- Email: [email protected]
Neighborhood Deprivation and Lung Cancer Incidence and Mortality: A Multilevel Analysis from Sweden
original article
Neighborhood Deprivation and Lung Cancer Incidence and Mortality A Multilevel Analysis from Sweden Xinjun Li, MD, PhD,* Jan Sundquist, MD, PhD,*† Bengt Zöller, MD, PhD* and Kristina Sundquist, MD, PhD*†
Background: Neighborhood deprivation has been implicated in lung cancer but no study has simultaneously analyzed the potential effect of neighborhood deprivation on both lung cancer incidence and mortality, after adjusting for individual-level socioeconomic factors, and comorbidities. The aim of this study was to analyze whether there is an association between neighborhood deprivation and incidence and mortality rates of lung cancer, beyond individual-level characteristics. Design: The incident and mortality cases of lung cancer were determined in the entire Swedish population aged over 50 (3.2 million individuals) between 2000 and 2010. Multilevel logistic regression was used in the analysis with individual-level characteristics (age, marital status, family income, education, immigration status, urban/ rural status, mobility, and comorbidities) at the first level and level of neighborhood deprivation at the second level. A neighborhood deprivation index, constructed from the variables education, income, unemployment, and welfare assistance, was used to assess the level of neighborhood deprivation. Results: There was a strong association between level of neighborhood deprivation and incidence and mortality of lung cancer. In the fully adjusted model, the odds of lung cancer were 1.27 and 1.32, respectively, in the most deprived neighborhood. The betweenneighborhood variance (i.e., the random intercept) was over 1.96 times the standard error in all models, indicating that there were significant differences in incidence and mortality rates of lung cancer between neighborhoods. Conclusions: Results suggest that neighborhood deprivation is associated with incident and mortality cases of lung cancer in Sweden, independently of individual-level characteristics. Key Words: Neighborhood deprivation, Lung cancer, Risk factors, Sweden. (J Thorac Oncol. 2015;10: 256–263) *Center for Primary Health Care Research, Lund University/Region Skåne, Malmö, Sweden; and †Stanford Prevention Research Center, Stanford University School of Medicine, Stanford, CA. Disclosure: The authors declare no conflict of interest. Author for Correspondence: Xinjun Li, MD, PhD, Center for Primary Health Care Research, Lund University, Jan Waldenströms gata 35, Skåne University Hospital, 205 02 Malmö, Sweden. Email: [email protected] DOI: 10.1097/JTO.0000000000000417 Copyright © 2014 by the International Association for the Study of Lung Cancer ISSN: 1556-0864/15/1002-0256
256
L
ung cancer is considered to be one of the major public health challenges, as the most common cancer in terms of both incidence and mortality worldwide. Many risks factors for lung cancer are known, such as tobacco smoking, air pollution, and family history. Studies have also shown associations between individual-level socioeconomic status (SES) and lung cancer risk and survival.1–8 During the first decade of this millennium an increasing number of studies have described the separate influences of individual- and neighborhood-level SES on health.9–13 The concept that disease determinants are in part environmental was described by Durkheim over a century ago. He stated that a population is more than only the sum of all individuals.14 This concept was further developed in Rose’s idea of the importance of distinguishing between the causes of individual cases of disease within a population, and the causes of differences in the rates of disease across populations.15 For example, if people within the same neighborhood share the same socioeconomic environment, access to healthcare resources, norms settings, and lifestyles, they may shape a common level of cardiovascular health beyond individual characteristics. Thus, in addition to individual-level sociodemographic factors, neighborhoodlevel factors may also increase the risk of disease. However, only a few studies have documented the potential effects of neighborhood-level SES on lung cancer risk.16–19 In a study from Taiwan between 2002 and 2007, lung cancer survival was significantly associated with low income but only among individuals younger than 65 years.14 Neighborhood deprivation, defined by income, had, however, no significant influence on lung cancer survival when adjusting for individual income. In a Danish study between 2004 and 2008, lung cancer incidence was higher in neighborhoods with high unemployment rates, independently of individual-level socioeconomic factors.15 A Swiss study conducted between 2001 and 2008 showed hazard ratios of 1.49 of lung cancer mortality in neighborhood with the lowest socioeconomic position compared with the highest position.18 The results were adjusted for individual-level socioeconomic factors. Moreover, a study from the US conducted between 1995 and 2007 showed that a decreased survival of non–small-cell lung cancer is associated with neighborhood deprivation.19 However, some studies have found no association between lung cancer survival and individual socioeconomic factors.20–22 Moreover, a Swedish
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
study conducted between 1990 and 2004 found no association between neighborhood deprivation and lung cancer mortality.23 To the best of our knowledge, no study to date has simultaneously analyzed the effect of neighborhood-level SES on both lung cancer incidence and mortality, after adjusting for individual-level socioeconomic factors and comorbidities, such as chronic pulmonary disease (COPD) and alcoholism. The first aim of this study was to investigate whether there is an association between neighborhood deprivation and incident and mortality rates of lung cancer between 2000 and 2010. The second aim was to investigate whether this possible difference remains after accounting for individual-level socio demographic characteristics, i.e., age, marital status, family income, education, immigration status, urban/rural status, mobility, and comorbidities (chronic obstructive pulmonary disease [COPD], tobacco abuse, and alcoholism or alcoholrelated liver disease).
MATERIALS AND METHODS The data sources were several national Swedish data registers, such as the Swedish National Population and Housing Census (1960–1990), the Total Population Register, the MultiGeneration register, the Swedish Cancer Registry (1958– 2010), the Swedish Hospital Discharge Register (1964–2010), and the Swedish Out-patient Register (2001–2010), provided to us by Statistics Sweden and the National Board of Health and Welfare. We used the primary diagnoses for lung cancer in the Swedish Cancer Register. Additional linkages were carried out to national census data to obtain individual-level SES, occupation, geographical region of residence, to the National Registry of Causes of Death (1961–2010) to identify date and cause of death, and to the Immigration Registry to identify date of emigration. All linkages were performed by the use of an individual national identification number that is assigned to each person in Sweden for their lifetime. This number was replaced by a serial number for each person to provide anonymity. The study period started on January 1, 2000 and proceeded until first incident of lung cancer, mortality of lung cancer, death from any other cause, emigration or the end of the study period on December 31, 2010.
Outcome (Dependent) Variable The outcome (dependent) variable was incident (yes/ no) and mortality (yes/no) cases from lung cancer. The unit of observation was individuals. We used the Swedish Cancer Registry to identify primary diagnoses of lung cancer in the study population during the study period. We then linked this information with records in the Cause of Death Register to identify deaths among lung cancer patients during the same period. All cases of cancer in Sweden must be registered in the Swedish Cancer Registry. The completeness of cancer registration is currently considered to be close to 100%. Only primary neoplasms of the lung classified according to the 7th revision of the International Classification of Diseases (ICD-7) (The Swedish Cancer Registry has transferred all the cancer ICD codes into ICD7) (codes 162, 163) were studied.
Neighborhood deprivation and lung cancer
The 10th revisions of the International Classification of Diseases (ICD-10 C33 to C34) were used to define the outcome variable of mortality due to lung cancer in the Cause of Death register.
Independent Variables Independent variables included sex, age at the start of the study, marital status, family income, educational attainment, immigration status, geographical region, mobility, diagnosis of COPDs, tobacco abuse, and alcoholism or alcohol-related liver disease of the subjects. Sex. Men and women. Age. Age was greater than or equal to 50 years and was divided into 10-year category. Marital status. Individuals were classified as married/ cohabitating or single. Family income by quartile. Information on family income in 2000 came from the Total Population Register, which was provided by Statistics Sweden. We used this information to determine the distribution of family incomes in Sweden, and then used the distribution to calculate empirical quartiles. Educational attainment. Educational attainment was classified as completion of compulsory school or less (less than or equal to 9 years), practical high school or some theoretical high school (10–12 years), or theoretical high school and/or college (greater than 12 years). Immigration status. (1) Born in Sweden and (2) Born Outside Sweden. Urban/rural status. Individuals were classified as living in a large city, a middle-sized town, or a small town/rural area. This variable was included because urban/rural status may be associated with access to care. Large cities were those with a population of greater than or equal to 200,000 (Stockholm, Gothenburg and Malmö). Middle-sized towns were towns with a population of greater than or equal to 90,000 but less than 200,000. Small towns were towns with a population of greater than or equal to 27,000 and less than 90,000; rural areas were areas with populations smaller than those of small towns. Mobility. Length of time lived in the neighborhood, categorized as lived in the neighborhood less than 5 years or greater than 5 years. Comorbidities. COPD and tobacco abuse: Patients’ previous diagnosis of COPD, which was suspected to be one important prognostic factor for lung cancer and used as a surrogate of smoking, was identified in the Hospital Registry 10 years before the follow-up period and Out-patient Register accordingly (COPD: ICD-91990–1996 = 490–496; ICD-10 1997– 2010 = J40–J49. The same approach was used to identify tobacco abuse: ICD-9 = 305.1, 292.0, 292.1, 292.2, 292.8, 292.9, V15.8, V65.3, V65.8; ICD-10 = F17, T65.2, Z71.6, Z72.0). Patients’ COPD and tobacco abuse in the patient Registries was individually linked to their lung cancer status using a serial number. Alcoholism and alcohol-related liver disease was identified in the Hospital Registry and Out-patient Register according to the International Classification of Diseases codes (ICD-9 = 291, 303, 571; ICD-10 = F10 and K70).
Copyright © 2014 by the International Association for the Study of Lung Cancer
257
Li et al.
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Neighborhood deprivation index. A summary measure was used to characterize neighborhood-level deprivation. We identified deprivation indicators used by past studies to characterize neighborhood environments and then used a principal components analysis to select deprivation indicators in the Swedish national database. The following four variables were selected for those aged 25–74: low educational status (less than 10 years of formal education), low income (income from all sources, including that from interest and dividends, defined as less than 50% of individual median income)24, unemployment (not employed, excluding full-time students, those completing compulsory military service, and early retirees), and social welfare assistance. Each of the four variables loaded on the first principal component with similar loadings (+0.47 to +0.53) and explained 52% of the variation between these variables. A z score was calculated for each small area market statistics neigh borhood. Small area market statistics are geographic units covering all Sweden with homogenous types of buildings. The z scores, weighted by the coefficients for the eigenvectors, were then summed to create the index.25 The index was categorized into three groups: below one standard deviation (SD) from the mean (low deprivation), above one SD from the mean (high deprivation), and within one SD of the mean (moderate deprivation). Higher scores reflect more deprived neighborhoods.
