Background: Physical activity (PA) protects against cancer and enhances cancer survivorship. Given high inactivity rates nationwide, population-level physical activity facilitators are needed. Several authoritative bodies have recognized that zoning and planning helps create activity-friendly environments. This study examined the association between activity-friendly zoning, inactivity, and cancer in 478 of the most populous U.S. counties.

Methods: County geocodes linked county-level data: cancer incidence and smoking (State Cancer Profiles), inactivity (Behavioral Risk Factor Surveillance System), 11 zoning measures (compiled by the study team), and covariates (from the American Community Survey and NAVTEQ). For each zoning measure, single mediation regression models and Sobel tests examined whether activity-friendly zoning was associated with reduced cancer incidence, and whether inactivity mediated those associations. All models were clustered on state with robust SEs and significance at the P < 0.05 level.

Results: Zoning for crosswalks, bike–pedestrian connectivity, and bike–pedestrian trails/paths were associated with reduced cancer incidence (β between −0.71 and −1.27, P < 0.05), about 1 case per 100,000 for each 10 percentage-point increase in county population exposure to zoning. Except for crosswalks, each association was mediated by inactivity. However, county smoking attenuated these results, with only crosswalks remaining significant. Results were similar for males (with zoning for bike–pedestrian connectivity, street connectivity, and bike–pedestrian trails/paths), but not females, alone.

Conclusions: Zoning can help to create activity-friendly environments that support decreased inactivity, and possibly reduced cancer incidence.

Impact: Given low physical activity levels nationwide, cross-sectoral collaborations with urban planning can inform cancer prevention and public health efforts to decrease inactivity and cancer. Cancer Epidemiol Biomarkers Prev; 26(4); 578–86. ©2017 AACR.

See all the articles in this CEBP Focus section, “Geospatial Approaches to Cancer Control and Population Sciences.”

Cancer incidence continues to rank among one of the top 10 leading causes of death in the United States and accounts for 23% of all deaths (1, 2), but cancer-related deaths are declining. The age-adjusted cancer-related death rate decreased by 15% between 2003 and 2013, from 190.9 to 163.2 deaths per 100,000 population (1). Although year-to-year changes over the last 5 years in cancer-related death rates have remained stable among women and decreased slightly among men (1.8%; ref. 2), longer time trends suggest that all-cause death rates are decreasing faster for women (11% decrease), compared with 6% among men ages 55–64 years due to lower cancer- and heart disease–related deaths (1). In addition, the number of cancer survivors is continuing to grow, with the number of people who have lived 5 years or more after their cancer diagnosis projected to increase approximately 37% to 11.9 million (3). These simultaneous trends of increasing cancer incidence and decreasing cancer mortality have contributed to a need for more focused research on modifiable lifestyle factors, including physical activity levels that can protect against cancer and improve post-diagnostic quality of life.

Current research supports physical activity as a protective factor against cancer development and as a factor in improved prognosis for cancer survivors. Cancer and physical activity research are of critical importance for the general population, as approximately one-third of adults worldwide are currently inactive, with the trend starting early in life (4, 5). Current research supports an association between physical activity and a reduction in risk of certain cancers in the general population (6, 7). Leisure time physical activity (LTPA) has been related to reductions in postmenopausal breast cancer (6, 8–10), colon (6, 11, 12), and endometrial cancers (13, 14). Mixed results support associations between moderate levels of physical activity (higher metabolic levels than LTPA) and pancreatic (15–17) and rectal cancer (6, 11) risk reductions. One recent systematic review and meta-analysis also found an association between LTPA and an 11% reduction in pancreatic cancer risk, with the strongest effects in younger populations (18). A recent study also found significant reductions in esophageal adenocarcinoma, liver, lung, kidney, myeloid leukemia, myeloma, gastric cardia, bladder, breast, colon, rectal, endometrial, and head and neck cancers, but increased risk in prostate and malignant melanoma associated with LTPA (6). Increased physical activity, especially after cancer treatment, has been related to improved prognosis, better physical functioning, reduced fatigue and bodily pain, and overall quality of life (7, 19–22), underscoring the need for physical activity promotion among cancer survivors (5, 23).

Studies vary widely in both the examination of and findings on gender differences in the relationship between physical activity and cancer risk. One study found a positive dose–response between age and inactivity among men, who in general had higher levels of activity, but not among women (24). However, many studies find consistent results for both men and women, particularly for colon cancer (12). Although the link between physical activity and cancer risk has received increasing attention, virtually all published studies indicate a need for additional research on ways to encourage physical activity among those affected by cancer and increase LTPA among healthy adults and gender-specific targeted programs may be most effective in some cases.

The implementation of policies to improve infrastructure for physical activity have been shown to be an effective means for improving physical activity, which is critically important among the general population where physical activity remains stagnant and among cancer survivors where only 4.5% are meeting physical activity recommendations of 150 minutes/week (25). Community and street-scale urban design policies have been shown to impact people's lifestyles and physical activity levels (26–30). Such policies are typically incorporated into a community's zoning code and design standards. Examples of the types of activity-friendly infrastructure that are typically included in zoning codes are requirements for more compact development with a mix of residential, commercial, retail, and recreational destinations; traditional neighborhood design that provides street and sidewalk connectivity; transportation infrastructure; and proximity to recreational areas/facilities (26). A previous study conducted by the current study team found that all of the activity-friendly zoning measures examined here, except crosswalks, were associated with reduced physical inactivity among adults ages 18–64 years (31). Reductions ranged from 11% to 16% for the individual zoning measures after adjusting for a variety of individual and county characteristics using multivariate generalized linear models. This was consistent with other previous studies by the study team finding positive associations between activity-friendly zoning and adult walking/biking (32) and active travel to work (33, 34).

This study is the first to our knowledge to examine the associations among activity-friendly zoning provisions, physical inactivity, and cancer incidence. It was conducted between May 2012 and August 2016. The University of Illinois at Chicago (UIC) Institutional Review Board deemed that this study did “not involve human subjects” (research protocol #2011-0880). The study design is cross-sectional and links 2010 zoning code data, aggregate 2009–2013 data on cancer incidence, inactivity, and county characteristics, aggregate 2008–2010 data on smoking, and aggregate 2007–2011 data on walkability.

