Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Oct 2.
Published in final edited form as: Cancer Epidemiol. 2026 Aug 18;104:103199. doi: 10.1016/j.canep.2026.103199

Wildfire smoke PM2.5 and cancer history prevalence in the contiguous United States

Hiwot Y Zewdie 1,2, Seigi Karasaki 1,3, Malia Cortez 3, Jiayu Lin 3, Caroline Nondin 3, Sam LS Chao 3, Sidney M Donzella 1,3, Amanda I Phipps 1,3, Trang VoPham 1,3
PMCID: PMC13628883  NIHMSID: NIHMS2205310  PMID: 42612421

Abstract

Background:

Little is known about the role of wildfire smoke PM2.5 in cancer burden, despite ambient PM2.5 being classified as a human carcinogen.

Methods:

Daily wildfire smoke PM2.5 estimates were aggregated into cumulative (2006–2020) and recent (2018–2020) averages, and linked to 2022 tract-level cancer history prevalence estimates for all tracts in the contiguous US. Quasi-Poisson generalized additive models with random effects and spatial splines estimated prevalence ratios and 95% CIs for each exposure window, modeling smoke PM2.5 continuously (per IQR) and across tertiles. Models adjusted for tract-level sociodemographic factors, health behaviors, health status, population density, climate, and Census region. Effect modification was assessed using interaction terms and stratified analyses.

Results:

Smoke PM2.5 exposure was not associated with cancer history prevalence in fully adjusted models across either exposure window. However, we observed effect modification by Census region. Higher cumulative and 3-year smoke PM2.5 exposure was associated with higher cancer history prevalence in the Midwest (IQR-difference in 3-year exposure PR: 1.020; 95% CI: 1.013–1.026) and Northeast (PR: 1.040; 95% CI: 1.032–1.049).

Conclusion:

Wildfire smoke PM2.5 was not associated with cancer history prevalence at the tract level in this cross-sectional ecological study, although there may be potential effect modification by region. Further investigation using individual-level data on smoke PM2.5 exposure and cancer outcomes is needed to confirm these findings and better understand the implications of wildfires for cancer burden.

Keywords: wildfire smoke, particulate matter, cancer

Background:

In 2013, the International Agency for Research on Cancer classified outdoor air pollution, including fine particulate matter (PM2.5), as a Group 1 human carcinogen, citing sufficient evidence of its carcinogenicity in humans.1,2 The mechanisms linking PM2.5 and cancer development are complex.3,4 In brief, PM2.5 is thought to induce a pro-tumorigenic environment through mechanisms including oxidative stress, genotoxicity, and chronic inflammation.3–5 These factors together prompt a cascade of biological pathways that disrupt cell function, activate oncogenic signaling, and promote uncontrolled cell proliferation, survival, angiogenesis, and metastasis.3,6 While PM2.5 primarily deposits in the lower lungs, smaller particles may diffuse into the bloodstream and induce a similar series of biological mechanisms in other organs, though these mechanisms are still being understood.7,8

PM2.5 levels in the United States (US) have seen significant declines over the last several decades following the establishment of National Ambient Air Quality Standards in 1970 through the Clean Air Act.9,10 However, decades of progress are now stagnating, and, in some cases, completely reversing, due to the increase of PM2.5 directly and indirectly emitted from wildfire smoke.10 Burke et al. found that since 2016, wildfire smoke has contributed to 25% of total ambient PM2.5 concentrations.11 Currently, the National Ambient Air Quality Standards for PM2.5 set by the U.S. Environmental Protection Agency include an annual average limit of 9 μg/m3 and a 24-hour (daily) average limit of 35 μg/m3.12 Between 2016 and 2020, an estimated 16 million people lived in areas that experienced at least one day with wildfire smoke PM2.5 levels exceeding 50 μg/m3-- a fourfold increase from 2006 to 2010.13 Of those, 1.5 million resided in areas with at least one day above 200 μg/m3, compared to almost none in 2006 to 2010.

Emerging literature highlights the distinct chemical composition of wildfire smoke PM2.5 (hereafter referred to as smoke PM2.5) that makes it potentially more carcinogenic compared to non-smoke PM2.5.14,15 Specifically, smoke PM2.5 is more carbonaceous-- containing more organic particles such as polycyclic aromatic hydrocarbons-- and thus has greater oxidative potential.16,17 The biological implications relevant to carcinogenic processes include heightened activation of biomolecules such as the aryl hydrocarbon receptor, which is sensitive to the carbon components of smoke PM2.5. When activated, this receptor can inhibit DNA repair and dysregulate growth signaling pathways.17–19 Additionally, the increased production of reactive oxygen species generated from smoke PM2.5 particles can promote genomic instability and chronic inflammation, together contributing to tumor-promoting mechanisms.17,20

Despite the strong biological plausibility linking smoke PM2.5 with cancer, epidemiologic research remains limited.21,22 With growing climate-driven incidence of wildfires,10,23,24 developing epidemiologic evidence is critical for characterizing the potential implications for population-level cancer burden. To that end, this study aimed to 1) investigate the association between smoke PM2.5 exposure (2006–2020) and cancer burden proxied using small-area estimates of cancer history prevalence (2022) in the US, and 2) examine potential effect modification by population age, sex, race and ethnicity, population density, and region. We leverage nationally available census tract-level cancer prevalence data as a first step toward building this evidence base and generating hypotheses for subsequent epidemiological investigation.

Methods

Exposure

Wildfire smoke PM2.5 exposure was measured using the Environmental Change and Health Outcomes (ECHO) dataset on daily estimates available for the contiguous US at a 10 km2 spatial resolution between 2006–2020.13 Model development and validation are explained in detail elsewhere;13 in brief, daily predicted smoke PM2.5 10 km2 surfaces were produced using gradient-boosted tree models trained on ground monitoring, satellite, and reanalysis inputs. Spatial cross-validation was used to develop and test the model.