Statistical Analysis Age-standardized incidence and mortality rates were calculated by direct age standardization using 10-year age groups specific to women or men, with the entire Swedish population of women or men in 2000 as the standard population. Multilevel (hierarchical) logistic regression models were calculated based on incidence proportions or mortality proportions (cumulative incidence or cumulative mortality) of lung cancer (yes/no). The analyses were performed using MLwiN, version 2.27. First, a neighborhood model was calculated that included only neighborhood-level deprivation to determine the crude odds of lung cancer incidence and mortality by level of neighborhood deprivation (model 1). Next, a model that included neighborhood-level deprivation and sex, age (model 2) and the rest of the individual-level sociodemographic variables, added simultaneously to the model (model 3), was calculated. Finally, a full model was calculated that also included comorbidities (model 4). These full models tested whether neighborhood-level deprivation was significantly associated with lung cancer incidence and mortality after adjusting for the sociodemographic characteristics and comorbidities, and whether there were differential effects of neighborhood-level deprivation on lung cancer incidence and mortality across sociodemographic characteristics.26 To check whether collinearity was a problem, we examined the correlation between the included variables. There was a low degree of correlation between the included variables in the models (0.4 or lower). The correlation between the neighborhood variables and the other socioeconomic factors was less than 0.3. The highest correlation was observed between individual income and individual education (r = 0.4).
258
Random effects. The between-neighborhood variance was estimated both with and without a random intercept. It was regarded as significant if it was larger than 1.96 times the standard error, indicating that there were significant differences in incident and mortality rates of lung cancer between neighborhoods. This is in accord with the precedent set in previous studies.12,27,28 The choice of 1.96 was based on the normal curve where 95% of the area under a normal curve lies within 1.96 SDs of the mean.29 For comparison, we also calculated Cox regression models and logistic regression models using the SAS statistical package (version 9.3; SAS Institute, Cary, NC).
Ethical Considerations This study was approved by the Ethics Committee of Lund University, Sweden.
RESULTS Table 1 shows population sizes and characteristics in the year 2000 by neighborhood-level deprivation. The total number of neighborhoods was 8361. Of the total population, 23%, 61%, and 16% lived in low, moderate, and high-deprivation neighborhoods, respectively. During the study period from January 1, 2000 to December 31, 2010, 33,704 (1.7%) of individuals were diagnosed with lung cancer, and 31,287 died due to lung cancer. Age-adjusted lung cancer incidence rates increased from 9.2 per 1000 in neighborhoods with low deprivation to 10.4 per 1000 in neighborhoods with moderate deprivation and 13.3 per 1000 in neighborhoods with high deprivation. The mortality rate of lung cancer was 8.1, 9.6, and 13.0 per 1000, respectively. There were more men than women with lung cancer in the study population. A similar pattern of higher incidence and mortality rates with each increasing level of neighborhood-level deprivation was observed across all individual-level sociodemographic categories and comorbidities. All categories showed a gradient effect across level of neighborhood deprivation. The odds ratio (OR) of lung cancer incidence for individuals living in a high versus low deprivation neighborhood was 1.45 (1.39–1.50) in the crude neighborhood-level model (Table 2). Neighborhood-level deprivation remained significantly associated with lung cancer incidence after adjusting for the individual-level sociodemographic variables and comorbidities (OR = 1.27; 95% confidence interval [CI]: 1.22–1.32). The highest odds of lung cancer incidence were found for individuals who were men, immigrants, had the lowest educational attainments, high mobility, or were affected with comorbidities. Individuals living in middle-sized towns or in small towns and rural areas had lower odds of lung cancer morbidity. Interestingly, individuals with middle-level family income had higher ORs. The OR of mortality due to lung cancer for individuals living in high versus low deprivation neighborhoods was 1.60 (1.54–1.67) in the crude neighborhood-level model (Table 3). Neighborhood-level deprivation remained significantly associated with lung cancer mortality after adjusting for the individual-level sociodemographic variables and comorbidities (OR = 1.32; 95% CI: 1.27–1.38). The highest
Copyright © 2014 by the International Association for the Study of Lung Cancer
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Neighborhood deprivation and lung cancer
TABLE 1. Distribution of Population, Number of Lung Cancer Events, and Age-Standardized Incidence (per 1000) and Mortality by Neighborhood-Level Deprivation Lung Cancer Events
Population No. Total population (%)
(%)
Lung Cancer Events No.
%
Incidence Rates by Neighborhood Deprivation Low
Moderate
High
33,704
721,772 (23%) 9.2
1,957,595 (61%) 10.4
510,599 (16%) 13.3
Low
Moderate
High
1,957,595 (61%) 9.6
510,599 (16%) 13.0
46.5 18,614 55.2 53.5 15,090 44.8
10.6 8.0
12.3 8.8
16.0 11.2
17,256 55.2 14,031 44.8
9.5 7.0
11.7 8.0
15.6 11.0
38.5 9523 28.3 25.6 12,010 35.6 21.8 9945 29.5 12.1 2163 6.4 2.0 63 0.2
6.2 12.0 14.3 6.6 1.3
7.7 14.7 13.6 5.4 0.9
10.9 18.6 16.6 5.5 0.8
7538 10,138 10,264 3199 148
24.1 32.4 32.8 10.2 0.5
4.6 9.6 13.8 9.5 3.5
6.0 12.3 14.2 7.9 2.1
9.3 16.6 17.5 8.8 1.9
25.0 25.0 25.0 25.0
23.7 29.6 26.6 20.2
10.1 9.9 9.9 8.2
10.1 11.8 11.1 8.9
13.0 14.7 13.7 11.8
8224 9756 7839 5468
26.3 31.2 25.1 17.5
9.4 9.0 8.6 7.0
9.7 11.2 10.3 7.9
12.5 14.7 13.4 11.0
57.6 18,931 56.2 42.4 14,773 43.8
9.0 10.0
10.1 11.4
12.5 14.4
16,755 53.6 14,532 46.4
7.8 9.1
9.2 10.8
12.0 14.2
88.9 29,053 86.2 11.1 4651 13.8
9.1 9.9
10.1 12.7
13.0 14.7
27,059 86.5 4228 13.5
8.0 8.8
9.4 11.9
12.7 14.4
54.3 19,983 59.3 29.2 10,228 30.3 16.5 3493 10.4
11.4 9.2 6.4
11.6 10.0 7.2
14.1 12.9 9.4
20,035 64.0 8473 27.1 2779 8.9
10.2 7.4 5.0
10.8 8.6 6.2
13.7 12.2 8.0
47.7 17,494 51.9 34.9 11,474 34.0 17.4 4736 14.1
9.6 9.0 6.8
11.6 10.0 8.5
15.1 13.1 9.5
16,099 51.5 10,722 34.3 4466 14.3
8.4 7.9 6.3
10.6 9.4 7.9
14.8 12.4 9.6
85.1 28,360 84.1 14.9 5344 15.9
9.2 8.9
10.2 11.5
13.0 14.7
26,395 84.4 4892 15.6
8.1 7.9
9.4 10.7
12.7 13.9
93.4 26,528 78.7 6.6 7176 21.3
7.9 30.2
8.8 32.4
11.0 39.9
24,909 79.6 6378 20.4
7.0 23.5
8.3 27.2
10.8 36.7
97.6 32,260 95.7 2.4 1444 4.3
9.0 15.0
10.2 16.5
13.0 20.6
29,948 95.7 1339 4.3
7.9 14.8
9.5 15.9
12.6 20.7
99.9 33,656 99.9 0.1 48 0.1
9.1 20.2
10.4 25.9
13.3 22.8
31,254 99.9 33 0.1
8.1 14.3
9.6 17.6
13.0 15.2
7989 9965 8958 6792
Copyright © 2014 by the International Association for the Study of Lung Cancer
No.
%
Morality Rate of Lung Cancer by Neighborhood Deprivation
721,772 (23%) 8.1
3,189,966
Total lung cancer events Sex Male 1,484,372 Female 1,705,594 Age (years) 50–59 1,227,614 60–69 816,845 70–79 696,612 80–89 384,662 ≥90 64,233 Family income Low income 797,795 Middle–low income 797,963 Middle–high income 797,899 High income 796,309 Marital status Married/cohabiting 1,837,019 Never married, Widowed, 1,352,947 or divorced Immigrant status Sweden 2,837,342 Other countries 352,624 Educational attainment ≤9 years 1,733,209 10–12 years 931,641 >12 years 525,116 Urban/rural status Large cities 1,521,470 Middle-sized towns 1,113,363 Small towns/rural areas 555,133 Move Not moved 2,715,639 Moved 474,327 Chronic lower respiratory disease No 2,978,654 Yes 211,312 Alcoholism and related liver disease No 3,111,900 Yes 78,066 Tobacco abuse No 3,188,197 Yes 1769
Mortality of Lung Cancer Mortality of Lung Cancer
31,287
259
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Li et al.