Study sample

The initial sample frame for this study included a purposeful sample of the 496 most populous U.S. counties for which zoning data were compiled, which collectively represented approximately 74.26% of the U.S. population based on 2010 Census estimates. County-level data on cancer incidence were not available for the states of Kansas, Minnesota, and Nevada, which eliminated 16 counties, and data from Michigan may have been limited but was not excluded (did not include cases diagnosed in other states). The final analytic sample (N = 478 counties in 46 states and DC), male-only sample (N = 432 counties in 46 states and DC), and female-only sample (N = 395 counties in 42 states and DC) excluded counties with missing childhood cancer data, as it was impossible to compute adult cancer incidence for those counties. Analyses controlling for smoking excluded one additional county due to missing data, which reduced state coverage by one. Excluded counties were similar to those included in the analytic sample in terms of race, income, poverty, median age, and walkability, with the exception that excluded counties for the combined-gender analyses had a significantly lower percentage of non-Hispanic blacks, and excluded counties for the female-only analyses had significantly higher percentages of non-Hispanic whites, lower percentages of non-Hispanic blacks and Hispanics, and lower walkability scales. The analytic sample covered 68%–72% of the U.S. population, depending on the specific analysis being conducted.

Data sources and measures

Cancer incidence.

Pooled, age-adjusted annual cancer incidence rates per 100,000 for years 2009–2013 at the county level were obtained from the State Cancer Profiles website, a collaboration of the National Cancer Institute (NCI) and the Centers for Disease Control and Prevention (CDC; ref. 35). These rates included invasive cancer only (except bladder cancer, which was invasive or in situ) at all sites. Cancer incidence rates for adults ages 20 years and older were computed by subtracting the incidence of childhood cancers from incidence rates for all ages. The time period for the incidence data was the only one available through our data source, and the most current publicly available data. These county-aggregated incidence rates are based on non-publicly available individual-level incidence data from the CDC's National Program of Cancer Registries Cancer Surveillance System (NPCR-CSS) and the NCI's Surveillance, Epidemiology, and End Results (SEER) Program. The NPCR and SEER jointly collect data for the entire U.S. population (36). Medical facilities such as hospitals report cancer cases to their local registry, which reports to central state cancer registries that in turn submit data annually to NPCR and SEER. Data collected included the type, extent, and location of the cancer, the type of initial treatment, and outcomes, together with demographic information. Information regarding primary or secondary site was not available. In addition, due to the low incidence of specific cancers by county, all cancers were combined for the primary analysis. Sensitivity analyses detailed below describe cancer type–specific results, as a secondary analysis. Age-adjusted incidence rates were calculated using SEER*Stat with population counts for denominators based on Census populations as modified by NCI.

Inactivity.

Average inactivity rates across the 5-year period 2009–2013 (to correspond with the available cancer incidence data) were computed based on yearly data on age-adjusted LTPA level prevalence among adults age 20 and older at the county level obtained from the CDC Diabetes Data and Statistics County Data Indicators (37). These data were model-based estimates derived from data from the CDC's Behavioral Risk Factor Surveillance System (BRFSS) and the U.S. Census Bureau's Population Estimates Program (38). The BRFSS is a state-based monthly telephone survey of adults designed to collect data on behavioral health risks and use of preventive health services which collects data in all 50 states, the District of Columbia, and three U.S. territories, completing more than 400,000 interviews per year (39). BRFSS respondents were counted as inactive if they answered no to the question “During the past month, other than your regular job, did you participate in any physical activities or exercises such as running, calisthenics, golf, gardening, or walking for exercise?”

A systematic review supported the findings of BRFSS with substantial (k > 0.6) agreement found when assessing reliability among persons who were categorized at the vigorous level of physical activity (40). And although there was some variation in validity as assessed by different data collection methods, trends were similar when comparing survey results over time (40).

Smoking.

Rates of current smoking among adults ages 18+ years at the county level were obtained from the State Cancer Profiles (41). Model-based estimates were obtained at the county level based on 2008–2010 BRFSS and National Health Interview Survey (NHIS) data.

Zoning codes.

Zoning codes were collected as part of a larger project to examine the association between zoning code reforms and activity-friendly zoning provisions nationwide and adult LTPA and commuting-related physical activity (32, 33). Zoning codes were collected by Master's level urban planners via Internet research with 100% verification via telephone and/or electronic mail with all jurisdictions of interest. To allow an implementation lag for the larger study for which these data were collected (32), we obtained the version in effect as of 2010 if the zoning code had been updated more recently. Although the effective date was 2010, a review of the zoning codes suggests that the vast majority of the provisions examined herein were effective at least several years prior to 2010. Zoning codes were compiled for all municipal jurisdictions and unincorporated areas in the 478 counties in the sample, excluding only those areas representing less than 0.5% of the given county population. The 0.5% population restriction reduced the sample from 6,515 municipal jurisdictions and unincorporated areas to 4,344. However, only very small jurisdictions were excluded that represented 2.12% of the initial sample population, and that would make little difference in the population-weighted county-aggregated measures used in this analysis. We could not obtain 151 municipal zoning codes and 2 county zoning codes (applicable to unincorporated areas), again generally in places with very small populations, so our final set of zoning data covered 4,191 municipal jurisdictions and unincorporated areas representing 97.63% of the population of the 478 counties in the sample.

Master's level urban planners assessed the zoning codes using an audit tool (see the Supplementary Material) and detailed coding protocol developed by the study team to evaluate the zoning codes for the presence of the following 11 markers of activity-friendly infrastructure across eight types of districts/zones within each jurisdiction's zoning code: sidewalks; crosswalks; bike lanes; bike parking; bike–pedestrian trails/paths; bike–pedestrian connectivity; street connectivity; mixed use; active recreation (e.g., playgrounds, athletic fields, recreation facilities); passive recreation (e.g., open space, parks); and other general walkability provisions (e.g., traffic calming or pedestrian measures). Coders were tested for inter-rater reliability and were not allowed to code independently until reaching at least a 90% rate of agreement. Two Research Electronic Data Capture (REDCap) databases were developed to record policy collection and coding (42). For this analysis, 11 dichotomous (yes/no) measures were created to reflect the presence of each of the activity-friendly markers across any district/zone.