To align exposure estimates with the spatial resolution of the outcome data (i.e., 2020 US Census tract boundaries), we aggregated gridded smoke PM2.5 predictions to 2020 census tracts. Tract-level predictions were derived using area-weighted aggregation of gridded smoke PM2.5 estimates.13 More specifically, gridded smoke PM2.5 predictions from Childs et al. were intersected with 2020 Census tract cartographic boundary files to calculate area of overlap between each tract-grid pair. For each tract and day, tract-level smoke PM2.5 exposure was computed as the area-weighted mean of grid-cell predictions intersecting the tract, with weights equal to the area of overlap between the tract and each grid cell. We compared our tract aggregation approach with the 2010-boundary approach used by Childs et al. to ensure that differences in aggregation did not meaningfully influence exposure estimates (Supplemental Material 1).

Daily tract-level values were then aggregated to compute annual averages, which were used to construct two exposure windows: a cumulative annual average over 2006–2020 and a 3-year annual average over 2018–2020. We use these exposure windows to align with prior work examining smoke PM2.5 and other chronic disease outcomes, given the absence of an established etiologic window for smoke PM2.5 and cancer.25–27 We considered the cumulative window as the more etiologically relevant metric for de novo carcinogenesis, while the 3-year window was included as an additional metric to capture potential shorter-term pathways such as acceleration of existing cancer progression. Further, since the exposure data were available only through 2020, whereas cancer history prevalence was measured in 2022, our analysis includes a structurally induced two-year lag between exposure and outcome assessment.

Outcome

We used Centers for Disease Control and Prevention (CDC) Population Level Analysis and Community Estimates (PLACES) to obtain 2022 tract-level prevalence estimates of cancer burden.28 Cancer incidence data are not available at the census tract level nationally; as such, cancer history prevalence represents the most spatially comprehensive measure of cancer burden currently accessible for ecological analyses of this scope.

CDC PLACES provides model-based estimates of various health outcomes, including cancer, and cancer risk factors based on data from Behavioral Risk Factor Surveillance System (BRFSS), Census decennial population counts, and the American Community Survey (ACS) 5-year estimates. CDC PLACES technical documentation describes the small-area estimation methodology in detail;29,30 in brief, cancer prevalence estimates were generated using a multi-level regression and post-stratification approach applied to BRFSS and ACS data to compute a probability of having cancer (i.e., reporting ‘yes’ to the question: “Have you ever been told by a doctor, nurse, or other health professional that you had melanoma or any other types of cancer?” and “no” to the question: “Have you ever been told by a doctor, nurse, or other health professional that you had skin cancer that is not melanoma?”). The probability was applied to tract population estimates to generate an estimated tract-level (non-skin) cancer prevalence for those aged 18 years and older.

Covariates

We used a directed acyclic graph to select covariates (Supplemental Material 2). We used 5-year 2022 ACS data on tract-level proportions of individuals aged 65 years and older, female, self-identifying as a racial minority (i.e., proportion of individuals that are not Non-Hispanic White), living below the federal poverty line, and population density (population/m2). We note that tract proportion racial minority was included as an imperfect proxy for structural racism for the purposes of confounding adjustment. CDC PLACES data were used to compile tract-level estimates of the proportion of individuals who self-reported as current smokers, as having been diagnosed with diabetes, having a body mass index (BMI) ≥30.0 kg/m2 (BMI threshold for obesity), engaging in no leisure-time physical activity, engaging in binge drinking (report having ≥5 drinks [men] or ≥4 drinks [women] on ≥1 occasion during the previous 30 days), and not having current health insurance.31 All continuous covariates were modeled linearly.

We used US Census Regions and Divisions designations to characterize tracts into one of four regions: Northeast, Midwest, South, and West. We used Köppen-Geiger climate classifications to classify tracts into one of five climates: equatorial, arid, warm temp, continental, and polar.32,33

Statistical Analysis

We used adjusted quasi-Poisson generalized additive models (GAM) with a log link to estimate the association between smoke PM2.5 exposure and modeled cancer history prevalence. To construct the model outcome, we multiplied the PLACES tract-level prevalence estimate by the corresponding adult population denominator to reconstruct pseudo-counts of prevalent cancer history cases. The quasi-Poisson models were used to account for overdispersion and potential variance misspecification in tract-level count data.34,35 All models included an offset for the log of the adult (age ≥18) population in a tract. To account for unmeasured state-level heterogeneity (e.g., differences in state-level cancer surveillance), we included a state-level random-effect. Spatial confounding was addressed using bivariate thin-plate splines constructed using tract centroid latitude and longitude (k = 5). Sensitivity analyses testing robustness of effect estimates to spline dimensions were conducted. Confounders were selected a priori using a directed acyclic graph and were restricted to variables available nationally at a shared spatial scale with the exposure and outcome.

We used a staged modelling approach to estimate associations with increasing level of confounding control. Namely, Model 1 estimated the association between smoke PM 2.5 and prevalent cancer history cases after adjustment for Census tract-level proportion aged 65+ years, female, and racial minority. Model 2 included all covariates in Model 1, with additional adjustment for tract-level sociodemographic variables of proportion living below the federal poverty level and population density. Model 3 built on Model 2 and included additional adjustment for behavioral and lifestyle factors, including tract-level prevalences of smoking, diabetes, obesity, physical inactivity, binge drinking, and no health insurance. We subsequently fit several separate models examining effect modification by age, sex, race and ethnicity, population density, Census region, and climate zone by fitting a multiplicative term between each modifier and smoke PM2.5. All continuous effect modifiers (i.e., age, sex, race and ethnicity, population density) were dichotomized at the median and fit as a binary variable. We used a Wald-type test to test for significance of the interaction, and fit stratified models. We fit the staged, interaction, and stratified models for both exposure windows (i.e., cumulative and 3-year smoke PM2.5). As a post-hoc sensitivity analysis, we refit the region-specific models replacing the state-level random effect with county-level random effects.

