Disparities in cancer survival and incidence by metropolitan versus rural residence in Utah

Abstract Cancer disparities in rural and frontier communities are an important issue in Utah because much of Utah is sparsely populated. The aims of this study were to investigate whether there are differences in the cancer incidence and 5‐year survival rates in Utah by metropolitan/rural residence and to investigate disparities in distributions of cancer risk factors. We used cancer registry records to identify patients diagnosed with a first primary cancer in Utah between 2004 and 2008. We estimated 5‐year survival and incidence rates. The Cox proportional hazards model was used to estimate hazard ratios (HRs) for the risk of death. There were 32,498 (86.9%) patients with cancer who lived in metropolitan counties and 4906 (13.1%) patients with cancer who lived in rural counties at the time of cancer diagnosis. Patients with cancer from rural counties were more likely to be older, American Indian/Alaskan Native, non‐Hispanic, male, and diagnosed at higher stage. Rural residents had a five‐year relative survival that was 5.2% lower than metropolitan residents and a 10% increase in risk of death (HR = 1.10, 95% CI = 1.03, 1.18) after adjustment for multiple factors. Overall, the cancer incidence rates in rural counties were lower by 11.9 per 100,000 per year (449.2 in rural counties vs. 461.1 in metropolitan counties). Cancer patients living in rural counties of Utah had different demographic characteristics as well as differences in incidence and survival rates. Further studies with individual‐level data are necessary to investigate the reasons behind these differences in cancer incidence and survival to reduce disparities.

Introduction mortality rates similar to the urban areas over a 4-year period. A study of 3562 endometrial cancer patients from the Kentucky cancer registry reported that rural endometrial cancer survivors had higher proportions of uninsured, black race, and unknown stage compared to urban survivors [3]. They were less likely to have received combination therapy of surgery and radiation, and more likely to receive treatment in small hospitals. Disease-specific survival however did not differ between rural and urban endometrial cancer patients.
Utah is an ideal location to assess cancer disparities among individuals living in rural and frontier communities because the vast majority of Utah is sparsely populated or considered "frontier." Approximately 70% of Utah's 80,000 square miles have less than seven persons per square mile which is considered frontier [4]. Although previous studies have reported on differences in incidence and death rates by metropolitan and rural regions, the Utah population is unique with lower tobacco use, lower alcohol drinking prevalence, higher educational attainment, lower obesity rates, and a younger population compared to other states in the United States [5]. With decreasing tobacco prevalence in the United States, we might expect that the Utah population represents where the general US population will be in another 10 years regarding tobacco-related cancers. For colorectal cancer in Utah, the incidence was higher among nonmetropolitan women between 2006 and 2010 and nonmetropolitan women also did not experience increased survival [6]. Additionally, to our knowledge, previous studies have not incorporated tobacco, obesity, and other factors to investigate survival differences between rural vs. urban cancer patients.
Although previous studies found potential disparities in cancer survival and incidence between rural and urban residents, it is critical to further examine whether there are differences in the cancer incidence and 5-year survival rates in Utah by metropolitan/rural residence and to investigate possible factors such as stage at diagnosis, cigarette smoking, obesity, and screening behaviors for any differences in survival observed using SEER data. County-level variables from SEER have not been utilized extensively for these purposes.

Methods
We used the SEER 18 database [7] and included cancer patients with the following eligibility criteria: (1) diagnosed between 2004 and 2008 to allow for estimations of minimum 5-year survival time since follow-up was available up to 2013; (2) diagnosis in Utah as identified by the Utah cancer registry; (3) first primary cancer diagnosis; and (4) age at diagnosis 20+ years old. We used the county characteristics discussed below to group cancer patients in metropolitan and rural residence at cancer diagnosis. There were a total of 37,404 eligible cancer patients. We also assessed age at diagnosis, race, Hispanic ethnicity, sex, and stage at diagnosis (derived SEER summary stage 2000 for 2004+ diagnoses). Insurance status was available for patients with cancer diagnosed in 2007-2008 for part of the cancer patient group (n = 15,711).
The 29 counties in Utah were categorized as metropolitan or rural. The "Metropolitan Statistical Area" designation is determined by the Office of Management and Budget as "based on urbanized areas of 50,000 or more population." "A Metropolitan Statistical Area containing a single urbanized area with a population of at least 2.5 million may be subdivided to form smaller groupings of counties referred to as Metropolitan Divisions. A county qualifies as a "main county" of a Metropolitan Division if 65% or more of workers living in the county also work within the county and the ratio of the number of workers working in the county to the number of workers living in the county is at least .75. [8]" From the SEER data, metropolitan counties included (1) counties in metro areas of 1 million population or more, (2) counties in metro areas of 250,000 to 1 million population, and (3) counties in metro areas of fewer than 250,000 population. Rural counties included (1) urban population of 20,000 or more, adjacent to a metro area, (2) urban population of 20,000 or more, not adjacent to a metro area, (3) urban population of 2500 to 19,999, adjacent to a metro area, (4) urban population of 2500 to 19,999, not adjacent to a metro area, (5) completely rural or less than 2500 urban population, adjacent to a metro area, and (6) completely rural or less than 2500 urban population, not adjacent to a metro area. There were 32,498 patients with cancer who lived in metropolitan counties and 4906 patients with cancer who lived in rural counties at the time of cancer diagnosis.
For county-level variables, we selected the data available that were closest to the years of diagnosis for the cancer patient cohort (2004)(2005)(2006)(2007)(2008)

