Te

Aa

b

c

d

a

ARR3A

KLRSTSDPPP

1

oa2at

ST

a

0h

Lung Cancer 83 (2014) 401– 407

Contents lists available at ScienceDirect

Lung Cancer

j ourna l ho me page: www.elsev ier .com/ locate / lungcan

reatment and survival disparities in lung cancer: The effect of socialnvironment and place of residence

sal Mohamadi Johnsona,c,∗, Robert B. Hinesb, James Allen Johnson III c, A. Rana Bayaklyd

Georigia Southern University, Center for International Studies, United StatesUniversity of Kansas School of Medicine-Wichita, Department of Preventive Medicine and Public Health, United StatesGeorgia Southern University, Jiann-Ping Hsu College of Public Health, United StatesGeorgia Department of Public Health, Georgia Comprehensive Cancer Registry, United States

r t i c l e i n f o

rticle history:eceived 25 August 2013eceived in revised form0 December 2013ccepted 12 January 2014

eywords:ung canceruralityocial environmentreatmenturvivalisparitylace of residenceovertylace

a b s t r a c t

Objective: The purpose of this study was to measure the extent to which geographic residency status andthe social environment are associated with disease stage at diagnosis, receipt of treatment, and five-yearsurvival for patients diagnosed with non-small cell lung cancer (NSCLC).Methods and materials: This study was a retrospective cohort study of the Georgia Comprehensive Can-cer Registry (GCCR) for incident cases of NSCLC diagnosed in the state. Multilevel logistic models wereemployed for five outcome variables: unstaged and late stage disease at diagnosis; receipt of treatment(surgery, chemotherapy, and radiation); and survival following diagnosis. The social and geographicalvariables of interest were census tract (CT) poverty level, CT-level educational attainment, and CT-levelgeographic residency status.Results: Compared to urban residents, rural and suburban residents had increased odds of unstageddisease (suburban OR = 1.23, 95% CI: 1.11–1.37; rural OR = 1.63, 95% CI: 1.45–1.83). In this study, ruralparticipants had lower odds of receiving radiotherapy (OR = 0.89, 95% CI: 0.82–0.96) and chemotherapy(OR = 0.92, 95% CI: 0.85–0.99). Living in CTs with lower educational levels was associated with decreasingodds of receiving both surgery (lowest educational level OR = 0.67, 95% CI: 0.59–0.75) and chemother-apy (lowest educational level OR = 0.74, 95% CI: 0.68–0.81). Living in areas with higher concentration ofdeprivation (high level of deprivation HR = 1.04, 95% CI: 1.01–1.09) and lower levels of education (lowesteducational level HR = 1.12, 95% CI: 1.07–1.17) was associated with poorer survival. Rural residents did

not show poorer survival when treatment was controlled and they even presented a lower risk of deathfor early stage disease (HR = 0.90, 95% CI: 0.82–0.99).Conclusion: This study concludes that where NSCLC patients live can, to some extent, explain treatmentand prognostic disparities. Public health practitioners and policy makers should be cognizant of theimportance of where people live and shift their efforts to improve lung cancer outcomes in rural areasand neighborhoods with concentrated poverty.

. Introduction

Among all cancers in the United States, lung cancer ranks sec-nd in incidence and first in mortality [1,2]. It is estimated thatpproximately 160,000 Americans will die from lung cancer in

013, accounting for 26% of all female cancer deaths and 28% ofll male cancer deaths [2]. Non-small cell lung cancer (NSCLC) ishe most common type of lung cancer [3]. It is also one of the few

∗ Corresponding author at: Georgia Southern University, Center for Internationaltudies, Forest Drive Building, Room 1325, Statesboro, GA 30460-8106, Unite States.el.: +1 912 478 1365; fax: +1 912 478 0824.

E-mail addresses: amohamadi@georgiasouthern.edu,salmohamadi@gmail.com (A.M. Johnson).

169-5002/$ – see front matter © 2014 Elsevier Ireland Ltd. All rights reserved.ttp://dx.doi.org/10.1016/j.lungcan.2014.01.008

© 2014 Elsevier Ireland Ltd. All rights reserved.

cancers with high rates of unknown disease stage at diagnosis [4].As stage of disease is used in making treatment decisions and isan important predictor of prognosis following a diagnosis of can-cer [5,6], it is important to examine predictors of unstaged disease.Previous studies [7,8] suggest more research is needed to addressunstaged cancer for several reasons: (1) an important proportion ofcancer in populations consists of unstaged cancer, (2) these patientsare less likely to receive treatment, and (3) unstaged patients havepoorer health outcomes [7].

In the delivery of care for patients diagnosed with NSCLC, dispar-ities that pertain to individual characteristics such as race [3,9,10],

marital status, education, and age have been reported both inreceipt of treatment and survival [11–14]. However, the impact ofarea-level social factors on NSCLC treatment and survival, espe-cially in the U.S., has yet to be determined [15]. Furthermore, the

4 g Can

ippanvre[wt

llfoSabalrofs

gnduiiv(Ciof

2

2

ttott(eonodlA(BS

2

r

02 A.M. Johnson et al. / Lun

mpact of geographical location on the utilization and success ofost-cancer care remains understudied. The spatial aspects of aatient’s residence may explain patterns of receiving treatmentnd impact survival. Area-level socioeconomic status (SES) is usefulot only because it can reflect individual SES [16], but it also pro-ides information about the greater environment in which a patientesides [16,17]. Living in economically impoverished areas or ruralnvironments are often associated with adverse health outcomes18–20] and may reflect lack of access to care or the absence of aell-connected infrastructure to support cancer patients living in

hese areas [21].Among the few studies that examined the association of area

evel characteristics with the receipt of treatment or survival forung cancer patients, Pozet et al. [20] found higher risk of deathor rural NSCLC patients. Another study that investigated a cohortf lung cancer patients, who received treatment in Duke Healthystem, found poorer survival for individuals that reside in low SESreas [21]. Shugarman and colleagues [22] found no relationshipetween rurality and survival although living in a low SES area wasssociated with poorer survival among Medicare beneficiaries. Forung and colorectal cancer patients, Campbell and colleagues [23]eported that as the distance to a cancer center increased, the oddsf late stage disease at diagnosis increased. Likewise, Liff et al. [4]ound an increased likelihood of late stage diagnosis for all cancerites in rural patients, particularly for lung cancer.