Rates estimated from the quasi-Poisson models were interpreted as prevalence ratios (PRs). Estimated prevalence ratios (PRs) corresponded to an IQR difference in smoke PM2.5 and a tertile difference in smoke PM2.5.

This study used administrative and public health data that are publicly available and was therefore exempt from human subjects review. All analyses were conducted using R (version 4.5.0).

Results

Our analytic sample included 82,684 Census tracts. The overall and region-specific smoke PM2.5 trends over our study period are shown in Figures 1 and 2. The median tract-level cumulative smoke PM2.5 exposure and 3-year smoke PM2.5 exposure was 4.09 μg/m3 [IQR:3.63–4.77] and 3.19 μg/m3 [IQR:2.84–3.76], respectively. The geographic distributions of smoke PM2.5 tertiles are presented in Figure 3. For cumulative smoke PM2.5 exposure (2006–2020), census tracts in the highest exposure tertile were more likely to be located in the West and the Northeast (Supplemental Material 3: Median Cumulative Smoke pm [IQR] West: 4.58 [3.97–6.24]; Northeast: 4.78 [4.37–5.39]). For the 3-year exposure window (2018–2020), tracts in the highest tertile of smoke PM2.5 exposure were again concentrated across the West (Median 3-year smoke PM2.5: 6.05 [4.89–12.90]), in addition to the South (Median 3-year smoke PM2.5: 3.19 [2.84–3.50]).

Figure 1.

Figure 1.

Open in a new tab

Longitudinal trends of smoke PM2.5 exposure (2006–2020)

Annual median tract-level smoke PM2.5 exposure (μg/m3) for US from 2006–2020. Shaded areas represent the interquartile range (IQR).

Figure 2.

Figure 2.

Open in a new tab

Longitudinal trends of smoke PM2.5 exposure (2006–2020) by region

Annual median tract-level smoke PM2.5 exposure (μg/m3) across US Census regions from 2006–2020. Shaded areas represent the interquartile range (IQR).

Figure 3.

Figure 3.

Open in a new tab

Smoke PM2.5 exposure across the continental US: a) cumulative annual average (2006–2020); b) 3-year annual average (2018–2020)

Tract-level cumulative and 3-year smoke PM2.5 exposure across the contiguous US, divided into tertiles. Data are projected using the USA Contiguous Albers Equal Area Conic (USGS) coordinate reference system (EPSG: 5070).

Tract characteristics were relatively similar across tertiles of smoke PM2.5, except for proportion racial minority, population density, and Census region (Table 1). More specifically, we observe that tracts in the highest tertile of cumulative smoke PM2.5 exposure have a greater proportion of racial minority populations (mean: 42%, SD: 26%), have higher population density (mean: 3,793 population/km2, SD: 7,141population/km2), and/or were mostly located in the Northeast and West (67% of Northeast tracts, 51% of Western tracts). Mean tract-level cancer history prevalence was 7.4% (SD: 2.7%) overall, and 8.7% (lowest tertile), 7.8 % (second tertile), 7.3% (highest tertile) across cumulative smoke PM2.5 tertiles.

Table 1.

Descriptive characteristics of tracts across cumulative (2006–2020) smoke PM2.5 tertiles

Overall
(n=82,684)		 Cumulative Smoke
PM2.5 Tertile 1
n= 27,636
[0.964 – 3.788 μg/m3]		 Cumulative Smoke
PM2.5 Tertile 2
n= 27,649
[3.789 – 4.522 μg/m3]		 Cumulative Smoke
PM2.5 Tertile 3
n= 27,399
[4.523 – 15.733 μg/m3]
Sociodemographic variables, mean (SD)
 Female, % 50 (5) 50 (5) 51 (5) 51 (5)
 Aged 65+ years, % 17 (9) 19 (9) 17 (9) 16 (8)
 Racial minority, % 34 (26) 24 (21) 35 (26) 42 (26)
 Poverty, % 13 (11) 13 (10) 14 (10) 13 (10)
 No health insurance, % 12 (7) 11 (6) 12 (8) 10 (7)
 Pop density, pop/km2 2,027 (4,597) 695 (1,218) 1,607 (2,478) 3,793 (7,141)
Behavioral & lifestyle variables, mean (SD)
 Smoking, % 15 (5) 16 (5) 15 (6) 13 (5)
 Obesity, % 35 (7) 37 (6) 35 (7) 32 (7)
 Binge drink, % 17 (3) 17(3) 17 (3) 18 (3)
 Diabetes, % 12 (4) 13 (4) 13 (4) 12 (4)
 No leisure physical activity, % 25 (8) 26 (7) 26 (8) 24 (8)
Census region * , n (%)
 Midwest 18,402 (22) 12,113 (66) 6,185 (34) 104 (0.1)
 Northeast 14,514 (18) 1,314 (9) 3,494 (24) 9,706 (67)
 South 31,748 (38) 10,548 (33) 12,909 (41) 8,291 (26)
 West 18,020 (22) 3,661 (20) 5,061 (28) 9,298 (51)
Climate zones * , n (%)
 Arid 7,594 (9) 3,555 (46) 2,273 (29) 1,766 (24)
 Equatorial 1,664 (1) 385 (23) 874 (53) 405 (0)
 Continental 21,201(26) 11,476 (54) 6,845 (32) 2,880 (14)
 Warm Temp 52,225 (63) 12,220 (23) 17,657 (34) 22,348 (43)
 Polar 0 (0) 0 (0) 0 (0) 0 (0)
Open in a new tab

