Comments
Abstract
Rationale: There is increased lung cancer mortality in rural areas of the United States. However, it remains unclear to what extent rural–urban differences in disease incidence, stage at diagnosis, or treatment explain this finding.
Objectives: To explore the relationship between smoking rates, lung cancer incidence, and lung cancer mortality in populations across the rural–urban continuum and to determine whether survival is decreased in rural patients diagnosed with lung cancer and whether this is associated with rural–urban differences in stage at diagnosis or the treatment received.
Methods: We conducted a retrospective cohort study of 348,002 patients diagnosed with lung cancer between 2000 and 2006. Data from metropolitan, urban, suburban, and rural areas in the United States were obtained from the Surveillance, Epidemiology, and End Results program database. County-level population estimates for 2003 were obtained from the U.S. Census Bureau, and corresponding estimates of smoking prevalence were obtained from published literature. The exposure was rurality, defined by the rural–urban continuum code area linked to each cohort participant by county of residence. Outcomes included lung cancer incidence, mortality, diagnostic stage, and treatment received.
Measurements and Main Results: Lung cancer mortality increased with rurality in a dose-dependent fashion across the rural–urban continuum. The most rural areas had almost twice the smoking prevalence and lung cancer incidence of the largest metropolitan areas. Rural patients diagnosed with stage I non–small cell lung cancer underwent fewer surgeries (69% vs. 75%; P < 0.001) and had significantly reduced median survival (40 vs. 52 mo; P = 0.0006) compared with the most urban patients. Stage at diagnosis was similar across the rural–urban continuum, as was median survival for patients with stages II–IV lung cancer.
Conclusions: Higher rural smoking rates drive increased disease incidence and per capita lung cancer mortality in rural areas of the United States. There were no rural–urban discrepancies in diagnostic stage, suggesting similar access to diagnostic services. Rural patients diagnosed with stage I non–small cell lung cancer had shorter survival, which may reflect disparities in access to surgical care. No survival difference for patients with advanced-stage lung cancer is attributed to lack of effective treatment during the time period of this study.
Lung cancer is the leading cause of cancer death in the United States (1). Population data suggest that the risk of lung cancer death is not uniform across the United States and that per capita lung cancer mortality is 18 to 20% greater in rural areas than in urban areas (2). The cause for this observed disparity is not clearly defined in existing literature.
Increased per capita rural lung cancer mortality might be explained by a population-based factor such as increased rural smoking prevalence leading to increased rural lung cancer incidence. There is some evidence to support this hypothesis; a national health interview survey suggests higher current smoking rates among rural residents (3), and the link between smoking and lung cancer is well known.
Alternatively, differences in access to diagnostic or treatment facilities for rural patients might account for increased rural lung cancer mortality. Identification of lung cancer at an early stage combined with surgical resection offers the greatest chance of cure. Therefore, lung cancer mortality should be sensitive to factors that limit timely diagnosis or availability of effective treatment. Researchers in studies done in the United States (4–8), the United Kingdom (9–11), France (12), and Australia (13) have looked at the effect of rurality on lung cancer stage at diagnosis, treatments received, and mortality. They have considered how certain aspects of rurality, such as geographic isolation and, in some cases, socioeconomic deprivation, might impact treatment received. Results vary widely, with no consistent association between rurality and mortality in patients diagnosed with lung cancer.
The drivers of U.S. rural–urban lung cancer mortality disparity have not been identified, and it is important to fill this knowledge gap to improve future lung cancer care in rural populations. If increased lung cancer mortality in rural areas is due solely to increased smoking, then a focus on smoking cessation in rural communities should be prioritized. However, if rural patients are not receiving the same diagnostic or treatment modalities that are available to patients in urban areas, then smoking cessation efforts alone will be inadequate and will need to be supplemented by efforts to increase access to cancer services for rural patients. Understanding the cause of this disparity in lung cancer outcomes is also timely because new diagnostic and treatment modalities, such as computed tomographic screening for lung cancer (14), are likely to be implemented in predominantly urban academic settings initially, with the potential to increase rural–urban disparities in care.
We designed this study to address two objectives: first, to explore the relationship between smoking rates, lung cancer incidence, and lung cancer mortality in populations across the rural–urban continuum; and second, to determine whether there is increased mortality in rural patients diagnosed with lung cancer and, if so, whether this is associated with presentation at a later stage or with disparity in the treatment received. Some of the results of these studies were previously reported in the form of an abstract (15).
Institutional review board approval was obtained from the Dartmouth-Hitchcock Medical Center Committee for the Protection of Human Subjects prior to beginning the study.