The purpose of this study was to measure the extent to whicheographic differences are associated with disease stage at diag-osis, receipt of treatment, and five-year survival for patientsiagnosed with non-small cell lung cancer (NSCLC). The odds ofnstaged and late stage disease at diagnosis, the odds of receiv-

ng treatment, and the risk of death constituted the outcomes ofnterest for this study. There were three social and geographicalariables as exposures of interest all measured at the census tractCT) level: CT-level poverty, CT-level educational attainment, andT-level geographic residency status. The results of this study will

dentify area-level characteristics as determinants of lung cancerutcomes and identify targets where future interventions shouldocus their efforts to reduce these disparities.

. Method

.1. Participants, data, and design

The Georgia Comprehensive Cancer Registry (GCCR) retrospec-ively collects data on all incident cases of cancer diagnosed inhe state of Georgia. The cohort for the current study consistedf incident NSCLC cases diagnosed in Georgia from January 2000o December 2009 (N = 57,120). Participants were excluded fromhe study if: (1) the age at diagnosis was <50 or >85 (n = 6060),2) they did not identify as black or white race (n = 372), (3) theirthnicity was identified as Hispanic (n = 401), (4) there were twor more primary tumors (n = 11,540), and (5) the tumor was diag-osed as small cell lung cancer (n = 5900). Individuals under the agef 50 were excluded because less than 10 percent of the disease isiagnosed in this age group where a considerable proportion of

ung cancer in this group can be due to hereditary factors [24,25].dditionally, patients with missing CT information were excluded

n = 136). This research was approved by the Institutional Reviewoards of the Georgia Department of Public Health and Georgiaouthern University.

.2. Study variables

The GCCR collects demographic, tumor-related, treatment-elated, and follow-up information on all cancer patients diagnosed

cer 83 (2014) 401– 407

in the state. The individual-level variables of interest included race,gender, age at diagnosis, date of diagnosis, tumor-related informa-tion (stage, grade), first course of treatment received, last date offollow-up, and vital status at last follow-up.

The GCCR also obtains the CT corresponding to the residentialaddress for all cancer patients. In order to capture multiple dimen-sions of the social environment, the data were merged with U.S.Census 2000 data. Consistent with previous studies [26–28], factoranalysis was utilized to create composite variables due to the highlycorrelated nature of variables comprising multiple dimensions ofsocioeconomic status (SES). The two composite variables indicatedeconomic deprivation and educational attainment. A higher scoreindicates lower level of educational attainment and higher levelof economic deprivation. As others have described [23,29], bothindicators were classified into four categories based on the quar-tile distribution. In addition to the CT-level indicators, we obtainedCT-level primary Rural Urban Commuting Area (RUCA) codes fromthe U.S. Department of Agriculture [30]. These codes have beendesigned to reflect population density and commuting patterns atthe CT level. It should be mentioned that definitions of rural, sub-urban, and urban vary across studies. As previously demonstratedby other investigators of health-related outcomes [31–33], RUCAcodes for each CT were applied to classify each study case as rural,suburban, or urban in the following manner: rural (RUCA codes7–10), suburban (RUCA codes 2–6), urban (RUCA code 1).

2.3. Statistical analysis

Descriptive statistics are presented as frequencies and percent-ages for the categorical study variables according to geography.Differences in categorical variables were compared across rural,suburban, and urban geographic categories by Chi-square tests. Allstatistical tests were two-sided. P < .05 was considered statisticallysignificant.

For the outcomes of unstaged disease at diagnosis, late stagedisease at diagnosis, and receipt of treatment by type (surgery,chemotherapy, and radiation), logistic regression was utilized toobtain odds ratios (ORs) with 95% confidence intervals (CIs). Toavoid residual confounding [34], patients with unstaged diagnosis(n = 3052) were excluded from all regression analyses except forthe dichotomous outcome of unstaged disease. In all models, theunadjusted effects (not displayed) were obtained followed by cal-culating adjusted effects controlling for tumor stage, tumor grade,age, race, and sex. The effects of each of two constructs of inter-est (social environment and rurality) were examined in separatemodels where the fully adjusted model included the effects of allarea-level variables together while controlling for individual-levelvariables.

In treatment and survival models, patients who died within 2weeks of diagnosis were excluded from the analysis (N = 1889) toremove those who were not recommended any treatment due topoor prognosis because of comorbid disease or advanced lung can-cer. This exclusion also removed cases diagnosed at autopsy. Toassess the association for the exposures of interest on NSCLC par-ticipants’ risk of death, a survival analysis was conducted. Five-yearsurvival time was calculated from the date of diagnosis of NSCLCuntil the last day of follow-up, the date of death, or the termina-tion of the study (December 31st, 2011). Patients who died afterfive years were censored at 5-years follow-up. Five-year survivalwas chosen as it is considered a benchmark for treatment successin cancer [15,35,36]. As mentioned for the logistic models, the sur-vival model measured the effect of exposures in separate models

while the final model included the effects of both exposure vari-ables together. The Cox proportional hazards model was used toobtain hazard rate ratios (HRs) with 95% CIs for the relative risk ofdeath.