Jurisdiction-level zoning data were aggregated to the county level to link with the other data. Each jurisdiction-level zoning variable was multiplied by the proportion of the county population represented by the jurisdiction, and then these weighted measures were summed across each county to produce the county-aggregated measures (32). The measures were multiplied by 10 so that a one-unit increase would correspond to a 10 percentage-point increase in population exposure to the zoning measure. For instance, if 25% of a given county's population lived in a jurisdiction with zoning for bike–pedestrian connectivity, then the bike–pedestrian connectivity zoning variable would equal 2.5 for that county. Jurisdictions for which the zoning code could not be obtained were not included in the county population denominator for purposes of this computation.

American Community Survey (ACS).

County controls were obtained from the Census Bureau's American Community Survey (ACS) 2009–2013 5-year estimates (again to correspond to the time frame for the cancer data). The 5-year estimates are the most precise available (43). ACS controls included tertiles of median household income (computed separately for each analytic sample), the percentage of households in poverty, percent non-Hispanic white, percent non-Hispanic black, percent Hispanic, median age, and region.

NAVTEQ.

ArcGIS 9.1 software was used to access NAVTEQ 2011 data with third quarter updates. NAVTEQ data provided county-level totals of 4-way intersections as well as all street level intersections. A standardized walkability scale was constructed using NAVTEQ 2011 and ACS 2007–2011 data. The walkability scale is standardized and adjusted by a factor of one to decrease negative scale values and is a summated scale of four density measures: the ratio of four-way intersections to all intersections (NAVTEQ), intersection density or the total number of intersections in the county divided by the county land area (NAVTEQ), housing unit density (ACS), and population density (ACS). The walkability scale is based on the scale created by Slater and colleagues (44) which was adapted from the scale created and updated by Reid Ewing and colleagues (45).

Statistical analysis

Using county-level Federal Information Processing Standards (FIPS) geocodes, data on cancer (2009–2013), inactivity (2009–2013), zoning (2010), ACS (2009–2013), smoking (2008–2010), and NAVTEQ (2011) were merged. The 2010 effective date for the zoning data was the only available data on activity-friendly zoning. As this analysis is cross-sectional and only intended to explore correlations rather than causation, we felt that the data periods were sufficient, particularly given our anecdotal knowledge that many of our jurisdictions' zoning provisions had been “on-the-books” for a number of years prior to our collection effective date of 2010.

Single mediation regression models were used to test whether activity-friendly zoning provisions were associated with reduced cancer incidence, and whether those associations were mediated by inactivity. Models were run separately for each zoning measure. Mediation was tested using two approaches: (i) causal steps (46) and (ii) the Sobel test (47). Testing the causal steps requires three separate regressions: a regression of inactivity on the zoning measure, a regression of cancer incidence on the zoning measure without controlling for inactivity, and a regression of cancer incidence on the zoning measure and inactivity. In addition, the magnitude of the coefficient relating the zoning measure to cancer incidence must be larger in the model that does not control for inactivity than in the model that does control for inactivity. For those zoning measures that met the causal steps criteria, the Sobel test was incorporated, which tests whether the indirect association of the independent variable of interest with the dependent variable through the mediator is significant. To explore the gender–cancer–activity connections, we also examined the mediation models using gender-specific cancer and inactivity measures. Also, due to the important role of smoking in cancer incidence, all models were run with and without controlling for gender-specific smoking.

For all models, coefficients on the zoning variables correspond to a 10 percentage-point increase in county-level population exposure to the zoning measure. All models were clustered on state with robust SEs. Statistical significance was determined at the P < 0.05 level. All analyses were conducted in Stata/SE 13.1 (48).

Table 1 shows descriptive statistics for the analytic sample. Average cancer incidence was nearly 435 cases per 100,000, with an average of just over one-fifth of the population being inactive and about one-fifth currently smoking. Average county-level population exposure to the zoning measures ranged from 17% for bike lanes to 81% for active and passive recreation. On average, counties were composed of mostly non-Hispanic white individuals (68%) with only 13% of households in poverty, and a median age of 37 years. Slightly more than 40% of the counties were located in the South, with the other counties being almost equally distributed across the West (19%), Midwest (20%), and Northeast (21%).

Table 1.

