Density of animal feeding operations, including concentrated animal feeding operations (CAFOs), and cancer incidence: A county-level ecological study across three U.S. states.

1.
Introduction
Despite growing evidence that exposure to animal feeding operations (AFOs),
including concentrated animal feeding operations (CAFOs), negatively impacts both
environmental quality and human health (Burkholder et
al. 2007; Son et al. 2024),
relatively few studies have comprehensively evaluated these impacts across diverse
geographic regions and populations. According to the EPA (2024), an AFO is defined as an operation in which animals are kept
and fed in a confined area for at least 45 days within a 12-month period without
sustaining vegetation or crop growth. A CAFO is an AFO that exceeds specific
animal-number thresholds (e.g., 1,000 beef cattle, 700 dairy cows, or 2,500 hogs)
and/or has the potential to release manure or wastewater into the waters of the
United States. AFOs/CAFOs produce substantial amounts of animal waste, which
generates harmful air pollutants (e.g., ammonia, hydrogen sulfide, particulate
matter) and contaminated runoff that contribute to environmental degradation and
increase the potential for human exposure through multiple pathways (Hribar 2010; Thorne
2007). Communities located near these operations, particularly
marginalized populations or those with limited resources, may experience
disproportionate environmental and health burdens (Son et al. 2021; Wing et al.
2000).
Epidemiological studies have reported a variety of adverse health effects
linked to exposure to AFOs/CAFOs, including respiratory symptoms, infectious
diseases, antimicrobial resistance, adverse pregnancy outcomes, and psychological
distress (Casey et al. 2015; Cole et al. 2000; Mirabelli et al. 2006; Quist et al.
2022; Van Dijk et al. 2016).
Emerging evidence also suggested that CAFO-related exposures may be associated with
increased cancer risk, potentially through air and water contamination (e.g.,
nitrates in drinking water, exposure to volatile organic compounds (VOCs),
endocrine-disrupting chemicals) (Kravchenko et al.
2020; Ward et al. 2005). For
example, high nitrate levels in drinking water have been associated with elevated
risk of various cancer types such as colorectal, bladder, and ovarian cancers (Burkholder et al. 2007; Schullehner et al. 2018). Residents living in areas with
intensive livestock farming may also be exposed to VOCs such as benzene and
formaldehyde through airborne emissions (Yuan et al.
2017). While many studies have focused on a specific cancer site or
region, broader evaluations across diverse locations and populations could enable
better understanding of differences in health risks. Regional variation in
agricultural practices, environmental conditions, regulatory policies, and
population vulnerability may contribute to substantial heterogeneity in exposure
levels and health impacts. Understanding these region-specific effects is critical,
as the distribution and intensity of AFOs/CAFOs and the susceptibility of nearby
communities can vary widely across different parts of the United States. These
differences may result from variations in population characteristics (e.g.,
racial/ethnic composition, socioeconomic status, baseline health status), local
healthcare access and infrastructure, and characteristics of AFO/CAFOs (e.g.,
facility type, size, regulation). Comparative, multi-state studies are therefore
essential to identify geographic disparities, assess differential risks among
populations, and inform targeted public health interventions and environmental
policies.
To address these gaps, this study aims to examine the associations between
exposure to AFOs/CAFOs and incidence rates of various cancers across three US
states: California, Iowa, and Texas. These states are among the most agriculturally
intensive in the country with diverse demographic composition and environmental
conditions. We selected California, Iowa, and Texas because they are among the few
states with SEER registries that provide population-based cancer incidence data with
broad population coverage, have a high prevalence and substantial spatial
variability of AFO/CAFOs, and represent diverse agricultural sectors (dairy in
California, swine in Iowa, and beef in Texas). These states also allow linkage to
publicly available, geocoded permitted animal feeding operation data over the study
period. Together, they encompass diverse geographic regions; environmental
conditions such as temperature, wind patterns, and topography that influence
pollutant dispersion and transport; agricultural systems; and population
characteristics, enhancing the generalizability of our findings while ensuring
consistent data availability across all study variables.
In this study, we examined overall cancer incidence and nine site-specific
cancers selected based on prior literature suggesting potential links to AFO/CAFO
exposures (Booth et al. 2017; Cha et al. 2014; Fisher
et al. 2020; Liu et al. 2023). By
linking county-level cancer incidence with AFO/CAFO density, we evaluated whether
higher exposure to AFOs/CAFOs is associated with an increased cancer burden. This
multi-state study provides insight into regional heterogeneity in potential cancer
risks associated with AFO/CAFO exposure.