TABLE 2. OR and 95% CI for Childhood Obesity; Results of Multilevel Logistic Regression Models Model 1 OR Neighborhood-level variable (ref. low) Moderate High Age Sex, men (ref. women) Family income (ref. high income) Middle–high income Middle–low income Low income Marital status (ref. married/cohabiting) Never married, widowed, or divorced Immigrant status (ref. born in Sweden) Education attainment (ref. > 12 years) ≤9 years 10–12 years Urban/rural status (ref. large cities) Middle-sized towns Small towns/rural areas Mobility (ref. not moved) Chronic lower respiratory disease (ref. No) Alcoholism and related liver disease (ref. No) Tobacco abuse (ref. No) Variance (SE) Explained variance (%)
1.13 1.45
Model 2
95% CI 1.10 1.39
1.17 1.50
OR 1.13 1.44 1.00 1.44
0.057 (0.004) 25
Model 3
95% CI 1.09 1.38 1.00 1.41
1.16 1.50 1.00 1.48
0.060 (0.004) 21
OR
Model 4
95% CI
OR
95% CI
p Value
1.08 1.31 1.00 1.46
1.05 1.26 1.00 1.43
1.12 1.36 1.00 1.49
1.08 1.27 1.00 1.44
1.04 1.22 1.00 1.41
1.11 1.32 1.00 1.47
<0.001 <0.001 0.003 <0.001
1.23 1.32 1.04
1.19 1.27 1.00
1.27 1.36 1.07
1.19 1.24 1.00
1.15 1.20 0.96
1.23 1.28 1.04
<0.001 <0.001 0.920
1.04 1.20
1.02 1.16
1.06 1.24
1.01 1.18
0.98 1.14
1.03 1.22
0.689 <0.001
1.66 1.57
1.59 1.51
1.72 1.63
1.59 1.51
1.53 1.46
1.66 1.57
<0.001 <0.001
0.87 0.72 1.05
0.85 0.69 1.02
0.90 0.74 1.09
0.90 0.74 1.03 3.69 1.30 1.76
0.88 0.93 0.72 0.77 1.00 1.06 3.59 3.79 1.23 1.38 1.32 2.35 0.034 (0.004) 58
<0.001 <0.001 0.057 <0.001 <0.001 <0.001
0.039(0.004) 50
Model 1: crude model. Model 2: adjusted for age and sex. Model 3: adjusted for age, sex, family income, marital status, country of birth, education, region of residence, and mobility. Model 4: adjusted for age, sex, family income, marital status, country of birth, education, region of residence, mobility, and comorbidities. OR, odds ratio; CI, confidence interval.
odds of mortality were found for individuals who were men, never married, widowed, or divorced, immigrants, those who had a middle-level family income, had the lowest educational attainment, or were affected with comorbidities. Individuals living in middle-sized towns or in small towns and rural areas had decreased odds of mortality. We performed an additional analysis using logistic regression models and the results were almost identical. In the full model, the OR was 1.28 (95% CI: 1.23–1.32) for lung cancer incidence and 1.33 (95% CI: 1.28–1.38) for mortality due to lung cancer for individuals living in the most deprived neighborhoods compared with those living in the least deprived neighborhoods (Supplementary Tables 1 and 2, Supplementary Digital Content, http://links.lww.com/JTO/ A747). We also performed an additional analysis using Cox regression models. In the full model, the hazard ratios were 1.29 (95% CI: 1.25–1.34) for lung cancer incidence and 1.35 (95% CI: 1.30–1.40) for mortality due to lung cancer among individuals living in the most deprived neighborhoods compared with those living in the least deprived neighborhoods (Supplementary Tables 3 and 4, Supplementary Digital Content, http://links.lww.com/JTO/A747). The test for cross-level interactions between the individual sociodemographic variables and neighborhood deprivation
260
on odds of lung cancer incidence and mortality showed no meaningful cross-level interactions or effect modification. The between-neighborhood variance (i.e., the random intercept) was over 1.96 times the standard error in all models, indicating that there were significant differences in incident and mortality rates of lung cancer between neighborhoods, after accounting for the sociodemographic and comorbidity variables. The neighborhood-level variable explained 25% of the incidence and 33% of the mortality in the between-neighborhood variance estimates (Tables 2, 23). After inclusion of all the independent variables, the explained variance was 58% and 66%, respectively.
DISCUSSION The main findings of this study are that the odds of both incident and mortality rates of lung cancer are higher among individuals living in deprived neighborhoods than among individuals living in affluent neighborhoods. This difference remained significant, after adjustment for the individuallevel sociodemographic variables and comorbidities (COPD, tobacco abuse, and alcoholism). This study represents a novel contribution to previous neighborhood studies on incident16,18 and mortality17,19,23 cases of lung cancer. No previous neighborhood study has, however, studied both incident and
Copyright © 2014 by the International Association for the Study of Lung Cancer
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Neighborhood deprivation and lung cancer
TABLE 3. OR and 95% CI for Mortality of Lung Cancer; Results of Multilevel Logistic Regression Models Model 1 OR Neighborhood-level variable (ref. low) Moderate High Age Sex, men (ref. women) Family income (ref. high income) Middle–high income Middle–low income Low income Marital status (ref. married/co-habiting) Never married, widowed, or divorced Immigrant status (ref. born in Sweden) Education attainment (ref. > 12 years) ≤9 years 10–12 years Urban/rural status (ref. large cities) Middle-sized towns Small towns/rural areas Mobility (ref. not moved) Chronic lower respiratory disease (ref. No) Alcoholism and related liver disease (ref. No) Tobacco abuse (ref. No) Variance (SE) Explained variance (%)
1.19 1.60
Model 2
95% CI 1.15 1.54
1.23 1.67
OR 1.16 1.55 1.02 1.49
0.057 (0.005) 33
Model 3
95% CI 1.12 1.49 1.01 1.45
1.20 1.62 1.02 1.52
0.057 (0.005) 33
OR
Model 4
95% CI
OR
95% CI
p Value
1.09 1.36 1.01 1.53
1.06 1.31 1.01 1.49
1.13 1.42 1.01 1.56
1.09 1.32 1.01 1.49
1.05 1.27 1.01 1.46
1.12 1.38 1.01 1.53
<0.001 <0.001 <0.001 <0.001
1.30 1.44 1.16
1.25 1.39 1.12
1.35 1.49 1.21
1.26 1.36 1.12
1.22 1.31 1.08
1.31 1.41 1.17
<0.001 <0.001 <0.001
1.10 1.20
1.08 1.16
1.13 1.24
1.07 1.18
1.04 1.14
1.09 1.22
<0.001 <0.001
1.75 1.57
1.67 1.51
1.83 1.64
1.68 1.52
1.60 1.46
1.75 1.59
<0.001 <0.001
0.88 0.73 1.05
0.86 0.70 1.02
0.91 0.76 1.09
0.91 0.75 1.03 3.31 1.38 1.41
0.88 0.93 0.73 0.78 1.00 1.06 3.22 3.41 1.30 1.46 1.00 2.00 0.030 (0.004) 66
<0.001 <0.001 0.072 <0.001 <0.001 0.057
0.035(0.004) 59
Model 1: crude model. Model 2: adjusted for age and sex. Model 3: adjusted for age, sex, family income, marital status, country of birth, education, region of residence, and mobility. Model 4: adjusted for age, sex, family income, marital status, country of birth, education, region of residence, mobility, and comorbidities. OR, odds ratio; CI, confidence interval.
mortality cases of lung cancer.16–19,23 In addition, to the best of our knowledge, no previous large-scale study has adjusted for smoking. This study was also unable to do so, although we adjusted for COPD, tobacco abuse, and alcoholism, in an attempt to adjust for potential surrogates for smoking. One previous Swedish study was based on data between 1990 and 2004 and it only included mortality cases and not incident cases of lung cancer. This study covers a more recent time period (2000–2010), and it shows that neighborhood deprivation, even after adjustment for individual-level characteristics and comorbidities, is associated with incident and mortality cases of lung cancer. Individual-level sociodemographic measures of socioeconomic inequality5,7,30 and education1,4,6 have been reported to be associated with lung cancer. However, the causal pathways between neighborhood socioeconomic deprivation and poor health outcomes are not fully understood; several possible mechanisms could thus lie behind our findings. One possible mediator could be psychological stress31,32 due to littered and unsafe environments, vandalism, isolation/alienation, and violent crime33 in deprived neighborhoods. Psychological stress could be harmful in itself as well as lead to harmful smoking habits. Data suggest that hyperactivity of the sympathetic branch of the autonomic nervous system caused by
chronic psychological stress or chronic exposure to nicotinic agonists in cigarette smoke may contribute to the development and progression of non–small-lung cancer.34 A clinical study has also reported improved survival outcomes with the incidental use of β-blockers among patients with non–small-lung cancer,35 which suppresses the sympathetic branch of the autonomic nervous system. It is possible that the lack of safe environments reduces the possibility to exercise, which aggravates a healthy life style. In addition socio-cultural norms regarding smoking and physical activity could vary between neighborhoods and affect the health of the residents and the risk for lung cancer. For example, studies show that environmental risk factors, such as neighborhood income,16 neighborhood education,19 and neighborhood unemployment17 are associated with lung cancer. Studies have also found associations between neighborhood deprivation and exposure to air pollution,36 smoking,37,38 and increased lung cancer mortality.39 Living in deprived neighborhoods can cause isolation from health-promoting milieus (e.g., safe places to exercise and decent housing) and services. In comparisons of wealthy nations, associations between neighborhood characteristics and different health outcomes were inconsistent.40 This implies that neighborhood determinants of health are complex. Such
Copyright © 2014 by the International Association for the Study of Lung Cancer
261
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Li et al.
determinants may include access to health care, education, and social services. Access to these services is uneven in the US, where the effects of income inequalities on health are more pronounced.41 For example, higher primary health care provider density was associated with lower risks of lung cancer mortality.42 Neighborhood-level inequities include unequal access to and quality of primary and secondary health care services.43 In Sweden, medical care is provided to all permanent residents, and primary health care clinics and hospitals are equally distributed and located centrally in all types of neighborhoods.43 However, the actual number of health care professionals working in primary health care clinics can vary considerably by neighborhood type. This is due to difficulties in recruiting and retaining health care personnel in high-deprivation neighborhoods.19,44 The uneven distribution of medical personnel across neighborhoods has also been documented in UK,45 another country with universal health care. Our study has some limitations. The most important one is the lack of smoking data; smoking might be a mediator between neighborhood deprivation and lung cancer. However, we adjusted for COPD, tobacco abuse, and alcoholism in an attempt to adjust for potential surrogates for smoking. We also adjusted for certain individual-level sociodemographic variables, such as income, education, marital status, and country of birth, which are strongly related to smoking, to minimize this potential confounder. Despite this, residual confounding is likely to exist. Previously published neighborhood studies were, however, also unable to adjust for smoking.16–19,23 Those studies did, however, not adjust for COPD, tobacco abuse, and alcoholism. In studies of neighborhood effects on health, selective residential mobility can cause compositional neighborhood differences. Selective residential mobility is the tendency of individuals to move to neighborhoods whose characteristics match their individual characteristics (for example, the tendency of individuals with low SES to move to low SES neighborhoods). However, we adjusted for family income, which improved our possibilities to differentiate between compositional and contextual effects on lung cancer incidence and mortality. There are a number of strengths in this study. The large cohort included practically all patients with lung cancer (50 years and older) in Sweden during the study period. This is due to the high coverage and validity of the Swedish cancer registry,46,47 which increases the generalizability of our results. Another strength is the use of personal identification numbers, which made it possible to include individuals in different registers, such as the Immigration Register, which permitted calculation of exact risk times.48,49 Second, the Swedish Total Population Register is highly complete, with very few missing data. For example, data for income were 99.4% complete.50 Finally, the use of multilevel modeling helped us to separate neighborhood- and individual-level effects and allowed us to consider both fixed and random effects in the analyses.