A.M. Johnson et al. / Lung Cancer 83 (2014) 401– 407 403

Table 1Characteristics of NSCLC participants from the Georgia Comprehensive Cancer Registry (2000–2009; N = 32, 711).

Characteristics Urban Suburban Rural P

n % n % n %

Study population 17,520 53.6 9522 29.1 5669 17.3Gender <0.001

Male 9982 57.0 5994 63.0 3644 64.2Female 7538 43.0 3528 37.0 2025 35.8

Race <0.001White 12,835 73.2 8272 86.9 4577 80.7Black 4685 26.8 1250 13.1 1092 19.3

Age at diagnosis <0.00150–64 6606 37.7 3631 38.1 2080 36.765–74 6104 34.8 3556 37.4 2178 38.475–85 4810 27.5 2335 24.5 1411 24.9

Tumor stage <0.001Localized (stage I) 3032 17.3 1766 18.6 922 16.3Regional (stage II) 1290 7.4 636 6.7 421 7.4Regional with lymph node involvement (stage III) 3144 17.8 1778 18.7 1001 17.7Distant (Stage IV) 8686 49.6 4407 46.3 2606 46.0Unknown 1398 8.0 935 9.8 719 12.7

Tumor Grade <0.001Well-differentiated 676 3.9 343 3.1 178 3.1Moderately differentiated 2689 15.4 1583 16.6 876 15.5Poorly differentiated 4481 25.6 2480 26.0 1484 26.2Undifferentiated 547 3.1 262 2.8 137 2.4Unknown 9127 52.1 4854 51.0 2994 52.8

Surgery <0.001No 12,715 72.6 6895 72.4 4148 73.2Yes 3968 22.7 2124 22.3 1115 19.7Unknown 837 4.8 503 2.3 406 7.2Chemotherapy <0.001No 9591 54.7 5342 56.1 3291 58.1Yes 6739 38.5 3502 36.8 1870 33.0Unknown 1190 6.8 678 7.1 508 9.0Radiation <0.001

No 10,040 57.3 5469 57.4 3345 59.0Yes 6577 37.5 3523 37.0 1905 33.6Unknown 903 5.2 530 5.6 419 7.4

Social environment (census tract level)Economic deprivationa <0.001

1st quartile 3064 20.6 3613 37.9 962 17.02 5093 29.0 2057 21.6 1034 18.23 4187 24.0 2440 25.6 1549 27.34th quartile 4635 26.4 1412 14.8 2124 37.5

Levels of educational attainmentb <0.0011st quartile 7657 43.7 539 5.7 3 0.12 5170 29.5 2155 22.6 851 15.03 2260 12.9 3470 36.4 2472 43.64th quartile 2432 13.9 3358 35.3 2343 41.3

privation =

3

3

TulA1pul

3

dti

a Lowest level of deprivation = 1; medium level of deprivation = 2; high level of deb Highest level of education = 1; medium level of education = 2; low level of educa

. Results

.1. Characteristics of the study population

The final study population consisted of 32,711 NSCLC cases.able 1 summarizes participant characteristics across urban, sub-rban, and rural categories. Approximately 54% of NSCLC cases

ived in urban CTs, whereas 17% resided in rural tracts. Africanmericans comprised 26.8% of urban residents, 19.3% of rural, and3.1% of suburban residents. Proportionately, more rural partici-ants had unknown disease stage (rural = 12.7% vs. suburban = 9.8%,rban = 8%). Most rural cases lived in census tracts at the highest

evel of deprivation (37.5%).

.2. Unstaged and late stage diagnosis

The effects of the social environment and geographic resi-ency status on unstaged diagnosis of NSCLC, when entered intohe model separately, were both statistically significant follow-ng adjustment for age, gender, race, and tumor grade. However,

tion = 3; highest level of deprivation = 4. 3; lowest level of education = 4.

in the fully adjusted model, economic deprivation was no longerassociated with unknown disease stage at diagnosis though geo-graphic residency status and lowest educational level remainedsignificant. Compared to urban residents, suburban residents expe-rienced a 23% increased odds (OR = 1.23, 95% CI: 1.11–1.37) ofhaving unstaged disease and rural residents experienced a 63%increased odds (OR = 1.63, 95% CI: 1.45–1.83). For educational level,compared to those living in CTs in the highest quartile, residents ofCTs with the lowest level of educational attainment experiencedan 18% increased odds (OR = 1.18, 95% CI: 1.03–1.35) of unstageddisease at diagnosis compared to participants living in areas withhighest level of education (Table 2).

After adjusting for individual variables, our study found nosignificant relationship between geography and the social environ-ment with late stage disease (results not shown).

3.3. Treatment

The results for the odds of receiving treatment are shown inTable 3. The association of geographic residency status and the

404 A.M. Johnson et al. / Lung Cancer 83 (2014) 401– 407

Table 2Partiallya and fully adjustedb associations with unstaged NSCLC diagnosis from the Georgia Comprehensive Cancer Registry (2000–2009; N = 32, 711, events = 3052).