2.
Methods
2.1
Data
We obtained county-level cancer incidence data from the Surveillance,
Epidemiology, and End Results (SEER) program using SEER*Stat software version
8.4.4 (Incidence - SEER Research Plus Limited-Field Data, 22 Registries, Nov
2023 Sub (2000–2021)). We considered overall cancers and nine
site-specific cancers (i.e., colon and rectum, lung and bronchus, breast,
lymphoma, non-Hodgkin lymphoma, pancreas, bladder, leukemia, and respiratory
system) based on previous literature suggesting plausible links to AFO/CAFO
exposure and the hazards emitted from these facilities. These exposures include
particulate matter, ammonia, hydrogen sulfide, and bioaerosols, which have been
linked to respiratory, hematologic, and gastrointestinal cancers (Booth et al. 2017; Cha
et al. 2014; Fisher et al.
2020; Liu et al. 2023). The
selection of these cancers reflects biologically plausible pathways through
which emissions from animal feeding operations could influence cancer risk.
The geocoordinates of permitted, operational animal facilities were
obtained from each state’s environmental department (i.e., California
Integrated Water Quality System, Iowa Department of Natural Resources, Texas
Commission on Environmental Quality). Based on the geographic location of each
AFO/CAFO, we calculated county-level AFO/CAFO densities by dividing the number
of AFO/CAFO by the land area of each county for each state.
We considered several county-level variables as potential confounders:
percentage of the population that is non-Hispanic Black (NHB), percentage of the
population that is Hispanic, percentage of the population with education less
than high school, median annual household income, percentage of adults who
currently smoke, and urbanicity. Variables were obtained from 2020 Census,
2018–2022 American Community Survey, 2024 Behavioral Risk Factor
Surveillance System, and 2023 Rural-Urban Continuum Codes.
To explore whether the observed associations between AFO/CAFO density
and county-level cancer incidence may reflect broader agricultural activity
rather than livestock operations specifically, we incorporated county-level
major cropland data in sensitivity analyses. Cropland data were obtained from
the USDA CropScape database (https://nassgeodata.gmu.edu/CropScape/).

2.2
Exposure assessment and statistical analysis
For each state, we assessed county-level AFO/CAFO exposure using the
AFO/CAFO density based on the number of operations per 100 square km. Due to
highly skewed distributions and different spatial patterns of AFO/CAFO across
the three states, we defined exposure counties as those with AFO/CAFO density
above the 75th percentile and no/low exposure counties as those at or below the
75th percentile based on state-specific distributions. This approach was used to
capture areas with relatively high AFO/CAFO density, where health impacts are
more likely to occur, and to enhance contrast between exposure categories. We
used propensity score matching to identify comparable high- and no/low- exposure
counties based on county-level covariates, aiming to minimize confounding in
this ecological analysis. To reduce the potential confounding, we conducted 1:1
greedy nearest-neighbor propensity score matching without replacement.
Propensity scores were estimated using a logistic regression model that includes
the following county-level covariates: percent of the population that is
non-Hispanic Black, percent of the population that is Hispanic, percent of the
population with education less than high school, median annual household income,
smoking, and urbanicity. Matched control counties were selected based on the
closest propensity score to each exposed county. We calculated standardized mean
differences (SMDs) of the covariates before and after matching to evaluate the
balance (to compare the distributions of the covariates between exposed and
matched control counties). Balance was evaluated using SMDs, with values below
0.1 considered indicative of adequate balance. We plotted SMDs to check the
imbalance of the covariates in the matched control counties and if there existed
imbalance after matching, additional adjustments were made in the model for the
unbalanced covariates to reduce the potential residual confounding. We ran the
Poisson regression model to generate incidence rate ratio (IRR) and 95%
confidence interval (CI) for each state and cancer type. Models included an
offset term for log population and were analyzed separately for each state and
cancer type. As a sensitivity analysis, we conducted a different matching scheme
(i.e., 1:2 rather than 1:1 matching) to check the robustness of the findings. To
assess potential spatial autocorrelation, Moran’s I was calculated using
the residuals from the final Poisson regression models for each state. As a
sensitivity analysis, we additionally included the county-level proportion of
residents aged ≥65 years in the propensity score matching to account for
potential confounding by age. This allowed us to evaluate whether differences in
the age distribution across counties influenced the association between AFO/CAFO
density and cancer incidence. All analyses and mapping were conducted using SAS
Version 9.4 (SAS Institute, Cary, NC, USA) and ArcGIS Pro 10.6.1 (ESRI,
Redlands, CA).