All variables are aggregated to the Census tract level

*

Proportions are calculated row-wise across smoke tertiles

Across all model specifications, we observed no evidence of an association between tract-level smoke PM2.5 exposure and cancer history prevalence. For both cumulative exposure (2006–2020) and the 3-year exposure window (2018–2020), prevalence ratios (PRs) per IQR increase were close to the null, with 95% confidence intervals that included 1.00 (fully adjusted model: cumulative exposure PR = 1.000, 95% CI: 0.999–1.001; 3-year exposure PR = 1.000, 95% CI: 0.999–1.001). Similarly, when smoke PM2.5 exposure was categorized into tertiles, PRs comparing higher exposure categories with the lowest category were again close to the null, and no consistent exposure-response pattern was observed across models or exposure windows. Results were similar with increasing spline dimensions (Supplemental Material 4).

Interaction tests indicated effect modification for all tested effect modifiers, except tract proportion female (Supplemental Material 5 & 6). However, given the large analytic sample, we use these interaction tests only to motivate stratified analyses and focus interpretation on the magnitude and direction of stratum-specific effect estimates.

The most pronounced differences in effect size were observed for Census region. Region-specific analyses showed an IQR increase in cumulative smoke PM2.5 exposure was associated with 3% higher cancer history prevalence in the Midwest (95% CI: 1.026–1.041), and 2% higher cancer history prevalence in the Northeast (95% CI: 1.013–1.026). These effects were sustained across the 3-year exposure definition (Midwest: 2% (95% CI: 1.013–1.026); Northeast: 4% (95% CI: 1.032–1.049)) and across tertiles of exposure for these regions. Associations in the South and West were either null or showed an inverse association. Region-specific results were similar using county-level random-effects (Supplemental Material 7).

Discussion

This ecological study examined the cross-sectional association between smoke PM2.5 exposure and modeled cancer history prevalence in 2022 in the contiguous US using cumulative (2006–2020) and 3-year (2018–2020) exposure definitions. We did not observe an overall association between smoke PM2.5 and cancer history prevalence at the tract-level using either exposure definitions. We did, however, observe statistically significant effect modification by tract-level characteristics including Census region. We found that Census tracts exposed to higher smoke PM2.5 values in Northeastern and Midwestern US had higher cancer burden when compared to tracts with lower smoke PM2.5 exposure, though differences were modest.

Given that smoke PM2.5 is composed of many carcinogenic species,14,15 we hypothesized that exposure to smoke PM2.5 would be associated with a higher cancer burden. We did not observe the expected positive overall association. Interpretation of these findings in the context of existing literature is challenging, as the few epidemiologic studies examining wildfire-related PM2.5 and cancer outcomes have focused on cancer mortality rather than prevalence. These mortality-based studies have generally reported small but statistically significant positive associations between wildfire-related smoke exposure and cancer mortality.25,27,36,37 For example, Gao et al. found that in the UK Biobank study population, a 10 ug/m3 increase in 3-year cumulative residential exposure to model-estimated smoke PM2.5 was associated with a 0.4 % (95% CI (1.001–1.007) greater risk of overall cancer mortality.25 Similarly, a study using death registry data in Brazil reported that municipalities with higher modeled smoke PM2.5 concentrations experienced higher all-cancer mortality, with a 2% increase in mortality risk per 1 μg/m3 increase in smoke PM2.5.37 To our knowledge, only one prior study has examined wildfire exposure in relation to cancer incidence.22,27 In a prospective Canadian cohort, residential proximity to wildfire-burned areas was associated with higher incidence of lung and brain cancers, but not hematologic malignancies.27

To our knowledge, this is the first study to examine associations between wildfire-related PM2.5 exposure and cancer history prevalence. Since prevalence reflects a combination of disease incidence, survival, and prognosis, it is possible that our findings do not translate directly to the cancer incidence and mortality findings because wildfire exposure influences cancer progression or survival differently than disease onset. For example, if wildfire-related PM2.5 primarily affects cancer progression or survival and not incidence, our associations observed for cancer history prevalence could be attenuated. In the absence of more detailed data on cancer onset and prognosis, it is difficult to disentangle these components in our study. Future work using prospective cohort data could help clarify these mechanisms.

There are several additional factors related to our study limitations that may also explain the null overall findings, and frame future research directions. First, this study used an ecological study design, which limited our ability to adjust for multi-level confounding, which could have made it difficult to detect the isolated effect of smoke PM2.5 on cancer outcomes. Environmental effects, particularly wildfire smoke PM2.5, are relatively small,25,26,37 which makes confounding control even more pertinent.

Additionally, cancer development following exposure to a carcinogen often occurs over a long latency period. The specific exposure window relevant to smoke-related PM2.5 is not yet well-established, and it is possible that this study did not capture the etiologically relevant period. This may be particularly important for regions like the South, which we find to be experiencing more smoke exposure in recent years.38,39 In such cases, any cancer burden attributable to wildfire smoke PM2.5 may not yet be observable in the population. Cohort studies with detailed residential histories would be better suited for discerning etiologically relevant exposure windows.