Data were obtained from the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute. We obtained county-level linked data for the SEER 18 areas of the United States (16). The SEER 18 areas comprise a variety of metropolitan, urban, suburban, and rural locations. It includes the states of Alaska, California, Connecticut, Georgia, Hawaii, Iowa, Kentucky, Louisiana, New Mexico, Utah, and the metropolitan areas of Detroit and Seattle-Puget Sound. SEER data are obtained from state cancer registries, and case-finding strategies are used to ensure a comprehensive dataset (17).
A database containing SEER data for a cohort of 348,002 patients diagnosed with lung cancer between 2000 and 2006 was created (SEER*Stat software [https://seer.cancer.gov/seerstat/] version 8.1.5, Surveillance Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, MD; Excel 2010, Microsoft Corporation, Redmond, WA). The patient-linked data obtained from SEER included patient age at diagnosis within a 5-year range, patient sex, patient race, county of residence, county rural–urban continuum area (RUCA) code (18), county median income, county high school graduation percentage, county estimate of percentage of ever smokers, year of diagnosis, stage of cancer at diagnosis, cancer histology, surgical treatment received, radiation treatment received or refused, survival in months from diagnosis, and whether lung cancer was recorded as a cause of death.
Additionally, 2003 county-level population estimates from the U.S. Census Bureau (19) and 2003 county-level estimates of smoking prevalence derived from published data (20) were obtained. The 2000–2006 time frame was chosen to maximize the number of SEER areas reporting data in a uniform manner and to ensure survival could be tracked to at least 5 years. The follow-up period extended to the end of 2011, meaning every patient was followed for between 5 and 12 years, depending on date of diagnosis. RUCA codes, which use population density and proximity to metropolitan areas to classify every county in the United States on a 1–9 scale, were used as the measure of rurality (18).
Lung cancer incidence was defined as the number of new cases per 100,000 population per year and was calculated as follows: the total number of cases in a given RUCA during the 7-year period, divided by 7, divided by the calculated 2003 population estimate of the sum of all counties in that RUCA category, multiplied by 100,000. Incidence rate ratios with 95% confidence intervals (CIs) were calculated to compare lung cancer incidence in the most rural and most urban areas. Lung cancer mortality was defined as the number of deaths attributable to lung cancer per year of entrance into the cohort per 100,000 population. Lung cancer mortality rates were calculated in a similar fashion to incidence rates. Smoking prevalence was calculated as the population weighted mean of the published smoking prevalence in the counties within that RUCA.
The cohort of patients diagnosed with lung cancer was then analyzed by RUCA for differences in patient- and county-level demographics, stage at presentation, treatment received, and mortality. Stage- and type-specific analyses were preplanned to see if associations between treatment and mortality existed for both curable and incurable disease.
The chi-square test was used to compare nominal data between groups. The Mann-Whitney U test was used to compare median survival between groups.
We compared the relative odds of receiving no treatment for stage I non–small cell lung cancer (NSCLC) in different RUCA regions using logistic regression. After determining the unadjusted odds ratios (ORs), we analyzed a model that included age, sex, race, and marital status to determine the association of rurality with nontreatment independent of other factors that might reasonably impact access to therapy. To compare relative survival durations for stage I NSCLC, we used quantile regression to determine the difference in median survival by RUCA classification, first without potential confounders and then controlling for age, sex, race, and marital status. All statistical analyses were performed using STATA 12 software (StataCorp LP, College Station, TX). Institutional review board approval was obtained prior to beginning the study (protocol number STUDY00028051).
The cohort consisted of 348,002 patients with lung cancer from the 18 geographical regions in the SEER program (Table 1). RUCA 1 counties are the most urban, and RUCA 9 counties the most rural. Of the patients in the cohort, 200,212 resided in a highly urban area (RUCA 1), with fewer patients living in more rural areas. Rural patients (RUCA 9), compared with the most urban patients (RUCA 1), were more likely to be white (96.5% vs. 81.9%) and male (61.9% vs. 52.9%). There was a trend toward a younger mean age of patients in rural areas (67.5 vs. 69.1 yr) (Table 2). Rural patients lived in counties with lower median annual income ($25,607 vs. $50,444), a lower level of high school–level educational attainment (65.8% vs. 80.5%), a higher proportion of ever smokers (54.6% vs. 41.9%), and a higher prevalence of current smokers (24.8% vs. 14.5%). Figure 1 shows 2003 estimated county-level smoking prevalence in the SEER 18 geographical region.