Census tract (CT)-level variables Partially adjusted ORf (95% CI) Fully adjusted ORf (95% CI)

Geographic residencystatusc

1 Refg Refg

2 1.30 (1.18, 1.42) 1.23(1.11, 1.37)3 1.74(1.57, 1.92) 1.63 (1.45,1.83)

Social environment (census tract level)Economic deprivationd

1st quartile Refg

2 1.13(1.01,1.27) 1.09(0.98, 1.23)3 1.01(0.90, 1.12) 0.96(0.85, 1.07)4th quartile 1.07(0.95,1.20) 1.01(0.90, 1.15)

Levels of educational attainmente

1st quartile Refg

2 1.22(1.08, 1.37) 1.09(0.97, 1.24)3 1.44(1.28, 1.61) 1.13(0.99, 1.29)4th quartile 1.49(1.33, 1.68) 1.18(1.03, 1.35)

a CT level variables were included separately (either geographic residency status or social environment) while adjustment was made for age, gender, race, and tumor grade.b Both CT level variables were included in the model while adjustment was made for age, gender, race, and tumor grade.c Urban: 1; suburban: 2; rural: 3.d Lowest level of deprivation: 1; medium level of deprivation: 2; high level of deprivation: 3; highest level of deprivation: 4.

tion: 3

strCordi

cssle9(t

TAG

e Highest level of education: 1; medium level of education: 2; low level of educaf OR, odds ratio; CI, confidence interval.g Ref, referent group.

ocial environment with the odds of receiving the first course ofreatment regardless of the type of treatment was examined. Ruralesidency was associated with 13% decreased odds (OR = 0.87, 95%I: 0.79–0.95) of receiving any treatment. Similar patterns werebserved for the association of social environment and the odds ofeceiving treatment. Participants living in areas with higher level ofeprivation and lower level of education had lower odds of receiv-

ng treatment.In order to show the pattern of treatment disparity, the asso-

iation of geographic residency status with the odds of receivingurgery, chemotherapy, and radiation were examined and arehown in Table 3. Compared to participants living in areas withowest level of deprivation, those living in areas with the high-

st, high, and medium levels of deprivation had 23% (OR = 0.77,5% CI: 0.69–0.86), >21%, (OR = 0.79, 95% CI: 0.71–0.87), and 16%OR = 0.84, 95% CI = 0.76–0.92) decreased odds of surgery respec-ively. Living in areas with lower educational attainment was also

able 3djusteda associations with odds of receiving any treatment regardless of type, and by tyeorgia Comprehensive Cancer Registry (2000–2009; N = 27,785).

CT-level variable Treatment Surgery

Adjusted ORe (95% CI) Adjusted ORe (95

Events = 21,010 (75.6%) Events = 7063 (2Geographic residency statusb

1 Reff Reff

2 0.98 (0.91, 1.06) 0.94 (0.86, 1.03)

3 0.87 (0.79, 0.95) 0.96 (0.86, 1.07)

Social environment (CT level)Economic deprivationc

1st quartile Reff Reff

2 0.91 (0.83, 0.99) 0.84 (0.76, 0.92)3 0.84 (0.77, 0.91) 0.79 (0.71, 0.87)4the quartile 0.82 (0.75, 0.90) 0.77 (0.69, 0.86)

Levels of educational attainmentd

1st quartile Reff Reff

2 0.77 (0.70, 0.84) 0.78 (0.71, 0.86)3 0.71 (0.65, 0.78) 0.69 (0.62, 0.78)4th quartile 0.62 (0.56, 0.68) 0.67 (0.59, 0.75)

a Both CT level variables were included in the model while adjustment was made for ab Urban: 1; suburban: 2; rural: 3.c Lowest level of deprivation: 1; medium level of deprivation: 2; high level of deprivatd Highest level of education: 1; medium level of education: 2; low level of education: 3e OR, odds ratio; CI, confidence interval.f Ref, referent group.

; lowest level of education: 4.

associated with decreased odds of surgery. Because surgery is thestandard treatment for individuals diagnosed with stage I and IIdisease [3], we ran the same model for only early stage partici-pants (stages I and II). The results showed participants living inareas with lowest (OR = 0.52, 95% CI = 0.43–0.6), low (OR = 0.63, 95%CI: 0.53–0.75), and medium (OR = 0.70, 95% CI: 0.60–0.81) levels ofeducation had lower odds of receiving surgery compared to thoseliving in areas with the highest educational attainment. The samepattern was observed for deprivation. As deprivation increased,the odds of receiving surgery decreased (level 4 OR = 0.75, 95% CI:0.64–0.89; level 3 OR = 0.80, 95% CI = 0.69–0.093; level 2 OR = 0.82,95% CI: 0.71–0.96). Geographic residency status was not significantfor receipt of surgery.

Compared to urban residents, rural residence was associatedwith an 8% decreased odds (OR = 0.92, 95% CI: 0.85–0.99) of receiv-ing chemotherapy and an 11% decreased odds (OR = 0.89, 95% CI:0.82–0.96) of receiving radiotherapy. A similar graded effect was

pe of treatment; surgery, chemotherapy, and radiation for NSCLC patients from the

Chemotherapy Radiotherapy% CI) Adjusted OR (95% CI) Adjusted OR (95% CI)

5.4%) Events = 11,817 (42.5%) Events = 11,630 (41.8)

Ref Ref1.01 (0.95, 1.09) 1.03 (0.96, 1.10)0.92 (0.85, 0.99) 0.89 (0.82, 0.96)

Ref Ref 0.99 (0.92, 1.07) 0.99 (0.92, 1.07) 0.90 (0.83, 0.97) 1.02 (0.95, 1.10) 0.81 (0.75,0.88) 1.04 (0.96, 1.12)

Ref Ref 0.91 (0.85, 0.98) 0.92 (0.85, 0.99) 0.86 (0.79, 0.93) 0.95 (0.88, 1.03) 0.74 (0.68, 0.81) 0.94(0.86, 1.02)

ge, gender, race, tumor stage, and tumor grade.

ion: 3; highest level of deprivation: 4.; lowest level of education: 4

A.M. Johnson et al. / Lung Cancer 83 (2014) 401– 407 405

Table 4Partially (without/with treatment), and fully adjusted hazard ratios for death overall and by disease stage for NSCLC patients from the Georgia Comprehensive Cancer Registry(2000–2009; N = 27,785).