Characteristics of the study samplea

Variable% (or mean)SDMinimumMaximum
Outcome and mediator variables 
 Cancer incidenceb (mean) 434.91 38.78 268.20 522.90 
 Male cancer incidence (mean) 488.56 49.58 328.80 606.30 
 Female cancer incidence (mean) 398.56 34.15 280.40 468.80 
 Inactivity 22.51 4.24 10.10 35.30 
 Male inactivity 20.79 3.83 9.32 31.86 
 Female inactivity 24.02 4.72 10.80 38.38 
Zoning provisions addressed 
 Sidewalks 69.48 32.52 100 
 Crosswalks 25.87 33.48 100 
 Bike–pedestrian connectivity 42.50 37.54 100 
 Street connectivity 40.35 37.32 100 
 Bike lanes 16.64 29.14 100 
 Bike parking (proxy for street furniture) 40.42 38.62 100 
 Bike–pedestrian trails/paths 56.82 37.97 100 
 Other walkability (e.g, pedestrian plazas) 67.71 33.29 100 
 Mixed use 64.81 33.99 100 
 Active recreation 81.22 28.83 100 
 Passive recreation 81.10 29.11 100 
County controls 
 West 18.62 38.97   
 Midwest 20.08 40.10   
 Northeast 20.92 40.72   
 South 40.38 49.12   
 Median household income tertiles: 
  Low ($29,806.00–$48,450.00) 33.05 47.09   
  Middle (>$48,450.00–$58,745.00) 33.47 47.24   
  High (>$58,745.00–$122,238.00) 33.47 47.24   
 % Households in poverty 13.41 4.71 3.18 31.67 
 % Non-Hispanic white 67.62 19.15 3.57 95.65 
 % Non-Hispanic black 12.04 12.70 0.27 69.45 
 % Hispanic 13.44 14.74 0.87 95.50 
 Median age (mean) 37.40 4.22 24.40 56.70 
 Walkability scale (mean) 1.00 1.02 0.72 18.04 
 Current smoking rate 20.17 5.16 1.80 41.60 
 Male current smoking rate 22.61 5.25 6.70 44.50 
 Female current smoking rate 18.12 5.37 0.90 38.30 
Variable% (or mean)SDMinimumMaximum
Outcome and mediator variables 
 Cancer incidenceb (mean) 434.91 38.78 268.20 522.90 
 Male cancer incidence (mean) 488.56 49.58 328.80 606.30 
 Female cancer incidence (mean) 398.56 34.15 280.40 468.80 
 Inactivity 22.51 4.24 10.10 35.30 
 Male inactivity 20.79 3.83 9.32 31.86 
 Female inactivity 24.02 4.72 10.80 38.38 
Zoning provisions addressed 
 Sidewalks 69.48 32.52 100 
 Crosswalks 25.87 33.48 100 
 Bike–pedestrian connectivity 42.50 37.54 100 
 Street connectivity 40.35 37.32 100 
 Bike lanes 16.64 29.14 100 
 Bike parking (proxy for street furniture) 40.42 38.62 100 
 Bike–pedestrian trails/paths 56.82 37.97 100 
 Other walkability (e.g, pedestrian plazas) 67.71 33.29 100 
 Mixed use 64.81 33.99 100 
 Active recreation 81.22 28.83 100 
 Passive recreation 81.10 29.11 100 
County controls 
 West 18.62 38.97   
 Midwest 20.08 40.10   
 Northeast 20.92 40.72   
 South 40.38 49.12   
 Median household income tertiles: 
  Low ($29,806.00–$48,450.00) 33.05 47.09   
  Middle (>$48,450.00–$58,745.00) 33.47 47.24   
  High (>$58,745.00–$122,238.00) 33.47 47.24   
 % Households in poverty 13.41 4.71 3.18 31.67 
 % Non-Hispanic white 67.62 19.15 3.57 95.65 
 % Non-Hispanic black 12.04 12.70 0.27 69.45 
 % Hispanic 13.44 14.74 0.87 95.50 
 Median age (mean) 37.40 4.22 24.40 56.70 
 Walkability scale (mean) 1.00 1.02 0.72 18.04 
 Current smoking rate 20.17 5.16 1.80 41.60 
 Male current smoking rate 22.61 5.25 6.70 44.50 
 Female current smoking rate 18.12 5.37 0.90 38.30 

aN = 478 counties, except 477 for smoking, 432 for male cancer, and 395 for female cancer.

bCancer incidence rates are annual rates per 100,000.

First, analyses examined the separate association between each individual zoning measure and cancer rates adjusting for county controls, but without controlling for inactivity, and these results are displayed in Table 2. Without the smoking control, zoning for crosswalks (β = −1.27; 95% CI, −2.33 to −0.22), bike–pedestrian connectivity (β = −0.88; 95% CI, −1.73 to −0.03), and bike–pedestrian trails/paths (β = −0.71; 95% CI, −1.41 to −0.01) were significantly associated with reduced cancer incidence. These results were attenuated once county smoking was included in the model. Only zoning for crosswalks (β = −1.14; 95% CI, −2.25 to −0.03) remained significantly associated with reduced cancer incidence.

Table 2.

Multivariate regressions of cancer incidence on zoning provisions addresseda

Without smoking controlWith smoking control
Zoning provision addressedCoefficient95% CICoefficient95% CI
Sidewalks −0.24 −1.41–0.93 0.00 −1.21–1.21 
Crosswalks −1.27b −2.33 to −0.22 −1.14b −2.25 to −0.03 
Bike–pedestrian connectivity −0.88b −1.73 to −0.03 −0.73 −1.65–0.18 
Street connectivity −0.68 −1.50–0.13 −0.59 −1.41–0.23 
Bike lanes −0.60 −1.80–0.60 −0.40 −1.62–0.82 
Bike parking 0.13 −0.84–1.11 0.40 −0.58–1.38 
Bike−pedestrian trails/paths −0.71b −1.41 to −0.01 −0.49 −1.23–0.25 
Other walkability 0.05 −0.97–1.06 0.30 −0.68–1.29 
Mixed use 0.46 −0.63–1.55 0.74 −0.35–1.83 
Active recreation 0.89 −0.51–2.29 1.09 −0.37–2.55 
Passive recreation 0.98 −0.43–2.40 1.23 −0.24–2.70 
Without smoking controlWith smoking control
Zoning provision addressedCoefficient95% CICoefficient95% CI
Sidewalks −0.24 −1.41–0.93 0.00 −1.21–1.21 
Crosswalks −1.27b −2.33 to −0.22 −1.14b −2.25 to −0.03 
Bike–pedestrian connectivity −0.88b −1.73 to −0.03 −0.73 −1.65–0.18 
Street connectivity −0.68 −1.50–0.13 −0.59 −1.41–0.23 
Bike lanes −0.60 −1.80–0.60 −0.40 −1.62–0.82 
Bike parking 0.13 −0.84–1.11 0.40 −0.58–1.38 
Bike−pedestrian trails/paths −0.71b −1.41 to −0.01 −0.49 −1.23–0.25 
Other walkability 0.05 −0.97–1.06 0.30 −0.68–1.29 
Mixed use 0.46 −0.63–1.55 0.74 −0.35–1.83 
Active recreation 0.89 −0.51–2.29 1.09 −0.37–2.55 
Passive recreation 0.98 −0.43–2.40 1.23 −0.24–2.70 

aN = 478 counties without smoking control and N = 477 counties with smoking control containing 72% of the U.S. population located in 46 states (45 with the smoking control) and the District of Columbia. Cancer incidence rates are annual rates per 100,000. Each row corresponds to two separate regression models, with and without a smoking control, but including all other county controls except inactivity. Coefficients correspond to a 10 percentage-point increase in county-level population exposure to the given zoning provision. All models were clustered on state with robust SEs.

bP < 0.05.