| RUCA | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | |
| RUCA classification | |||||||||
| Description | Metropolitan | Metropolitan | Metropolitan | Urban | Urban | Urban | Urban | Rural | Rural |
| Population size | >1 million | 250,000–1 million | <250,000 | >20,000 | >20,000 | 2,500–20,000 | 2,500–20,000 | <2,500 | <2,500 |
| Adjacent to metropolitan area | Yes | No | Yes | No | Yes | No | |||
| County-level demographics | |||||||||
| Estimated 2003 population | 50.6 million | 15.9 million | 6.8 million | 2.2 million | 1.1 million | 2.7 million | 1.9 million | 0.48 million | 0.53 million |
| Median income, U.S. dollars | $50,444 | $45,600 | $37,793 | $36,231 | $35,239 | $32,065 | $27,922 | $30,863 | $25,607 |
| High school education | 80.5% | 79.7% | 78.7% | 77.5% | 80.7% | 72.4% | 70.0% | 70.5% | 65.8% |
| Percentage of ever smokers | 42% | 44% | 47% | 49% | 50% | 50% | 51% | 54% | 55% |
| Smoking prevalence | 14% | 16% | 18% | 20% | 21% | 22% | 24% | 24% | 25% |
| Incidence | |||||||||
| Diagnosed with lung cancer between 2000 and 2006 | 200,212 | 61,583 | 32,399 | 12,155 | 6,284 | 16,652 | 11,479 | 3,179 | 3,623 |
| Lung cancer incidence per 100,000 population | 56.56 | 55.28 | 67.82 | 79.70 | 80.50 | 87.00 | 87.17 | 94.35 | 97.57 |
| Mortality | |||||||||
| Cohort members who died as a result of lung cancer during follow-up period* | 124,247 | 39,096 | 20,790 | 7,907 | 4,125 | 11,227 | 7,705 | 2,141 | 2,492 |
| Annual mortality attributable to lung cancer per 100,000 population† | 35.10 | 35.09 | 43.52 | 51.85 | 52.84 | 58.65 | 58.51 | 63.56 | 67.11 |
| Patient Demographics | RUCA | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | |
| Mean age, yr | 69.1 | 69.1 | 68.7 | 68.4 | 69.2 | 68.2 | 67.8 | 68.0 | 67.5 |
| White race | 82% | 85% | 90% | 90% | 84% | 89% | 95% | 93% | 96% |
| Male sex | 53% | 54% | 57% | 57% | 57% | 61% | 61% | 63% | 62% |
| Diagnosis and staging | |||||||||
| Small cell lung cancer | 12% | 13% | 14% | 14% | 15% | 16% | 16% | 16% | 16% |
| Stage I | 15% | 15% | 15% | 14% | 14% | 14% | 14% | 13% | 13% |
| Stage II | 3% | 3% | 3% | 3% | 3% | 3% | 3% | 3% | 3% |
| Stage III | 24% | 24% | 25% | 25% | 25% | 24% | 24% | 23% | 24% |
| Stage IV | 39% | 38% | 37% | 35% | 39% | 36% | 35% | 37% | 35% |
| Stage unknown | 13% | 15% | 15% | 16% | 13% | 17% | 18% | 18% | 19% |
| Treatment | |||||||||
| Any-stage surgery | 21.8% | 21.0% | 20.3% | 19.2% | 19.7% | 18.4% | 19.9% | 17.9% | 20.1% |
| Stage I NSCLC surgery | 75% | 75% | 72% | 70% | 70% | 68% | 71% | 65% | 69% |
| Any-stage radiation | 34.8% | 36.5% | 36.9% | 36.8% | 38.5% | 37.0% | 33.9% | 36.4% | 36.5% |
| Refused radiation | 1.7% | 1.7% | 1.8% | 2.0% | 2.4% | 1.6% | 2.5% | 2.3% | 2.6% |
| Median survival,* mo | |||||||||
| All stages, all types | 8 | 8 | 8 | 8 | 8 | 8 | 7 | 7 | 8 |
| Stage I | 51 | 47 | 43 | 39 | 40.5 | 38 | 35 | 35 | 38.5 |
| Stage II | 26 | 24 | 22 | 25 | 28 | 23 | 22 | 29.5 | 23.5 |
| Stage III | 9 | 9 | 8 | 9 | 9 | 9 | 8 | 8 | 8 |
| Stage IV | 4 | 4 | 4 | 4 | 4 | 4 | 3 | 4 | 4 |
| Stage I NSCLC only | 52 | 49 | 44 | 40 | 42 | 40 | 37 | 35 | 40 |
| Cause of death | |||||||||
| Lung cancer recorded as cause of death | 62% | 63% | 64% | 65% | 66% | 67% | 67% | 67% | 69% |
Figure 1. Calendar year 2003 estimates of smoking prevalence at the county level for the Surveillance, Epidemiology, and End Results 18 areas used in our cohort.