CT-level variable Partially adjusteda Partially adjustedb Fully adjustedc Adjustedc Stage I and II Adjusted.c Stage IIIHRg (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)

Geographic residency statusd

1 Refh Ref Ref Ref Ref2 1.04 (1.01, 1.07) 1.01 (0.98, 1.04) 0.98 (0.94, 1.01) 0.91 (0.84, 0.98) 0.98 (0.91, 1.06)3 1.07 (1.03, 1.11) 1.00 (0.97, 1.04) 0.96 (0.92, 1.00) 0.90 (0.82, 0.99) 0.97 (0.89, 1.07)

Social environment (CT level)Economic deprivatione

1st quartile Ref Ref Ref Ref Ref2 1.06(1.02, 1.10) 1.04(0.99, 1.08) 1.04 (1.00, 1.08) 1.08 (1.00, 1.17) 1.08 (0.99, 1.18)3 1.10(1.06, 1.14) 1.04(1.004, 1.08) 1.04 (1.01, 1.09) 1.06 (0.97, 1.15) 1.10 (1.01, 1.20)4th quartile 1.10(1.06, 1.14) 1.04(0.99, 1.08) 1.04 (1.00, 1.08) 1.03 (0.94, 1.12) 1.12 (1.03, 1.23)

Levels of educational attainmentf

1st quartile Ref Ref Ref Ref Ref2 1.10(1.06, 1.14) 1.07(1.03, 1.11) 1.08 (1.04, 1.12) 1.18 (1.08, 1.29) 1.17 (1.07, 1.28)3 1.11(1.07, 1.15) 1.04(1.002, 1.08) 1.06 (1.02, 1.11) 1.17 (1.07, 1.29) 1.12 (1.02, 1.23)4th quartile 1.19(1.15, 1.24) 1.10(1.06, 1.14) 1.12 (1.07, 1.17) 1.36 (1.24, 1.50) 1.13 (1.02, 1.25)

a CT level variables were included separately (either geographic residency status or social environment) while adjustment was made for age, gender, race, tumor grade,and tumor stage. Treatment is not included in this model.

b CT level variables were included separately (either geographic residency status or social environment) while adjustment was made for age, gender, race, tumor grade,and tumor stage. These models also adjust for treatment (surgery, chemotherapy, radiation).

c Both CT level variables were included in the model while adjustment was made for age, gender, race, and tumor grade, and treatment (surgery, chemotherapy, radiation).These models are stratified by tumor stage.

d Urban: 1; suburban: 2; rural: 3.e Lowest level of deprivation: 1; medium level of deprivation: 2; high level of deprivation: 3; highest level of deprivation: 4

ion: 3

fae0slr

3

aotvarsga(tasAgtrrlt

Tmwwa

f Highest level of education: 1; medium level of education: 2; low level of educatg HR, hazard ratio; CI, confidence interval.h Ref, referent group.

ound for the highest two levels of deprivation and the associ-tion with receipt of chemotherapy. Census tracts with mediumducational level had an 8% decreased odds (OR = 0.92, 95% CI:.85–0.99) of receiving radiotherapy; although, there was no con-istent pattern according to educational level. Living in CTs withower educational levels was associated with decreasing odds ofeceiving chemotherapy.

.4. Lung cancer survival

The results of the survival analysis for participants with NSCLCre depicted in Table 4. Because earlier analysis revealed decreaseddds of treatment for rural participants, we examined the associa-ion between the risk of death both with and without treatmentariables in the model. In Table 4, we report (A) the partiallydjusted HRs for geographic residency status and the social envi-onment, separately, while adjusting for age, gender, race, diseasetage, and tumor grade; (B) the partially adjusted results for geo-raphic residency and the social environment, separately, withdjustment for treatment and the individual variables in A; andC) the fully adjusted results for all stages with both CT variables inhe model. The fully adjusted results were also stratified by staget diagnosis. When treatment was not controlled for in the analy-is, rural and suburban participants showed higher risks of death.fter we controlled for the treatment variables, the association witheographic residency disappeared. There was no evidence of sta-istical interaction between the social environment and geographicesidency status. In the fully adjusted model, when the social envi-onment variables were included, participants living in areas withower levels of education had increased risks of death compared tohose living in areas with the highest level of education.

In the stage-stratified models, more information was revealed.here was no association with the risk of death for patients with

etastatic disease (information not shown). Geographic residenceas not associated with the risk of death for stage III patients;hereas, living in suburban and rural areas was significantly associ-

ted with lower risks of death for patients with early stage disease.

; lowest level of education: 4.

For participants diagnosed with disease stage I and II, living in sub-urban and rural areas decreased the risk of death by 9% and 10%,respectively, compared with urban residents. The effect of eco-nomic deprivation was significant for patients with stage III disease.The two most deprived CTs had 10–12% increased risks of death. Forparticipants living in CTs with lower levels of education, there wereincreased risks of death for patients with node negative and nodepositive disease. The effect of living in CTs with the lowest level ofeducation was nearly 3 times as great for patients with early diseaseversus node positive disease.