3.
Results
Supplemental Figure
1 illustrates the spatial distribution of permitted animal facilities in
each state. AFO/CAFOs in California and Texas were generally clustered in specific
areas (e.g., California Central Valley), whereas AFO/CAFOs in Iowa are spread
throughout the state.
Supplemental Table
1 and Supplemental
Figure 2 present the distribution of AFO/CAFO density by state and across
all states. County-level AFO/CAFO density varied substantially by state. In Iowa,
all counties had AFOs exposure, with the highest average AFO density (7.82 per 100
km2). In contrast, approximately 60% of Texas counties and 40% of
California counties had no CAFOs. Among Texas counties with exposure, 75% had fewer
than four facilities. Both states exhibited highly right-skewed distributions, with
most counties having few or no CAFOs and a small number of counties exhibiting
moderate to high densities, including exceptionally high values.
The distribution of community-level characteristics varied by state (Table 1). Texas had the highest average
percentage of non-Hispanic Black residents and individuals with less than a high
school education. California had the highest average median household income and the
lowest current smoking rate. Additionally, 64% of counties in California were
classified as metropolitan areas.
Supplemental Table
2 shows summary statistics of incidence rates for various cancers for the
study period. State-level crude incidence rates for all types of cancer investigated
in this study were highest in Iowa.
Figure 1 shows the distribution of
exposed and matched control counties in each state, and Supplemental Table 3 summarizes the
matching variables before and after matching for each state. Propensity score
matching resulted in 8, 24, and 60 exposed counties being matched to an equal number
of matched control counties in California, Iowa, and Texas, respectively. After
matching, substantial improvements in balance were observed across all states, with
most covariates falling within the negligible difference range (exhibiting SMDs
close to zero or within the threshold of 0.1), indicating improved comparability
between exposed and matched control counties. Although small residual imbalances
remained for some covariates, these were notably reduced relative to the unmatched
samples (Supplemental Figure
3). This demonstrates that the selected matches effectively reduced
baseline differences between high- and no/low-exposure counties, supporting the
validity of subsequent Poisson regression analyses using the matched sample.
However, in California, % Non-Hispanic Black and % Hispanic remained imbalanced
after matching. To address this issue, we conducted two sensitivity analyses. First,
we implemented an alternative matching that excluded these two variables from the
matching variables and adjusted for them in the model. Second, we restricted the
sample by excluding counties with extreme racial/ethnic compositions to address
residual imbalance in % Non-Hispanic Black and % Hispanic. Propensity score matching
and outcome analyses were re-run within this subset of counties, representing the
population in which exposed and control counties were comparable. Covariate balance
improved for both % Non-Hispanic Black and % Hispanic (Supplemental Figure 4).
Incidence rate ratios and 95% confidence intervals were estimated to assess
the association between county-level AFO/CAFO density and cancer incidence rates
(Table 2). Compared to control counties,
exposed counties had significantly higher rates of all-cancer incidence in all three
states. Moran’s I statistics based on the Poisson model residuals indicated
no significant spatial autocorrelation in any state (California: I = 0.0834, p =
0.236; Iowa: I = −0.0332, p = 0.614; Texas: I = −0.0084, p = 0.500).
Stratified analyses by cancer type indicated the highest and statistically
significant associations for bladder cancer, colon and rectum cancer, and lung and
bronchus cancer in California, Iowa, and Texas, respectively. In California, all
other cancer types we investigated also showed significant positive associations. In
Iowa, although some cancer types (i.e., breast cancer, bladder cancer, and leukemia)
did not exhibit statistically significant associations, all investigated cancer
types showed positive associations. In Texas, we observed statistically significant
positive associations for all cancer types except breast cancer. These findings
suggest a potential link between higher AFO/CAFO exposure and increased cancer
incidence rates, although further investigation is warranted to explore causal
mechanisms and potential confounding factors.