[More] [Minimize]Lung cancer mortality increased steadily with rural residency (Figure 2). RUCA 9 regions had an annual lung cancer mortality almost twice that of RUCA 1 regions (67.11 vs. 35.10 per 100,000 population). Differences in lung cancer mortality, smoking prevalence, and lung cancer incidence in the State of Georgia can be seen in Figure 3. Mortality is lowest in the Greater Atlanta metropolitan region.
Figure 2. Lung cancer mortality, incidence, and estimated smoking prevalence all increase with rurality. RUCA = rural–urban continuum area.
[More] [Minimize]
Figure 3. Calendar year 2003 estimates of (A) smoking prevalence, (B) lung cancer incidence, and (C) lung cancer mortality at the county level in the State of Georgia. There is considerable variation in all three variables with the state, but there is low smoking prevalence, lung cancer incidence, and lung cancer mortality in the Greater Atlanta metropolitan area (northwestern Georgia).
[More] [Minimize]The incidence of lung cancer increased with rural residency in a consistent fashion across all RUCAs (Figure 2). The most rural counties (RUCA 9) had an annual lung cancer incidence almost twice that of the largest metropolitan areas (RUCA 1) (97.57 vs. 56.56 per 100,000 population; incidence rate ratio, 1.72; 95% CI, 1.67–1.78; P < 0.0001). Smoking prevalence increased with rurality in a similar pattern to lung cancer incidence (Figure 2).
Stage at diagnosis was similar, regardless of RUCA. Stage data were unknown more often for rural residents (18.9% vs. 13.3%; P < 0.001). Small cell lung cancer accounted for a greater percentage of the total lung cancer incidence in rural areas than in the most urban areas (15.8% vs. 12.1%; P < 0.001).
The median survival for all stages and types of lung cancer was 8 months, with no significant differences across RUCAs. Lung cancer was recorded as the cause of death more frequently in rural patients (69% vs. 62%; P < 0.001). Rural residents and urban residents received surgery (20.1% vs. 21.8%) and radiation treatment (36.5% vs. 34.8%) with similar frequencies (Table 2).
Rural patients with stage I lung cancer had a median survival 12 months shorter than that of the most urban patients (38.5 vs. 51 mo; P = 0.0005). When cases of small cell lung cancer were excluded, this survival difference was preserved (40 vs. 52 mo; P = 0.0006, Figure 4). Fewer surgeries were performed for patients with stage I NSCLC living in rural areas than for those in the most urban areas (69% vs. 75%, Figure 4), but radiation treatment (14.7% vs. 15.1%) and refusal of radiation treatment (2% vs. 1%) were similar. The distribution of treatments for stage I NSCLC in the most rural and most urban areas is detailed in Table E1 in the online supplement.
Figure 4. Median survival for patients with stage I non–small cell lung cancer and percentage of patients receiving surgery both decrease as rurality increases. RUCA = rural–urban continuum area.
[More] [Minimize]When no treatment versus any treatment for stage I NSCLC was compared, we found that rural patients were more likely to receive no treatment (13.2% vs. 17.6%; OR, 1.40; P = 0.007). When adjusted for age, sex, race, and marital status, rural patients with stage I NSCLC remained less likely to receive treatment (OR, 1.74; P < 0.001). A similarly adjusted survival model showed that, for every stepwise increase in rurality as measured by RUCA, median survival for patients with stage I NSCLC decreased by 1.63 months (Table 3). Lung cancer was more frequently recorded as the cause of death in rural residents presenting with stage I NSCLC (37% vs. 29%; P < 0.001).