Relatedly, cancer itself is a heterogenous disease, with its origins and development complex.40,41 While environmental conditions, including PM2.5, are implicated in some cancers,2,3,42 smoke PM2.5 may not be relevant for the etiology of all cancers. Our aggregate cancer history prevalence measure, while provided national scope, did not allow us to examine specific cancer types, a limitation of the CDC PLACES dataset. Null findings in analyses of aggregate cancer prevalence may therefore obscure site-specific associations that would be detectable with more granular outcome data. Future analyses leveraging cancer registry or cohort data with site-specific incidence information, or cause-specific mortality using data sources such as the National Death Index, would allow for a more targeted evaluation of smoke PM2.5 in relation to cancers with stronger biological plausibility for smoke-related carcinogenesis.

The cross-sectional nature of the study also raises the possibility of survivorship bias. Individuals who may have developed cancer due to earlier wildfire smoke PM2.5 exposure but did not survive long enough to be included in the cross-sectional sample could be missed, potentially leading to an attenuation of the estimated association.

A further limitation is that we were unable to propagate uncertainty from the PLACES small area estimation process into our regression models. While the quasi-Poisson specification accounts for overdispersion in the regression residuals, it does not capture the estimation uncertainty inherent in the SAE-derived outcome. Because we do not expect this uncertainty to be related to smoke PM2.5 exposure, point estimates are likely approximately unbiased. However, confidence intervals may be anti-conservative.

Finally, we did not adjust for non-wildfire carcinogenic exposures, including non-smoke PM2.5, traffic-related air pollution, industrial emissions, ozone, and radon. Adjustment for total or non-smoke PM2.5 was not pursued because our exposure measure is a modeled decomposition of total ambient PM2.5, and conditioning on its constituent components simultaneously would introduce collinearity. We therefore cannot rule out residual confounding by these exposures. Future studies with more comprehensive environmental exposure data, including co-pollutant estimates, would be better positioned to isolate the independent contribution of smoke PM2.5 to cancer burden.

Despite these limitations, to our knowledge this is the first study to examine wildfire smoke PM2.5 in relation to population cancer prevalence nationally, and the regional patterns we observe provide hypothesis-generating evidence for future studies with individual-level incidence data. Specifically, we find heterogeneity in effects across geographic regions, particularly in the Northeast and the Midwest. The strongest association is seen in the Northeast, where cancer history prevalence in Census tracts within the highest tertile of 3-year smoke PM2.5 exposure is 4% higher than those within the lowest tertile. These suggested regional differences may reflect both shifting patterns in the distribution of smoke exposure and contextual factors that influence population vulnerability. In recent years, Northeast and Midwest regions of the US have experienced increased exposure to wildfire smoke, driven by both the rising frequency of wildfires and the broader geographic reach of smoke plumes from wildfire activity within and outside the US.11,38,39,43 As regions that have been historically less affected by wildfire activity,44–46 the higher cancer burden observed in these regions may also reflect limited preparedness and adaptive capacity, though not directly measured in the present study. Populations in regions less accustomed to wildfire smoke may have limited public health infrastructure, awareness, or behavioral protocols for mitigating exposure during smoke events.47,48 As such, even comparable levels of smoke PM2.5 may have more pronounced health impacts in areas with less familiarity with managing air quality hazards, which may help explain the regional effect modification seen in this study. We encourage future work to formally test these hypotheses.

Our study has several strengths. We used nationwide wildfire smoke data with high spatial and temporal resolution. This allowed for the use of an objective measure of wildfire exposure derived from multiple instruments and analytic tools, which likely captures exposure more accurately than single measure approaches (e.g., area-burned). Use of these national data also supported a nationwide ecological analysis, providing the large sample size needed to detect effects. Third, the temporal scope of the data also enabled us to average smoke exposure over varying periods, allowing us to evaluate multiple potentially etiologically relevant exposure windows for cancer burden--an important area for continued research. Finally, we were able to adjust for several area-level covariates across multiple domains (e.g., social, behavioral, and climatic) that could confound the relationship between smoke PM2.5 and cancer burden.

Continued investigation into the relationship between wildfire smoke PM2.5 exposure and cancer outcomes is needed. Future studies can address current study limitations by leveraging georeferenced cohort or cancer registry data, which allow for improved control of both individual- and area-level confounders. Such data sources can also provide more comprehensive information on cancer site and histology, as well as allow for more accurate exposure assessment through the incorporation of residential histories.

Conclusion:

This ecological study examined the association between wildfire smoke PM2.5 exposure and cancer history prevalence across all census tracts in the contiguous US. We did not observe an overall association between smoke PM2.5 and cancer history prevalence, but did observe regional heterogeneity in effects, with higher smoke PM2.5 exposure associated with greater cancer burden in the Northeast and Midwest-- regions that have not historically experienced frequent wildfire activity and may have limited capacity for mitigating exposure. Future work leveraging individual data is needed to confirm these overall and regional findings. As wildfire activity continues to intensify, strengthening the epidemiological evidence linking wildfire smoke exposure to cancer outcomes is important for guiding effective prevention strategies and public health interventions.

Supplementary Material

1

Table 2.