To assess the robustness of our findings, we conducted a sensitivity
analysis using 1:2 propensity score matching, pairing each exposed county with two
control counties (Supplemental
Table 4). Some exposed counties had fewer than the specified two matched
controls due to limited availability of suitable controls (California: 8 exposed, 14
control counties; Iowa: 24 exposed, 40 control counties; Texas: 60 exposed, 118
control counties). This approach increased the matched sample size. Covariate
balance assessed using SMDs was similar to the main analysis, with most covariates
achieving acceptable balance after matching. The estimated incidence rate ratio for
all-cancer incidence associated with high AFO/CAFO exposure remained consistent with
the main analysis using 1:1 matching, supporting the robustness of the association.
Stratified results by cancer type were generally consistent with main analysis;
however, associations for some cancer types in Texas lost statistical significance,
and breast cancer showed a significantly protective association. These results
suggest that the observed associations are not sensitive to the matching ratio and
remain stable across different matching specifications.
Because % Non-Hispanic Black and % Hispanic were poorly balanced after
initial matching in California, in sensitivity analyses, they were excluded from the
propensity score model and instead adjusted for in regression. The results of the
sensitivity analysis were generally consistent with the original findings (Supplemental Table 5). In the
sensitivity analysis excluding counties with extreme racial/ethnic compositions, the
estimated associations remained generally consistent with the main analysis (Supplemental Table 6).
After adjusting for the county-level proportion of land that is major
cropland, the associations between high AFO/CAFO density and overall cancer
incidence remained generally consistent in all states, suggesting that the observed
effects are not solely driven by general agricultural activity. Results for
site-specific cancers showed similar patterns (Supplemental Table 7). In the
sensitivity analysis including the county-level proportion of residents aged
≥65 years in the propensity score matching, the results were generally
similar to the original findings, indicating that age differences across counties
did not significantly affect the observed associations (Supplemental Table 8).

4.
Discussion
This investigation of county-level AFO/CAFO exposure and cancer incidence
rates in three US states observed significant positive associations between higher
AFO/CAFO density and increased incidence rates of multiple cancers. Counties with
higher AFO/CAFO exposure had significantly elevated all-cancer incidence rates
compared to control counties. All cancer types except breast cancer in Texas showed
positive associations with high AFO/CAFO exposure in all three states. Stronger
associations were observed for specific cancers in California (e.g., bladder
cancer), Iowa (e.g., colorectal cancer), and Texas (e.g., lung and bronchus cancer),
while some cancer types showed no significant association. In California, all cancer
types we investigated showed significant positive associations, while no significant
links were found for breast cancer in Texas and Iowa, or for bladder cancer and
leukemia in Iowa. These results suggest a potential relationship between AFO/CAFO
exposure and cancer incidence with regional variation, warranting further
investigation.
Consistent with our findings, previous studies reported positive
associations between AFO/CAFO exposure with several cancer outcomes. Booth et al. (2017) found significant positive
associations between county-level incidence of childhood leukemia (e.g., acute
myeloid leukemia, acute lymphoblastic leukemia) and densities of animals or
operations (e.g., densities of broiler chicken or hog operation) in nine US states.
A study in Iowa observed that residential proximity to intensive animal agriculture
was positively associated with risk of non-Hodgkin lymphoma and leukemia (Fisher et al. 2020). Another study in Korea
reported increases in childhood leukemia mortality in rural areas, and in counties
with both higher farming index and pesticide exposure index (Cha et al. 2014). Our findings incorporating data from
multiple states suggest that communities with greater AFO/CAFO density may
experience an elevated incidence of certain cancers, raising public health concerns
about the environmental consequences of intensive animal agriculture.
AFO/CAFOs emit a wide range of harmful pollutants that can adversely affect
human health. These include gaseous emissions such as ammonia and hydrogen sulfide,
particulate matter (PM), volatile organic compounds, and bioaerosols containing
endotoxins and antibiotic-resistant bacteria (Hribar
2010; Mitloehner and Schenker
2007). Chronic exposure to these pollutants has been linked to inflammation,
oxidative stress, and immunosuppression which may contribute to the cancer
development (Heederik et al. 2007; Reuter et al. 2010). For example, hydrogen
sulfide has been shown to induce DNA damage and impair cellular respiration, whereas
particulate matter can provoke systemic inflammation and facilitate the delivery of
carcinogenic compounds into the lungs (IARC
2016; Jin et al. 2020).