| Odds Ratio for No Treatment | 95% Confidence Interval | P Value | Adjusted Odds Ratio for No Treatment | 95% Confidence Interval | P Value | |||
|---|---|---|---|---|---|---|---|---|
| RUCA | ||||||||
| 2 | 0.93 | 0.87 | 1.00 | 0.046 | 0.94 | 0.87 | 1.02 | 0.123 |
| 3 | 1.12 | 1.03 | 1.27 | 0.009 | 1.17 | 1.07 | 1.29 | 0.001 |
| 4 | 1.09 | 0.95 | 1.26 | 0.227 | 1.13 | 0.97 | 1.31 | 0.119 |
| 5 | 1.16 | 0.96 | 1.40 | 0.123 | 1.18 | 0.97 | 1.44 | 0.105 |
| 6 | 1.34 | 1.19 | 1.52 | <0.001 | 1.46 | 1.29 | 1.66 | <0.001 |
| 7 | 1.21 | 1.05 | 1.40 | 0.010 | 1.47 | 1.26 | 1.71 | <0.001 |
| 8 | 1.33 | 1.03 | 1.75 | 0.032 | 1.56 | 1.18 | 2.07 | 0.002 |
| 9 | 1.40 | 1.09 | 1.80 | 0.007 | 1.74 | 1.33 | 2.26 | <0.001 |
| Age (+1 yr) | 1.06 | 1.06 | 1.06 | <0.001 | ||||
| Male sex | 1.21 | 1.14 | 1.28 | <0.001 | ||||
| Race | ||||||||
| Black | 2.07 | 1.90 | 2.26 | <0.001 | ||||
| Other nonwhite | 1.27 | 1.12 | 1.43 | <0.001 | ||||
| Median Survival in Months | 95% Confidence Interval | P Value | Median Survival in Months, Adjusted for Age, Sex, and Race | 95% Confidence Interval | P Value | |||
| Rurality (+1 RUCA) | −2.43 | −2.89 | −1.97 | <0.001 | −1.63 | −1.94 | −1.31 | <0.001 |
| Age (+1 yr) | −1.34 | −1.40 | −1.28 | <0.001 | ||||
| Male sex | −14.94 | −16.10 | −13.80 | <0.001 | ||||
| Race: all nonwhite | −0.94 | −2.05 | 0.17 | 0.098 | ||||
There were relatively few cases of stage II NSCLC in our cohort and only 86 cases of stage II NSCLC in the RUCA 9 group. When stage II cases were added to stage I cases, there was no change in the pattern of fewer surgeries, similar frequency of radiation treatment, and reduced survival for rural patients (Table E2).
Stage-specific median survival was similar across all RUCAs and for stage III (9 mo) and stage IV (4 mo) disease. Surgical and radiation treatment was received with similar frequency across all RUCAs for stages III and IV disease. Stages III and IV disease are frequently treated with chemotherapy, and we have no data regarding this treatment modality.
Consistent with prior studies, we observed an association of rurality with increased lung cancer mortality, but we further characterized this disparity as larger than previously recognized and with a dose–response relationship. We describe an increase in lung cancer incidence that tracks closely with rural smoking rates. Finally, we did not find evidence of delayed diagnosis in rural populations, but we did observe significant differences in management of early-stage disease that may have contributed to increased case fatality rates in rural communities.
Published data based on 2007 U.S. national mortality statistics showed age-adjusted lung cancer mortality rates of 49 per 100,000 population in metropolitan areas and 59 per 100,000 population in nonmetropolitan areas (2). We found that lung cancer mortality rates in the most rural areas were almost double those of the most urban areas, and we additionally have demonstrate a previously undocumented dose–response association between rurality and mortality. Our results suggest that prior work may have underestimated the magnitude of disparity between the most urban and most rural areas.
The observed association of rurality with increased per capita lung cancer mortality is opposite to that reported in the United Kingdom, where both the incidence of lung cancer and lung cancer mortality are greater in urban areas than in rural areas (9). A parsimonious hypothesis for this U.S.–U.K. difference is that smoking prevalence, lung cancer incidence, and lung cancer mortality are all more common in areas of socioeconomic deprivation. These areas may be urban in the United Kingdom; however, our data show that, at the county level, it is rural areas in the United States that have lower median incomes and higher prevalence of smoking (9, 20). In the United States, other smoking-related diseases, including coronary artery disease, chronic obstructive pulmonary disease, and stroke, are also more prevalent in rural areas (21, 22).
To our knowledge, this is the first study to link county-level smoking data, lung cancer incidence, and lung cancer mortality across a rural–urban continuum. The incidence ranged from 56 to 97 cases per 100,000 population and was consistent with the reported national U.S. incidence of 60.6 per 100,000 population in 2003 (1). Differences in genetics or radon exposure could theoretically contribute to increased incidence of rural lung cancer; however, the nearly direct correlation between smoking prevalence and lung cancer incidence across RUCA classes observed in our data suggests that these factors are unlikely to be the primary explanation. Occupational exposures such as coal mining that track with rural communities in Appalachia are known to be associated with lung cancer mortality (23), but patients from neither Pennsylvania nor Virginia are included in the SEER 18 dataset, and we know of no other clear occupational exposures in the regions included in the data.
Differences in lung cancer mortality do not seem to be due to a failure to detect cancer at a curable stage. We suspected that rural patients might present later if their access to health care were diminished, but there were no significant differences in stage at diagnosis across RUCAs, a finding consistent with prior studies (4, 12). These results remained unchanged even after we excluded cases of small cell lung cancer, which were more common in rural areas, likely owing to higher rates of current tobacco use in those regions. There was a greater proportion of stage-unknown patients in rural areas, but the median survival of the unknown-stage patients was similar across all RUCAs (data not shown), and differences in the proportion of unknown-stage patients did not influence stage-specific mortality.