Next, for the zoning measures that were significantly associated with cancer rates, mediation analyses tested for potential mediation of these associations by inactivity. With the smoking control, there were no significant mediations, and zoning for crosswalks was not significantly associated with cancer incidence when also controlling for inactivity. Models that had significant results for both causal steps and Sobel tests without the smoking control are displayed in Fig. 1A and B. Without the smoking control, zoning for crosswalks had a significant direct association with reduced cancer incidence even when controlling for inactivity (β = −1.09; 95% CI, −2.15 to −0.03), but as it was not significantly associated with inactivity it failed the first of the causal steps criteria for testing mediation. Figure 1A and B show mediation analysis path diagrams for the remaining two zoning provisions that were significantly associated with cancer (bike–pedestrian connectivity and bike–pedestrian trails/paths). For each figure, path a shows the association of the zoning measure with inactivity; path b shows the association of inactivity with cancer; path c shows the direct association of the zoning measure with cancer incidence; and path c′ shows the association of the zoning measure with cancer incidence after accounting for inactivity. In both cases, the remaining three causal steps are satisfied and the Sobel test confirmed significant mediation. Bike–pedestrian connectivity and bike–pedestrian trails/paths were indirectly associated with reduced cancer incidence through an association with inactivity. In both instances, inactivity fully mediated the relationship between zoning and cancer as shown with a significant c path but nonsignificant c′ path. In each case, the total association was about one fewer case of cancer per 100,000 for a 10 percentage-point increase in county-level population exposure to the zoning measure.

Figure 1.

Mediation analysis path diagrams showing separate regression coefficients for each model.

Figure 1.

Mediation analysis path diagrams showing separate regression coefficients for each model.

Close modal

Gender-specific analyses were also conducted. Table 3 presents the results of regressions of cancer incidence on the zoning measures and county controls, excluding inactivity, separately by gender. Among males, bike–pedestrian connectivity, street connectivity, and bike/pedestrian trails and paths were associated with reduced cancer incidence. Mediation by inactivity was found for all three zoning measures, as shown in Fig. 2A–C. However, these findings were completely attenuated by the inclusion of county smoking. Among females, interestingly, although no zoning measures were originally associated with cancer, once county smoking was included in the model, bike parking, other walkability, mixed use, and passive recreation was positively associated with cancer incidence, although mediation by inactivity was not established. Sensitivity analysis conducted by cancer type revealed an increased risk of melanoma associated with LTPA. It is possible that zoning directed at increasing outdoor types of LTPA, while reducing some types of cancers (our sensitivity analysis found reduced incidence of cervical, colon, kidney, and lung associated with LTPA), may also increase exposure to the sun and risk of melanoma. This finding is consistent with Moore and colleagues 2016 (6), which found a positive association between LTPA and melanoma.

Table 3.

Multivariate regressions of cancer incidence on zoning provisions addressed, separately by gendera

MalesFemales
Without smokingWith smokingWithout smokingWith smoking
Zoning provision addressedCoefficient95% CICoefficient95% CICoefficient95% CICoefficient95% CI
Sidewalks −0.90 −2.59–0.78 −0.50 −2.29–1.29 0.03 −1.10–1.17 0.14 −1.04–1.32 
Crosswalks −1.25 −2.66–0.15 −1.01 −2.50–0.48 −0.29 −1.15–0.56 −0.24 −1.09–0.62 
Bike–pedestrian connectivity −1.30b −2.40 to −0.20 −1.09 −2.20–0.02 −0.06 −0.88–0.77 0.02 −0.82–0.86 
Street connectivity −1.22b −2.31 to −0.12 −1.08 −2.17–0.01 −0.03 −0.87–0.81 0.00 −0.84–0.84 
Bike lanes −0.93 −2.68–0.81 −0.69 −2.43–1.04 0.18 −0.88–1.25 0.30 −0.73–1.33 
Bike parking −0.32 −1.61–0.97 0.02 −1.27–1.31 0.79 −0.05–1.64 0.95b 0.11–1.79 
Bike–pedestrian trails/paths −1.11b −2.19 to −0.03 −0.73 −1.78–0.32 −0.13 −0.78–0.53 −0.05 −0.73–0.64 
Other walkability −0.60 −2.20–1.00 −0.28 −1.85–1.30 0.84 −0.03–1.71 0.95b 0.06–1.84 
Mixed use −0.39 −1.94–1.17 0.00 −1.48–1.48 0.89 −0.08–1.86 1.06b 0.11–2.01 
Active recreation −0.38 −2.38–1.62 0.02 −2.01–2.05 1.18 −0.06–2.42 1.26 −0.01–2.53 
Passive recreation −0.23 −2.26–1.80 0.25 −1.81–2.30 1.22 −0.02–2.46 1.31b 0.04–2.58 
MalesFemales
Without smokingWith smokingWithout smokingWith smoking
Zoning provision addressedCoefficient95% CICoefficient95% CICoefficient95% CICoefficient95% CI
Sidewalks −0.90 −2.59–0.78 −0.50 −2.29–1.29 0.03 −1.10–1.17 0.14 −1.04–1.32 
Crosswalks −1.25 −2.66–0.15 −1.01 −2.50–0.48 −0.29 −1.15–0.56 −0.24 −1.09–0.62 
Bike–pedestrian connectivity −1.30b −2.40 to −0.20 −1.09 −2.20–0.02 −0.06 −0.88–0.77 0.02 −0.82–0.86 
Street connectivity −1.22b −2.31 to −0.12 −1.08 −2.17–0.01 −0.03 −0.87–0.81 0.00 −0.84–0.84 
Bike lanes −0.93 −2.68–0.81 −0.69 −2.43–1.04 0.18 −0.88–1.25 0.30 −0.73–1.33 
Bike parking −0.32 −1.61–0.97 0.02 −1.27–1.31 0.79 −0.05–1.64 0.95b 0.11–1.79 
Bike–pedestrian trails/paths −1.11b −2.19 to −0.03 −0.73 −1.78–0.32 −0.13 −0.78–0.53 −0.05 −0.73–0.64 
Other walkability −0.60 −2.20–1.00 −0.28 −1.85–1.30 0.84 −0.03–1.71 0.95b 0.06–1.84 
Mixed use −0.39 −1.94–1.17 0.00 −1.48–1.48 0.89 −0.08–1.86 1.06b 0.11–2.01 
Active recreation −0.38 −2.38–1.62 0.02 −2.01–2.05 1.18 −0.06–2.42 1.26 −0.01–2.53 
Passive recreation −0.23 −2.26–1.80 0.25 −1.81–2.30 1.22 −0.02–2.46 1.31b 0.04–2.58 

aN = 432/431 counties for males without and with smoking control, and N = 395/394 counties for females with and without smoking control, containing 68–70% of the U.S. population located in 41–46 states and the District of Columbia. Cancer incidence rates are annual rates per 100,000. Each row corresponds to four separate regression models, by gender, including all county controls except inactivity. Models were run separately using cancer incidence, inactivity, and smoking among males and females, respectively. Coefficients correspond to a 10 percentage-point increase in county-level population exposure to the given zoning provision. All models were clustered on state with robust SEs.

bP < 0.05.