Statistical analysis
We used SAS Cary, North Carolina, US version 9.1 to analyze the data for survival. We used the Chi-square test to assess differences in the distribution of demographic, clinical, and county-level variables between metropolitan residents and rural residents. The Kaplan-Meier (KM) curve was used to compare the probability of survival between the two groups, and the log-rank test was used to compare overall survival functions in the KM curve. The Cox proportional hazards model was used to estimate hazard ratios (HRs) for the risk of death. Time to event was calculated as the time from cancer diagnosis to cancer death or end of follow-up. Five-year relative survival rates were estimated using SEER*Stat version 8.3.4. We selected cancer sites with at least 100 metropolitan cancer cases  and 50 rural cancer cases to assure sufficient sample sizes to calculate relative survival rates. We adjusted for age, sex, and race/ethnicity when assessing the potential association between rural residence at cancer diagnosis and the risk of death as those factors are risk factors for death and may impact decisions on living in a rural county (i.e., they meet the three properties of confounders [9]). Other prognostic factors such as stage at cancer diagnosis, cancer treatment, and county-level education, income, obesity, and smoking were adjusted on as sensitivity analyses to assess the impact on the hazard ratio.
For incidence analysis, we included patients diagnosed between 2009 and 2013 to investigate a recent cohort. The eligibility criteria were as follows: (1) diagnosis in Utah, (2) first primary cancer diagnosis, and (3) age at diagnosis 20+ years old. We used SEER*Stat version 8.3.4 for incidence rates that were age-adjusted, using the 2000 US standard population from the census with 19 age groups.

Results
Demographic and clinical characteristics for the patients with cancer from metropolitan vs. rural counties are shown in Table 1. Patients with cancer from rural counties were more likely to be older, American Indian/Alaskan Native, non-Hispanic, male, and diagnosed at higher stage. For cancer treatment, rural residents were more likely to have no radiation/surgery and have missing data for treatment. The proportion of cancer patients with insurance was lower among the rural residents (73.8% in rural vs. 77.4% in metropolitan).
For the cancer patients living in rural counties, the proportion completing a bachelor or college degree in the year 2000 was lower, median income was lower, proportions of families living below poverty were higher, percent obese was higher, and percent current smoking was higher (Table 2). For cancer screening, both pap smears (within 3 years) and mammography (within 2 years) were less likely in rural counties.
For all cancer sites combined, rural residents had a fiveyear relative survival that was 5.2% lower than metropolitan residents ( Table 3). The cancers for which rural residents experienced lower survival were brain cancer (6.6% lower), endometrial cancer (6.6% lower), oral cavity and pharyngeal cancer (5.7% lower), and kidney and renal pelvis cancer (5.1% lower). Slightly higher five-year relative survival rates were experienced by rural residents for ovarian cancer (2.0% higher), thyroid cancer (1.2% higher), and pancreatic cancer (0.8% higher). Survival curves for all cancers combined and the most common cancers are shown in Figure  S1 for metropolitan and rural cancer patients. Rural residence was associated with an 7% increase in risk of death ( Table 4). The increased risk of death was persistent after adjustment for most factors and all factors together (10% increase). Other prognostic factors associated with increased risks of death among the entire cohort of cancer patients included older age at diagnosis, black or unknown race/ethnicity, being male, and higher stage at diagnosis. Compared to patients with cancer who did not receive radiation or surgery, patients who had received any combination of radiation and surgery were protected against the risk of death. Increased risk of death was observed for patients who lived in counties with a higher percentage of current smokers (HR = 1.06, 95% CI = 1.02-1.11; Table 5). We observed a decreased risk of death for women in counties with higher mammography rates.
Overall, the cancer incidence rates in rural counties were lower by 11.9 per 100,000 per year (449.2 in rural counties vs. 461.1 in metropolitan counties; Table 6). Higher cancer incidence rates were observed for metropolitan counties for cancers of the pancreas (2.1 per 100,000 higher), thyroid (5.1 per 100,000 higher), female breast (6.3 per 100,000 higher), and prostate (21.1 per 100,000 higher). Higher cancer incidence rates were observed in rural counties for lung cancer (5.6 per 100,000 higher) and colorectal cancer (4.5 per 100,000 higher).