Water contamination is another major pathway of exposure, particularly in
rural areas that rely on private wells. Animal waste from AFO/CAFOs, often stored in
large open-air lagoons or applied to fields as fertilizer, can lead to nitrate
leaching into groundwater. Elevated nitrate levels in drinking water have been
associated with gastrointestinal cancers, especially colorectal and gastric cancers,
due to the formation of N-nitroso compounds known carcinogens (Burkholder et al. 2007; Ward et al. 2005). Other studies also reported higher nitrate levels in
the water near farms and found various adverse health outcomes including colon,
bladder, and thyroid cancers (Essien et al.
2020; Ward et al. 2018).
Importantly, private wells are largely unregulated, and many households may remain
unaware of contamination risks or lack resources to treat their water.
Although a generally consistent pattern of increased cancer incidence was
observed in relation to higher AFO/CAFO exposure in California, Iowa, and Texas, the
magnitude and statistical significance of these associations varied by state. These
differences may reflect variation in sample size, exposure distributions, and
covariate structures, highlighting the importance of region-specific analysis in
environmental health research. One potential source of this heterogeneity may result
from differences in the dominant types of animal agriculture. In California,
large-scale dairy operations, particularly concentrated in the Central Valley,
employ manure lagoons and open-lot systems that generate large volumes of ammonia,
hydrogen sulfide, methane, and volatile organic compounds (Emmett Institute 2024; Gerber et al. 2013). These pollutants can promote the formation of
secondary particulate matter and reduce local air quality, potentially increasing
cancer risks through inhalation of carcinogens including nitrogen-containing
compounds and polycyclic aromatic hydrocarbons (PAHs). In Iowa, the livestock
industry is dominated by swine operations, which commonly use deep-pit slurry
storage and extensive land application of liquid manure. These practices can lead to
nitrate contamination of groundwater, a particular concern in rural areas that rely
on private wells, since high nitrate levels have been associated with elevated risks
of various cancers (Ward et al. 2005; Weyer et al. 2001). In contrast, Texas is
dominated by expansive beef cattle feedlots and poultry operations, particularly in
the Panhandle and eastern regions. The combination of their scale and the
state’s relatively dry climate conditions may enhance the long-range
transport of airborne pollutants such as bioaerosols, endotoxins, and
particulate-bound contaminants, complicating both exposure assessment and regulatory
control (Emert et al. 2023).
Variations in state-level regulatory approaches may contribute to regional
differences in exposure to AFO/CAFOs. Some states enforce stricter agricultural
emission controls (e.g., comprehensive monitoring and mandatory mitigation
strategies), while others rely more on less prescriptive compliance. For example,
California maintains relatively rigorous regulations through its state and regional
air quality agencies, requiring large-scale operations to obtain permits and
implement emission reduction strategies. In contrast, Texas generally regulates
through general permits focused on wastewater and manure management and has fewer
statewide requirements specific to agricultural air emissions, although some local
jurisdictions may impose their own rules. Iowa, as a major livestock producer,
primarily focuses on nutrient management and water quality regulations, with
comparatively limited direct regulation of air emissions from animal facilities.
These regulatory differences can influence both the intensity and duration of
community exposures near AFOs/CAFOs.
The sociodemographic characteristics of populations living near AFO/CAFOs
may contribute to regional disparities. For example, many communities with
low-income or racial/ethnic minority populations often experience cumulative
environmental exposures and structural barriers to healthcare access (Morello-Frosch et al. 2011). Rural populations
that are often older, lower-income, and less politically empowered may experience
insufficient environmental monitoring and fewer opportunities for resources such as
healthcare, increasing their susceptibility. Differences across states may be partly
driven by historical zoning and land use policies, which have often placed
large-scale livestock operations in underserved communities, leading to increased
exposure to environmental hazards (Pew Commission
2008; Wilson et al. 2002).
These interrelated factors such as differences in facility type and density,
pollutant profiles, regulatory enforcement, and population vulnerability may
contribute to the heterogeneity in cancer risks observed across states.
Understanding these contextual variations is critical for interpreting state-level
findings and for developing targeted public health interventions that reflect the
complex nature of environmental health disparities.