Figure 2.

Mediation analysis path diagrams for males only showing separate regression coefficients for each model.

Figure 2.

Mediation analysis path diagrams for males only showing separate regression coefficients for each model.

Close modal

To our knowledge, this was the first nationwide study to examine the associations among activity-friendly zoning provisions, physical inactivity, and cancer incidence. Our findings for both genders combined suggest that zoning for crosswalks was associated with reduced cancer incidence, although mediation by inactivity could not be established, while two other zoning provisions, for bike–pedestrian connectivity and bike–pedestrian trails/paths, were associated with reduced cancer incidence with full mediation by inactivity. Although the overall reduction was small at the county level, about one fewer case of cancer per 100,000 for each 10 percentage-point increase in county-level population exposure to the zoning measure, summated across all U.S. counties this could correspond to a significant change in population level cancer rates. Results were substantially attenuated after county smoking was included in the model, but zoning for crosswalks remained significant.

Additionally, analyses were examined separately by gender and results were similar for males compared with the gender combined models, but results were insignificant for females alone. One possible explanation is that the zoning provisions associated with reduced inactivity and cancer were primarily focused on infrastructure that would support bicycling. Several studies have documented that men cycle for recreational and active travel purposes significantly more than women (24, 49, 50). Relatedly, while zoning permits specific types of retail outlets (e.g., commercial gyms, studios, etc.), this study focused specifically on the physical infrastructure that would be conducive for activity. Thus, it could be speculated that the zoning measures captured herein largely focused on outdoor exercise which may be more likely to decrease inactivity levels among males, whereas females may be more inclined to choose indoor LTPA that is not covered by the type of zoning measures captured for this study. It is also possible that the female cancers captured in our combined cancer measure were not as strongly associated with physical activity as the male cancers examined. Our study supports previous research that reducing cancer though physical activity may need to tailor programs to gender-specific needs (24).

The findings in this study are subject to several limitations. First, this study is cross-sectional and findings should be interpreted as associations rather than causation. Second, the aggregate nature of the data at the county level allows for the possibility of an ecological fallacy and future studies will need to examine these pathways with individual level analysis as well. Third, although we attempted to include a policy lag it was not possible to tease out timing of the individual elements. Fourth, while self-reported data are typically deemed less reliable, that was less of a concern here given that our outcome measure, inactivity, is less likely to lead to respondent bias (40); however population level results need to be considered with some caution (51). Fifth, although our sample of counties spans the United States, it focused entirely on the most populous counties that cover approximately 72% of the U.S. population and therefore results cannot be generalized beyond the counties studied. Sixth, due to the small incidence of individual cancers by county and the ecological nature of the study we were unable to completely examine the relationships among zoning, LTPA, and specific cancer, although these secondary results do suggest similar findings on the inactivity-cancer relationship to a recent study (6) with respect to colon, kidney, and lung cancers, and melanoma. Lastly, although we controlled for community walkability infrastructure using proven and reliable methods (44), we did not have objective measures of the built environment directly corresponding to each zoning measure.

Yet this study is the first of its kind to examine a linkage between zoning provisions, inactivity levels, and cancer. We found that living in a county with a higher proportion of population-level exposure to more activity-friendly zoning (especially for crosswalks, bike–pedestrian connectivity and bike–pedestrian trails/paths) was associated with reduced inactivity levels, and in turn, may reduce cancer, especially among males. Given the low physical activity level in the United States, these findings identify an opportunity for the public health and cancer communities to work with urban planners to design communities to be more activity-friendly.

No potential conflicts of interest were disclosed.

The findings and conclusions of this report are those of the authors and do not necessarily represent the official positions of the NIH or NCI.

Conception and design: L.M. Nicholson, J. Leider, J.F. Chriqui

Development of methodology: L.M. Nicholson, J. Leider, J.F. Chriqui

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Leider, J.F. Chriqui

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.M. Nicholson, J. Leider, J.F. Chriqui

Writing, review, and/or revision of the manuscript: L.M. Nicholson, J. Leider, J.F. Chriqui

Study supervision: J.F. Chriqui

The authors would like to gratefully acknowledge the research and zoning coding assistance provided by Emily Thrun, MUPP, Haytham Abu Zayd, MAPSS, Anthony Pelikan, MUPP, Sunny Bhat, MUPP, Erika Strauss, MUPP, Brad Gregorka, MUPP, April Jackson, PhD, MUPP, Nija Fountano, Carmen Aiken, MUPP, and Jennifer Nalbantyan, MUPP.