4. Discussion and conclusion

This study revealed the existence of area-level inequitiesattributed to the geography of residence and the social environ-ment for the outcomes of unstaged diagnosis, receipt of treatment,and survival following diagnosis of NSCLC. For the outcome ofunstaged disease, consistent with the findings of Liff et al. [4] andElliott et al. [37], rural and suburban participants were more likelyto have unstaged disease at diagnosis compared to their urbancounterparts. Furthermore, living in areas with the lowest educa-tional attainment increased the risk of having unstaged disease.Liff and colleagues [4] reported that such differences might be dueto less accurate diagnostic staging of disease or inconsistent docu-mentation in rural areas. In either case, this result is an indicator ofa disparity in the quality of cancer care in rural and lower SES (mea-sured by educational attainment) areas especially as the prognosisof unstaged tumors is similar to that of more advanced tumors [38].

For late stage disease at diagnosis, contrary to the analysis oflung cancer patients conducted by Campbell and colleagues [39],we did not find a significant relationship between social depriva-tion and late stage disease. Consistent with our results, in theirrecent analysis of lung cancer patients at a teaching hospital in the

United Kingdom, Cheyne and colleagues [40] found no associationbetween SES and late stage lung cancer.

For receipt of treatment, we found significant associationsby geographic residence and the social environment. Rural

4 g Can

pcptrm[Ctli

ttcftlrbnUwaffcgtsdrsaer

tlpaiotegicsb

foptbsw[ft3va

g

[

06 A.M. Johnson et al. / Lun

articipants had lower odds of receiving radiotherapy andhemotherapy. Baldwin and colleagues [41] also found that ruralatients with stage IIIA NSCLC had lower rates of receiving radio-herapy. One possible explanation for such a variation might beelated to distance to radiotherapy clinics requiring daily treat-ent for several weeks [41]; therefore, distance to cancer clinics

38] could be one possible explanation for rural/urban disparity.onsistent with the study of Earl and colleagues, which examinedhe receipt of chemotherapy for lung cancer patients [42], we foundower rates for receipt of chemotherapy for patients who residedn areas with disadvantaged social environments.

We found that rural residents were less likely to receivereatment, and in the partially adjusted survival model withoutreatment, rural and suburban residents had higher risks of deathompared to their urban counterparts. The increased risks of deathor rural and suburban residents vanished after accounting forreatment differences. In contrast with our results, Singh and col-eagues [26], in their national county-level analysis, found ruralesidents had higher lung cancer mortality. This difference mighte explained by differences in receipt of treatment which wasot accounted for in the Singh et al. study. Several studies in the.K. reported adverse survival outcomes for lung cancer patientsho live in remote areas far from cancer clinics, in impoverished

reas, and patients who were seen by a non-specialist physician oracility [39,43,44]. Thus, lack of access to facilities is still presentor rural residents even in countries that provide universal healthare. However, in our U.S.-based study, differences in sociodemo-raphic factors explained the increased risk of death accordingo geographic residence and remained significant determinants ofurvival for all stages combined. After accounting for treatmentifferences, rural residency was not associated with survival andural residents even had a survival advantage for patients with earlytage disease. However, this study does indicate that rural residentsre less likely to receive treatment and that this treatment differ-nce could account for the increased risk of death for rural residentseported in the partially adjusted survival model.

Consistent with the literature [22,26,29,45], this study suggestshat living in areas with higher concentrations of deprivation andower levels of education is associated with poorer survival amongatients with NSCLC. Areas with higher concentrations of povertynd lower educational levels are more likely to lack social and phys-cal infrastructures [46], which negatively impacts survival. Groupsf individuals aggregated in impoverished areas often lack accesso public transportation systems, social support, social capital, andnvironmental amenities such as health care facilities or parks andreen spaces [46–48]. In our study, the decreased odds of receiv-ng treatment for rural patients and the poorer survival for lungancer patients who live in areas with higher concentrations ofocial deprivation might be explained by the presence of structuralarriers.

This study has several strengths. While other investigators haveocused on survival, treatment, or stage at diagnosis as their soleutcome, we analyzed several outcomes to elucidate a more com-rehensive view for the impact of area-level exposures. Moreover,here is a dearth of US-based studies examining the relationshipetween lung cancer treatments, survival, and area-level expo-ures. We used census tracts as our area level unit of analysishich can meaningfully capture SES variations within a county

49]. Additionally, utilizing factor analysis through which six dif-erent variables were employed to explain different dimension ofhe social environment and categorizing geographic residency into

categories rather than two (urban and rural) also adds to the

alidity of the study. Furthermore, because the GCCR has high casescertainment [50], the impact of selection bias is reduced.

Our study had some limitations. Our individual-level demo-raphic variables were limited to race, gender, and age. Therefore,

[

cer 83 (2014) 401– 407

we were not able to control for other variables such individual SES,comorbidity, or insurance status. We also lacked information on thecause of death for those participants who died during the follow-upperiod. Another limitation concerns information on the first courseof treatment. According to Du and colleagues [51], when using can-cer registry data, it is likely that treatment information is missingor incomplete for some participants. Rural patients were slightlymore likely to have missing information on receipt of chemother-apy (6.8% vs. 7.1%, and 9% for suburban and urban, respectively) andradiation (5.2%, vs. 5.6%, and 7.4% for suburban and urban, respec-tively). As it has been previously noted [52], if rural patients werealso more likely to have been misclassified as not having receivedtreatment, the effect measures in the analysis for the odds of treat-ment would be biased away from the null.

Overall, this study concludes that where individuals live hasan impact on the receipt of treatment and prognosis of lung can-cer. As suggested by Yen and Syme [53] and Kaplan [46], placeof residence affects health not only through an individual’s socio-economic status and behaviors, but the social environment, itself,plays a significant role in determining health outcomes. In sum-mary, public health practitioners and policy makers alike shouldbe cognizant of the importance of where people live and focusefforts (such as cancer awareness campaigns, anti-tobacco cam-paigns, increasing access to treatment, etc.) toward populationgroups aggregated in disadvantaged areas to assure the most vul-nerable groups receive the appropriate level of health literacy forlung cancer and cancer care.

