Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2018 Oct 1.
Published in final edited form as: Cancer Epidemiol. 2017 Aug 31;50(Pt A):99–106. doi: 10.1016/j.canep.2017.08.009

Association between the neighborhood obesogenic environment and colorectal cancer risk in the Multiethnic Cohort

Alison J Canchola a, Salma Shariff-Marco a,b, Juan Yang a, Cheryl Albright c, Andrew Hertz a, Song-Yi Park d, Yurii B Shvetsov d, Kristine R Monroe e, Loïc Le Marchand d, Scarlett Lin Gomez a,b, Lynne R Wilkens d, Iona Cheng a,b
PMCID: PMC5660936  NIHMSID: NIHMS907947  PMID: 28850934

Abstract

Background

Information on the role of the neighborhood environment and colorectal cancer risk is limited. We investigated the association between a comprehensive suite of possible obesogenic neighborhood attributes (socioeconomic status, population density, restaurant and retail food environments, numbers of recreational facilities and businesses, commute patterns, traffic density, and street connectivity) and colorectal cancer risk in the Multiethnic Cohort Study.

Methods

Among 81,197 eligible participants living in California (35,397 males and 45,800 females), 1,973 incident cases (981 males and 992 females) of invasive colorectal cancer were identified between 1993 and 2010. Separately for males and females, multivariable Cox regression models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for colorectal cancer risk overall and by racial/ethnic group (African American, Japanese American, Latino, white).

Results

In males, higher traffic density was associated with an increased risk of colorectal cancer (HR=1.29, 95% CI: 1.03–1.61, p=0.03, for quintile 5 vs. quintile 1; p-trend=0.06). While this association may be due to chance, this pattern was seen (albeit non-statistically significant) in all racial/ethnic groups except whites. There were no other significant associations between other neighborhood obesogenic attributes and colorectal cancer risk.

Conclusion

Findings from our large racial/ethnically diverse cohort suggest neighborhood obesogenic characteristics are not strongly associated with the risk of colorectal cancer.

Keywords: neighborhood, colorectal cancer, race/ethnicity

1. Introduction

It is estimated that 45% of U.S. colorectal cancer (CRC) cases could be prevented by maintaining a healthy diet, regular physical activity, and healthy weight1. Within the Multiethnic Cohort Study (MEC), obesity, smoking, alcohol, and a number of dietary factors have been associated with the risk of CRC2–4. In addition, there is evidence that the neighborhood environment can impact diet, obesity, and physical activity, and can influence obesity-related health disparities5–11. In the MEC, neighborhood socioeconomic status (nSES) has been associated with obesity in African Americans, Latinos, and whites8.

While individual-level factors such as obesity and level of physical activity are associated with CRC risk, what is less clear is the effect of the neighborhood environment, and whether its role is independent of these individual-level risk factors. No cohort studies have examined neighborhood-level factors other than socioeconomic status (SES) in relation to CRC risk12, 13. In the MEC, we investigated the association between a comprehensive suite of ten a priori selected neighborhood obesogenic attributes and risk of CRC, assessing whether associations were independent of individual-level factors and varied by racial/ethnic group.

2. Methods

2.1 Study subjects

The MEC is a large population-based cohort of U.S. adults of five racial/ethnic groups. Methodological details of this study have been described previously14. In brief, participants from Hawaii and California completed a baseline questionnaire in 1993–1996 that included information on sociodemographics, height, weight, medical history, family history of cancer, smoking, physical activity, medications, diet, alcohol, and vitamin use.

Of 105,759 African American, Japanese American, Latino, Native Hawaiian, and white MEC participants from California who completed the baseline questionnaire, we excluded participants, hierarchically, who had a history of CRC (n=1,308); were Native Hawaiian (n=171); had no follow-up time (n=8); were an incident, invasive CRC case with a non-adenocarcinoma histology (n=77 carcinoid, n=7 squamous cell, and n=25 other tumors; and n=6 missing); had a residential address that was not geocodable (n=2,155), had missing BMI (n=2,247), or had missing or invalid covariate data (n=18,558)4. Thus, 81,197 MEC participants were eligible for analysis.

2.2 Follow-up and case identification

Incident CRC cases were identified through linkage of the cohort to the California Cancer Registry. Deaths were determined through linkages with California death certificate files and the National Death Index.

Follow-up time was calculated as the number of days between the date of completion of the baseline questionnaire and the earliest of: the first diagnosis of invasive CRC (International Classification of Diseases for Oncology-3 [ICD-O-3] site codes C18.0–C18.9, C26.0, C19.9, and C20.9), death, or December 31, 2010. Over a median follow-up time of 16.6 years, 1,973 incident CRC cases were identified.

2.3 Residential neighborhood obesogenic attributes

Baseline residential addresses were geocoded to latitude and longitude coordinates, using parcel data (96%) and street centerline data for those that failed to geocode to a parcel (4%).

Geocodes were linked with the California Neighborhoods Data System, an integrated system of small area-level measures of the social and built environments for California15. Census 1990 block group-level data were utilized to ascertain: neighborhood SES (nSES), a validated composite measure16; population density (per square mile); commute patterns; and street connectivity17, which was defined as the ratio of actual number of street segments to the maximum possible number of links between nodes (intersections and cul-de-sacs). These measures were categorized based on the distribution of Los Angeles (LA) County block groups (≥90% of the sample resided in LA County). Business, farmers’ market18, and park data were used to quantify the amenities within a one-mile pedestrian network distance of the participant’s residence: the Restaurant Environment Index (REI), defined as the ratio of the number of fast-food restaurants to other restaurants19; the Retail Food Environment Index (RFEI), defined as the ratio of the number of convenience stores, liquor stores, and fast-food restaurants to supermarkets and farmers’ markets; and total number of recreational facilities, parks, and businesses. Traffic density was based on traffic counts within a 500 meter radius of a participant’s residence. These business and traffic-related attributes were categorized according to the study participant distributions (Supplementary Table 1). These ten neighborhood attributes were selected a priori for their potential associations with obesity or colorectal cancer risk (all but street connectivity and number of recreational facilities were associated with obesity8 and all but the number of parks were univariately associated with CRC risk).

2.4 Statistical analyses

Hazard rate ratios (HR) and 95% confidence intervals (CI) were estimated using multivariable Cox regression models with age as the time metric. Sex and race/ethnicity-specific models were run given the heterogeneity in CRC incidence between these subgroups4. Multivariable models were adjusted for the following individual-level CRC risk factors: age, race/ethnicity, body mass index, family history of CRC, history of intestinal polyps, education, cigarette smoking, multivitamin use, nonsteroidal anti-inflammatory medication use, alcohol consumption, vigorous physical activity, history of diabetes, average energy intake, red and processed meat, dietary fiber, calcium, folacin, Vitamin D, and use of hormone therapy (females). These covariates were selected a priori as they were associated with CRC risk in the literature or in this cohort1, 4. Distributions of these covariates are presented in Supplementary Table 2.