J.F. Chriqui received a grant from the NCI, NIH (grant number R01CA158035), and the REDCap databases used for policy collection and coding were funded by the University of Illinois at Chicago Center for Clinical and Translational Science located within the Institute for Health Research and Policy at UIC (grant number UL1RR029879).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
National Center for Health Statistics
.
Health, United States, 2014: with special feature on adults aged 55–64
.
Hyattsville, MD
: NCHS; 
2015
.
2.
Siegel
R
,
Miller
K
,
Jemal
A
. 
Cancer statistics, 2015
.
CA Cancer J Clin
2015
;
65
:
29
.
3.
De Moor
JS
,
Mariotto
AB
,
Parry
C
,
Alfano
CM
,
Padgett
L
,
Kent
EE
, et al
Cancer survivors in the United States: prevalence across the survivorship trajectory and implications for care
.
Cancer Epidemiol Biomarkers Prev
2013
;
22
:
561
70
.
4.
Hallal
PC
,
Andersen
LB
,
Bull
FC
,
Guthold
R
,
Haskell
W
,
Ekelund
U
, et al
Global physical activity levels: surveillance progress, pitfalls, and prospects
.
Lancet
2012
;
380
:
247
57
.
5.
Sanchis-Gomar
F
,
Lucia
A
,
Yvert
T
,
Ruiz-Casado
A
,
Pareja-Galeano
H
,
Santos-Lozano
A
, et al
Physical inactivity and low fitness deserve more attention to alter cancer risk and prognosis
.
Cancer Prev Res
2015
;
8
:
105
10
.
6.
Moore
SC
,
Lee
I-M
,
Weiderpass
E
,
Campbell
PT
,
Sampson
JN
,
Kitahara
CM
, et al
Association of leisure-time physical activity with risk of 26 types of cancer in 1.44 million adults
.
JAMA Intern Med
2016
;
176
:
816
25
.
7.
Petrelli
NJ
,
Winer
EP
,
Brahmer
J
,
Dubey
S
,
Smith
S
,
Thomas
C
, et al
Clinical Cancer Advances 2009: major research advances in cancer treatment, prevention, and screening–a report from the American Society of Clinical Oncology
.
J Clin Oncol
2009
;
27
:
6052
69
.
8.
Gonçalves
AK
,
Dantas Florencio
GL
,
Maisonnette de Atayde Silva
MJ
,
Cobucci
RN
,
Giraldo
PC
,
Cote
NM
. 
Effects of physical activity on breast cancer prevention: a systematic review
.
J Phys Act Health
2014
;
11
:
445
54
.
9.
Loprinzi
PD
,
Cardinal
BJ
,
Smit
E
,
Winters-Stone
KM
. 
Physical activity and breast cancer risk
.
J Exerc Sci Fit
2012
;
10
:
1
7
.
10.
Wu
Y
,
Zhang
D
,
Kang
S
. 
Physical activity and risk of breast cancer: a meta-analysis of prospective studies
.
Breast Cancer Res Treat
2013
;
137
:
869
82
.
11.
Harriss
DJ
,
Atkinson
G
,
Batterham
A
,
George
K
,
Cable
NT
,
Reilly
T
, et al
Lifestyle factors and colorectal cancer risk (2): a systematic review and meta-analysis of associations with leisure-time physical activity
.
Colorectal Dis
2009
;
11
:
689
701
.
12.
Martínez
ME
,
Giovannucci
E
,
Spiegelman
D
,
Hunter
DJ
,
Willett
WC
,
Colditz
GA
. 
Leisure-time physical activity, body size, and colon cancer in women. Nurses' Health Study Research Group
.
J Natl Cancer Inst
1997
;
89
:
948
55
.
13.
Keum
N
,
Ju
W
,
Lee
DH
,
Ding
EL
,
Hsieh
CC
,
Goodman
JE
, et al
Leisure-time physical activity and endometrial cancer risk: dose-response meta-analysis of epidemiological studies
.
Int J Cancer
2014
;
135
:
682
94
.
14.
McElroy
JA
,
Bloom
T
,
Moore
K
,
Geden
B
,
Everett
K
,
Bullock
LF
. 
Perinatal mortality and adverse pregnancy outcomes in a low-income rural population of women who smoke
.
Birth Defects Res Part A - Clin Mol Teratol
2012
;
94
:
223
9
.
15.
Behrens
G
,
Jochem
C
,
Schmid
D
,
Keimling
M
,
Ricci
C
,
Leitzmann
MF
. 
Physical activity and risk of pancreatic cancer: a systematic review and meta-analysis
.
Eur J Epidemiol
2015
;
30
:
279
98
.
16.
O'Rorke
MA
,
Cantwell
MM
,
Cardwell
CR
,
Mulholland
HG
,
Murray
LJ
. 
Can physical activity modulate pancreatic cancer risk? a systematic review and meta-analysis
.
Int J Cancer
2010
;
126
:
2957
68
.
17.
Bao
Y
,
Michaud
DS
. 
Physical activity and pancreatic cancer risk: a systematic review
.
Cancer Epidemiol Biomarkers Prev
2008
;
17
:
2671
82
.
18.
Farris
MS
,
Mosli
MH
,
McFadden
AA
,
Friedenreich
CM
,
Brenner
DR
. 
The association between leisure time physical activity and pancreatic cancer risk in adults: a systematic review and meta-analysis
.
Cancer Epidemiol Biomarkers Prev
2015
;
24
:
1462
73
.
19.
Alfano
CM
,
Smith
AW
,
Irwin
ML
,
Bowen
DJ
,
Sorensen
B
,
Reeve
BB
, et al
Physical activity, long-term symptoms, and physical health-related quality of life among breast cancer survivors: a prospective analysis
.
J Cancer Surviv
2007
;
1
:
116
28
.
20.
Bellizzi
KM
,
Rowland
JH
,
Jeffery
DD
,
McNeel
T
. 
Health behaviors of cancer survivors: examining opportunities for cancer control intervention
.
J Clin Oncol
2005
;
23
:
8884
93
.
21.
Kendall
AR
,
Mahue-Giangreco
M
,
Carpenter
CL
,
Ganz
PA
,
Bernstein
L
. 
Influence of exercise activity on quality of life in long-term breast cancer survivors
.
Qual Life Res
2005
;
14
:
361
71
.
22.
Lee
C-D
,
Sui
X
,
Hooker
SP
,
Hébert
JR
,
Blair
SN
. 
Combined impact of lifestyle factors on cancer mortality in men
.
Ann Epidemiol
2011
;
21
:
749
54
.
23.
Kim
RB
,
Phillips
A
,
Herrick
K
,
Helou
M
,
Rafie
C
,
Anscher
MS
, et al
Physical activity and sedentary behavior of cancer survivors and non-cancer individuals: results from a national survey
.
PLoS One
2013
;
8
:
e57598
.
24.
Azevedo
MR
,
Araújo
CLP
,
Reichert
FF
,
Siqueira
FV
,
da Silva
MC
,
Hallal
PC
. 
Gender differences in leisure-time physical activity
.
Int J Public Health
2007
;
52
:
8
15
.
25.
Smith
WA
,
Nolan
VG
,