This study has several limitations that should be considered when
interpreting the findings. While the use of county-level AFO/CAFO density allowed
for a broad geographic comparison, it may have introduced exposure
misclassification. Given the long-time frame of the development of cancer, long-term
cohort studies that include residential mobility are needed to improve exposure
estimates. Detailed information on AFO/CAFO facility characteristics was not
consistently available across the three states. Facility size, animal type,
operational capacity by animal unit, manure management practices, and local
meteorological factors may influence actual exposure levels, and these are not
captured in aggregate metrics. Counties with similar numbers of permitted operations
may differ in total animal unit capacity, and facilities raising different types of
animals may have varying emission profiles. Exposure to AFO/CAFOs using county-level
AFO/CAFO density may not capture the full range of environmental emissions,
historical exposures relevant to cancer development, or individual-level exposure
pathways such as drinking water contamination, dietary intake, or occupational
contact. Additionally, the use of facility-level data may not reflect temporal
changes in operation size, emission levels, or regulatory compliance over time.
Future research would benefit from incorporating detailed characteristics of
AFO/CAFOs including animal type, operational capacity, dispersion modeling, or
environmental monitoring data to refine exposure assessment. Additionally, linkage
with cancer registry data at the individual level would enable better control for
confounders such as smoking, occupation, and family history. The ecological design
limits the ability to make causal inferences at the individual level. Exposure to
AFO/CAFO density and cancer outcomes were assessed at the county level, which
precludes inference at the individual level and raises the possibility of ecological
fallacy. The observed associations therefore reflect population-level associations
rather than individual risk and should be interpreted as exploratory rather than
causal. Cancer incidence at the county level may obscure finer-scale spatial
variation in exposure and outcome. Residual confounding by unmeasured or imperfectly
measured factors remains a concern. Although county-level sociodemographic and
environmental covariates were included, unmeasured factors such as healthcare
access, cancer screening practices, population mobility, occupational exposures, and
co-exposure to other industrial pollutants may have influenced cancer risk. Residual
confounding by other individual-level cancer risk factors such as age, BMI, physical
activity, and dietary behaviors may still exist, and future studies incorporating
more detailed risk factor data could help better characterize exposure-outcome
associations. The use of propensity score matching restricts the analysis to a
subset of high- and no/low-exposure counties, leading to the exclusion of some
counties and potentially limiting generalizability. Finally, heterogeneity in
state-level regulations and monitoring practices may have affected both exposure
estimation and the comparability of findings across regions. Despite these
limitations, this study provides important insight into potential environmental
health risks associated with AFO/CAFO exposure and highlights the need for more
granular, longitudinal research using individual-level data and direct exposure
assessments that can disentangle the various exposures from these facilities (e.g.,
water contamination, air pollution, odor).
Our findings suggest that high-density AFO/CAFO exposure is associated with
increased cancer incidence, independent of general cropland intensity. Although
counties with high livestock density are often located in areas of intensive
agriculture, sensitivity analyses adjusting for the proportion of land that is major
cropland indicate that the observed associations are unlikely to be driven solely by
broader agricultural activity. Nevertheless, unmeasured aspects of agricultural
exposures may still contribute to observed patterns and warrant further
investigation.
While this study focused on residential exposure, individuals working in or
near AFO/CAFOs may face greater health risks due to more direct and frequent contact
such as with airborne pollutants, contaminated water, pesticides, and biological
agents. Occupational exposure levels are often substantially higher than those
experienced by the general population and may contribute to higher cancer risks
among agricultural workers. Several studies reported increased cancer risks among
workers in animal farms (Freeman et al. 2012;
Fritschi et al. 2002; Svec et al. 2005). Future research should incorporate
occupational histories and workplace exposure assessments to better understand the
burden of health risk associated with AFO/CAFO exposure.

5.
Conclusion
In conclusion, our findings provide suggestive evidence linking higher
exposure to AFO/CAFOs to increased cancer incidence in diverse U.S. states. While
the ecological design limits causal interpretation, the observed associations raise
important concerns about potential health risks in communities located near
large-scale animal operations. Continued research that incorporates individual-level
data, exposure monitoring, and longitudinal designs will be critical to clarifying
these relationships and guiding effective policy responses that protect public
health in agricultural regions.

Supplementary Material
1