All models were additionally adjusted for clustering by block group, using a sandwich estimator of the covariance structure that accounts for intracluster dependence20. As a sensitivity analysis, gamma frailty models were run with block group as a random effect21. As the random effect term was not statistically significant and the CIs for the neighborhood attributes did not change, the fixed effect models are presented here. Wald tests for trend across neighborhood characteristic categories (excluding no restaurants for REI, no retail food for RFEI, and missing data categories) were conducted using quantile number as an ordinal variable. Wald Type 3 tests for heterogeneity of the trend parameter across neighborhood characteristic categories by race/ethnicity, BMI, and nSES were computed using cross-product terms. Based on correlation tests of time versus scaled Schoenfeld residuals, no neighborhood or adjustment variables violated the proportional hazards assumption.

The ten neighborhood characteristic variables were first entered separately into models, minimally adjusted for age, race/ethnicity (if applicable), and clustering by block group (Supplementary Tables 3 and 4). Nine of the ten neighborhood variables (all except the number of parks) had at least one category or trend that had a p-value<0.10 in sex and race/ethnicity-specific models. Thus, for the final multivariable models, all the neighborhood attributes except the number of parks were included.

3. Results

This study population (males: 26.0% African American, 14.8% Japanese American, 47.2% Latino, 12.0% white; females: 34.7% African American, 11.8% Japanese American, 36.8% Latino, 16.7% white; Supplementary Table 2) was followed for a median of 16.6 years. The mean age at entry into the cohort was 60 for males and 59 for females. Only 33.2% of males and 30.5% of females lived in high SES neighborhoods (quintiles 4 and 5) (Supplementary Table 1).

The MEC participants in this analysis resided at baseline in 7,348 unique block groups predominantly in LA County. The median number of participants in each block group was five (interquartile range 2 to 13). Of the 7,348 block groups, 19.5% included one participant and 11.6% included two; the largest block group included 432 participants. The neighborhood attributes in these block groups were moderately correlated (Supplementary Table 5; correlations <|0.72|). For example, high SES neighborhoods tended to have a lower population density (r=−0.43) and more commuting (r=0.40).

When each of the neighborhood obesogenic attributes were entered individually into minimally adjusted sex-specific models, only higher traffic density was associated with CRC risk in males (HR=1.24, 95% CI: 1.03–1.51, p=0.025, for quintile 5 vs. quintile 1, p-trend=0.092, Supplementary Tables 3 and 4). In minimally adjusted race/ethnicity-specific models, two trend tests reached statistical significance at the 0.05 level. White males in neighborhoods with more unhealthy retail food options were at increased CRC risk (HR=1.79, 95% CI: 0.95–3.37, p=0.073, for quartile 4 versus quartile 1, p-trend=0.038). In addition, Latino males living in neighborhoods with fewer businesses were at increased CRC risk (HR=1.43, 95% CI: 1.04–1.95, p=0.026, for quintile 1 versus quintile 5, p-trend=0.010). This association differed by race/ethnicity (p-heterogeneity of the trend=0.023), as white males living in neighborhoods with fewer businesses were at decreased risk (HR=0.53, 95% CI: 0.31–0.91, p=0.022, for quintile 1 versus quintile 5, p-trend=0.054), with no association seen in the other racial/ethnic groups (p-trends >=0.55).

Similar to the minimally adjusted results (Supplementary Tables 3 and 4), when adjusting for individual-level CRC risk factors as well as the other neighborhood attributes, none of the neighborhood obesogenic attributes were associated with CRC risk in males or females (Tables 1 and 2), except for higher traffic density, which was associated with increased risk of CRC in males (HR=1.29, 95% CI: 1.03–1.61, p=0.026, for quintile 5 vs. quintile 1, p-trend=0.056). In males, higher traffic density was associated with non-statistically significant increased risk in quintile 5 for African Americans, Japanese Americans, and Latinos (with hazard ratios above 1.2), while there was no association among whites, with no evidence of heterogeneity in associations by race/ethnicity (p-heterogeneity of the trend=0.78). Similarly, there was no heterogeneity in associations detected by BMI, and only one interaction was seen between nSES and number of recreational facilities in females (p-heterogeneity of the trend =0.011), whereby no versus 3+ recreational facilities was associated with a decreased risk in low nSES areas and an increased risk in high nSES areas (data not shown).

Table 1.

Associations between colorectal cancer risk and body mass index and the neighborhood obesogenic environment in males, Multiethnic Cohort, 1993 – 2010

