Article Text

Download PDFPDF
Urban–rural differences in cancer-directed surgery and survival of patients with non-small cell lung cancer
  1. Audrey Alforque Thomas1,
  2. Alison Pearce2,
  3. Ciaran O'Neill3,
  4. Michal Molcho4,
  5. Linda Sharp5
  1. 1Health Promotion Research Centre, National University of Ireland, Galway, Ireland
  2. 2National Cancer Registry, Cork, Ireland
  3. 3School of Business and Economics, National University of Ireland, Galway, Ireland
  4. 4School of Health Sciences, National University of Ireland, Galway, Ireland
  5. 5Institute of Health and Society, Newcastle University, Newcastle upon Tyne, UK
  1. Correspondence to Dr Audrey Alforque Thomas, Office of Planning and Analysis, University of California, Berkeley, University Hall, Berkeley, CA 94702, USA; audrey{at}


Background Lung cancer is the leading cause of cancer death worldwide. Clinically appropriate cancer-directed surgery is an influential and significant prognostic factor. In a population-based study, we determined how urban/rural residence was related to surgery receipt for patients with non-small cell lung cancer. We assessed the relationship between relative survival and patients' area of residence, taking into account surgery receipt and area socioeconomic level.

Methods We extracted data from the National Cancer Registry Ireland on patients with non-small cell lung cancer diagnosed during 1994–2011 and linked to area-level data on socioeconomic indicators and urban/rural categories. We calculated ORs for receipt of cancer-directed surgery using logistic regression with postestimation of adjusted proportions. Relative survival estimates with follow-up to 31 December 2012 were calculated for all cases and stratified by surgery receipt, adjusting for clinical variables, area socioeconomic level and other sociodemographic characteristics.

Results 15 031 people diagnosed with non-small cell lung cancer were included in the analysis. On the basis of the multiple logistic regression model, a significantly larger proportion of urban patients (adjusted proportion 23%) as compared with rural patients (adjusted proportion 21%) received surgery (p<0.001). In multivariate analysis, rural residence was significantly related to a decrease in excess mortality for all cases (HR 0.90, 95% CI 0.87 to 0.94, p<0.001) and for non-surgical cases (HR 0.88, 95% CI 0.85 to 0.92, p<0.001).

Conclusions The findings point to the need for targeted policies addressing access to treatment for rural patients with non-small cell lung cancer.

  • Cancer epidemiology
  • Health inequalities

Statistics from


Lung cancer is the fourth most common cancer in Ireland,1 as it is in Europe in general.2 Poor survival makes it the most common cause of cancer death in both sexes in Ireland,3 and among men in Europe as a whole.2 Several demographic,4 ,5 clinical6 and area socioeconomic factors7–11 have been independently associated with lung cancer treatment and survival. Reducing inequalities in treatment access and patient outcomes is a key objective of cancer control programmes.12 Such inequalities may arise from remoteness from treatment centres, exposure to air pollution, or other factors related to geography. Measuring the effect of urban or rural area could inform new cancer control strategies and initiatives.

An urban–rural disparity in lung cancer outcomes has been detected, but is in need of further investigation. Living far from treatment facilities has been associated with lower surgery rates, even controlling for area-level deprivation13 or patient financial status.14 However, the literature on survival is mixed: while some find a rural disadvantage,9 ,15–17 others report no difference by area of residence,8 ,11 ,18 or a qualified rural advantage.7 ,19 The present study is the first to address urban/rural differences in lung cancer outcomes in Ireland, which has a distinctively high rural population (37%) compared with other countries in the World Bank High Income category (average 19%).20

In this population-based study, we investigated individual and area-level factors influencing treatment and mortality of patients with non-small cell lung cancer in Ireland, focusing on urban–rural differences.


Study population

The population-based National Cancer Registry Ireland21 reported 18 423 patients diagnosed with a primary non-small cell lung tumour during the period January 1994 to December 2011 (figure 1). Follow-up for death from all causes was until 31 December 2011. Patients were excluded if they had: a previous tumour except non-melanoma skin cancer (N=1289), multiple primary lung tumours (N=19), missing values for any factors, in the relative survival model, except stage and smoking status (N=1848) or survived <1 day after diagnosis (N=236). The urban:rural ratio for cases excluded because of a previous invasive cancer (68% urban:32% rural) were distributed in proportion with all cases (68% urban:32% rural). For those who survived <1 day, there was a higher proportion of urban patients (77%) compared with all cases (68% urban). The analysis was conducted on an anonymised data set provided by the National Cancer Registry Ireland, which has permission, under the Health (Provision of Information) Act 1997 to collect and hold data on all persons diagnosed with cancer in Ireland. The use of those data for research is covered by the Statutory Instrument which established the Registry Board in 1991.

Figure 1

Flow chart showing selection of patients with non-small cell lung cancer included in analysis.

