide partnerships with regional outreach staff to support community-based participation and bi-directional communication between rural residents, communities and the academic cancer center. A representative Community Advisory Committee advises on strategic partnerships with health systems, rural care providers, churches, and community organizations to deliver culturally adapted, evidence-based programming, intervention, and screening services.
The etiology of the increased incidence of cancers in rural Appalachia is likely multifactorial and related to health care disparities and environmental factors. Certain environmental exposures, such as living in a rural or farming community, have been associated with an increased risk for the development of AML. 17 The Iowa Women's Study found that women who lived on a farm or in a rural area were twice as likely to develop AML as compared with women who lived in a city (relative risk 2.38). 18 Further analysis demonstrated that even women who did not identify as a farmer but who lived on farms or in rural communities still had an increased risk of the development of AML. 19 The effect of geographic residence on genetic characteristics of AML is poorly understood. In addition, the impact of geographic residence on molecular characteristics is not well studied. In a study evaluating a cohort of AML patients from Florida and Arizona, several lifestyle and environmental factors, such as rural or farm habitats, obesity, and smoking, were evaluated to determine their effect on cytogenetics and outcomes of AML. 17 Although only 5% (n = 14) of the cohort resided in a farm habitat, more than half of those patients had poor risk cytogenetics.
Farm residence was not found to be a statistically significant predictor of poor risk cytogenetics (odds ratio: 2.65, p value: .087). 17 Molecular characteristics were not included in this study. 17 Other patient characteristics such as race or ethnicity have also been demonstrated to impact outcomes. A retrospective analysis of 2570 patients with de novo AML found that African American patients (10.5% of the study population) presented at a younger age and were more likely to have favorable risk cytogenetics. Despite these good prognostic factors, African American men had lower complete remission (CR) rates as compared with African American women or white patients. 20 The authors of this study surmised the lower CR rates were due to biologic factors as there were no differences in other prognostic markers, induction regimens, or treatment-related toxicities. 20 Additionally, poverty has been associated with worse outcomes.
A study evaluating the impact of non-biological factors on AML outcomes demonstrated that patients who were within the lower three quintiles of median household income had worse OS. 21 In an analysis of patients with AML, acute lymphoblastic leukemia (ALL), and myelodysplastic syndrome (MDS), patients with higher levels of poverty were less likely to undergo allogeneic hematopoietic cell transplantation (HCT). 22 In this study, there was no difference in HCT based on rural or urban residency. 22 In our study, the molecular characteristics of patients with AML in rural and urban Virginia were compared. The effect of socioeconomic factors, including geographic residency, on AML biology and outcomes was also investigated.

| METHODS
We performed a single center retrospective study of AML patients diag- data. For zip codes that belong to more than one county, the characteristics of the dominant county (based on population density) were used. Rural-Urban Commuting Area codes were used to identify the urban-rural residential areas based on zip codes as previously described. [23][24][25] Geographic-level socioeconomic data were extrapolated from the individual counties based on the Area Health Resource Files (AHRF) database and used to determine percent ratio of income to poverty and percent high school or above education. The distance from the cancer center to the zip code of the residence of the patients was calculated using the SAS procedure "zipcitydistance." This program calculates the geodetic distance in miles between two zip code locations. The centroid of each zip code is used in the calculation.

| STATISTICAL ANALYSIS
Bivariate associations were calculated to determine if each variable in this study was associated with either the groups (urban or rural area of residency). Time to event data were analyzed using log-rank tests, Kaplan-Meier method, and Cox proportional hazard regression model.
The OS was estimated using Kaplan-Meier product limit by area of residency, treatment type, and transplant. Multivariable analyses based on Cox Proportional Hazards regression model were used to examine the effect of the various covariates on the risk of dying by area of residence. Statistical tests were two-sided at .05 significance level. All analyses were performed using SAS version 9.4.