CONCLUSIONS Neighborhood deprivation contributes to lung cancer incidence and mortality. The individual- and
262
neighborhood-level variables cumulatively load against individuals so that the most at-risk individuals would be those who have both individual- and neighborhood-level risk factors. These findings raise important clinical and public health concerns, and indicate that both individual- and neighborhood-level approaches are important in health care policies.
Acknowledgements
The authors thank Science Editor Stephen Gilliver for his useful comments on the text. This study was supported by ALF funding awarded to Jan and Kristina Sundquist and by grants from the Swedish Research Council (awarded to Kristina Sundquist), the Swedish Freemasons Foundation (awarded to Jan Sundquist), and the Swedish Heart-Lung Foundation (Bengt Zöller). The registers used in this study are maintained by Statistics Sweden and the National Board of Health and Welfare. REFERENCES 1. Albano JD, Ward E, Jemal A, et al. Cancer mortality in the United States by education level and race. J Natl Cancer Inst 2007;99:1384–1394. 2. Clegg LX, Li FP, Hankey BF, Chu K, Edwards BK. Cancer survival among US whites and minorities: A SEER (surveillance, epidemiology, and end results) program population-based study. Arch Intern Med 2002;162:1985–1993. 3. Du XL, Lin CC, Johnson NJ et al. Effects of individual-level socioeconomic factors on racial disparities in cancer treatment and survival: Findings from the National Longitudinal Mortality Study, 1979–2003. Cancer 2011;117:3242–3251. 4. Hemminki K, Li X. Level of education and the risk of cancer in Sweden. Cancer Epidemiol Biomarkers Prev 2003;12:796–802. 5. Hemminki K, Zhang H, Czene K. Socioeconomic factors in cancer in Sweden. Int J Cancer 2003;105:692–700. 6. Hussain SK, Lenner P, Sundquist J, Hemminki K. Influence of education level on cancer survival in Sweden. Ann Oncol 2008;19:156–162. 7. Mackillop WJ, Zhang-Salomons J, Groome PA, Paszat L, Holowaty E. Socioeconomic status and cancer survival in Ontario. J Clin Oncol 1997;15:1680–1689. 8. Vinnakota S, Lam NS. Socioeconomic inequality of cancer mortality in the United States: A spatial data mining approach. Int J Health Geogr 2006;5:9. 9. Diez Roux AV, Merkin SS, Arnett D, et al. Neighborhood of residence and incidence of coronary heart disease. N Engl J Med 2001;345:99–106. 10. Martikainen P, Kauppinen TM, Valkonen T. Effects of the characteristics of neighbourhoods and the characteristics of people on cause specific mortality: A register based follow up study of 252,000 men. J Epidemiol Community Health 2003;57:210–217. 11. Sampson RJ, Raudenbush SW, Earls F. Neighborhoods and violent crime: A multilevel study of collective efficacy. Science 1997;277:918–924. 12. Sundquist K, Ahlen H. Neighbourhood income and mental health: A multilevel follow-up study of psychiatric hospital admissions among 4.5 million women and men. Health Place 2006;12:594–602. 13. Winkleby M, Sundquist K, Cubbin C. Inequities in CHD inci dence and case fatality by neighborhood deprivation. Am J Prev Med 2007;32:97–106. 14. Durkheim E. The Rules of Sociological Method. New York: Free Press, 1982. 15. Rose GA. Individuals and Populations. The Strategy of Preventive Medicine, Oxford University Press: Oxford, 1992. 16. Chang CM, Su YC, Lai NS, et al. The combined effect of individual and neighborhood socioeconomic status on cancer survival rates. PLoS One 2012;7:e44325. 17. Meijer M, Bloomfield K, Engholm G. Neighbourhoods matter too: The association between neighbourhood socioeconomic position, population density and breast, prostate and lung cancer incidence in Denmark between 2004 and 2008. J Epidemiol Community Health 2013;67:6–13.
Copyright © 2014 by the International Association for the Study of Lung Cancer
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
18. Panczak R, Galobardes B, Voorpostel M, Spoerri A, Zwahlen M, Egger M; Swiss National Cohort and Swiss Household Panel. A Swiss neighbourhood index of socioeconomic position: Development and association with mortality. J Epidemiol Community Health 2012;66:1129–1136. 19. Erhunmwunsee L, Joshi MB, Conlon DH, Harpole DH Jr. Neighborhoodlevel socioeconomic determinants impact outcomes in nonsmall cell lung cancer patients in the Southeastern United States. Cancer 2012;118:5117–5123. 20. Bouchardy C, Fioretta G, De Perrot M, Obradovic M, Spiliopoulos A. Determinants of long term survival after surgery for cancer of the lung: A population-based study. Cancer 1999;86:2229–2237. 21. Tammemagi CM, Neslund-Dudas C, Simoff M, Kvale P. In lung cancer patients, age, race-ethnicity, gender and smoking predict adverse comorbidity, which in turn predicts treatment and survival. J Clin Epidemiol 2004;57:597–609. 22. Stavraky KM, Skillings JR, Stitt LW, Gwadry-Sridhar F. The effect of socioeconomic status on the long-term outcome of cancer. J Clin Epidemiol 1996;49:1155–1160. 23. Sundquist J, Li X, Sundquist K. Neighborhood deprivation and mortality in individuals with cancer: A multilevel analysis from Sweden. Eur J Cancer Prev 2012;21:387–394. 24. EurostatYearbook 2004. http://www.scb.se/templates/Standard____36500. asp. Accessed December 2014. 25. Gilthorpe MS. The importance of normalisation in the construction of deprivation indices. J Epidemiol Community Health 1995;49:S45–S50. 26. Goldstein H. Multilevel Statistical Models, 3rd ed. London: Hodder Arnold Publishers, 2003. 27. Johnell K, Lindström M, Melander A, Sundquist J, Eriksson C, Merlo J. Anxiolytic-hypnotic drug use associated with trust, social participation, and the miniaturization of community: A multilevel analysis. Soc Sci Med 2006;62:1205–1214. 28. Johnell K, Lindström M, Sundquist J, Eriksson C, Merlo J. Individual characteristics, area social participation, and primary non-concordance with medication: A multilevel analysis. BMC Public Health 2006;6:52. 29. Rees DG. Foundations of Statistics. New York: CRC Press, 1987. 30. Sharpe KH, McMahon AD, McClements P et al. Socioeconomic inequalities in incidence of lung and upper aero-digestive tract cancer by age, tumour subtype and sex: A population-based study in Scotland (2000– 2007). Cancer Epidemiol 2012;36:e164–e170. 31. Black PH, Garbutt LD. Stress, inflammation and cardiovascular disease. J Psychosom Res 2002;52:1–23. 32. Holden L, Scuffham P, Hilton M, Vecchio N, Whiteford H. Psychological distress is associated with a range of high-priority health conditions affecting working Australians. Aust N Z J Public Health 2010;34:304–310. 33. Sundquist K, Theobald H, Yang M, Li X, Johansson SE, Sundquist J. Neighborhood violent crime and unemployment increase the risk of coronary heart disease: A multilevel study in an urban setting. Soc Sci Med 2006;62:2061–2071.
Neighborhood deprivation and lung cancer
34. Schuller HM. Impact of neuro-psychological factors on smoking-associated lung cancer. Cancers (Basel) 2014;6:580–594. 35. Wang HM, Liao ZX, Komaki R, et al. Improved survival outcomes with the incidental use of beta-blockers among patients with non-smallcell lung cancer treated with definitive radiation therapy. Ann Oncol 2013;24:1312–1319. 36. Naess O, Piro FN, Nafstad P, Smith GD, Leyland AH. Air pollution, social deprivation, and mortality: A multilevel cohort study. Epidemiology 2007;18:686–694. 37. Hiscock R, Bauld L, Amos A, Platt S. Smoking and socioeconomic status in England: The rise of the never smoker and the disadvantaged smoker. J Public Health (Oxf) 2012;34:390–396. 38. Chuang YC, Cubbin C, Ahn D, Winkleby MA. Effects of neighbourhood socioeconomic status and convenience store concentration on individual level smoking. J Epidemiol Community Health 2005;59:568–573. 39. Hayanga AJ, Zeliadt SB, Backhus LM. Residential segregation and lung cancer mortality in the United States. JAMA Surg 2013;148:37–42. 40. Lynch J, Smith GD, Hillemeier M, Shaw M, Raghunathan T, Kaplan G. Income inequality, the psychosocial environment, and health: Comparisons of wealthy nations. Lancet 2001;358:194–200. 41. Lochner K, Pamuk E, Makuc D, Kennedy BP, Kawachi I. State-level income inequality and individual mortality risk: A prospective, multilevel study. Am J Public Health 2001;91:385–391. 42. Backhus LM, Hayanga AJ, Au D, Zeliadt SB. The effect of provider density on lung cancer survival among blacks and whites in the United States. J Thorac Oncol 2013;8:549–553. 43. van Doorslaer E, Wagstaff A, Bleichrodt H, et al. Income-related inequalities in health: Some international comparisons. J Health Econ 1997;16:93–112. 44. Hardy D, Chan W, Liu CC, et al. Racial disparities in the use of hospice services according to geographic residence and socioeconomic status in an elderly cohort with nonsmall cell lung cancer. Cancer 2011;117:1506–1515. 45. Taylor DH Jr, Leese B. General practitioner turnover and migration in England 1990–94. Br J Gen Pract 1998;48:1070–1072. 46. Barlow L, Westergren K, Holmberg L et al. The completeness of the Swedish Cancer Register: A sample survey for year 1998. Acta Oncol 2009;48:27–33. 47. Socialstyrelsen. Cancerförekomst i Sverige 2011. Stockholm: Socialstyrelsen, 2012. 48. Ludvigsson JF, Otterblad-Olausson P, Pettersson BU, Ekbom A. The Swedish personal identity number: Possibilities and pitfalls in healthcare and medical research. Eur J Epidemiol 2009;24:659–667. 49. Ludvigsson JF, Andersson E, Ekbom A, et al. External review and validation of the Swedish national inpatient register. BMC Public Health 2011;11:450. 50. Rosen M, Hakulinen T. Use of disease registers. In W Ahrens, I Pigeot (Eds.) Handbook of Epidemiology. Berlin: Springer-Verlag, 2005. Pp. 232–251.