Males Overall
N=35,397		 African Americans
N=9,190		 Japanese Americans
N=5,222		 Latinos
N=16,724		 Whites
N=4,261
No. HRa 95% CI p-intb No. HRc 95% CI No. HRc 95% CI No. HRc 95% CI No. HRc 95% CI
Body mass index (kg/m2)
  <18.5 8 1.73 (0.86–3.49) 5 3.25d (1.24–8.54) 2 1.14 (0.28–4.66) 1 0.88 (0.13–6.04) 0
  18.5–24.9 315 1.00 76 1.00 102 1.00 108 1.00 29 1.00
  25.0–29.9 472 1.01 (0.87–1.17) 124 0.95 (0.71–1.27) 72 0.97 (0.70–1.34) 218 0.94 (0.74–1.19) 58 1.57 (0.98–2.51)
  ≥30.0 186 1.18 (0.98–1.43) 80 1.42e (1.03–1.96) 11 1.21 (0.65–2.26) 80 0.97 (0.73–1.31) 15 1.21 (0.62–2.37)
Neighborhood socioeconomic statuse 0.52
  Q5 high 126 1.00 20 1.00 43 1.00 36 1.00 27 1.00
  Q4 209 1.21 (0.96–1.53) 53 1.45 (0.84–2.51) 62 1.19 (0.75–1.88) 66 1.17 (0.79–1.75) 28 1.14 (0.63–2.07)
  Q3 219 1.21 (0.95–1.56) 47 1.11 (0.62–1.96) 42 1.09 (0.68–1.76) 108 1.42 (0.95–2.12) 22 0.96 (0.50–1.85)
  Q2 230 1.09 (0.83–1.44) 79 1.12 (0.62–2.03) 28 1.42 (0.78–2.59) 109 1.12 (0.73–1.74) 14 0.83 (0.35–1.98)
  Q1 low 196 1.06 (0.77–1.45) 86 1.07 (0.56–2.02) 12 1.36 (0.57–3.26) 87 1.07 (0.64–1.78) 11 1.33 (0.48–3.72)
  p-trendf 0.95 0.52 0.38 0.94 0.95
Population densitye 0.67
  Q5 high 171 1.00 62 1.00 23 1.00 73 1.00 13 1.00
  Q4 225 1.02 (0.82–1.28) 82 0.89 (0.62–1.28) 29 0.80 (0.40–1.58) 95 1.17 (0.82–1.68) 19 1.20 (0.56–2.56)
  Q3 229 0.99 (0.78–1.27) 61 0.81 (0.53–1.22) 50 0.85 (0.42–1.73) 96 1.10 (0.74–1.62) 22 1.04 (0.44–2.43)
  Q2 189 1.05 (0.81–1.37) 38 0.81 (0.50–1.33) 50 0.92 (0.46–1.85) 84 1.28 (0.85–1.93) 17 0.75 (0.31–1.84)
  Q1 low 167 0.97 (0.72–1.31) 42 0.88 (0.51–1.51) 35 0.72 (0.32–1.62) 59 1.05 (0.66–1.70) 31 1.29 (0.55–3.02)
  p-trendf 0.96 0.53 0.69 0.72 0.87
Restaurant environment indexg 0.65
  0h 176 1.00 35 1.00 35 1.00 85 1.00 21 1.00
  Q1 low 198 0.97 (0.76–1.23) 42 1.23 (0.71–2.14) 45 0.79 (0.44–1.43) 86 0.95 (0.67–1.34) 25 1.27 (0.57–2.81)
  Q2 190 0.95 (0.75–1.20) 51 1.08 (0.65–1.80) 41 0.94 (0.55–1.61) 75 0.86 (0.61–1.22) 23 1.21 (0.57–2.57)
  Q3 184 0.92 (0.73–1.17) 55 0.88 (0.53–1.48) 30 1.06 (0.60–1.90) 79 0.87 (0.62–1.23) 20 1.21 (0.54–2.72)
  Q4 & <1 170 0.96 (0.76–1.23) 79 1.12 (0.68–1.83) 23 1.25 (0.71–2.23) 59 0.85 (0.58–1.24) 9 0.80 (0.31–2.04)
  Q4 & ≥1 19 0.85 (0.54–1.33) 9 1.32 (0.62–2.81) 1 0.31 (0.04–2.25) 9 1.02 (0.52–1.97) 0
  No restaurants 44 1.22 (0.79–1.87) 14 0.59 (0.21–1.70) 12 2.58d (1.13–5.89) 14 1.54 (0.86–2.77) 4 0.63 (0.24–1.69)
  p-trendf 0.55 0.80 0.44 0.29 0.41
Retail food environment indexi 0.27
  0i 33 0.92 (0.62–1.38) 8 1.96 (0.80–4.81) 8 0.90 (0.40–2.04) 12 0.65 (0.35–1.20) 5 1.42 (0.52–3.90)
  Q1 low 206 1.00 48 1.00 36 1.00 109 1.00 13 1.00
  Q2 226 1.15 (0.95–1.39) 74 1.24 (0.85–1.80) 31 0.84 (0.50–1.40) 99 1.19 (0.89–1.59) 22 1.58 (0.77–3.21)
  Q3 208 1.04 (0.85–1.28) 53 0.84 (0.56–1.27) 47 1.06 (0.63–1.76) 83 1.02 (0.75–1.39) 25 1.89 (0.87–4.12)
  Q4 high 273 1.11 (0.91–1.36) 86 1.15 (0.77–1.73) 57 0.97 (0.59–1.59) 97 0.95 (0.69–1.30) 33 2.14d (1.00–4.57)
  No retail food 35 1.08 (0.69–1.68) 16 2.68d (1.02–7.03) 8 0.83 (0.30–2.33) 7 0.73 (0.31–1.71) 4 1.00 (0.31–3.22)
  p-trendf 0.32 0.77 0.81 0.94 0.056
Number of recreational facilitiesk 0.50
  3+ 302 1.00 69 1.00 82 1.00 113 1.00 38 1.00
  2 189 1.10 (0.91–1.34) 68 1.28 (0.91–1.80) 25 0.85 (0.53–1.38) 83 1.14 (0.84–1.54) 13 0.89 (0.45–1.79)
  1 242 1.15 (0.95–1.39) 62 1.15 (0.77–1.70) 47 1.29 (0.84–1.99) 107 1.02 (0.76–1.39) 26 1.70 (0.94–3.08)
  0 248 1.04 (0.85–1.29) 86 1.06 (0.72–1.57) 33 0.99 (0.58–1.70) 104 0.98 (0.70–1.36) 25 1.37 (0.75–2.49)
  p-trendf 0.45 0.95 0.74 0.42 0.75
Commute by car/motorcyclee 0.72
  Q1 low 177 1.00 65 1.00 20 1.00 77 1.00 15 1.00
  Q2 162 0.83 (0.67–1.05) 66 1.23 (0.86–1.78) 24 0.96 (0.49–1.88) 60 0.65d (0.45–0.94) 12 0.57 (0.26–1.25)
  Q3 202 0.97 (0.77–1.22) 52 1.21 (0.83–1.78) 41 1.15 (0.57–2.31) 86 0.77 (0.54–1.12) 23 1.12 (0.50–2.50)
  Q4 217 0.84 (0.65–1.08) 51 1.16 (0.75–1.80) 44 0.91 (0.44–1.89) 92 0.66d (0.44–1.00) 30 0.85 (0.37–1.95)
  Q5 high 222 0.85 (0.66–1.11) 51 1.22 (0.75–1.97) 58 1.20 (0.60–2.41) 91 0.66 (0.43–1.01) 22 0.60 (0.26–1.38)
  p-trendf 0.35 0.59 0.58 0.17 0.39
Total number of businessesl 0.048
  Q5 high 206 1.00 46 1.00 64 1.00 66 1.00 30 1.00
  Q4 202 1.12 (0.91–1.38) 69 1.17 (0.77–1.77) 48 1.19 (0.77–1.85) 69 1.16 (0.80–1.68) 16 0.87 (0.44–1.71)
  Q3 172 1.02 (0.81–1.28) 54 0.79 (0.49–1.28) 23 0.92 (0.53–1.60) 72 1.21 (0.83–1.75) 23 1.34 (0.67–2.67)
  Q2 195 1.13 (0.89–1.43) 56 1.03 (0.64–1.64) 22 0.91 (0.49–1.67) 100 1.34 (0.91–1.95) 17 0.98 (0.48–2.00)
  Q1 low 206 1.15 (0.89–1.49) 60 0.91 (0.53–1.56) 30 1.17 (0.62–2.20) 100 1.45 (0.96–2.21) 16 0.74 (0.33–1.63)
  p-trendf 0.37 0.48 0.96 0.07 0.63
Traffic densitym 0.78
  Q1 low 199 1.00 53 1.00 38 1.00 85 1.00 23 1.00
  Q2 205 1.15 (0.92–1.43) 55 1.06 (0.66–1.71) 41 1.46 (0.88–2.40) 89 1.08 (0.78–1.49) 20 1.10 (0.56–2.18)
  Q3 174 1.02 (0.81–1.29) 57 0.98 (0.61–1.57) 27 1.14 (0.64–2.02) 75 1.01 (0.72–1.42) 15 0.81 (0.39–1.69)
  Q4 182 1.10 (0.87–1.39) 64 1.08 (0.67–1.74) 42 1.38 (0.83–2.28) 54 0.92 (0.63–1.34) 22 0.95 (0.45–1.98)
  Q5 high 221 1.29d (1.03–1.61) 56 1.22 (0.76–1.96) 39 1.47 (0.84–2.57) 104 1.23 (0.89–1.70) 22 0.95 (0.47–1.91)
  p-trendf 0.056 0.40 0.30 0.34 0.77
Street connectivitye, n 0.31
  Q5 high 152 1.00 66 1.00 27 1.00 46 1.00 13 1.00
  Q4 178 1.17 (0.93–1.45) 79 1.42d (1.01–1.99) 24 0.91 (0.52–1.60) 59 1.01 (0.68–1.49) 16 1.32 (0.63–2.75)
  Q3 193 1.18 (0.94–1.49) 54 1.26 (0.86–1.85) 36 0.99 (0.55–1.79) 82 1.23 (0.84–1.82) 21 1.41 (0.70–2.83)
  Q2 193 0.93 (0.74–1.18) 34 0.73 (0.46–1.15) 46 0.82 (0.45–1.47) 93 1.07 (0.74–1.56) 20 0.98 (0.47–2.05)
  Q1 low 265 1.10 (0.86–1.41) 52 1.23 (0.77–1.95) 54 0.83 (0.45–1.55) 127 1.19 (0.81–1.76) 32 1.10 (0.52–2.30)
  p-trendf 0.89 0.80 0.49 0.37 0.86
Open in a new tab
a