Staged quasi-Poisson GAMs estimating association between cumulative and 3-year wildfire smoke PM2.5 exposure and 2022 cancer prevalence

Model 1
PR (95% CI)		 Model 2
PR (95% CI)		 Model 3
PR (95% CI)
Cumulative (2006–2020)
IQR [3.638–4.769 μg/m3] 0.981 (0.980–0.983) 0.988 (0.986–0.990) 1.000 (0.999–1.001)
Smoke Tertile
 Tertile 1 [0.964 – 3.788 μg/m3] Ref Ref Ref
 Tertile 2 [3.789 – 4.522 μg/m3] 0.975 (0.972–0.978) 0.973 (0.970–0.976) 0.997 (0.995–0.999)
 Tertile 3 [4.523 – 15.733 μg/m3] 0.954 (0.950–0.958) 0.961 (0.957–0.965) 0.993 (0.990–0.996)
3-year(2018–2020)
IQR [2.840–3.760 μg/m3] 0.996 (0.996–0.997) 0.998 (0.997–0.999) 1.000 (0.999–1.001)
Smoke Tertile
 Tertile 1 [1.195 – 2.956 μg/m3] Ref Ref Ref
 Tertile 2 [2.957 – 3.450 μg/m3] 0.977 (0.974–0.980) 0.988 (0.985–0.991) 1.004 (1.002–1.006)
 Tertile 3 [3.451 – 42.696 μg/m3] 0.967 (0.963–0.971) 0.967 (0.963–0.972) 0.993 (0.990–0.997)
Open in a new tab

PR: Prevalence Ratio

All models include a random effect for state and thin plate splines using tract centroids (k=5)

Model 1: adjusted for tract percent 65+ years old, female, racial minority

Model 2: adjusted for Model 1 variables and tract percent below poverty, population density

Model 3: adjusted for Model 2 variables and tract percent self-report current smoker, no health insurance, diabetes, no leisure physical activity, obesity, binge drinking

Table 3.

Region-stratified fully adjusted models for wildfire smoke PM2.5 exposure and cancer prevalence

South West Midwest Northeast
PR (95% CI) PR (95% CI) PR (95% CI) PR (95% CI)
Cumulative (2006–2020)
IQR [3.638–4.769 μg/m3] 0.988 (0.985–0.991) 1.003 (1.001–1.005) 1.033 (1.026–1.041) 1.020 (1.013–1.026)
Smoke Tertile
 Tertile 1 [0.964 – 3.788 μg/m3] Ref Ref Ref Ref
 Tertile 2 [3.789 – 4.522 μg/m3] 0.997 (0.993–1.000) 0.990 (0.983–0.997) 1.010 (1.007–1.014) 1.025 (1.018–1.033)
 Tertile 3 [4.523 – 15.733 μg/m3] 0.989 (0.984–0.993) 0.988 (0.980–0.997) 1.035 (1.012–1.059) 1.032 (1.023–1.041)
3-year(2018–2020)
IQR [2.840–3.760 μg/m3] 0.987 (0.984–0.990) 1.001 (1.000–1.001) 1.020 (1.013–1.026) 1.040 (1.032–1.049)
Smoke Tertile
 Tertile 1 [1.195 – 2.956 μg/m3] Ref Ref Ref Ref
 Tertile 2 [2.957 – 3.450 μg/m3] 0.995 (0.992–0.999) 0.964 (0.951–0.976) 1.006 (1.002–1.010) 1.011 (1.004–1.017)
 Tertile 3 [3.451 – 42.696 μg/m3] 0.988 (0.984–0.992) 0.949 (0.937–0.961) 1.013 (1.005–1.020) 1.043 (1.032–1.054)
Open in a new tab

Models adjusted for tract percent 65+ years old, female, racial minority, below poverty, population density, current smoker, no health insurance, diabetes, no leisure physical activity, obesity, binge drinking, and include a random effect for state and thin plate splines using tract centroids (k=5)

Highlights.

  • Wildfire smoke PM2.5 and cancer burden remain understudied despite carcinogenicity

  • No overall association with tract cancer history prevalence across exposure windows

  • Higher smoke PM2.5 linked to higher cancer history prevalence in Midwest and Northeast

Acknowledgements

HYZ was supported by the following NIH training grants: T32CA094880, UL1TR002319, and T32ES015459. SK was supported by T32CA094880. TV was supported by NIDDK K01DK125612