| Patient characteristics
A total of 163 patients were identified who met criteria for inclusion in this study. Table 1 demonstrates the patient characteristics stratified by area of residence. Median age of the entire cohort was 65.4 years old. 58.9% (n = 96) were male. 41.7% (n = 68) of patients were from a rural area; 58.3% (n = 95) were from an urban area. 15.34% (n = 25) of patients were from a county in Appalachia; 33.33% (n = 13) were from rural Appalachia; 9.68% (n = 12) were from urban Appalachia. A map of the area of residency stratified by rural versus urban is demonstrated in Figure 1. Patients from a rural area were more likely to live below the poverty level as compared to the urban cohort (13.55 vs. 11.57%, p = .0304). Patients from a rural area were less likely to have graduated high school as compared to the urban cohort (85.57 vs. 87.79%, p = .004). There was no difference in median age, race, prior MDS or MPN, number of patients who underwent chemotherapy or number of patients who underwent HCT between the rural and urban cohorts, as demonstrated in Table 1.  Table 2). The most frequently identified mutations were DNMT3a (urban cohort: n = 19, 20%; rural cohort: n = 9, 13.2%) and NPM1 (urban cohort: n = 18, 19%; rural cohort: n = 9, 13.2%). There was similar frequency in ASXL-1, a poor prognosis mutation, in both cohorts: ASXL-1 was identified in 2% (n = 2) and 5.9% (n = 4) of the urban and rural cohorts, respectively.

| Socioeconomic characteristics
Data on the median income and high school graduation rate was available at the county level. In the rural cohort, 13.5% of patients lived in a county with a median income below the poverty level; in the urban cohort, 11.6% of patients lived in a county with a median income below the poverty level (p = .03). The rate of high school graduation from county data was lower in the rural cohort as compared to the urban cohort (85.6 vs. 87.8%, p = .0004). Rural patients traveled an average of 80.3 miles for treatment; urban patients traveled 58.05 miles (p = .029).

| DISCUSSION
To our knowledge, this is the first report comparing AML mutations and outcomes for rural versus urban patients, including 15% from counties in Appalachia. Studies of outcomes of other cancers for patients from Appalachia suggest outcomes are worse compared to non-Appalachia patients. [12][13][14]26 Our retrospective study did not demonstrate inferior outcomes for rural patients. Our study included 42% of patients from a rural area of whom 33.33% live within rural Appalachia. No differences were seen between the urban and rural cohorts, specifically in age, sex, race, and smoking status, history of prior MPN obesity, diabetes) than the rest of Virginia. 27 We did not collect information on chronic diseases other than obesity, which did not predict for survival. The rural patients treated at the cancer center could have lower rates of comorbidities as compared to that described by Huttlinger and colleagues which may be one factor to explain their better outcomes. Future studies on AML outcomes from our center should include comorbidities to evaluate whether this contributes to the similar outcomes between rural and urban cohorts.
Lower socioeconomic status also contributes to cancer care disparities. 26,29 Rural patients in our study had a lower education level and were more likely to live below the poverty level. However, it is important to recognize that socioeconomic status from our study was extrapolated from county data and does not reflect the individual patient socioeconomic status. We did not have access to individual patient education level, occupation or income level. Further studies will include these factors in addition to race to more completely assess the influence of social determinants of health. 30,31 Traveling a distance to receive care has been associated with worse outcomes. Rotz et al found living distances greater than 50 miles from a treatment center was associated with inferior outcomes in patients with acute lymphoblastic leukemia but not in patients with AML. 32  Our center has a commitment to community outreach and engagement, and to serve rural Appalachia and other rural areas in our state. Outreach and engagement activities that will have short-or long-term impact are informed by the catchment area through formal and informal interactions with regional community groups. There is extensive collaboration including community action teams to focus on cancer screening, tobacco reduction, and healthy behaviors. Our Cancer Center Without Walls represents our commitment and efforts to reach and collaborate with our catchment area, shares information, provides access to research and tertiary care. In addition, we participate in many statewide initiatives such as the Virginia Rural Health Association to improve care in rural areas.
In conclusion, this study evaluated the patient characteristics, genetics, and survival outcomes of AML patients in a single academic center treated in Virginia. The initial hypothesis was that rural patients would have worse cytogenetics, mutations, and OS; however, there were no differences observed between patients from an urban or rural area. In addition, patients from urban and rural areas, despite differences in poverty level and education, had similar rates of receiving induction chemotherapy and proceeding with HCT. The similar OS rates may be attributed to many factors including intense telehealth, frequent communication with referring physicians, and patient treatment preferences. This highlights the importance of continuing to strengthen collaborative efforts between academic oncologists and local oncologists. Future studies incorporating the collection of geographic and socioeconomic data in a prospective study are planned to provide additional insight into treatment patterns and survival outcomes for patients from rural areas.