Adjusted for clustering effect of block group, age, race/ethnicity, body mass index, family history of colorectal cancer, history of intestinal polyps, education, pack-years of smoking, multivitamin use, nonsteroidal anti-inflammatory medication use, alcohol consumption, vigorous physical activity, history of diabetes, calories, red meat, dietary fiber, calcium, folacin, vitamin D, and all of the neighborhood variables presented in the table; age was the time metric.

b

Heterogeneity of the trend across race/ethnicity, excluding missing values, no restaurants for restaurant environment index, and no retail food for retail food environment.

c

Adjusted for clustering effect of block group, age, body mass index, family history of colorectal cancer, history of intestinal polyps, education, pack-years of smoking, multivitamin use, nonsteroidal anti-inflammatory medication use, alcohol consumption, vigorous physical activity, history of diabetes, calories, red meat, dietary fiber, calcium, folacin, vitamin D, and all of the neighborhood variables presented in the table; age wast the time metric.

d

p-value 0.01–<0.05

e

Census 1990 block-group level measures; quintiles based on Los Angeles County distribution.

f

p-trend analysis excluding missing values, no restaurants, and no retail food.

g

Residential buffer measure within a 1600 meter network distance; ratio of the average number of fast food restaurants to other restaurants; quartiles based on sample distribution.

h

No fast food restaurants, but other restaurants.

i

Residential buffer measure within a 1600 meter network distance; ratio of the average number of convenience stores, liquor stores, and fast food restaurants to supermarkets and farmers' markets; quartiles based on sample distribution.

j

No convenience stores, liquor stores, or fast food restaurants, but supermarkets or farmers' markets

k

Within a 1600 meter network distance.

l

Annual average number of businesses that were active during a 3-year window within a 1600 meter network distance; quintiles based on sample distribution.l

m

Traffic density within a 500 meter network distance; vehicle kilometers traveled (VKmT); quintiles based on sample distribution.

n

Ratio of actual number of street segments to maximum possible number of intersections.

Abbreviations: No., number of cases of invasive colorectal cancer; HR, hazard rate ratio; CI, confidence interval

Table 2.

Associations between colorectal cancer risk and body mass index and the neighborhood obesogenic environment in females, Multiethnic Cohort, 1993 – 2010