Footnotes

Declaration of interests

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

Hiwot Y. Zewdie reports financial support was provided by National Institutes of Health. Seigi Karasaki reports financial support was provided by National Institutes of Health. Trang VoPham reports financial support was provided by National Institutes of Health. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • 1.IARC. Outdoor Air Pollution. Accessed April 22, 2025. https://publications.iarc.fr/Book-And-Report-Series/Iarc-Monographs-On-The-Identification-Of-Carcinogenic-Hazards-To-Humans/Outdoor-Air-Pollution-2015 [Google Scholar]
  • 2.Loomis D, Grosse Y, Lauby-Secretan B, et al. The carcinogenicity of outdoor air pollution. Lancet Oncol. 2013;14(13):1262–1263. doi: 10.1016/S1470-2045(13)70487-X [DOI] [PubMed] [Google Scholar]
  • 3.Wang M, Kim RY, Kohonen-Corish MRJ, Chen H, Donovan C, Oliver BG. Particulate matter air pollution as a cause of lung cancer: epidemiological and experimental evidence. Br J Cancer. Published online April 4, 2025:1–11. doi: 10.1038/s41416-025-02999-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Thangavel P, Park D, Lee YC. Recent Insights into Particulate Matter (PM2.5)-Mediated Toxicity in Humans: An Overview. Int J Environ Res Public Health. 2022;19(12):7511. doi: 10.3390/ijerph19127511 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Li R, Zhou R, Zhang J. Function of PM2.5 in the pathogenesis of lung cancer and chronic airway inflammatory diseases. Oncol Lett. 2018;15(5):7506–7514. doi: 10.3892/ol.2018.8355 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Hill W, Lim EL, Weeden CE, et al. Lung adenocarcinoma promotion by air pollutants. Nature. 2023;616(7955):159–167. doi: 10.1038/s41586-023-05874-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Moreno-Ríos AL, Tejeda-Benítez LP, Bustillo-Lecompte CF. Sources, characteristics, toxicity, and control of ultrafine particles: An overview. Geosci Front. 2022;13(1):101147. doi: 10.1016/j.gsf.2021.101147 [DOI] [Google Scholar]
  • 8.Terzano C, Di Stefano F, Conti V, Graziani E, Petroianni A. Air pollution ultrafine particles: toxicity beyond the lung. Eur Rev Med Pharmacol Sci. 2010;14(10):809–821. [PubMed] [Google Scholar]
  • 9.Clean Air Act: A Summary of the Act and Its Major Requirements. Accessed April 22, 2025. https://www.congress.gov/crs-product/RL30853
  • 10.Burke M, Childs ML, de la Cuesta B, et al. The contribution of wildfire to PM2.5 trends in the USA. Nature. 2023;622(7984):761–766. doi: 10.1038/s41586-023-06522-6 [DOI] [PubMed] [Google Scholar]
  • 11.Burke M, Driscoll A, Heft-Neal S, Xue J, Burney J, Wara M. The changing risk and burden of wildfire in the United States. Proc Natl Acad Sci. 2021;118(2):e2011048118. doi: 10.1073/pnas.2011048118 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.US Environmental Protection Agency. National Ambient Air Quality Standards (NAAQS) for PM. August 19, 2025. Accessed September 17, 2025. https://www.epa.gov/pm-pollution/national-ambient-air-quality-standards-naaqs-pm
  • 13.Childs ML, Li J, Wen J, et al. Daily Local-Level Estimates of Ambient Wildfire Smoke PM2.5 for the Contiguous US. Environ Sci Technol. 2022;56(19):13607–13621. doi: 10.1021/acs.est.2c02934 [DOI] [PubMed] [Google Scholar]
  • 14.O’Dell K, Hornbrook RS, Permar W, et al. Hazardous Air Pollutants in Fresh and Aged Western US Wildfire Smoke and Implications for Long-Term Exposure. Environ Sci Technol. 2020;54(19):11838–11847. doi: 10.1021/acs.est.0c04497 [DOI] [PubMed] [Google Scholar]
  • 15.Krasovich Southworth E, Qiu M, Gould CF, et al. The Influence of Wildfire Smoke on Ambient PM2.5 Chemical Species Concentrations in the Contiguous US. Environ Sci Technol. 2025;59(6):2961–2973. doi: 10.1021/acs.est.4c09011 [DOI] [PubMed] [Google Scholar]
  • 16.Henkler F, Stolpmann K, Luch A. Exposure to polycyclic aromatic hydrocarbons: bulky DNA adducts and cellular responses. Exp Suppl 2012. 2012;101:107–131. doi: 10.1007/978-3-7643-8340-4_5 [DOI] [PubMed] [Google Scholar]
  • 17.Holme JA, Vondráček J, Machala M, et al. Lung cancer associated with combustion particles and fine particulate matter (PM2.5) - The roles of polycyclic aromatic hydrocarbons (PAHs) and the aryl hydrocarbon receptor (AhR). Biochem Pharmacol. 2023;216:115801. doi: 10.1016/j.bcp.2023.115801 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Bersten DC, Sullivan AE, Peet DJ, Whitelaw ML. bHLH–PAS proteins in cancer. Nat Rev Cancer. 2013;13(12):827–841. doi: 10.1038/nrc3621 [DOI] [PubMed] [Google Scholar]
  • 19.Tsay JJ, Tchou-Wong KM, Greenberg AK, Pass H, Rom WN. Aryl Hydrocarbon Receptor and Lung Cancer. Anticancer Res. 2013;33(4):1247–1256. [PMC free article] [PubMed] [Google Scholar]
  • 20.Srinivas US, Tan BWQ, Vellayappan BA, Jeyasekharan AD. ROS and the DNA damage response in cancer. Redox Biol. 2019;25:101084. doi: 10.1016/j.redox.2018.101084 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Black C, Tesfaigzi Y, Bassein JA, Miller LA. Wildfire smoke exposure and human health: Significant gaps in research for a growing public health issue. Environ Toxicol Pharmacol. 2017;55:186–195. doi: 10.1016/j.etap.2017.08.022 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Gould CF, Heft-Neal S, Johnson M, Aguilera J, Burke M, Nadeau K. Health Effects of Wildfire Smoke Exposure. Annu Rev Med. 2024;75(Volume 75, 2024):277–292. doi: 10.1146/annurev-med-052422-020909 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Abatzoglou JT, Williams AP. Impact of anthropogenic climate change on wildfire across western US forests. Proc Natl Acad Sci. 2016;113(42):11770–11775. doi: 10.1073/pnas.1607171113 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Jones MW, Abatzoglou JT, Veraverbeke S, et al. Global and Regional Trends and Drivers of Fire Under Climate Change. Rev Geophys. 2022;60(3):e2020RG000726. doi: 10.1029/2020RG000726 [DOI] [Google Scholar]