Robison
LL
,
Hudson
MM
,
Ness
KK
. 
Physical activity among cancer survivors and those with no history of cancer—a report from the National Health and Nutrition Examination Survey 2003–2006
.
Am J Transl Res
2011
;
3
:
342
50
.
26.
Heath
GW
,
Brownson
RC
,
Kruger
J
,
Miles
R
,
Powell
KE
,
Ramsey
LT
. 
The effectiveness of urban design and land use and transport policies and practices to increase physical activity: a systematic review
.
J Phys Act Health
2006
;
3
Suppl 1
:
S55
76
.
27.
Aytur
SA
,
Rodriguez
DA
,
Evenson
KR
,
Catellier
DJ
. 
Urban containment policies and physical activity. A time-series analysis of metropolitan areas, 1990–2002
.
Am J Prev Med
2008
;
34
:
320
32
.
28.
Sallis
JF
,
Cervero
RB
,
Ascher
W
,
Henderson
KA
,
Kraft
MK
,
Kerr
J
. 
An ecological approach to creating active living communities
.
Annu Rev Public Health
2006
;
27
:
297
322
.
29.
Sallis
JF
,
Bauman
A
,
Pratt
M
. 
Environmental and policy interventions to promote physical activity
.
Am J Prev Med
1998
;
15
:
379
97
.
30.
Waxman
A
. 
WHO's global strategy on diet, physical activity and health. Response to a worldwide epidemic of non-communicable diseases
.
Food Nutr Res
2004
;
48
:
58
60
.
31.
Leider
J
,
Chriqui
JF
,
Thrun
E
. 
Associations between active living-oriented zoning and no adult leisure-time physical activity in the US
.
Prev Med.
2016
Jun 27. [Epub ahead of print].
32.
Chriqui
JF
,
Nicholson
LM
,
Thrun
E
,
Leider
J
,
Slater
SJ
. 
More active living-oriented county and municipal zoning is associated with increased adult leisure time physical activity—United States, 2011
.
Env Behav
2016
;
48
:
111
30
.
33.
Chriqui
JF
,
Leider
J
,
Thrun
E
,
Nicholson
L
,
Slater
S
. 
Communities on the move: pedestrian-oriented zoning as a facilitator of adult active travel to work in the United States
.
Front Public Health
2016
;
4
:
71
.
34.
Thrun
E
,
Leider
J
,
Chriqui
JF
. 
Exploring the cross-sectional association between transit-oriented development zoning and active travel and transit usage in the United States, 2010–2014
.
Front Public Health
2016
;
4
:
113
.
35.
National Cancer Institute and Centers for Disease Control and Prevention
. 
Incidence rates tables
.
Available from
: http://www.statecancerprofiles.cancer.gov/incidencerates/.
36.
Centers for Disease Control and Prevention
. 
About the program
.
Available from
: http://www.cdc.gov/cancer/npcr/about.htm.
37.
Centers for Disease Control and Prevention
. 
County data indicators
.
Available from
: http://www.cdc.gov/diabetes/atlas/countydata/County_ListofIndicators.html.
38.
Centers for Disease Control and Prevention
. 
Methods and references for county-level estimates and ranks
.
Available from
: http://www.cdc.gov/diabetes/pdfs/data/calculating-methods-references-county-level-estimates-ranks.pdf.
39.
Centers for Disease Control and Prevention
. 
About BRFSS
.
Available from
: http://www.cdc.gov/brfss/about/.
40.
Pierannunzi
C
,
Hu
SS
,
Balluz
L
. 
A systematic review of publications assessing reliability and validity of the Behavioral Risk Factor Surveillance System (BRFSS), 2004–2011
.
BMC Med Res Methodol
2013
;
13
:
49
.
41.
National Cancer Institute and Centers for Disease Control and Prevention
. 
Screening and risk factors table
.
Available from
: https://statecancerprofiles.cancer.gov/risk/index.php.
42.
UIC Center for Clinical and Translational Science
. 
REDCap: Research electronic data capture
; 
2015
.
Available from
: http://www.ccts.uic.edu/contents/redcap-research-electronic-data-capture.
43.
U.S. Census Bureau
. 
American Community Survey: when to use 1-year, 3-year, or 5-year estimates
; 
2015
.
Available from
: http://www.census.gov/programs-surveys/acs/guideance/estiamtes.html.
44.
Slater
SJ
,
Ewing
R
,
Powell
LM
,
Chaloupka
FJ
,
Johnston
LD
,
O'Malley
PM
. 
The association between community physical activity settings and youth physical activity, obesity, and body mass index
.
J Adolesc Health
2010
;
47
:
496
503
.
45.
Ewing
R
,
Hamidi
S
.
Measuring urban sprawl and validating sprawl measures
.
Metropolitan Research Center
; 
2014
.
Available from
: http://gis.cancer.gov/tools/urban-sprawl/.
46.
Baron
RM
,
Kenny
DA
. 
The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations
.
J Pers Soc Psychol
1986
;
51
:
1173
82
.
47.
Sobel
ME
. 
Asymptotic confidence intervals for indirect effects in structural equation models
.
Sociol Methodol
1982
;
13
:
290
312
.
48.
StataCorp
.
Stata statistical software: Release 13
.
College Station, TX
:
StataCorp LP
; 
2013
.
49.
Krizek
KJ
,
Johnson
PJ
,
Tilahun
N
. 
Gender differences in bicycling behavior and facility preferences
.
In
:
Rosenbloom
S
,
editor
.
Research on women's issues in transportation: Report of a conference
. In: Proceedings of the
Transportation Research Board Conference; 2004 Nov 18–20
; Chicago, IL.
Washington, DC
:
Transportation Research Board
; 
2005
.
p.
31
40
.
50.
Heesch
KC
,
Sahlqvist
S
,
Garrard
J
. 
Gender differences in recreational and transport cycling: a cross-sectional mixed-methods comparison of cycling patterns, motivators, and constraints
.
Int J Behav Nutr Phys Act
2012
;
9
:
106
.
51.
Steene-Johannessen
J
,
Anderssen
SA
,
van der Ploeg
HP
,
Hendriksen
IJM
,
Donnelly
AE
,
Brage
S
, et al
Are self-report measures able to define individuals as physically active or inactive?
Med Sci Sports Exerc
2016
;
48
:
235
44
.