Females Overall
N=45,800		 African Americans
N=15,903		 Japanese Americans
N=5,415		 Latinas
N=16,836		 Whites
N=7,646
No. HRa 95% CI p-intb No. HRc 95% CI No. HRc 95% CI No. HRc 95% CI No. HRc 95% CI
Body mass index (kg/m2)
  <18.5 30 1.71d (1.17–2.50) 8 1.95 (0.92–4.12) 15 1.87e (1.07–3.27) 3 1.63 (0.53–5.03) 4 1.52 (0.55–4.22)
  18.5–24.9 326 1.00 97 1.00 91 1.00 73 1.00 65 1.00
  25.0–29.9 332 1.06 (0.91–1.25) 157 1.03 (0.80–1.34) 25 0.97 (0.60–1.57) 97 1.06 (0.79–1.43) 53 1.12 (0.77–1.63)
  ≥30.0 304 1.30d (1.09–1.55) 162 1.13 (0.87–1.47) 10 1.83 (0.92–3.62) 89 1.41e (1.01–1.97) 43 1.45 (0.96–2.19)
Neighborhood socioeconomic statusf 0.57
  Q5 high 111 1.00 22 1.00 34 1.00 21 1.00 34 1.00
  Q4 197 1.14 (0.90–1.43) 67 1.13 (0.72–1.77) 45 1.14 (0.67–1.94) 29 0.69 (0.39–1.23) 56 1.50 (0.96–2.33)
  Q3 216 1.12 (0.88–1.43) 74 1.07 (0.67–1.69) 41 1.07 (0.63–1.81) 61 0.94 (0.56–1.59) 40 1.23 (0.75–2.02)
  Q2 258 1.12 (0.87–1.46) 129 1.19 (0.75–1.89) 13 0.62 (0.29–1.36) 89 0.98 (0.57–1.68) 27 1.22 (0.64–2.32)
  Q1 low 210 1.06 (0.78–1.43) 132 1.14 (0.70–1.84) 8 1.07 (0.38–2.98) 62 0.83 (0.43–1.57) 8 0.90 (0.34–2.40)
  p-trendg 0.81 0.67 0.55 0.88 0.92
Population densityf 0.68
  Q5 high 187 1.00 89 1.00 13 1.00 68 1.00 17 1.00
  Q4 244 0.97 (0.78–1.20) 135 1.15 (0.86–1.55) 26 1.08 (0.50–2.35) 62 0.71 (0.48–1.05) 21 0.69 (0.33–1.43)
  Q3 238 0.97 (0.77–1.24) 97 1.12 (0.77–1.62) 41 1.38 (0.61–3.11) 53 0.60e (0.38–0.94) 47 1.04 (0.53–2.06)
  Q2 178 1.00 (0.77–1.30) 53 1.02 (0.66–1.56) 32 1.19 (0.49–2.91) 45 0.73 (0.46–1.17) 48 1.06 (0.53–2.10)
  Q1 low 145 0.91 (0.68–1.21) 50 1.04 (0.63–1.73) 29 1.46 (0.58–3.68) 34 0.74 (0.45–1.21) 32 0.60 (0.28–1.28)
  p-trendg 0.70 0.96 0.46 0.37 0.46
Restaurant environment indexh 0.07
  0i 163 1.00 47 1.00 25 1.00 45 1.00 46 1.00
  Q1 low 194 0.98 (0.77–1.24) 73 1.17 (0.77–1.78) 45 1.66 (0.87–3.16) 48 0.79 (0.50–1.24) 28 0.70 (0.39–1.26)
  Q2 213 1.08 (0.85–1.35) 78 0.99 (0.66–1.47) 40 2.02e (1.10–3.70) 55 0.96 (0.63–1.48) 40 1.07 (0.64–1.81)
  Q3 191 0.97 (0.76–1.23) 86 0.92 (0.61–1.38) 18 1.48 (0.75–2.95) 62 1.06 (0.69–1.63) 25 0.76 (0.42–1.37)
  Q4 & <1 178 0.93 (0.72–1.19) 115 1.01 (0.67–1.52) 9 1.18 (0.51–2.68) 39 0.99 (0.62–1.59) 15 0.54 (0.27–1.07)
  Q4 & ≥1 27 1.29 (0.84–1.99) 12 1.16 (0.58–2.32) 1 0.67 (0.09–5.23) 11 2.21e (1.11–4.40) 3 0.63 (0.20–2.02)
  No restaurants 26 0.71 (0.45–1.13) 13 0.74 (0.40–1.37) 3 0.37 (0.10–1.38) 2 0.74 (0.17–3.26) 8 0.75 (0.34–1.62)
  p-trendg 0.94 0.78 0.79 0.15 0.14
Retail food environment indexj 0.49
  0j 26 0.93 (0.62–1.37) 9 1.45 (0.70–2.99) 6 1.67 (0.71–3.94) 3 0.38 (0.12–1.19) 8 1.04 (0.44–2.48)
  Q1 low 226 1.00 90 1.00 31 1.00 75 1.00 30 1.00
  Q2 218 0.87 (0.71–1.05) 100 0.88 (0.65–1.18) 22 0.56 (0.30–1.03) 67 1.01 (0.70–1.45) 29 1.03 (0.61–1.76)
  Q3 222 0.85 (0.70–1.03) 92 0.76 (0.56–1.04) 38 0.81 (0.47–1.39) 57 0.95 (0.65–1.37) 35 1.09 (0.66–1.81)
  Q4 high 274 0.90 (0.74–1.10) 121 0.89 (0.66–1.20) 38 0.73 (0.41–1.30) 60 0.85 (0.56–1.28) 55 1.34 (0.79–2.27)
  No retail food 26 1.00 (0.62–1.64) 12 1.16 (0.51–2.63) 6 2.22 (0.66–7.40) 0 8 1.08 (0.35–3.35)
  p-trendg 0.41 0.27 0.32 0.76 0.25
Number of recreational facilitiesl 0.33
  3+ 323 1.00 117 1.00 70 1.00 83 1.00 53 1.00
  2 183 0.97 (0.80–1.18) 88 0.98 (0.72–1.33) 18 0.82 (0.48–1.40) 53 0.94 (0.64–1.37) 24 1.10 (0.66–1.83)
  1 245 1.02 (0.85–1.24) 106 0.98 (0.73–1.32) 34 1.38 (0.84–2.28) 70 0.85 (0.59–1.22) 35 1.20 (0.74–1.95)
  0 241 0.92 (0.75–1.12) 113 0.82 (0.60–1.12) 19 0.86 (0.44–1.69) 56 0.73 (0.50–1.07) 53 1.64e (1.02–2.65)
  p-trendg 0.82 0.93 0.22 0.23 0.19
Commute by car/motorcyclef 0.62
  Q1 low 176 1.00 110 1.00 12 1.00 42 1.00 12 1.00
  Q2 192 1.09 (0.88–1.36) 92 1.00 (0.74–1.34) 26 1.31 (0.61–2.79) 49 1.29 (0.80–2.06) 25 1.12 (0.54–2.32)
  Q3 203 1.21 (0.96–1.53) 68 1.00 (0.72–1.41) 27 0.99 (0.45–2.19) 64 1.68e (1.06–2.65) 44 1.22 (0.61–2.44)
  Q4 207 1.07 (0.84–1.36) 80 1.16 (0.84–1.60) 42 1.01 (0.45–2.23) 48 1.28 (0.75–2.19) 37 0.65 (0.31–1.36)
  Q5 high 214 1.17 (0.90–1.50) 74 1.28 (0.90–1.82) 34 0.84 (0.38–1.87) 59 1.71e (1.00–2.92) 47 0.73 (0.34–1.55)
  p-trendg 0.42 0.12 0.20 0.12 0.044
Total number of businessesm 0.65
  Q5 high 213 1.00 76 1.00 60 1.00 43 1.00 34 1.00
  Q4 205 0.94 (0.76–1.17) 84 0.81 (0.57–1.17) 32 0.88 (0.54–1.46) 61 1.41 (0.93–2.15) 28 1.09 (0.63–1.87)
  Q3 205 0.94 (0.75–1.17) 105 0.92 (0.65–1.32) 19 0.98 (0.52–1.85) 45 1.02 (0.65–1.60) 36 1.14 (0.64–2.04)
  Q2 191 0.94 (0.74–1.19) 79 0.91 (0.63–1.34) 14 0.71 (0.35–1.43) 63 1.22 (0.78–1.92) 35 1.12 (0.64–1.98)
  Q1 low 178 0.85 (0.65–1.13) 80 0.85 (0.54–1.34) 16 0.69 (0.32–1.51) 50 1.18 (0.71–1.96) 32 0.95 (0.45–2.00)
  p-trendg 0.35 0.81 0.32 0.87 0.94
Traffic densityn 0.78
  Q1 low 205 1.00 77 1.00 34 1.00 60 1.00 34 1.00
  Q2 211 0.97 (0.78–1.19) 87 0.94 (0.68–1.32) 26 0.77 (0.44–1.33) 56 0.81 (0.55–1.20) 42 1.70e (1.04–2.77)
  Q3 188 0.83 (0.67–1.02) 73 0.69e (0.48–0.99) 28 0.87 (0.51–1.48) 44 0.65e (0.44–0.98) 43 1.66e (1.03–2.68)
  Q4 196 0.81 (0.65–1.01) 102 0.84 (0.60–1.19) 26 0.62 (0.34–1.11) 43 0.76 (0.50–1.16) 25 1.10 (0.59–2.03)
  Q5 high 192 0.90 (0.72–1.12) 85 1.02 (0.71–1.46) 27 0.78 (0.44–1.40) 59 0.89 (0.60–1.32) 21 0.89 (0.49–1.62)
  p-trendg 0.16 0.94 0.31 0.62 0.36
Street connectivityf, o 0.32
  Q5 high 194 1.00 116 1.00 18 1.00 39 1.00 21 1.00
  Q4 189 1.06 (0.86–1.30) 103 1.09 (0.84–1.42) 21 1.36 (0.70–2.64) 46 1.06 (0.67–1.68) 19 0.85 (0.46–1.58)
  Q3 163 0.94 (0.76–1.17) 68 0.98 (0.73–1.31) 26 1.16 (0.63–2.11) 47 0.85 (0.55–1.30) 22 0.81 (0.41–1.60)
  Q2 219 1.07 (0.87–1.31) 72 1.10 (0.81–1.51) 38 1.01 (0.57–1.80) 62 0.94 (0.63–1.40) 47 1.13 (0.64–1.98)
  Q1 low 227 1.03 (0.82–1.30) 65 1.20 (0.83–1.75) 38 0.98 (0.51–1.88) 68 0.85 (0.55–1.31) 56 1.21 (0.69–2.14)
  p-trendg 0.78 0.45 0.58 0.38 0.24
Open in a new tab
a