Copyright © 2014 by the International Association for the Study of Lung Cancer
263
Neighborhood Deprivation and Lung Cancer Incidence and Mortality A Multilevel Analysis from Sweden Xinjun Li, MD, PhD,* Jan Sundquist, MD, PhD,*† Bengt Zöller, MD, PhD* and Kristina Sundquist, MD, PhD*†
Background: Neighborhood deprivation has been implicated in lung cancer but no study has simultaneously analyzed the potential effect of neighborhood deprivation on both lung cancer incidence and mortality, after adjusting for individual-level socioeconomic factors, and comorbidities. The aim of this study was to analyze whether there is an association between neighborhood deprivation and incidence and mortality rates of lung cancer, beyond individual-level characteristics. Design: The incident and mortality cases of lung cancer were determined in the entire Swedish population aged over 50 (3.2 million individuals) between 2000 and 2010. Multilevel logistic regression was used in the analysis with individual-level characteristics (age, marital status, family income, education, immigration status, urban/ rural status, mobility, and comorbidities) at the first level and level of neighborhood deprivation at the second level. A neighborhood deprivation index, constructed from the variables education, income, unemployment, and welfare assistance, was used to assess the level of neighborhood deprivation. Results: There was a strong association between level of neighborhood deprivation and incidence and mortality of lung cancer. In the fully adjusted model, the odds of lung cancer were 1.27 and 1.32, respectively, in the most deprived neighborhood. The betweenneighborhood variance (i.e., the random intercept) was over 1.96 times the standard error in all models, indicating that there were significant differences in incidence and mortality rates of lung cancer between neighborhoods. Conclusions: Results suggest that neighborhood deprivation is associated with incident and mortality cases of lung cancer in Sweden, independently of individual-level characteristics. Key Words: Neighborhood deprivation, Lung cancer, Risk factors, Sweden. (J Thorac Oncol. 2015;10: 256–263) *Center for Primary Health Care Research, Lund University/Region Skåne, Malmö, Sweden; and †Stanford Prevention Research Center, Stanford University School of Medicine, Stanford, CA. Disclosure: The authors declare no conflict of interest. Author for Correspondence: Xinjun Li, MD, PhD, Center for Primary Health Care Research, Lund University, Jan Waldenströms gata 35, Skåne University Hospital, 205 02 Malmö, Sweden. Email: [email protected] DOI: 10.1097/JTO.0000000000000417 Copyright © 2014 by the International Association for the Study of Lung Cancer ISSN: 1556-0864/15/1002-0256
256
L
ung cancer is considered to be one of the major public health challenges, as the most common cancer in terms of both incidence and mortality worldwide. Many risks factors for lung cancer are known, such as tobacco smoking, air pollution, and family history. Studies have also shown associations between individual-level socioeconomic status (SES) and lung cancer risk and survival.1–8 During the first decade of this millennium an increasing number of studies have described the separate influences of individual- and neighborhood-level SES on health.9–13 The concept that disease determinants are in part environmental was described by Durkheim over a century ago. He stated that a population is more than only the sum of all individuals.14 This concept was further developed in Rose’s idea of the importance of distinguishing between the causes of individual cases of disease within a population, and the causes of differences in the rates of disease across populations.15 For example, if people within the same neighborhood share the same socioeconomic environment, access to healthcare resources, norms settings, and lifestyles, they may shape a common level of cardiovascular health beyond individual characteristics. Thus, in addition to individual-level sociodemographic factors, neighborhoodlevel factors may also increase the risk of disease. However, only a few studies have documented the potential effects of neighborhood-level SES on lung cancer risk.16–19 In a study from Taiwan between 2002 and 2007, lung cancer survival was significantly associated with low income but only among individuals younger than 65 years.14 Neighborhood deprivation, defined by income, had, however, no significant influence on lung cancer survival when adjusting for individual income. In a Danish study between 2004 and 2008, lung cancer incidence was higher in neighborhoods with high unemployment rates, independently of individual-level socioeconomic factors.15 A Swiss study conducted between 2001 and 2008 showed hazard ratios of 1.49 of lung cancer mortality in neighborhood with the lowest socioeconomic position compared with the highest position.18 The results were adjusted for individual-level socioeconomic factors. Moreover, a study from the US conducted between 1995 and 2007 showed that a decreased survival of non–small-cell lung cancer is associated with neighborhood deprivation.19 However, some studies have found no association between lung cancer survival and individual socioeconomic factors.20–22 Moreover, a Swedish
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
study conducted between 1990 and 2004 found no association between neighborhood deprivation and lung cancer mortality.23 To the best of our knowledge, no study to date has simultaneously analyzed the effect of neighborhood-level SES on both lung cancer incidence and mortality, after adjusting for individual-level socioeconomic factors and comorbidities, such as chronic pulmonary disease (COPD) and alcoholism. The first aim of this study was to investigate whether there is an association between neighborhood deprivation and incident and mortality rates of lung cancer between 2000 and 2010. The second aim was to investigate whether this possible difference remains after accounting for individual-level socio demographic characteristics, i.e., age, marital status, family income, education, immigration status, urban/rural status, mobility, and comorbidities (chronic obstructive pulmonary disease [COPD], tobacco abuse, and alcoholism or alcoholrelated liver disease).
MATERIALS AND METHODS The data sources were several national Swedish data registers, such as the Swedish National Population and Housing Census (1960–1990), the Total Population Register, the MultiGeneration register, the Swedish Cancer Registry (1958– 2010), the Swedish Hospital Discharge Register (1964–2010), and the Swedish Out-patient Register (2001–2010), provided to us by Statistics Sweden and the National Board of Health and Welfare. We used the primary diagnoses for lung cancer in the Swedish Cancer Register. Additional linkages were carried out to national census data to obtain individual-level SES, occupation, geographical region of residence, to the National Registry of Causes of Death (1961–2010) to identify date and cause of death, and to the Immigration Registry to identify date of emigration. All linkages were performed by the use of an individual national identification number that is assigned to each person in Sweden for their lifetime. This number was replaced by a serial number for each person to provide anonymity. The study period started on January 1, 2000 and proceeded until first incident of lung cancer, mortality of lung cancer, death from any other cause, emigration or the end of the study period on December 31, 2010.
Outcome (Dependent) Variable The outcome (dependent) variable was incident (yes/ no) and mortality (yes/no) cases from lung cancer. The unit of observation was individuals. We used the Swedish Cancer Registry to identify primary diagnoses of lung cancer in the study population during the study period. We then linked this information with records in the Cause of Death Register to identify deaths among lung cancer patients during the same period. All cases of cancer in Sweden must be registered in the Swedish Cancer Registry. The completeness of cancer registration is currently considered to be close to 100%. Only primary neoplasms of the lung classified according to the 7th revision of the International Classification of Diseases (ICD-7) (The Swedish Cancer Registry has transferred all the cancer ICD codes into ICD7) (codes 162, 163) were studied.
Neighborhood deprivation and lung cancer
The 10th revisions of the International Classification of Diseases (ICD-10 C33 to C34) were used to define the outcome variable of mortality due to lung cancer in the Cause of Death register.
Independent Variables Independent variables included sex, age at the start of the study, marital status, family income, educational attainment, immigration status, geographical region, mobility, diagnosis of COPDs, tobacco abuse, and alcoholism or alcohol-related liver disease of the subjects. Sex. Men and women. Age. Age was greater than or equal to 50 years and was divided into 10-year category. Marital status. Individuals were classified as married/ cohabitating or single. Family income by quartile. Information on family income in 2000 came from the Total Population Register, which was provided by Statistics Sweden. We used this information to determine the distribution of family incomes in Sweden, and then used the distribution to calculate empirical quartiles. Educational attainment. Educational attainment was classified as completion of compulsory school or less (less than or equal to 9 years), practical high school or some theoretical high school (10–12 years), or theoretical high school and/or college (greater than 12 years). Immigration status. (1) Born in Sweden and (2) Born Outside Sweden. Urban/rural status. Individuals were classified as living in a large city, a middle-sized town, or a small town/rural area. This variable was included because urban/rural status may be associated with access to care. Large cities were those with a population of greater than or equal to 200,000 (Stockholm, Gothenburg and Malmö). Middle-sized towns were towns with a population of greater than or equal to 90,000 but less than 200,000. Small towns were towns with a population of greater than or equal to 27,000 and less than 90,000; rural areas were areas with populations smaller than those of small towns. Mobility. Length of time lived in the neighborhood, categorized as lived in the neighborhood less than 5 years or greater than 5 years. Comorbidities. COPD and tobacco abuse: Patients’ previous diagnosis of COPD, which was suspected to be one important prognostic factor for lung cancer and used as a surrogate of smoking, was identified in the Hospital Registry 10 years before the follow-up period and Out-patient Register accordingly (COPD: ICD-91990–1996 = 490–496; ICD-10 1997– 2010 = J40–J49. The same approach was used to identify tobacco abuse: ICD-9 = 305.1, 292.0, 292.1, 292.2, 292.8, 292.9, V15.8, V65.3, V65.8; ICD-10 = F17, T65.2, Z71.6, Z72.0). Patients’ COPD and tobacco abuse in the patient Registries was individually linked to their lung cancer status using a serial number. Alcoholism and alcohol-related liver disease was identified in the Hospital Registry and Out-patient Register according to the International Classification of Diseases codes (ICD-9 = 291, 303, 571; ICD-10 = F10 and K70).
Copyright © 2014 by the International Association for the Study of Lung Cancer
257
Li et al.
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Neighborhood deprivation index. A summary measure was used to characterize neighborhood-level deprivation. We identified deprivation indicators used by past studies to characterize neighborhood environments and then used a principal components analysis to select deprivation indicators in the Swedish national database. The following four variables were selected for those aged 25–74: low educational status (less than 10 years of formal education), low income (income from all sources, including that from interest and dividends, defined as less than 50% of individual median income)24, unemployment (not employed, excluding full-time students, those completing compulsory military service, and early retirees), and social welfare assistance. Each of the four variables loaded on the first principal component with similar loadings (+0.47 to +0.53) and explained 52% of the variation between these variables. A z score was calculated for each small area market statistics neigh borhood. Small area market statistics are geographic units covering all Sweden with homogenous types of buildings. The z scores, weighted by the coefficients for the eigenvectors, were then summed to create the index.25 The index was categorized into three groups: below one standard deviation (SD) from the mean (low deprivation), above one SD from the mean (high deprivation), and within one SD of the mean (moderate deprivation). Higher scores reflect more deprived neighborhoods.