  • 25.Gao Y, Huang W, Xu Z, et al. Wildfire-related PM2.5 and cause-specific cancer mortality. Ecotoxicol Environ Saf. 2024;285:117023. doi: 10.1016/j.ecoenv.2024.117023 [DOI] [PubMed] [Google Scholar]
  • 26.Elser H, Frankland TB, Chen C, et al. Wildfire Smoke Exposure and Incident Dementia. JAMA Neurol. 2025;82(1):40–48. doi: 10.1001/jamaneurol.2024.4058 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Korsiak J, Pinault L, Christidis T, Burnett RT, Abrahamowicz M, Weichenthal S. Long-term exposure to wildfires and cancer incidence in Canada: a population-based observational cohort study. Lancet Planet Health. 2022;6(5):e400–e409. doi: 10.1016/S2542-5196(22)00067-5 [DOI] [PubMed] [Google Scholar]
  • 28.Greenlund KJ, Lu H, Wang Y, et al. PLACES: Local Data for Better Health. Prev Chronic Dis. 2022;19:E31. doi: 10.5888/pcd19.210459 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Zhang X, Holt JB, Yun S, Lu H, Greenlund KJ, Croft JB. Validation of Multilevel Regression and Poststratification Methodology for Small Area Estimation of Health Indicators From the Behavioral Risk Factor Surveillance System. Am J Epidemiol. 2015;182(2):127–137. doi: 10.1093/aje/kwv002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Zhang X, Holt JB, Lu H, et al. Multilevel regression and poststratification for small-area estimation of population health outcomes: a case study of chronic obstructive pulmonary disease prevalence using the behavioral risk factor surveillance system. Am J Epidemiol. 2014;179(8):1025–1033. doi: 10.1093/aje/kwu018 [DOI] [PubMed] [Google Scholar]
  • 31.CDC. PLACES: Local Data for Better Health. Centers for Disease Control and Prevention. July 13, 2023. Accessed October 5, 2023. https://www.cdc.gov/places/index.html [Google Scholar]
  • 32.Kottek M, Grieser J, Beck C, Rudolf B, Rubel F. World map of the Köppen-Geiger climate classification updated. Published online 2006. Accessed October 8, 2025. https://opus.bibliothek.uni-augsburg.de/opus4/files/40083/metz_Vol_15_No_3_p259-263_World_Map_of_the_Koppen_Geiger_climate_classification_updated_55034.pdf [Google Scholar]
  • 33.Bryant C, Wheeler NR, Rubel F, French RH. Kgc: Koeppen-Geiger climatic zones. R package version 1.0. 0.2. Published online 2017. [Google Scholar]
  • 34.Richardson DB, Kinlaw AC, MacLehose RF, Cole SR. Standardized binomial models for risk or prevalence ratios and differences. Int J Epidemiol. 2015;44(5):1660–1672. doi: 10.1093/ije/dyv137 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Petersen MR, Deddens JA. A comparison of two methods for estimating prevalence ratios. BMC Med Res Methodol. 2008;8:9. doi: 10.1186/1471-2288-8-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.VoPham T, Liu T, Cortez M, et al. Exposure to outdoor air pollution, wildfires, and cancer survival in the United States. Cancer Epidemiol. 2025;98:102899. doi: 10.1016/j.canep.2025.102899 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Yu P, Xu R, Wu Y, et al. Cancer mortality risk from short-term PM2.5 exposure and temporal variations in Brazil. J Hazard Mater. 2024;473:134606. doi: 10.1016/j.jhazmat.2024.134606 [DOI] [PubMed] [Google Scholar]
  • 38.Brey SJ, Ruminski M, Atwood SA, Fischer EV. Connecting smoke plumes to sources using Hazard Mapping System (HMS) smoke and fire location data over North America. Atmospheric Chem Phys. 2018;18(3):1745–1761. doi: 10.5194/acp-18-1745-2018 [DOI] [Google Scholar]
  • 39.Larsen AE, Reich BJ, Ruminski M, Rappold AG. Impacts of fire smoke plumes on regional air quality, 2006–2013. J Expo Sci Environ Epidemiol. 2018;28(4):319–327. doi: 10.1038/s41370-017-0013-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Kandoth C, McLellan MD, Vandin F, et al. Mutational landscape and significance across 12 major cancer types. Nature. 2013;502(7471):333–339. doi: 10.1038/nature12634 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Swanton C, Bernard E, Abbosh C, et al. Embracing cancer complexity: Hallmarks of systemic disease. Cell. 2024;187(7):1589–1616. doi: 10.1016/j.cell.2024.02.009 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.White AJ, Hart JE, Quraishi SM, et al. Air Pollutants and Breast Cancer Risk: A Parallel Analysis of Five Large US Prospective Cohorts. Am J Public Health. 2025;115(12):2030–2043. doi: 10.2105/AJPH.2025.308247 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Bian Q, Ford B, Pierce JR, Kreidenweis SM. A Decadal Climatology of Chemical, Physical, and Optical Properties of Ambient Smoke in the Western and Southeastern United States. J Geophys Res Atmospheres. 2020;125(1):e2019JD031372. doi: 10.1029/2019JD031372 [DOI] [Google Scholar]
  • 44.Iglesias V, Balch JK, Travis WR. U.S. fires became larger, more frequent, and more widespread in the 2000s. Sci Adv. 2022;8(11):eabc0020. doi: 10.1126/sciadv.abc0020 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Donovan VM, Crandall R, Fill J, Wonkka CL. Increasing Large Wildfire in the Eastern United States. Geophys Res Lett. 2023;50(24):e2023GL107051. doi: 10.1029/2023GL107051 [DOI] [Google Scholar]
  • 46.O’Dell K, Bilsback K, Ford B, et al. Estimated Mortality and Morbidity Attributable to Smoke Plumes in the United States: Not Just a Western US Problem. GeoHealth. 2021;5(9):e2021GH000457. doi: 10.1029/2021GH000457 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Hertelendy AJ, Howard C, Sorensen C, et al. Seasons of smoke and fire: preparing health systems for improved performance before, during, and after wildfires. Lancet Planet Health. 2024;8(8):e588–e602. doi: 10.1016/S2542-5196(24)00144-X [DOI] [PubMed] [Google Scholar]
  • 48.Burke M, Heft-Neal S, Li J, et al. Exposures and behavioural responses to wildfire smoke. Nat Hum Behav. 2022;6(10):1351–1361. doi: 10.1038/s41562-022-01396-6 [DOI] [PubMed] [Google Scholar]

Associated Data

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

Supplementary Materials

1
NIHMS2205310-supplement-1.docx (1.2MB, docx)

RESOURCES