Adjusted for clustering effect of block group, age, race/ethnicity, body mass index, family history of colorectal cancer, history of intestinal polyps, education, pack-years of smoking, multivitamin use, nonsteroidal anti-inflammatory medication use, use of hormone therapy, alcohol consumption, vigorous physical activity, history of diabetes, calories, red meat, dietary fiber, calcium, folacin, vitamin D, and all of the neighborhood variables presented in the table; age was the time metric.

b

Heterogeneity of the trend across race/ethnicity, excluding missing values, no restaurants for restaurant environment index, and no retail food for retail food environment.

c

Adjusted for clustering effect of block group, age, body mass index, family history of colorectal cancer, history of intestinal polyps, education, pack-years of smoking, multivitamin use, nonsteroidal anti-inflammatory medication use, use of hormone therapy, alcohol consumption, vigorous physical activity, history of diabetes, calories, red meat, dietary fiber, calcium, folacin, vitamin D, and all of the neighborhood variables presented in the table; age was the time metric.

d

p-value 0.001–<0.01

e

p-value 0.01–<0.05

f

Census 1990 block-group level measures; quintiles based on Los Angeles County distribution.

g

p-trend analysis excluding missing values, no restaurants, and no retail food.

h

Residential buffer measure within a 1600 meter network distance; ratio of the average number of fast food restaurants to other restaurants; quartiles based on sample distribution.

i

No fast food restaurants, but other restaurants.

j

Residential buffer measure within a 1600 meter network distance; ratio of the average number of convenience stores, liquor stores, and fast food restaurants to supermarkets and farmers' markets; quartiles based on sample distribution.

k

No convenience stores, liquor stores, or fast food restaurants, but supermarkets or farmers' markets.

l

Within a 1600 meter network distance.

m

Annual average number of businesses that were active during a 3-year window within a 1600 meter network distance; quintiles based on sample distribution.

n

Traffic density within a 500 meter network distance; vehicle kilometers traveled (VKmT); quintiles based on sample distribution.

o

Ratio of actual number of street segments to maximum possible number of intersections.

Abbreviations: No., number of cases of invasive colorectal cancer; HR, hazard rate ratio; CI, confidence interval

In race/ethnicity-specific analyses in males and females, when adjusting for individual-level CRC risk factors and the other neighborhood attributes, the significant trends seen in the minimally adjusted models (Supplementary Tables 3 and 4) were no longer statistically significant at the p<0.05 level (Tables 1 and 2), and an additional trend test in white females reached statistical significance at this level. Among white females, those living in neighborhoods with more commuting were at decreased risk of CRC (p-trend=0.044), although the number of cases in the reference category was small and the association was not statistically significant for any quintile. Similar to the minimally adjusted results, only the association with total number of businesses in males differed by race/ethnicity in fully adjusted models (p-heterogeneity of the trend=0.048), with a lower number of businesses associated with a non-statistically significant increased risk of CRC in Latino males (HR=1.45, 95% CI: 0.96–2.21, p=0.079, for quintile 1 vs. quintile 5, p-trend=0.069), and a non-statistically significant decreased risk in white males (HR=0.74, 95% CI: 0.33–1.63, p=0.45, for quintile 1 versus quintile 5, p-trend=0.63) and no association seen in the other racial/ethnic groups (p-trends >=0.48).

To address the possibility of chance due to the number of hypotheses tested (9 neighborhood variables × 2 sexes × 5 racial/ethnic groups including overall), we applied a conservative Bonferroni significance threshold of p<5.5 ×10−4 (alpha=0.05/90). No associations reached this conservative significance threshold.

4. Discussion

Our study found a modest association between higher traffic density and increased CRC risk among MEC males overall, residing predominately in LA County, with a similar pattern of association among African Americans, Japanese Americans, and Latinos. Higher traffic density may be an indicator of a less walkable environment22 and have a negative impact on physical activity23, which is an established health behavior associated with obesity and CRC risk1, 24. In our cohort, traffic density was weakly negatively correlated with moderate (males: r=−0.02, p<0.0001; females: r=−0.04, p<0.0001) but not vigorous physical activity. No other associations between the neighborhood obesogenic environment and CRC risk were observed.

Neighborhood-level SES and CRC risk has been examined previously in two prospective studies12, 13. The National Institutes of Health-AARP Diet and Health Study (NIH-AARP) assessed in men and women combined baseline education and a measure of baseline census-tract level neighborhood socioeconomic deprivation12. Lower levels of education and living in a more deprived neighborhood were independently associated with higher CRC risk, even after adjustment for individual-level behavioral factors. However, the association with neighborhood deprivation was limited to rectal cancer risk. The Nurses’ Health Study (NHS) examined in women baseline education and a measure of baseline census-tract level nSES13. Lower nSES was associated with increased rectal cancer risk for all women and increased colon cancer risk for higher-educated women, adjusted for neighborhood factors, demographics, education, and family history. A path analysis found this association to be mediated by individual-level behavioral factors, including BMI. These results are not consistent with our finding of no association between baseline block group-level nSES and CRC risk in models adjusted for other neighborhood attributes, with and without adjustment for individual-level behavioral factors. In addition, we found no association between nSES and either risk of colon or rectal cancer (data not shown). These differing results could be due to substantial differences in study populations and/or study designs. The NIH-AARP Study and NHS were both national samples in which over 90% were white and 29.7% and 23.7%, respectively, were in the quintile representing those in the least deprived or highest SES neighborhood. In contrast, our study includes predominantly African Americans and Latinos in LA County14, and only 14.7% were white and 13.0% were in the highest nSES quintile. In addition, our study may support the hypothesis that the effect of SES differs by race/ethnicity, as certain racial/ethnic groups at the same level of SES may not share the same benefits as non-Hispanic whites25, 26. Moreover, the NIH-AARP Study and NHS used a larger geographical area of census tract that typically contain 1,200 to 8,000 people to characterize nSES, whereas we used census block groups, which typically contain 600 to 3,000 people. Census block groups are more homogenous than census tracts and better represent where individuals likely engage in healthy behaviors, access services, and receive health care; and have been shown to be useful for defining neighborhoods for health studies27. Also, the NIH-AARP study used education and a measure of neighborhood deprivation, which included different components than were used to define nSES in our study, so results may not be directly comparable.