Discussion
Cancer patients living in rural counties of Utah had different distributions of demographic characteristics, lower cancer incidence rates, and lower survival rates. The greatest differences in survival between metropolitan and rural residents were observed for brain cancer, endometrial cancer, oral cavity and pharyngeal cancers, and kidney and pelvis cancer, with higher survival rates for all of these cancers among metropolitan residents. The greatest differences in incidence were observed for breast and prostate cancer, with higher rates among metropolitan residents. Overall, rural residence was associated with a 10% increase in the risk of death among cancer patients in Utah, which persisted after adjustment for various factors such as stage of cancer, treatment, obesity, education, income, and smoking.
Rural residents were diagnosed with cancer at an older age and later stage, perhaps due to differences in access to care by distance to primary care, major cancer hospitals as well as differences in insurance status. The screening rates for pap smears and mammography were also very different, with the majority of rural residents living in counties with low pap smear rates (<75%) and low mammography rates (<65%) among women. The corresponding incidence rates of screening-related cancers including breast and prostate cancers were much higher in metropolitan regions. The survival rates for breast and prostate cancer were higher in metropolitan regions. These results suggest that the lower screening rates in rural counties may contribute to lower cancer incidence for some cancers, but the later diagnosis may result in lower survival.
Differences in treatment were observed between metropolitan and rural residents, with a higher proportion in the latter cohort receiving no surgery or radiation which may be appropriate if they were diagnosed at later stage and palliation is indicated. Adjustment for treatment did not impact the increase in the risk of death observed among rural residents, although we were only able to take into account radiation and surgery and not systemic therapies. Treatment differences alone may not explain the increase in the risk of death among metropolitan residents. However, treatment differences by rural and metropolitan counties are a concern, and further areas of study could include differences in adherence to treatment guidelines and barriers to obtaining treatment for patients.
Smoking prevalence was higher in the rural counties of Utah where patients with cancer lived. Cancer incidence rates of tobacco-related cancers were generally higher in the rural counties. However, the overall cancer incidence rates were lower in rural counties in Utah, in contrast to results from the Appalachian rural region where cancer incidence rates were higher [2]. The difference in incidence rates in rural Utah vs. rural Appalachia compared to their urban counterparts may be due to the lower tobacco prevalence in Utah, although tobacco use was higher in the Utah rural counties than Utah metropolitan counties. The lower screening mentioned earlier also may contribute to the lower overall cancer incidence rates in the rural areas of Utah compared to the metropolitan areas.
Obesity-related cancers were not all higher in incidence rates among rural residents. For example, endometrial cancer (corpus and uterus) is one of the cancers most strongly associated with obesity, but there was no difference in the incidence rates between metropolitan and rural residents. In contrast, the largest difference in survival rates between rural and metropolitan regions was observed for endometrial cancer (6.6% higher in metropolitan regions). Survival rates for two other obesity-related cancers, kidney cancer, and colorectal cancer were also higher in metropolitan regions. Overall risk of death for cancer patients was not clearly associated with obesity rates in the counties; this may be because we estimated the risk of death for cancer overall and the association of obesity with cancer survival appears to differ by cancer site [10]. Higher obesity in rural counties needs to be addressed as a public health issue for cancer survivors especially given the data regarding better outcomes with weight loss.
The 10% increase in the risk of death among rural cancer patients persisted after adjustment for various factors. However, as some of the adjustment variables were at a county level and not individual level, residual confounding may be a limitation for this estimate. The estimate is also fairly modest, although it is still supportive of an increase in risk. Further studies focusing on individual cancers, with individual-level data on education, obesity, smoking, and income, would be beneficial to clarify the contributions of those factors to the suggested increase in the risk of death among cancer patients living in rural regions.
The strengths of our study include the large sample size with investigation of various cancers. The SEER data provide the advantage of investigating metropolitan vs. rural differences with a population-based design that covers the entire state of Utah. We explored obesity, income, and education as possible predictors for the differences in survival between rural vs. urban residents. We were able to investigate the unique Utah population which includes approximately 13% rural residents and has very low rates of tobacco use and obesity.
Our study also has some limitations such as ecologic bias due to the use of county-level characteristics for obesity, income, and education. However, the county-level data allowed us to explore the potential role of these factors in the risk of death among patients with cancer. The associations were largely in the direction expected showing that the county-level variables are useful proxy measures. Additionally, although the overall sample size was large, we were unable to explore rarer cancers, and for some of the cancer investigated, the sample sizes in the rural populations were limited. We also assumed that rural residence at cancer diagnosis reflects general residential history without changes, but some patients with cancer may relocate due to treatment or for other reasons from rural to urban areas. More detailed studies including residential history would be beneficial to understand the contribution of rural residence to survival differences.
In summary, our study showed important differences in survival and cancer incidence between rural and metropolitan residents in Utah that were different from other rural populations in the United States. Further studies are necessary to investigate the reasons behind these differences with individual-level data in cancer incidence and survival. The reasons for differences in treatment for rural vs. metropolitan cancer patients need to be studied in terms of treatment guideline adherence, access to care, and insurance status. The role of obesity and education in differences in survival by residence should be studied for individual cancers. Understanding the role of individual-level factors in cancer survival differences will be crucial in eliminating these disparities for the future.