Stage I NSCLC is a potentially curative disease, is treated surgically where possible, and has by far the best prognosis of all primary lung cancers. We found that median survival was reduced by 12 months among the most rural patients diagnosed with stage I NSCLC, and significantly fewer patients in this group underwent surgical intervention than in the most urban patient groups with the same stage of disease. We believe this to be a new finding; the largest previous U.S. study showed no difference in mortality or in the receipt of surgical treatment between urban and rural patients, perhaps because of fewer patients overall, and particularly in the most rural areas (4).
It could be hypothesized that fewer surgeries for rural patients reflects limited access to the health care system in general. However, despite lower median incomes, rural patients are diagnosed at the same stage as urban patients and receive similar amounts of radiation. We suspect that, with safety nets available for the most financially disadvantaged, rural residents do have access to health care in general but have impaired access to specialist interventions such as thoracic surgery, perhaps owing to geographic isolation.
It is possible that fewer surgeries for rural patients with stage I NSCLC in rural areas occur because of increased comorbidities among rural patients, which prevent them from undergoing surgery. If this were the case, we might expect to see that rural patients received more radiation treatment, which is the usual therapy for patients with stage I NSCLC not amenable to surgery, but we did not. If the comorbidities in the rural population were severe enough to prevent them from having surgery, then we might expect these patients to die as a result of their comorbidities, but in fact they were more likely than their urban counterparts to have lung cancer recorded as the cause of death.
It is possible that fewer surgeries were performed for rural patients because of patient refusal; we do not have data on refusal of surgery in our cohort. However, we think refusal of surgery is a less likely explanation of our findings because, although a belief that surgery leads to the spread of cancer has been reported, this belief was more prevalent in the African American population, and the rural population in our cohort was predominantly white and less racially diverse than the urban population (24).
Lung cancer mortality for later-stage disease (stages II–IV) was similar across all RUCAs. The very poor survival statistics, particularly for stage III disease (median survival, around 9 mo) and for stage IV disease (median survival, 4 mo), are consistent with the paucity of highly effective chemotherapy for lung cancer in this time period (25, 26). The fact that we observed differences in survival for early-stage cancer only, the subset of lung cancer with proven effective therapy, is compatible with the hypothesis that rurality is a barrier to access for some types of cancer treatment.
The rural–urban disparities in treatment and mortality may not be limited to lung cancer; similar disparities have been reported in breast and colorectal cancer mortality (9, 27). If new cancer treatments are offered primarily in urban academic medical centers, it is possible that rural–urban disparities will increase in a number of diseases. There is evidence that the rural–urban disparity in life expectancy is increasing (21). Targeted interventions to improve rural cancer care using local community resources and telemedicine have been trialed with some success in the fields of smoking cessation, colorectal and cervical cancer screening, and cancer-related genetic testing (28–33).
This study has many strengths. The SEER database provides a large number of cases with uniform data reporting across a geographically representative population. This increases the applicability of findings to the wider U.S. population. To our knowledge, this is the largest cohort study to explore the impact of rurality on lung cancer and the first to examine rurality, smoking rates, lung cancer incidence, and mortality on such a scale. A large sample of patients is particularly important when analyzing the most rural residents, who account for a relatively small proportion of the overall population, and when examining stage-specific treatment and survival effects.
The study also has important limitations. The SEER database does not include direct measures of insurance status or comorbidities. Use of an alternative database such as the SEER-linked Medicare database would have provided greater information regarding these features, but it would have excluded patients under the age of 65 years and would have lowered overall numbers in the cohort. Our database also does not include a direct measurement of smoking history. We were able, however, to link geographical regions to regional measures of smoking prevalence at the county level as a proxy for individual smoking history. The use of an older cohort is a limitation. Our cohort is from the 2000–2006 time period, chosen to ensure a minimum of a 5-year follow-up period for all patients and to allow the calculation of median survival in early-stage cancer. This leaves unanswered the question whether the observed disparities narrowed in the intervening time period. This important question should be the basis of further research. We do not have data on the receipt of chemotherapy, although there were no detected rural–urban differences in mortality for patients with advanced-stage disease in our cohort.
We have used RUCAs, a measurement of population density and proximity to a metropolitan area, as our measure of rurality. This may not be the best measure of rurality when considering access to care. Australian data have shown that greater distance from a specialist hospital reduces the chances of admission to that hospital, reduces the chance of having surgical resection, and increases the chance of death as a result of lung cancer (13). An assessment of distance to the nearest hospital that provides comprehensive diagnostic and therapeutic cancer management could provide additional insights into the impact of geographic isolation and access to care, and we hope to explore this further in our future work.