Conflict of interest statement

None declared

Acknowledgments

The authors would like to thank the anonymous reviewers andeditors for their helpful comments and feedback. This work wassupported by a Faculty Research Award from Georgia Southern Uni-versity. Final thanks to the Graduate Student Association at GeorgiaSouthern University.

References

[1] Alberg AJ, Ford JG, Samet JM. Epidemiology of Lung Cancer*: ACCPevidence-based clinical practice guidelines. CHEST Journal 2007;132(3(Suppl.)):29S–55S.

[2] Siegel R, Naishadham D, Jemal A. Cancer statistics: 2013. CA: A Cancer Journalfor Clinicians 2013;63(1):11–30.

[3] Kim AW, Liptay MJ, Higgins RS. Contemporary review on the inequities in themanagement of lung cancer among the African-American population. Journalof the National Medical Association 2008;100(6):683.

[4] Liff JM, Chow WH, Greenberg RS. Rural–urban differences in stage at diagnosis.Possible relationship to cancer screening. Cancer 2006;67(5):1454–9.

[5] Rami-Porta R, Crowley JJ, Goldstraw P. Review: the revised TNM staging sys-tem for lung cancer. Annals of Thoracic and Cardiovascular Surgery: OfficialJournal of the Association of Thoracic and Cardiovascular Surgeons of Asia2009;15(1):5.

[6] Yoh Watanabe M. TNM classification for lung cancer. Annals of Thoracic andCardiovascular Surgery: Official Journal of the Association of Thoracic and Car-diovascular Surgeons of Asia 2003;9:6.

[7] Lengerich EJ, Chase GA, Beiler J, Darnell M. Increased risk of unknown stage can-cer from residence in a rural area: health disparities with poverty and minoritystatus; 2006.

[8] Shavers VL, Brown ML. Racial and ethnic disparities in the receipt of cancertreatment. Journal of the National Cancer Institute 2002;94(5):334–57.

[9] Hede K. Drilling down to the causes of racial disparities in lung cancer. Journalof the National Cancer Institute 2010;102(18):1385–7.

10] Bach PB, Cramer LD, Warren JL, Begg CB. Racial differences in thetreatment of early-stage lung cancer. New England Journal of Medicine

1999;341(16):1198–205.

11] Greenberg ER, Chute CG, Stukel T, Baron JA, Freeman DH, Yates J, et al. Socialand economic factors in the choice of lung cancer treatment. A population-based study in two rural states. The New England journal of medicine1988;318(10):612–7.

g Can

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

A.M. Johnson et al. / Lun

12] Williams DR, Kontos EZ, Viswanath K, Haas JS, Lathan CS, MacConaill LE, et al.Integrating multiple social statuses in health disparities research: the case oflung cancer. Health Services Research 2012;47(3 (pt 2)):1255–77.

13] Duncan GJ, Daly MC, McDonough P, Williams DR. Optimal indicators ofsocioeconomic status for health research. American Journal of Public Health2002;92(7):1151–7.

14] Du XL, Lin CC, Johnson NJ, Altekruse S. Effects of individual-level socioeco-nomic factors on racial disparities in cancer treatment and survival. Cancer2011;117(14):3242–51.

15] Greenwald HP, Polissar NL, Borgatta EF, McCorkle R, Goodman G. Social factors:treatment, and survival in early-stage non-small cell lung cancer. AmericanJournal of Public Health 1998;88(11):1681–4.

16] Krieger N. Socioeconomic data in cancer registries. American Journal of PublicHealth 2001;91(1), 156-156.

17] Macintyre S, Ellaway A, Cummins S. Place effects on health: how can weconceptualise, operationalise and measure them? Social Science & Medicine2002;55(1):125–39.

18] Bosma H, van de Mheen HD, Borsboom GJ, Mackenbach JP. Neighborhood socio-economic status and all-cause mortality. American Journal of Epidemiology2001;153(4):363–71.

19] Monroe AC, Ricketts TC, Savitz LA. Cancer in rural versus urban populations: areview. The Journal of Rural Health 1992;8(3):212–20.

20] Pozet A, Westeel V, Berion P, Danzon A, Debieuvre D, Breton JL, et al. Ruralityand survival differences in lung cancer: a large population-based multivariateanalysis. Lung Cancer 2008;59(3):291–300.

21] Erhunmwunsee L, Joshi MBM, Conlon DH, Harpole DH. Neighborhood-levelsocioeconomic determinants impact outcomes in nonsmall cell lung cancerpatients in the Southeastern United States. Cancer 2012;118(20):5117–23.

22] Shugarman LR, Sorbero ME, Tian H, Jain AK, Ashwood JS. An exploration of urbanand rural differences in lung cancer survival among Medicare beneficiaries.American Journal of Public Health 2008;98(7):1280.

23] Campbell NC, Elliott AM, Sharp L, Ritchie LD, Cassidy J, Little J. Rural and urbandifferences in stage at diagnosis of colorectal and lung cancers. British Journalof Cancer 2001;84(7):910.

24] Spiro SG, Porter JC. Lung cancer—where are we today? Current advances instaging and nonsurgical treatment. American Journal of Respiratory and CriticalCare Medicine 2002;166(9):1166–96.

25] Li X, Hemminki K. Inherited predisposition to early onset lung cancer accordingto histological type. International Journal of Cancer 2004;112(3):451–7.