Independent variables

Electoral divisions (EDs), of which there are 3440 in Ireland, are the smallest area for which population statistics are published. Population density of the ED was used to categorise patients into urban (one or more person per hectare) or rural (less than one person per hectare); this categorisation has previously been shown to appropriately identify rural and more urban areas (intermediate and urban areas) in Ireland.22 Quartiles of socioeconomic levels were based on the Trutz Haase 2002 index of relative affluence and deprivation, which takes into account several factors including unemployment rate and housing quality.23 The 2002 measure for socioeconomic levels was selected for its relevance to patients across the diagnosis timeframe 1994–2011.

Variables collected by the Registry at the time of diagnosis and included in the analysis were: age, sex, marital status, smoking status, year of diagnosis and stage. Treatment information was available for the 12 months postdiagnosis.

Statistical analyses

For each variable of interest, group differences by urban or rural residence were compared using χ2 tests.

We used logistic regression models to assess the relationship between cancer-directed surgery and living in an urban or rural area, yielding ORs for receiving cancer-directed surgery. Potential confounders were determined a priori10 ,24 and were controlled in the model. We used likelihood-ratio tests to determine overall p values for categorical variables and calculated adjusted proportions for surgery receipt among urban and rural patients.

Unadjusted probability of 1-year survival was estimated, using Kaplan-Meier methods, for all cases, by urban or rural residence, and for other potential prognostic factors.

Relative survival models, estimated by comparing a group of patients with cancer to a relevant population, do not rely on cause-of-death data, which can be inaccurate or incomplete.25 Relative survival models were based on sex-specific and age-specific life tables for the general population. Excess mortality for rural compared with urban residents was estimated using Poisson regression on relative survival estimates, derived according to the Ederer II method.26 Models were estimated for all cases and stratified by surgery, with confounders determined a priori.4–11 ,27 ,28 We used postestimation Wald tests to determine p values for categorical variables. Stata V.12.0 (StataCorp. Stata Statistical Software: Release 12. College Station, Texas, USA: StataCorp LP 2011) was used for all analysis.


Patient characteristics

The analysis included 15 031 cases of non-small cell lung cancer. The urban–rural distribution of patients (table 1) was similar to the general Irish population in 2013, when the rural population was 37%.20 There were more male than female patients with lung cancer, most markedly in rural areas where almost two-thirds were male. In terms of area socioeconomic level, urban patients came from the most disadvantaged and the most advantaged areas. Rural patients were mostly found in intermediate areas. In urban compared with rural areas, there was a lower proportion of patients who had never smoked and a higher proportion of patients with unknown smoking status. Cancer-directed surgery was more common, and chemotherapy less common, in patients who lived in urban areas.

Table 1

Column per cent of baseline characteristics for all cases, by urban/rural residence, and by receipt of cancer-directed surgery

Table 1 also presents baseline characteristics for urban and rural patients stratified into surgical and non-surgical cases. Among those who received surgery, a higher proportion was married in rural versus urban areas (p=0.001). A notable difference in stage at diagnosis was also found among surgical patients: a greater proportion of rural patients were stage 4 (p=0.001) and a smaller proportion of rural patients were stage 1 (p=0.010). Non-receipt of surgery was higher among men in rural (67%) compared with urban areas (p<0.001).

Factors related to receiving tumour-directed surgery

In the multiple regression analysis, rural patients received cancer-directed surgery significantly less often than urban patients (adjusted OR 0.86, 95% CI 0.78 to 0.95) (table 2). Taking into account age, sex, marital status, area socioeconomic level, smoking status, year of diagnosis and stage at diagnosis, the proportion receiving surgery was 23% of urban patients compared with 21% of rural patients, a statistically significant advantage for urban patients (p<0.001). It is unclear whether this difference would be considered material by clinicians or how heterogeneous they might be in their assessments of clinical significance. The probability at which clinical significance is judged may differ from that conventionally used when reporting results in academic journals.

Table 2

Adjusted ORs of receiving cancer-directed surgery, with 95% CIs, and p values from likelihood ratio tests

Surgery was also more common among patients living in the third quartile for socioeconomic level (adjusted OR 1.31, 95% CI 1.15 to 1.49) and the most advantaged quartile (adjusted OR 1.50, 95% CI 1.32 to 1.71) compared with patients in the most disadvantaged areas. In trend analysis (not shown), increasing socioeconomic level was significantly related to surgery receipt. Using the youngest group as the reference and in trend analysis, likelihood of surgery decreased with increasing age. Female patients were significantly more likely to receive surgery than male patients. Married people had a significantly increased likelihood of surgery. Compared with patients who had never smoked, patients who were smokers at the time of diagnosis and those with unknown smoking status had a decreased likelihood of surgery. The likelihood of surgery was reduced significantly for tumours diagnosed beyond stage 1 and tumours of unknown stage. The trend analysis for stage (not shown), ordered stage 1 to unknown stage, was significant.