Statistical Analysis Age-standardized incidence and mortality rates were calculated by direct age standardization using 10-year age groups specific to women or men, with the entire Swedish population of women or men in 2000 as the standard population. Multilevel (hierarchical) logistic regression models were calculated based on incidence proportions or mortality proportions (cumulative incidence or cumulative mortality) of lung cancer (yes/no). The analyses were performed using MLwiN, version 2.27. First, a neighborhood model was calculated that included only neighborhood-level deprivation to determine the crude odds of lung cancer incidence and mortality by level of neighborhood deprivation (model 1). Next, a model that included neighborhood-level deprivation and sex, age (model 2) and the rest of the individual-level sociodemographic variables, added simultaneously to the model (model 3), was calculated. Finally, a full model was calculated that also included comorbidities (model 4). These full models tested whether neighborhood-level deprivation was significantly associated with lung cancer incidence and mortality after adjusting for the sociodemographic characteristics and comorbidities, and whether there were differential effects of neighborhood-level deprivation on lung cancer incidence and mortality across sociodemographic characteristics.26 To check whether collinearity was a problem, we examined the correlation between the included variables. There was a low degree of correlation between the included variables in the models (0.4 or lower). The correlation between the neighborhood variables and the other socioeconomic factors was less than 0.3. The highest correlation was observed between individual income and individual education (r = 0.4).
258
Random effects. The between-neighborhood variance was estimated both with and without a random intercept. It was regarded as significant if it was larger than 1.96 times the standard error, indicating that there were significant differences in incident and mortality rates of lung cancer between neighborhoods. This is in accord with the precedent set in previous studies.12,27,28 The choice of 1.96 was based on the normal curve where 95% of the area under a normal curve lies within 1.96 SDs of the mean.29 For comparison, we also calculated Cox regression models and logistic regression models using the SAS statistical package (version 9.3; SAS Institute, Cary, NC).
Ethical Considerations This study was approved by the Ethics Committee of Lund University, Sweden.
RESULTS Table 1 shows population sizes and characteristics in the year 2000 by neighborhood-level deprivation. The total number of neighborhoods was 8361. Of the total population, 23%, 61%, and 16% lived in low, moderate, and high-deprivation neighborhoods, respectively. During the study period from January 1, 2000 to December 31, 2010, 33,704 (1.7%) of individuals were diagnosed with lung cancer, and 31,287 died due to lung cancer. Age-adjusted lung cancer incidence rates increased from 9.2 per 1000 in neighborhoods with low deprivation to 10.4 per 1000 in neighborhoods with moderate deprivation and 13.3 per 1000 in neighborhoods with high deprivation. The mortality rate of lung cancer was 8.1, 9.6, and 13.0 per 1000, respectively. There were more men than women with lung cancer in the study population. A similar pattern of higher incidence and mortality rates with each increasing level of neighborhood-level deprivation was observed across all individual-level sociodemographic categories and comorbidities. All categories showed a gradient effect across level of neighborhood deprivation. The odds ratio (OR) of lung cancer incidence for individuals living in a high versus low deprivation neighborhood was 1.45 (1.39–1.50) in the crude neighborhood-level model (Table 2). Neighborhood-level deprivation remained significantly associated with lung cancer incidence after adjusting for the individual-level sociodemographic variables and comorbidities (OR = 1.27; 95% confidence interval [CI]: 1.22–1.32). The highest odds of lung cancer incidence were found for individuals who were men, immigrants, had the lowest educational attainments, high mobility, or were affected with comorbidities. Individuals living in middle-sized towns or in small towns and rural areas had lower odds of lung cancer morbidity. Interestingly, individuals with middle-level family income had higher ORs. The OR of mortality due to lung cancer for individuals living in high versus low deprivation neighborhoods was 1.60 (1.54–1.67) in the crude neighborhood-level model (Table 3). Neighborhood-level deprivation remained significantly associated with lung cancer mortality after adjusting for the individual-level sociodemographic variables and comorbidities (OR = 1.32; 95% CI: 1.27–1.38). The highest
Copyright © 2014 by the International Association for the Study of Lung Cancer
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Neighborhood deprivation and lung cancer
TABLE 1. Distribution of Population, Number of Lung Cancer Events, and Age-Standardized Incidence (per 1000) and Mortality by Neighborhood-Level Deprivation Lung Cancer Events
Population No. Total population (%)
(%)
Lung Cancer Events No.
%
Incidence Rates by Neighborhood Deprivation Low
Moderate
High
33,704
721,772 (23%) 9.2
1,957,595 (61%) 10.4
510,599 (16%) 13.3
Low
Moderate
High
1,957,595 (61%) 9.6
510,599 (16%) 13.0
46.5 18,614 55.2 53.5 15,090 44.8
10.6 8.0
12.3 8.8
16.0 11.2
17,256 55.2 14,031 44.8
9.5 7.0
11.7 8.0
15.6 11.0
38.5 9523 28.3 25.6 12,010 35.6 21.8 9945 29.5 12.1 2163 6.4 2.0 63 0.2
6.2 12.0 14.3 6.6 1.3
7.7 14.7 13.6 5.4 0.9
10.9 18.6 16.6 5.5 0.8
7538 10,138 10,264 3199 148
24.1 32.4 32.8 10.2 0.5
4.6 9.6 13.8 9.5 3.5
6.0 12.3 14.2 7.9 2.1
9.3 16.6 17.5 8.8 1.9
25.0 25.0 25.0 25.0
23.7 29.6 26.6 20.2
10.1 9.9 9.9 8.2
10.1 11.8 11.1 8.9
13.0 14.7 13.7 11.8
8224 9756 7839 5468
26.3 31.2 25.1 17.5
9.4 9.0 8.6 7.0
9.7 11.2 10.3 7.9
12.5 14.7 13.4 11.0
57.6 18,931 56.2 42.4 14,773 43.8
9.0 10.0
10.1 11.4
12.5 14.4
16,755 53.6 14,532 46.4
7.8 9.1
9.2 10.8
12.0 14.2
88.9 29,053 86.2 11.1 4651 13.8
9.1 9.9
10.1 12.7
13.0 14.7
27,059 86.5 4228 13.5
8.0 8.8
9.4 11.9
12.7 14.4
54.3 19,983 59.3 29.2 10,228 30.3 16.5 3493 10.4
11.4 9.2 6.4
11.6 10.0 7.2
14.1 12.9 9.4
20,035 64.0 8473 27.1 2779 8.9
10.2 7.4 5.0
10.8 8.6 6.2
13.7 12.2 8.0
47.7 17,494 51.9 34.9 11,474 34.0 17.4 4736 14.1
9.6 9.0 6.8
11.6 10.0 8.5
15.1 13.1 9.5
16,099 51.5 10,722 34.3 4466 14.3
8.4 7.9 6.3
10.6 9.4 7.9
14.8 12.4 9.6
85.1 28,360 84.1 14.9 5344 15.9
9.2 8.9
10.2 11.5
13.0 14.7
26,395 84.4 4892 15.6
8.1 7.9
9.4 10.7
12.7 13.9
93.4 26,528 78.7 6.6 7176 21.3
7.9 30.2
8.8 32.4
11.0 39.9
24,909 79.6 6378 20.4
7.0 23.5
8.3 27.2
10.8 36.7
97.6 32,260 95.7 2.4 1444 4.3
9.0 15.0
10.2 16.5
13.0 20.6
29,948 95.7 1339 4.3
7.9 14.8
9.5 15.9
12.6 20.7
99.9 33,656 99.9 0.1 48 0.1
9.1 20.2
10.4 25.9
13.3 22.8
31,254 99.9 33 0.1
8.1 14.3
9.6 17.6
13.0 15.2
7989 9965 8958 6792
Copyright © 2014 by the International Association for the Study of Lung Cancer
No.
%
Morality Rate of Lung Cancer by Neighborhood Deprivation
721,772 (23%) 8.1
3,189,966
Total lung cancer events Sex Male 1,484,372 Female 1,705,594 Age (years) 50–59 1,227,614 60–69 816,845 70–79 696,612 80–89 384,662 ≥90 64,233 Family income Low income 797,795 Middle–low income 797,963 Middle–high income 797,899 High income 796,309 Marital status Married/cohabiting 1,837,019 Never married, Widowed, 1,352,947 or divorced Immigrant status Sweden 2,837,342 Other countries 352,624 Educational attainment ≤9 years 1,733,209 10–12 years 931,641 >12 years 525,116 Urban/rural status Large cities 1,521,470 Middle-sized towns 1,113,363 Small towns/rural areas 555,133 Move Not moved 2,715,639 Moved 474,327 Chronic lower respiratory disease No 2,978,654 Yes 211,312 Alcoholism and related liver disease No 3,111,900 Yes 78,066 Tobacco abuse No 3,188,197 Yes 1769
Mortality of Lung Cancer Mortality of Lung Cancer
31,287
259
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Li et al.