Although we did not observe associations between obesogenic neighborhood attributes and CRC risk after applying the conservative Bonferroni correction, the neighborhood environment has been shown to impact obesity, levels of physical activity, and diet5–11, as well as cancer incidence and mortality28. In the MEC, after adjustment for individual- and neighborhood-level factors, obesity (versus normal weight) was associated with lower nSES in African American, Latino, and white men (odds ratio (OR)=1.48, 95% CI: 1.07–2.06 for nSES quintile 1 vs. quintile 5 for African Americans), whereas both overweight and obese were associated with lower nSES in African American, Latina, and white women (OR=1.31, 95% CI; 1.08–1.59 and OR=2.07, 95% CI: 1.62–2.65 for African Americans, respectively)8. A review paper identified six studies most of which found differences in cancer incidence by nSES and ethnic enclave28, including one which found lower incidence of CRC with higher nSES in whites29. In addition, individual-level behavioral factors, such as obesity and physical activity, may be on the causal pathway between the neighborhood environment and cancer risk. In fact, we observed some notable (albeit modest) neighborhood associations in the minimally adjusted models stratified by sex and race/ethnicity, which were no longer statistically significant (at the p<0.05 level) after adjusting for the individual-level factors. We did not evaluate a mediation effect, as we saw no associations between neighborhood attributes and CRC risk by sex, except for traffic density in males, and were limited by the small number of cases in the strata by sex and race/ethnicity. In addition, there was adequate variation in neighborhood attributes between block groups of MEC participants. For example, the mean value of nSES was −0.09 and the standard deviation (Std) was 1.05. Neighborhood SES quintile 1 values ranged from −3.31 to −1.11 and block groups in nSES quintile 1 had a mean percent of households below the poverty line of 32%. In comparison, nSES quintile 5 values ranged from 0.88 to 2.88 and block groups in nSES quintile 5 had a mean percent of households below the poverty line of 4%. For population density, the mean value was 4.53 and the Std was 3.14, with quintile 1 ranging from 0 to 1.98 and quintile 5 ranging from 6.39 to 47.25.

Strengths of our study included a large, racial/ethnically diverse prospective cohort; linkage to a cancer registry for virtually complete outcome ascertainment; the availability of detailed individual-level data; and the availability of comprehensive and geographically extensive, small area-level data on SES and built environments15. Limitations included multiple comparisons and a small number of cases in some race/ethnicity-specific analyses; thus, in these stratified analyses we focused on the trend across levels of the neighborhood attributes. In addition, a large number of participants were excluded due to missing covariate information. When models were rerun including participants with missing covariate data by including categories for missing data, results were similar (data not shown), although the association with traffic density in males was attenuated somewhat and no longer statistically significant at p<0.05 (HR=1.20, 95% CI: 0.98–1.48, p=0.076, for quintile 5 vs. quintile 1, p-trend=0.14) and the p-trend for unhealthy retail food in white males was statistically significant at p<0.05 (HR=2.11, 95% CI: 1.12–3.97, p=0.021, for quintile 4 vs. quintile 1, p-trend=0.041). A potential limitation is that we did not adjust for CRC screening, which can be lower in low SES groups30. However, results were similar when adjusting for CRC screening in the subcohort (75%) with information on screening (data not shown). Also, neighborhood boundaries, which were based on pre-defined census block groups, may not represent perceived neighborhood environments. However, it is plausible that perceived neighborhoods correlate well with census block groups and the use of pre-defined census areas is an efficient and cost-effective way to examine a large number of neighborhood attributes31, 32. Another limitation was the use of residential neighborhoods, as individuals may spend a substantial amount of time outside of their residence each day28. In addition, residential neighborhood was assessed at baseline, which reflected only one point in time and may not capture the critical exposure window for CRC development. When we repeated the analysis using a time-dependent approach based on the residential history for the entire study period, we found no associations with CRC risk in females and no meaningful associations in males, including no significant association with traffic density and an attenuation of the positive association (HR=1.05, 95% CI: 0.86–1.29 for quintile 5 vs. quintile 1). Future studies should include assessments of the work environment, account for changes in the neighborhood environment and changes in residence, and evaluate the neighborhood environment across the lifecourse28.

In conclusion, neighborhood obesogenic characteristics were not strongly associated with risk of CRC in the large racial/ethnically diverse MEC cohort. However, increased risk associated with traffic density warrants further follow-up.

Supplementary Material

Supplemental

Acknowledgments

Funding sources: This work was supported by National Cancer Institute grant R01 CA154644. The Multiethnic Cohort Study was supported by National Cancer Institute grant U01 CA164973. The development of the California Neighborhoods Data System was supported by National Cancer Institute grant R03 CA117324 and by a Rapid Response Surveillance Study from the SEER program under a modification to contract N01-PC-35136.