Factors related to mortality

Patients with non-small cell lung cancer had a 75% probability of surviving 65 days, a 50% probability of surviving 6 months, and a 25% probability of surviving 18 months (not shown in tables).

The unadjusted 1-year survival estimates by area-level, demographic and clinical variables are presented in table 3. While receipt of surgery was itself influenced by many of these factors, those who had surgery nonetheless had a significantly higher unadjusted survival than those who did not. The crude survival curves for surgery are shown in figure 2, illustrating the unadjusted survival based only on receipt of surgery.

Table 3

Unadjusted 1-year survival estimates with log-rank p values, HRs and 95% CIs

Figure 2

Unadjusted survival curve by receipt of surgery.

Crude 1-year survival estimates were higher for: those who lived in the most advantaged areas, patients who had never smoked, those diagnosed at stage 1. Crude 1-year survival estimates were also higher for patients who were younger than 55 when diagnosed, women, and married people (results not shown).

Multiple regression model of excess mortality

Table 3 summarises the multiple regression models for excess mortality for patients with non-small cell lung cancer, overall and stratified by surgery. The HRs presented in each model take into account the other variables in the model. Overall, rural patients had a small but statistically significantly lower risk of excess mortality (adjusted HR 0.90, 95% CI 0.87 to 0.94) than urban patients. Patients living in more advantaged areas had a significantly lower risk of excess mortality than those in the most disadvantaged areas. The trend analysis for socioeconomic levels (not shown) was significant. Those who had never smoked had significantly lower excess mortality than current smokers, former smokers and those with unknown smoking status. Stage at diagnosis was significantly related to survival, with excess mortality increasing with higher stage at diagnosis. The trend analysis for stage (not shown), ordered stage 1 to unknown stage, was significant. Cancer-directed surgery, chemotherapy and radiotherapy each had a negative relationship with excess mortality, with the effect of cancer-directed surgery having the largest magnitude. In results not shown, risk of excess mortality was higher for men and increased with increasing age, in categories and in trend analysis.

Many of the factors in the model with all cases had a similar magnitude of effect and significance in the models stratified by surgery. While among non-surgical cases there was a significantly lower excess mortality in rural patients (HR 0.88, 95% CI 0.85 to 0.92), for surgical cases there was no urban–rural difference in excess mortality. The effect of socioeconomic level was diminished in the model for surgical cases, but remained significant among non-surgical cases.


In this population-based study, there was a modest but statistically significant higher receipt of cancer-directed surgery among urban than rural patients with non-small cell lung cancer. In multiple regression analysis, cancer-directed surgery was a strong predictor of survival, as would be expected from the literature.29–31 It was somewhat surprising that, among non-surgical patients, urban patients had higher excess mortality than rural patients.

Rural patients received cancer-directed surgery less often than urban patients, even after considering stage at diagnosis and sociodemographic factors. While the difference was small in magnitude, it was statistically significant. This is an important finding since surgery is the main option for curative treatment in these patients. Rural patients with cancer may have different expectations for care.32 ,33 In addition, virtually all of the surgical resections for lung cancer in Ireland are performed in hospitals in three cities: Dublin, Cork and Galway,34 located in urban areas of the east, south and west of the country, respectively. This may place a heavy transportation and accommodation burden on rural patients. The negative relationship between living far from treatment facilities and receipt of surgery for non-small cell lung cancer has been shown in England13 and Australia.14 Another English study found that distance from treatment hospital, while having little or no effect on its own, amplified the effect of area-level disadvantage in terms of thoracic surgery for patients with non-small cell lung cancer,35 although we found no interactions between urban or rural residence and area socioeconomic level in our data set (not shown). In this study, rural patients had a higher receipt of chemotherapy, which is available at many hospitals geographically spread throughout Ireland.34 Since geographic factors are context-specific, our study identifies opportunities for policy and practice to address geographic inequalities in receipt of cancer treatment in Ireland.

When demographic and clinical variables were accounted for, rural patients had a lower excess mortality than urban patients. There is mixed evidence of urban–rural survival differences in the literature.7–9 ,11 ,15–19 While it did not show a statistically significant relationship between area of residence of survival for all cases, a study in the US state of Georgia found that among early-stage patients with non-small cell lung cancer, those in rural areas had a higher survival rate than urban patients.7 A study in Denmark found that those in peripheral rural areas, which was the most remote category, had higher survival than those in urban areas or rural areas.19 Several other studies describe a rural disadvantage in survival with lung cancer,9 ,15–17 but these did not distinguish between surgical and non-surgical cases.