TABLE 2. OR and 95% CI for Childhood Obesity; Results of Multilevel Logistic Regression Models Model 1 OR Neighborhood-level variable (ref. low) Moderate High Age Sex, men (ref. women) Family income (ref. high income) Middle–high income Middle–low income Low income Marital status (ref. married/cohabiting) Never married, widowed, or divorced Immigrant status (ref. born in Sweden) Education attainment (ref. > 12 years) ≤9 years 10–12 years Urban/rural status (ref. large cities) Middle-sized towns Small towns/rural areas Mobility (ref. not moved) Chronic lower respiratory disease (ref. No) Alcoholism and related liver disease (ref. No) Tobacco abuse (ref. No) Variance (SE) Explained variance (%)
1.13 1.45
Model 2
95% CI 1.10 1.39
1.17 1.50
OR 1.13 1.44 1.00 1.44
0.057 (0.004) 25
Model 3
95% CI 1.09 1.38 1.00 1.41
1.16 1.50 1.00 1.48
0.060 (0.004) 21
OR
Model 4
95% CI
OR
95% CI
p Value
1.08 1.31 1.00 1.46
1.05 1.26 1.00 1.43
1.12 1.36 1.00 1.49
1.08 1.27 1.00 1.44
1.04 1.22 1.00 1.41
1.11 1.32 1.00 1.47
<0.001 <0.001 0.003 <0.001
1.23 1.32 1.04
1.19 1.27 1.00
1.27 1.36 1.07
1.19 1.24 1.00
1.15 1.20 0.96
1.23 1.28 1.04
<0.001 <0.001 0.920
1.04 1.20
1.02 1.16
1.06 1.24
1.01 1.18
0.98 1.14
1.03 1.22
0.689 <0.001
1.66 1.57
1.59 1.51
1.72 1.63
1.59 1.51
1.53 1.46
1.66 1.57
<0.001 <0.001
0.87 0.72 1.05
0.85 0.69 1.02
0.90 0.74 1.09
0.90 0.74 1.03 3.69 1.30 1.76
0.88 0.93 0.72 0.77 1.00 1.06 3.59 3.79 1.23 1.38 1.32 2.35 0.034 (0.004) 58
<0.001 <0.001 0.057 <0.001 <0.001 <0.001
0.039(0.004) 50
Model 1: crude model. Model 2: adjusted for age and sex. Model 3: adjusted for age, sex, family income, marital status, country of birth, education, region of residence, and mobility. Model 4: adjusted for age, sex, family income, marital status, country of birth, education, region of residence, mobility, and comorbidities. OR, odds ratio; CI, confidence interval.
odds of mortality were found for individuals who were men, never married, widowed, or divorced, immigrants, those who had a middle-level family income, had the lowest educational attainment, or were affected with comorbidities. Individuals living in middle-sized towns or in small towns and rural areas had decreased odds of mortality. We performed an additional analysis using logistic regression models and the results were almost identical. In the full model, the OR was 1.28 (95% CI: 1.23–1.32) for lung cancer incidence and 1.33 (95% CI: 1.28–1.38) for mortality due to lung cancer for individuals living in the most deprived neighborhoods compared with those living in the least deprived neighborhoods (Supplementary Tables 1 and 2, Supplementary Digital Content, http://links.lww.com/JTO/ A747). We also performed an additional analysis using Cox regression models. In the full model, the hazard ratios were 1.29 (95% CI: 1.25–1.34) for lung cancer incidence and 1.35 (95% CI: 1.30–1.40) for mortality due to lung cancer among individuals living in the most deprived neighborhoods compared with those living in the least deprived neighborhoods (Supplementary Tables 3 and 4, Supplementary Digital Content, http://links.lww.com/JTO/A747). The test for cross-level interactions between the individual sociodemographic variables and neighborhood deprivation
260
on odds of lung cancer incidence and mortality showed no meaningful cross-level interactions or effect modification. The between-neighborhood variance (i.e., the random intercept) was over 1.96 times the standard error in all models, indicating that there were significant differences in incident and mortality rates of lung cancer between neighborhoods, after accounting for the sociodemographic and comorbidity variables. The neighborhood-level variable explained 25% of the incidence and 33% of the mortality in the between-neighborhood variance estimates (Tables 2, 23). After inclusion of all the independent variables, the explained variance was 58% and 66%, respectively.
DISCUSSION The main findings of this study are that the odds of both incident and mortality rates of lung cancer are higher among individuals living in deprived neighborhoods than among individuals living in affluent neighborhoods. This difference remained significant, after adjustment for the individuallevel sociodemographic variables and comorbidities (COPD, tobacco abuse, and alcoholism). This study represents a novel contribution to previous neighborhood studies on incident16,18 and mortality17,19,23 cases of lung cancer. No previous neighborhood study has, however, studied both incident and
Copyright © 2014 by the International Association for the Study of Lung Cancer
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Neighborhood deprivation and lung cancer
TABLE 3. OR and 95% CI for Mortality of Lung Cancer; Results of Multilevel Logistic Regression Models Model 1 OR Neighborhood-level variable (ref. low) Moderate High Age Sex, men (ref. women) Family income (ref. high income) Middle–high income Middle–low income Low income Marital status (ref. married/co-habiting) Never married, widowed, or divorced Immigrant status (ref. born in Sweden) Education attainment (ref. > 12 years) ≤9 years 10–12 years Urban/rural status (ref. large cities) Middle-sized towns Small towns/rural areas Mobility (ref. not moved) Chronic lower respiratory disease (ref. No) Alcoholism and related liver disease (ref. No) Tobacco abuse (ref. No) Variance (SE) Explained variance (%)
1.19 1.60
Model 2
95% CI 1.15 1.54
1.23 1.67
OR 1.16 1.55 1.02 1.49
0.057 (0.005) 33
Model 3
95% CI 1.12 1.49 1.01 1.45
1.20 1.62 1.02 1.52
0.057 (0.005) 33
OR
Model 4
95% CI
OR
95% CI
p Value
1.09 1.36 1.01 1.53
1.06 1.31 1.01 1.49
1.13 1.42 1.01 1.56
1.09 1.32 1.01 1.49
1.05 1.27 1.01 1.46
1.12 1.38 1.01 1.53
<0.001 <0.001 <0.001 <0.001
1.30 1.44 1.16
1.25 1.39 1.12
1.35 1.49 1.21
1.26 1.36 1.12
1.22 1.31 1.08
1.31 1.41 1.17
<0.001 <0.001 <0.001
1.10 1.20
1.08 1.16
1.13 1.24
1.07 1.18
1.04 1.14
1.09 1.22
<0.001 <0.001
1.75 1.57
1.67 1.51
1.83 1.64
1.68 1.52
1.60 1.46
1.75 1.59
<0.001 <0.001
0.88 0.73 1.05
0.86 0.70 1.02
0.91 0.76 1.09
0.91 0.75 1.03 3.31 1.38 1.41
0.88 0.93 0.73 0.78 1.00 1.06 3.22 3.41 1.30 1.46 1.00 2.00 0.030 (0.004) 66
<0.001 <0.001 0.072 <0.001 <0.001 0.057
0.035(0.004) 59
Model 1: crude model. Model 2: adjusted for age and sex. Model 3: adjusted for age, sex, family income, marital status, country of birth, education, region of residence, and mobility. Model 4: adjusted for age, sex, family income, marital status, country of birth, education, region of residence, mobility, and comorbidities. OR, odds ratio; CI, confidence interval.
mortality cases of lung cancer.16–19,23 In addition, to the best of our knowledge, no previous large-scale study has adjusted for smoking. This study was also unable to do so, although we adjusted for COPD, tobacco abuse, and alcoholism, in an attempt to adjust for potential surrogates for smoking. One previous Swedish study was based on data between 1990 and 2004 and it only included mortality cases and not incident cases of lung cancer. This study covers a more recent time period (2000–2010), and it shows that neighborhood deprivation, even after adjustment for individual-level characteristics and comorbidities, is associated with incident and mortality cases of lung cancer. Individual-level sociodemographic measures of socioeconomic inequality5,7,30 and education1,4,6 have been reported to be associated with lung cancer. However, the causal pathways between neighborhood socioeconomic deprivation and poor health outcomes are not fully understood; several possible mechanisms could thus lie behind our findings. One possible mediator could be psychological stress31,32 due to littered and unsafe environments, vandalism, isolation/alienation, and violent crime33 in deprived neighborhoods. Psychological stress could be harmful in itself as well as lead to harmful smoking habits. Data suggest that hyperactivity of the sympathetic branch of the autonomic nervous system caused by
chronic psychological stress or chronic exposure to nicotinic agonists in cigarette smoke may contribute to the development and progression of non–small-lung cancer.34 A clinical study has also reported improved survival outcomes with the incidental use of β-blockers among patients with non–small-lung cancer,35 which suppresses the sympathetic branch of the autonomic nervous system. It is possible that the lack of safe environments reduces the possibility to exercise, which aggravates a healthy life style. In addition socio-cultural norms regarding smoking and physical activity could vary between neighborhoods and affect the health of the residents and the risk for lung cancer. For example, studies show that environmental risk factors, such as neighborhood income,16 neighborhood education,19 and neighborhood unemployment17 are associated with lung cancer. Studies have also found associations between neighborhood deprivation and exposure to air pollution,36 smoking,37,38 and increased lung cancer mortality.39 Living in deprived neighborhoods can cause isolation from health-promoting milieus (e.g., safe places to exercise and decent housing) and services. In comparisons of wealthy nations, associations between neighborhood characteristics and different health outcomes were inconsistent.40 This implies that neighborhood determinants of health are complex. Such
Copyright © 2014 by the International Association for the Study of Lung Cancer
261
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
Li et al.
determinants may include access to health care, education, and social services. Access to these services is uneven in the US, where the effects of income inequalities on health are more pronounced.41 For example, higher primary health care provider density was associated with lower risks of lung cancer mortality.42 Neighborhood-level inequities include unequal access to and quality of primary and secondary health care services.43 In Sweden, medical care is provided to all permanent residents, and primary health care clinics and hospitals are equally distributed and located centrally in all types of neighborhoods.43 However, the actual number of health care professionals working in primary health care clinics can vary considerably by neighborhood type. This is due to difficulties in recruiting and retaining health care personnel in high-deprivation neighborhoods.19,44 The uneven distribution of medical personnel across neighborhoods has also been documented in UK,45 another country with universal health care. Our study has some limitations. The most important one is the lack of smoking data; smoking might be a mediator between neighborhood deprivation and lung cancer. However, we adjusted for COPD, tobacco abuse, and alcoholism in an attempt to adjust for potential surrogates for smoking. We also adjusted for certain individual-level sociodemographic variables, such as income, education, marital status, and country of birth, which are strongly related to smoking, to minimize this potential confounder. Despite this, residual confounding is likely to exist. Previously published neighborhood studies were, however, also unable to adjust for smoking.16–19,23 Those studies did, however, not adjust for COPD, tobacco abuse, and alcoholism. In studies of neighborhood effects on health, selective residential mobility can cause compositional neighborhood differences. Selective residential mobility is the tendency of individuals to move to neighborhoods whose characteristics match their individual characteristics (for example, the tendency of individuals with low SES to move to low SES neighborhoods). However, we adjusted for family income, which improved our possibilities to differentiate between compositional and contextual effects on lung cancer incidence and mortality. There are a number of strengths in this study. The large cohort included practically all patients with lung cancer (50 years and older) in Sweden during the study period. This is due to the high coverage and validity of the Swedish cancer registry,46,47 which increases the generalizability of our results. Another strength is the use of personal identification numbers, which made it possible to include individuals in different registers, such as the Immigration Register, which permitted calculation of exact risk times.48,49 Second, the Swedish Total Population Register is highly complete, with very few missing data. For example, data for income were 99.4% complete.50 Finally, the use of multilevel modeling helped us to separate neighborhood- and individual-level effects and allowed us to consider both fixed and random effects in the analyses.