Our study demonstrates that the disparity in lung cancer mortality among rural communities is likely greater than previously documented and that a dose–response association exists between rurality and lung cancer mortality. Furthermore, the disparity in mortality appears to be related to both increased smoking and lung cancer incidence in rural communities as well as less access to surgical care for stage I disease. Recent and future advances in lung cancer care, including computed tomographic screening for lung cancer, more widespread use of positron emission tomographic scanners, improved techniques for lymph node sampling, stereotactic body radiotherapy, tumor cell genetic analysis, and targeted chemotherapies, are more likely to be available at urban academic medical centers. This may further increase rural–urban disparities in lung cancer mortality. To improve lung cancer mortality in rural areas, policy makers should focus on smoking cessation interventions and improving access to effective treatments such as surgery for early-stage disease.
| 1 . | Surveillance, Epidemiology, and End Results (SEER) Program, National Cancer Institute (NCI). Cancer stat facts: lung and bronchus cancer. Bethesda, MD: NCI, National Institutes of Health; 2016 [accessed 2016 Feb 1]. Available from: http://seer.cancer.gov/statfacts/html/lungb.html |
| 2 . | Singh GK, Siahpush M, Williams SD. Changing urbanization patterns in US lung cancer mortality, 1950–2007. J Community Health 2012;37:412–420. |
| 3 . | Pleis JR, Ward BW, Lucas JW. Summary health statistics for US adults: National Health Interview Survey, 2009. Vital Health Stat 2010;10(249). |
| 4 . | Shugarman LR, Sorbero MES, Tian H, Jain AK, Ashwood JS. An exploration of urban and rural differences in lung cancer survival among Medicare beneficiaries. Am J Public Health 2008;98:1280–1287. |
| 5 . | Baldwin LM, Patel S, Andrilla CH, Rosenblatt RA, Doescher MP. Receipt of recommended radiation therapy among rural and urban cancer patients. Cancer 2012;118:5100–5109. |
| 6 . | Forrest LF, Adams J, Wareham H, Rubin G, White M. Socioeconomic inequalities in lung cancer treatment: systematic review and meta-analysis. PLoS Med 2013;10:e1001376. |
| 7 . | Johnson AM, Hines RB, Johnson JA III, Bayakly AR. Treatment and survival disparities in lung cancer: the effect of social environment and place of residence. Lung Cancer 2014;83:401–407. |
| 8 . | McLafferty S, Wang F. Rural reversal? Rural-urban disparities in late-stage cancer risk in Illinois. Cancer 2009;115:2755–2764. |
| 9 . | National Cancer Intelligence Network (NCIN). The effect of rurality on cancer incidence and mortality: NCIN data briefing. London: NCIN; 2011 [accessed 2015 Jan 1]. Available from: http://www.ncin.org.uk/publications/data_briefings/rurality |
| 10 . | O’Dowd EL, McKeever TM, Baldwin DR, Anwar S, Powell HA, Gibson JE, Iyen-Omofoman B, Hubbard RB. What characteristics of primary care and patients are associated with early death in patients with lung cancer in the UK? Thorax 2015;70:161–168. |
| 11 . | Riaz SP, Horton M, Kang J, Mak V, Lüchtenborg M, Møller H. Lung cancer incidence and survival in England: an analysis by socioeconomic deprivation and urbanization. J Thorac Oncol 2011;6:2005–2010. |
| 12 . | Pozet A, Westeel V, Berion P, Danzon A, Debieuvre D, Breton JL, Monnier A, Lahourcade J, Dalphin JC, Mercier M. Rurality and survival differences in lung cancer: a large population-based multivariate analysis. Lung Cancer 2008;59:291–300. |
| 13 . | Tracey E, McCaughan B, Badgery-Parker T, Young J, Armstrong B. Survival of Australian lung cancer patients and the impact of distance from and attendance at a thoracic specialist centre: a data linkage study. Thorax 2015;70:152–160. |
| 14 . | Aberle DR, Adams AM, Berg CD, Black WC, Clapp JD, Fagerstrom RM, Gareen IF, Gatsonis C, Marcus PM, Sicks JD; National Lung Screening Trial Research Team. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med 2011;365:395–409. |
| 15 . | Atkins GT, Munson JC. The effect of rural residence on lung cancer incidence and stage specific survival [abstract]. Am J Respir Crit Care Med 2015;191:A3595. |