Further analysis showed that stage at diagnosis was the confounder with the strongest influence on the relationships between area of residence and survival (not shown). Since early stage at diagnosis was more common among urban patients and late stage at diagnosis was more common among rural patients, controlling for stage allowed the rural survival advantage to emerge. A possible explanation for the rural advantage in survival in this study is differences in patterns of care. As noted above, surgical treatment in Ireland is concentrated in larger hospitals and cancer centres which are located in urban areas, so regardless of the area of residence, urban and rural patients undergoing surgery attend the same hospitals and are subject to the same quality of treatment, and hence have similar outcomes. Non-surgical patients might be seen at local hospitals, increasing the variability in quality of care. It is possible that, for non-surgical patients, a smaller and less crowded clinic is more beneficial than one that is large and busy. The rural advantage in survival could also be attributable to unobserved characteristics which stem directly from living in a rural area, such as the therapeutic value of nature,36 higher quality of life37 and lower anxiety.38 ,39 A previous study by the authors has shown that rural patients with cancer in Ireland have a higher quality of life,40 which has been shown to be an important prognostic factor.41 ,42

While this study found an urban–rural difference that persisted after adjustment for area-level deprivation, it also found a strong independent association between deprivation and survival, especially in non-surgical cases. Our findings therefore support calls for urgent action to address socioeconomic disparity in treatment7 ,10 ,24 ,35 and survival of patients with lung cancer.8–11 ,24

The major strength of this analysis is that it was based on high-quality, national, population-based data; completeness of registration in Ireland is in excess of 97%.43 A limitation is the use of national life tables, which are the only ones available for this population, in the relative survival models. Irish life tables do not contain urban/rural information, potentially contributing to differences in relative survival by urban/rural residence. There is evidence that urban residents have a higher mortality rate than rural residents; however, the estimated difference in life expectancy was small, was not tested for statistical significance, and did not control for area socioeconomic level.44 Additionally, the proportion of smokers in the general population is much lower than the proportion of smokers among those diagnosed with lung cancer, leading to underestimates of relative survival of patients with lung cancer because life expectancy of smokers is lower than that of non-smokers.45 There is some evidence from Canada15 and from Poland18 that more smokers are found in rural as compared with urban areas, but urban–rural differences in smoking in Ireland are not well characterised. Underlying differences between the urban and rural populations would affect the relative survival results, resulting in an underestimation of relative survival for the area that has a lower proportion of smokers. Another limitation is the lack of clinical information on private insurance, functional status, general health status and comorbidities of these patients. Urban–rural variation in these health-related factors could, at least in part, account for the differences in surgery receipt and survival. In addition, we did not have access to information about type of surgery, specifically palliative versus curative, and combinations of treatments. A further limitation is that the measure for area socioeconomic group was taken at 2002 for all cases, although patients were diagnosed between 1994 and 2011. Since this measure is grouped into quartiles, the recategorisation of EDs would tend to be equal over time. This would tend to attenuate the true association between deprivation and surgery or survival.

In conclusion, our data showed that rural residence at diagnosis was associated with a decreased likelihood of surgery for patients with non-small cell lung cancer in Ireland after adjusting for sociodemographic and clinical variables. Despite this disadvantage, rural patients had lower excess mortality than urban patients. This mixed picture of clinical outcomes for urban and rural patients with lung cancer points to the need for further investigation of the unique factors and processes that stem from area of residence. Improving equity of access to treatment, and treatment quality, could mitigate these urban–rural differences in treatment and survival.

What is already known on this subject

  • Literature on urban–rural disparity in non-small cell lung cancer survival is mixed. Exploring this issue in Ireland is of interest because of the country's high rural population in relation to other developed countries.

What this study adds

  • This study contributes to the literature that cancer-directed surgery is more common among urban versus rural residents with lung cancer. It also adds to the evidence that, despite issues of access, rural residence is positively related to relative survival.


The authors are grateful for statistical advice from Eamonn O'Leary and Chris Brown at the National Cancer Registry Ireland, and Susanna Cramb and Paramita Dasgupta at Cancer Council Queensland. Data and geographic information system support was provided by Sandra Deady and Neil McCluskey at the National Cancer Registry Ireland. They gratefully acknowledge Pamela Gallagher at Dublin City University and Aileen Timmons at the National Cancer Registry Ireland for input at early stages of the project. They are also grateful to the tumour registration officers and data team at the National Cancer Registry Ireland who collected and processed the data and the General Register Office who provide death certificates to the National Cancer Registry for linkage to cancer registrations.


View Abstract


  • Contributors All authors contributed to the study concepts, data analysis and interpretation, manuscript editing and manuscript review. AAT and LS contributed to data acquisition and quality control of data. In addition to AAT and LS, MM contributed to the study design. AAT conducted the statistical analysis and prepared the manuscript.

  • Funding This study was supported by a Health Research Board, Ireland, Interdisciplinary Capacity Enhancement Award (ICE/2012/09).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.