26] Singh GK, Williams SD, Siahpush M, & Mulhollen A. Socioeconomic,rural–urban, and racial inequalities in US Cancer mortality: part I—all cancersand lung cancer and part II—colorectal, prostate, breast, and cervical cancers.Journal of Cancer Epidemiology 2012;2011.

27] Miles R, Coutts C, Mohamadi A. Neighborhood urban form, social environment,and depression. Journal of Urban Health 2012:1–18.

28] Siegel R, Naishadham D, Jemal A. Cancer statistics, 2012. CA: A Cancer Journalfor Clinicians 2012;62(1):10–29.

29] Haynes R, Pearce J, Barnett R. Cancer survival in New Zealand: ethnic, socialand geographical inequalities. Social Science & Medicine 2006;67(6):928–37.

30] Documentation. Economic research, S., rural–urban continuum codes. UnitedStates Department of Agriculture; 2012.

31] Hall SA, Kaufman JS, Ricketts TC. Defining urban and rural areas in US epidemi-ologic studies. Journal of Urban Health 2006;83(2):162–75.

32] Weeks WB, Kazis LE, Shen Y, Cong Z, Ren XS, Miller D, et al. Differences in

health-related quality of life in rural and urban veterans. Journal Information2004;94:10.

33] Hao Y, Landrine H, Jemal A, Ward KC, Bayakly AR, Young JL, et al. Race: neigh-bourhood characteristics and disparities in chemotherapy for colorectal cancer.Journal of Epidemiology and Community Health 2011;65(3):211–7.

[

cer 83 (2014) 401– 407 407

34] Hines RB, Markossian TW. Differences in late-stage diagnosis. Treatment, andcolorectal cancer-related death between rural and urban African Americansand Whites in Georgia. The Journal of Rural Health 2011;28(3):296–305.

35] Molina JR, Yang P, Cassivi SD, Schild SE, Adjei AA. Non–small cell lung can-cer: epidemiology, risk factors, treatment, and survivorship. In: Mayo Clinicproceedings. Mayo Clinic. NIH Public Access.

36] Gordis L. Epidemiology. Saunders. Elsevier; 2009.37] Elliott TE, Elliott BA, Renier CM, Haller IV. Rural-urban differences in cancer

care: results from the Lake Superior Rural Cancer Care Project. MinnesotaMedicine 2004;87(9):44.

38] Voti L, Richardson LC, Reis IM, Fleming LE, MacKinnon J, Coebergh JWW. Treat-ment of local breast carcinoma in Florida. Cancer 2006;106(1):201–7.

39] Campbell NC, Elliott AM, Sharp L, Ritchie LD, Cassidy J, Little J. Impact of depri-vation and rural residence on treatment of colorectal and lung cancer. BritishJournal of Cancer 2002;87(6):585–90.

40] Cheyne L, Taylor A, Milton R, Fear J, Callister MEJ. Social deprivation doesnot affect lung cancer stage at presentation or disease outcome. Lung Cancer2013;81(2):247–51.

41] Baldwin LM, Andrilla CHA, Porter MP, Rosenblatt RA, Patel S, Doescher MP.Receipt of recommended radiation therapy among rural and urban cancerpatients. Cancer 2012;118(20):5100–9.

42] Earle CC, Venditti LN, Neumann PJ, Gelber RD, Weinstein MC, Potosky AL,et al. Who gets chemotherapy for metastatic lung cancer? CHEST Journal2000;117(5):1239–46.

43] Jack RH, Gulliford MC, Ferguson J, Møller H. Geographical inequalities in lungcancer management and survival in South East England: evidence of variationin access to oncology services? British Journal of Cancer 2003;88(7):1025–31.

44] Jack RH, Gulliford MC, Ferguson J, Møller H. Explaining inequalities inaccess to treatment in lung cancer. Journal of Evaluation in Clinical Practice2006;12(5):573–82.

45] Merletti F, Galassi C, Spadea T. The socioeconomic determinants of cancer.Environmental Health 2011;10(Suppl. 1):57.

46] Kaplan GA. People and places: contrasting perspectives on the associationbetween social class and health. International Journal of Health Services1996;26(3):507–19.

47] Ward E, Jemal A, Cokkinides V, Singh GK, Cardinez C, Ghafoor A, et al. Cancerdisparities by race/ethnicity and socioeconomic status. CA: A Cancer Journalfor Clinicians 2004;54(2):78–93.

48] Cattell V. Poor people, poor places, and poor health: the mediating role ofsocial networks and social capital. Social Science & Medicine 2001;52(10):1501–16.

49] Krieger N, Chen JT, Waterman PD, Soobader MJ, Subramanian SV, Carson R.Choosing area based socioeconomic measures to monitor social inequalitiesin low birth weight and childhood lead poisoning: The Public Health Dispar-ities Geocoding Project (US). Journal of Epidemiology and Community Health2003;57(3):186–99.

50] Clarkson LS. An evaluation of the georgia comprehensive cancer registry:improving an established system. Georgia Department of Human Resources;2007.

51] Du XL, Key CR, Dickie L, Darling R, Delclos GL, Waller K, et al. Informationon chemotherapy and hormone therapy from tumor registry had moderateagreement with chart reviews. Journal of Clinical Epidemiology 2006;59(1):53–60.

52] Hines R, Markossian T, Johnson A, Dong F, Bayakly R. Geographic Residency

Status and Census Tract Socioeconomic Status as Determinants of ColorectalCancer Outcomes. Am J Public Health 2014;104:3.

53] Yen IH, Syme SL. The social environment and health: a discussion ofthe epidemiologic literature. Annual Review of Public Health 1999;20(1):287–308.