CONCLUSIONS Neighborhood deprivation contributes to lung cancer incidence and mortality. The individual- and
262
neighborhood-level variables cumulatively load against individuals so that the most at-risk individuals would be those who have both individual- and neighborhood-level risk factors. These findings raise important clinical and public health concerns, and indicate that both individual- and neighborhood-level approaches are important in health care policies.
Acknowledgements
The authors thank Science Editor Stephen Gilliver for his useful comments on the text. This study was supported by ALF funding awarded to Jan and Kristina Sundquist and by grants from the Swedish Research Council (awarded to Kristina Sundquist), the Swedish Freemasons Foundation (awarded to Jan Sundquist), and the Swedish Heart-Lung Foundation (Bengt Zöller). The registers used in this study are maintained by Statistics Sweden and the National Board of Health and Welfare. REFERENCES 1. Albano JD, Ward E, Jemal A, et al. Cancer mortality in the United States by education level and race. J Natl Cancer Inst 2007;99:1384–1394. 2. Clegg LX, Li FP, Hankey BF, Chu K, Edwards BK. Cancer survival among US whites and minorities: A SEER (surveillance, epidemiology, and end results) program population-based study. Arch Intern Med 2002;162:1985–1993. 3. Du XL, Lin CC, Johnson NJ et al. Effects of individual-level socioeconomic factors on racial disparities in cancer treatment and survival: Findings from the National Longitudinal Mortality Study, 1979–2003. Cancer 2011;117:3242–3251. 4. Hemminki K, Li X. Level of education and the risk of cancer in Sweden. Cancer Epidemiol Biomarkers Prev 2003;12:796–802. 5. Hemminki K, Zhang H, Czene K. Socioeconomic factors in cancer in Sweden. Int J Cancer 2003;105:692–700. 6. Hussain SK, Lenner P, Sundquist J, Hemminki K. Influence of education level on cancer survival in Sweden. Ann Oncol 2008;19:156–162. 7. Mackillop WJ, Zhang-Salomons J, Groome PA, Paszat L, Holowaty E. Socioeconomic status and cancer survival in Ontario. J Clin Oncol 1997;15:1680–1689. 8. Vinnakota S, Lam NS. Socioeconomic inequality of cancer mortality in the United States: A spatial data mining approach. Int J Health Geogr 2006;5:9. 9. Diez Roux AV, Merkin SS, Arnett D, et al. Neighborhood of residence and incidence of coronary heart disease. N Engl J Med 2001;345:99–106. 10. Martikainen P, Kauppinen TM, Valkonen T. Effects of the characteristics of neighbourhoods and the characteristics of people on cause specific mortality: A register based follow up study of 252,000 men. J Epidemiol Community Health 2003;57:210–217. 11. Sampson RJ, Raudenbush SW, Earls F. Neighborhoods and violent crime: A multilevel study of collective efficacy. Science 1997;277:918–924. 12. Sundquist K, Ahlen H. Neighbourhood income and mental health: A multilevel follow-up study of psychiatric hospital admissions among 4.5 million women and men. Health Place 2006;12:594–602. 13. Winkleby M, Sundquist K, Cubbin C. Inequities in CHD inci dence and case fatality by neighborhood deprivation. Am J Prev Med 2007;32:97–106. 14. Durkheim E. The Rules of Sociological Method. New York: Free Press, 1982. 15. Rose GA. Individuals and Populations. The Strategy of Preventive Medicine, Oxford University Press: Oxford, 1992. 16. Chang CM, Su YC, Lai NS, et al. The combined effect of individual and neighborhood socioeconomic status on cancer survival rates. PLoS One 2012;7:e44325. 17. Meijer M, Bloomfield K, Engholm G. Neighbourhoods matter too: The association between neighbourhood socioeconomic position, population density and breast, prostate and lung cancer incidence in Denmark between 2004 and 2008. J Epidemiol Community Health 2013;67:6–13.
Copyright © 2014 by the International Association for the Study of Lung Cancer
Journal of Thoracic Oncology ® • Volume 10, Number 2, February 2015
18. Panczak R, Galobardes B, Voorpostel M, Spoerri A, Zwahlen M, Egger M; Swiss National Cohort and Swiss Household Panel. A Swiss neighbourhood index of socioeconomic position: Development and association with mortality. J Epidemiol Community Health 2012;66:1129–1136. 19. Erhunmwunsee L, Joshi MB, Conlon DH, Harpole DH Jr. Neighborhoodlevel socioeconomic determinants impact outcomes in nonsmall cell lung cancer patients in the Southeastern United States. Cancer 2012;118:5117–5123. 20. Bouchardy C, Fioretta G, De Perrot M, Obradovic M, Spiliopoulos A. Determinants of long term survival after surgery for cancer of the lung: A population-based study. Cancer 1999;86:2229–2237. 21. Tammemagi CM, Neslund-Dudas C, Simoff M, Kvale P. In lung cancer patients, age, race-ethnicity, gender and smoking predict adverse comorbidity, which in turn predicts treatment and survival. J Clin Epidemiol 2004;57:597–609. 22. Stavraky KM, Skillings JR, Stitt LW, Gwadry-Sridhar F. The effect of socioeconomic status on the long-term outcome of cancer. J Clin Epidemiol 1996;49:1155–1160. 23. Sundquist J, Li X, Sundquist K. Neighborhood deprivation and mortality in individuals with cancer: A multilevel analysis from Sweden. Eur J Cancer Prev 2012;21:387–394. 24. EurostatYearbook 2004. http://www.scb.se/templates/Standard____36500. asp. Accessed December 2014. 25. Gilthorpe MS. The importance of normalisation in the construction of deprivation indices. J Epidemiol Community Health 1995;49:S45–S50. 26. Goldstein H. Multilevel Statistical Models, 3rd ed. London: Hodder Arnold Publishers, 2003. 27. Johnell K, Lindström M, Melander A, Sundquist J, Eriksson C, Merlo J. Anxiolytic-hypnotic drug use associated with trust, social participation, and the miniaturization of community: A multilevel analysis. Soc Sci Med 2006;62:1205–1214. 28. Johnell K, Lindström M, Sundquist J, Eriksson C, Merlo J. Individual characteristics, area social participation, and primary non-concordance with medication: A multilevel analysis. BMC Public Health 2006;6:52. 29. Rees DG. Foundations of Statistics. New York: CRC Press, 1987. 30. Sharpe KH, McMahon AD, McClements P et al. Socioeconomic inequalities in incidence of lung and upper aero-digestive tract cancer by age, tumour subtype and sex: A population-based study in Scotland (2000– 2007). Cancer Epidemiol 2012;36:e164–e170. 31. Black PH, Garbutt LD. Stress, inflammation and cardiovascular disease. J Psychosom Res 2002;52:1–23. 32. Holden L, Scuffham P, Hilton M, Vecchio N, Whiteford H. Psychological distress is associated with a range of high-priority health conditions affecting working Australians. Aust N Z J Public Health 2010;34:304–310. 33. Sundquist K, Theobald H, Yang M, Li X, Johansson SE, Sundquist J. Neighborhood violent crime and unemployment increase the risk of coronary heart disease: A multilevel study in an urban setting. Soc Sci Med 2006;62:2061–2071.
Neighborhood deprivation and lung cancer
34. Schuller HM. Impact of neuro-psychological factors on smoking-associated lung cancer. Cancers (Basel) 2014;6:580–594. 35. Wang HM, Liao ZX, Komaki R, et al. Improved survival outcomes with the incidental use of beta-blockers among patients with non-smallcell lung cancer treated with definitive radiation therapy. Ann Oncol 2013;24:1312–1319. 36. Naess O, Piro FN, Nafstad P, Smith GD, Leyland AH. Air pollution, social deprivation, and mortality: A multilevel cohort study. Epidemiology 2007;18:686–694. 37. Hiscock R, Bauld L, Amos A, Platt S. Smoking and socioeconomic status in England: The rise of the never smoker and the disadvantaged smoker. J Public Health (Oxf) 2012;34:390–396. 38. Chuang YC, Cubbin C, Ahn D, Winkleby MA. Effects of neighbourhood socioeconomic status and convenience store concentration on individual level smoking. J Epidemiol Community Health 2005;59:568–573. 39. Hayanga AJ, Zeliadt SB, Backhus LM. Residential segregation and lung cancer mortality in the United States. JAMA Surg 2013;148:37–42. 40. Lynch J, Smith GD, Hillemeier M, Shaw M, Raghunathan T, Kaplan G. Income inequality, the psychosocial environment, and health: Comparisons of wealthy nations. Lancet 2001;358:194–200. 41. Lochner K, Pamuk E, Makuc D, Kennedy BP, Kawachi I. State-level income inequality and individual mortality risk: A prospective, multilevel study. Am J Public Health 2001;91:385–391. 42. Backhus LM, Hayanga AJ, Au D, Zeliadt SB. The effect of provider density on lung cancer survival among blacks and whites in the United States. J Thorac Oncol 2013;8:549–553. 43. van Doorslaer E, Wagstaff A, Bleichrodt H, et al. Income-related inequalities in health: Some international comparisons. J Health Econ 1997;16:93–112. 44. Hardy D, Chan W, Liu CC, et al. Racial disparities in the use of hospice services according to geographic residence and socioeconomic status in an elderly cohort with nonsmall cell lung cancer. Cancer 2011;117:1506–1515. 45. Taylor DH Jr, Leese B. General practitioner turnover and migration in England 1990–94. Br J Gen Pract 1998;48:1070–1072. 46. Barlow L, Westergren K, Holmberg L et al. The completeness of the Swedish Cancer Register: A sample survey for year 1998. Acta Oncol 2009;48:27–33. 47. Socialstyrelsen. Cancerförekomst i Sverige 2011. Stockholm: Socialstyrelsen, 2012. 48. Ludvigsson JF, Otterblad-Olausson P, Pettersson BU, Ekbom A. The Swedish personal identity number: Possibilities and pitfalls in healthcare and medical research. Eur J Epidemiol 2009;24:659–667. 49. Ludvigsson JF, Andersson E, Ekbom A, et al. External review and validation of the Swedish national inpatient register. BMC Public Health 2011;11:450. 50. Rosen M, Hakulinen T. Use of disease registers. In W Ahrens, I Pigeot (Eds.) Handbook of Epidemiology. Berlin: Springer-Verlag, 2005. Pp. 232–251.
Copyright © 2014 by the International Association for the Study of Lung Cancer
263