| 16 . | Surveillance, Epidemiology, and End Results (SEER) Program. SEER*Stat Database: populations - total US (1969–2013) <single ages to 85+, Katrina/Rita adjustment> - linked to county attributes - total US, 1969–2013 counties. Bethesda, MD: National Cancer Institute, DCCPS, Surveillance Research Program, Surveillance Systems Branch; 2014 [released 2014 Oct, accessed 2015 Feb 1]. Available from: www.seer.cancer.gov |
| 17 . | Surveillance, Epidemiology, and End Results (SEER) Program, National Cancer Institute (NCI). SEER training website. Bethesda, MD: NCI, National Institutes of Health; 2015 [accessed 2015 Feb 1]. Available from: http://training.seer.cancer.gov/ |
| 18 . | Hart LG, Larson EH, Lishner DM. Rural definitions for health policy and research. Am J Public Health 2005;95:1149–1155. |
| 19 . | Population Division, U.S. Census Bureau. Annual estimates of the population for counties (multiple states): April 1, 2000 to July 1, 2003 (CO-EST 2003). Washington, DC: Author; April 2004. |
| 20 . | Dwyer-Lindgren L, Mokdad AH, Srebotnjak T, Flaxman AD, Hansen GM, Murray CJL. Cigarette smoking prevalence in US counties: 1996-2012. Popul Health Metr 2014;12:5. |
| 21 . | Singh GK, Siahpush M. Widening rural–urban disparities in life expectancy, U.S., 1969–2009. Am J Prev Med 2014;46:e19–e29. |
| 22 . | O’Connor A, Wellenius G. Rural–urban disparities in the prevalence of diabetes and coronary heart disease. Public Health 2012;126:813–820. |
| 23 . | Hendryx M, O’Donnell K, Horn K. Lung cancer mortality is elevated in coal-mining areas of Appalachia. Lung Cancer 2008;62:1–7. |
| 24 . | Margolis ML, Christie JD, Silvestri GA, Kaiser L, Santiago S, Hansen-Flaschen J. Racial differences pertaining to a belief about lung cancer surgery: results of a multicenter survey. Ann Intern Med 2003;139:558–563. |
| 25 . | Non-Small Cell Lung Cancer Collaborative Group. Chemotherapy and supportive care versus supportive care alone for advanced non-small cell lung cancer. Cochrane Database Syst Rev 2010;5:CD007309. |
| 26 . | De Petris L, Crinò L, Scagliotti GV, Gridelli C, Galetta D, Metro G, Novello S, Maione P, Colucci G, de Marinis F. Treatment of advanced non-small cell lung cancer. Ann Oncol 2006;17:ii36–ii41. |
| 27 . | Singh GK, Williams SD, Siahpush M, 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;2011:107497. |
| 28 . | Schoenberg NE, Studts CR, Shelton BJ, Liu M, Clayton R, Bispo JB, Fields N, Dignan M, Cooper T. A randomized controlled trial of a faith-placed, lay health advisor delivered smoking cessation intervention for rural residents. Prev Med Rep 2016;3:317–323. |
| 29 . | Vander Weg MW, Cozad AJ, Howren MB, Cretzmeyer M, Scherubel M, Turvey C, Grant KM, Abrams TE, Katz DA. An individually-tailored smoking cessation intervention for rural Veterans: a pilot randomized trial. BMC Public Health 2016;16:811. |
| 30 . | Schlichting JA, Mengeling MA, Makki NM, Malhotra A, Halfdanarson TR, Klutts JS, Levy BT, Kaboli PJ, Charlton ME. Increasing colorectal cancer screening in an overdue population: participation and cost impacts of adding telephone calls to a FIT mailing program. J Community Health 2014;39:239–247. |
| 31 . | Thompson B, Vilchis H, Moran C, Copeland W, Holte S, Duggan C. Increasing cervical cancer screening in the United States-Mexico border region. J Rural Health 2014;30:196–205. |
| 32 . | Hitt WC, Low GM, Lynch CE, Gauss CH, Magann EF, Lowery CL, Eswaran H. Application of a telecolposcopy program in rural settings. Telemed J E Health 2016;22:816–820. |
| 33 . | Mette LA, Saldívar AM, Poullard NE, Torres IC, Seth SG, Pollock BH, Tomlinson GE. Reaching high-risk underserved individuals for cancer genetic counseling by video-teleconferencing. J Community Support Oncol 2016;14:162–168. |
Author Contributions: G.T.A. and J.M.: had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis; all authors: made substantial contributions to the conception or design of the work and to the acquisition, analysis, or interpretation of data for the work, were involved in drafting and revising the work, are accountable for its accuracy and integrity, and gave approval for submission of the work for publication.
This article has an online supplement, which is accessible from this issue’s table of contents at www.atsjournals.org
Author disclosures are available with the text of this article at www.atsjournals.org.