References

  • 1.World Cancer Research Fund/American Institute for Cancer Research. Policy and Action for Cancer Prevention. Food, Nutrition, and Physical Activity: a Global Perspective: American Institute for Cancer Research. 2009 [Google Scholar]
  • 2.Nomura AM, Hankin JH, Henderson BE, et al. Dietary fiber and colorectal cancer risk: the multiethnic cohort study. Cancer Causes Control. 2007;18:753–764. doi: 10.1007/s10552-007-9018-4. [DOI] [PubMed] [Google Scholar]
  • 3.Park SY, Murphy SP, Wilkens LR, Nomura AM, Henderson BE, Kolonel LN. Calcium and vitamin D intake and risk of colorectal cancer: the Multiethnic Cohort Study. Am J Epidemiol. 2007;165:784–793. doi: 10.1093/aje/kwk069. [DOI] [PubMed] [Google Scholar]
  • 4.Ollberding NJ, Nomura AM, Wilkens LR, Henderson BE, Kolonel LN. Racial/ethnic differences in colorectal cancer risk: the multiethnic cohort study. Int J Cancer. 2011;129:1899–1906. doi: 10.1002/ijc.25822. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Glass TA, McAtee MJ. Behavioral science at the crossroads in public health: extending horizons, envisioning the future. Soc Sci Med. 2006;62:1650–1671. doi: 10.1016/j.socscimed.2005.08.044. [DOI] [PubMed] [Google Scholar]
  • 6.Lovasi GS, Hutson MA, Guerra M, Neckerman KM. Built environments and obesity in disadvantaged populations. Epidemiol Rev. 2009;31:7–20. doi: 10.1093/epirev/mxp005. [DOI] [PubMed] [Google Scholar]
  • 7.Odoms-Young AM, Zenk SN, Karpyn A, Ayala GX, Gittelsohn J. Obesity and the food environment among minority groups. Current Obesity Reports. 2012;1:141–151. [Google Scholar]
  • 8.Conroy S, Shariff-Marco S, Yang J, et al. Characterizing the Neighborhood Obesogenic Environment in the Multiethnic Cohort: a Multi-level Resource for Cancer Research. Cancer Causes & Control. doi: 10.1007/s10552-017-0980-1. submitted. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Keegan THM, Hurley S, Goldberg D, et al. The association between neighborhood characteristics and body size and physical activity in the California Teachers Study cohort. Am J Public Health. 2012;102:689–697. doi: 10.2105/AJPH.2011.300150. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.King WC, Belle SH, Brach JS, Simkin-Silverman LR, Soska T, Kriska AM. Objective Measures of Neighborhood Environment and Physical Activity in Older Women. American Journal of Preventive Medicine. 2005;28:461–469. doi: 10.1016/j.amepre.2005.02.001. [DOI] [PubMed] [Google Scholar]
  • 11.Yang W, Spears K, Zhang F, Lee W, Himler HL. Evaluation of Personal and Built Environment Attributes to Physical Activity: A Multilevel Analysis on Multiple Population-Based Data Sources. Journal of Obesity. 2012;2012:548910. doi: 10.1155/2012/548910. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Doubeni CA, Laiyemo AO, Major JM, et al. Socioeconomic status and the risk of colorectal cancer: an analysis of more than a half million adults in the National Institutes of Health-AARP Diet and Health Study. Cancer. 2012;118:3636–3644. doi: 10.1002/cncr.26677. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Kim D, Masyn KE, Kawachi I, Laden F, Colditz GA. Neighborhood socioeconomic status and behavioral pathways to risks of colon and rectal cancer in women. Cancer. 2010;116:4187–4196. doi: 10.1002/cncr.25195. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Kolonel LN, Henderson BE, Hankin JH, et al. A multiethnic cohort in Hawaii and Los Angeles: baseline characteristics. Am J Epidemiol. 2000;151:346–357. doi: 10.1093/oxfordjournals.aje.a010213. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Gomez SL, Glaser SL, McClure LA, et al. The California Neighborhoods Data System: a new resource for examining the impact of neighborhood characteristics on cancer incidence and outcomes in populations. Cancer Causes Control. 2011;22:631–647. doi: 10.1007/s10552-011-9736-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Yost K, Perkins C, Cohen R, Morris C, Wright W. Socioeconomic status and breast cancer incidence in California for different race/ethnic groups. Cancer Causes Control. 2001;12:703–711. doi: 10.1023/a:1011240019516. [DOI] [PubMed] [Google Scholar]
  • 17.Berrigan D, Pickle LW, Dill J. Associations between street connectivity and active transportation. Int J Health Geogr. 2010;9:20. doi: 10.1186/1476-072X-9-20. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.California Certified Farmers' Market Database: California Department of Food and Agriculture. 2010 [Google Scholar]
  • 19.Designed for Disease: The Link Between Local Food Environments and Obesity and Diabetes: California Center for Public Health Advocacy, PolicyLink, and the UCLA Center for Health Policy Research. 2008 [Google Scholar]
  • 20.Lin DY, Wei LJ. The robust inference for the Cox Proportional Hazards Model. J Am Stat Assoc. 1989;84:1074–1078. [Google Scholar]
  • 21.Therneau TM, Grambsch PM, Pankratz VS. Penalized survival models and frailty. Journal of computational and graphical statistics. 2003;12:156–175. [Google Scholar]
  • 22.McGinn AP, Evenson KR, Herring AH, Huston SL, Rodriguez DA. Exploring Associations between Physical Activity and Perceived and Objective Measures of the Built Environment. Journal of Urban Health. 2007;84:162–184. doi: 10.1007/s11524-006-9136-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Keegan TH, Shariff-Marco S, Sangaramoorthy M, et al. Neighborhood influences on recreational physical activity and survival after breast cancer. Cancer Causes Control. 2014;25:1295–1308. doi: 10.1007/s10552-014-0431-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Haskell WL, Lee IM, Pate RR, et al. Physical activity and public health: updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Med Sci Sports Exerc. 2007;39:1423–1434. doi: 10.1249/mss.0b013e3180616b27. [DOI] [PubMed] [Google Scholar]
  • 25.Williams DR, Sternthal M. Understanding Racial/ethnic Disparities in Health: Sociological Contributions. J Health Soc Behav. 2010;51(Suppl):S14–S27. doi: 10.1177/0022146510383838. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Krieger N, Williams DR, Moss NE. Measuring social class in US public health research: concepts, methodologies, and guidelines. Annu Rev Public Health. 1997;18:341–378. doi: 10.1146/annurev.publhealth.18.1.341. [DOI] [PubMed] [Google Scholar]
  • 27.Krieger N, Chen JT, Waterman PD, Soobader MJ, Subramanian SV, Carson R. Geocoding and monitoring of US socioeconomic inequalities in mortality and cancer incidence: does the choice of area-based measure and geographic level matter?: the Public Health Disparities Geocoding Project. Am J Epidemiol. 2002;156:471–482. doi: 10.1093/aje/kwf068. [DOI] [PubMed] [Google Scholar]
  • 28.Gomez SL, Shariff-Marco S, DeRouen M, et al. The impact of neighborhood social and built environment factors across the cancer continuum: Current research, methodological considerations, and future directions. Cancer. 2015;121:2314–2330. doi: 10.1002/cncr.29345. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Ladabaum U, Clarke CA, Press DJ, et al. Colorectal Cancer Incidence in Asian Populations in California: Effect of Nativity and Neighborhood-Level Factors. Am J Gastroenterol. 2014;109:579–588. doi: 10.1038/ajg.2013.488. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Doubeni CA, Laiyemo AO, Reed G, Field TS, Fletcher RH. Socioeconomic and racial patterns of colorectal cancer screening among Medicare enrollees in 2000 to 2005. Cancer Epidemiol Biomarkers Prev. 2009;18(8):2170–2175. doi: 10.1158/1055-9965.EPI-09-0104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Cheng I, Shariff-Marco S, Koo J, et al. Contribution of the neighborhood environment and obesity to breast cancer survival: the California Breast Cancer Survivorship Consortium. Cancer Epidemiol Biomarkers Prev. 2015;24:1282–1290. doi: 10.1158/1055-9965.EPI-15-0055. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Diez-Roux AV. Neighborhoods and health: where are we and were do we go from here?: Environnement résidentiel et santé: état de la question et perspectives pour le futur. Revue d'epidemiologie et de sante publique. 2007;55:13–21. doi: 10.1016/j.respe.2006.12.003. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental
NIHMS907947-supplement-Supplemental.docx (70KB, docx)

RESOURCES