Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Sep 29.
Published in final edited form as: Environ Res. 2026 Jul 13;306(Pt 2):125250. doi: 10.1016/j.envres.2026.125250

Association between ambient air pollution and colorectal cancer risk: The Multiethnic Cohort Study

Ugonna Ihenacho a, Chiuchen Tseng a, Jun Wu b, Scott Fruin a, Timothy V Larson c,d, Salma Shariff-Marco e,f, Loïc Le Marchand g, Daniel O Stram a, Brian Z Huang a, Lynne R Wilkens g, Christopher A Haiman a, Beate Ritz h, Iona Cheng e,f, Anna H Wu a
PMCID: PMC13617549  NIHMSID: NIHMS2200316  PMID: 42442448

Abstract

Growing evidence links air pollution to colorectal cancer (CRC) incidence. We examined this association within the large Multiethnic Cohort Study (MEC).

Geocoded residential addresses for 98,675 California MEC participants were appended to ambient air pollution measures of PM2.5 (particulate matter [PM] with an aerodynamic diameter <2.5μm), PM10 (PM <10μm), nitrogen dioxide (NO2), nitrogen oxides (NOx), carbon monoxide (CO), and ozone (O3), generated from enrollment (1993–1996) to December 31, 2018. Multivariable-adjusted Cox proportional hazards models evaluated associations of time-varying air pollutants with CRC incidence (n=3,217 cases). We assessed heterogeneity in associations by demographics, tumor stage, and anatomical subsite.

CRC incidence increased with PM2.5 exposure (per 10μg/m3; hazard ratio [HR]=1.13, 95% confidence interval [CI]=0.96–1.33), mainly among female (HR=1.29, 95% CI=1.03–1.62) but not among male participants (Pheterogeneity=0.08). CRC incidence also increased with NOx exposure among female (HR=1.22, 95% CI=1.01–1.48) but not male participants (Pheterogeneity=0.07). Increased incidence associated with PM2.5 (HR=1.36, 95% CI=1.05–1.76), NO2 (per 20 parts per billion [ppb]; HR=1.32, 95% CI=1.05–1.68) and CO (per 1000 ppb; HR=1.36, 95% CI=1.01–1.84) exposures were observed for left colon and rectal cancers combined, but not right colon cancers (Pheterogeneity by site=0.08, 0.06 and 0.13, respectively). Associations of PM2.5 and NO2 with rectal cancer incidence differed by population group (Pheterogeneity=0.04 and 0.03, respectively), and was mostly driven by positive associations among Latino participants.

In summary, increasing PM2.5, NO2, NOx, and CO exposures were suggestively associated with increased CRC incidence, particularly among female and Latino participants and for left colon and rectal cancers.

Keywords: air pollution, colorectal cancer, multiethnic cohort, incidence, epidemiology

1. Introduction

In 2013, the International Agency for Research on Cancer (IARC) classified outdoor air pollution as a Group 1 human carcinogen, based primarily on studies of fine particulate matter (PM2.5; particulate matter [PM] with an aerodynamic diameter <2.5μm) and lung cancer (Hamra et al., 2014; Speizer, 1983). Growing evidence suggests that air pollution may be a risk factor for cancer at other organ sites, including colorectal cancer (CRC) (Pritchett et al., 2022; Turner et al., 2020). In rodent models, PM exposure has been found to increase intestinal permeability and levels of pro-inflammatory factors in colonic epithelia cells, suggesting a potential biological pathway to carcinogenesis in humans (Cheng et al., 2024; Kish et al., 2013; Mutlu et al., 2018).

CRC is the fourth most common cancer in the United States (US) with 154,270 incident cases estimated in 2025 (Siegel et al., 2025). Although the incidence of CRC has decreased in the past four decades due to increased screening and changes in risk factors, including reduced smoking and increased nonsteroidal anti-inflammatory drug (NSAID) use (Edwards et al., 2010), CRC incidence in 2017–2021 remained higher among African American (40.4 per 100,000) than in non-Hispanic (NH) White (35.0 per 100,000) and Asian American and Pacific Islander (AAPI; 27.9 per 100,000) individuals (Siegel et al., 2025). Moreover, aggregate AAPI statistics may mask important heterogeneity. For example, although AAPI individuals tended to have the lowest incidence of CRC when grouped together, in the Multiethnic Cohort Study (MEC), CRC incidence among Japanese American men and women was 1.27 and 1.49 times that of NH White men and women, respectively (Ollberding et al., 2011). The persistence of racial and ethnic disparities in CRC has been attributed to multifactorial structural and social drivers of health that result in differential exposure to environmental hazards, which influence the gut microbiome and promote inflammatory and immune pathways associated with cancer risk with different effects across sociodemographic groups defined by race and ethnicity, sex, or socioeconomic status (SES) (Carethers, 2021; Godoy-Vitorino et al., 2026; He et al., 2026; Polite et al., 2017). Although health behaviors such as diet, smoking, physical activity and screening practices are commonly evaluated in CRC and cancer disparities research (Harmon et al., 2014; Islami et al., 2018; Le Marchand et al., 1997; Sninsky et al., 2022), few studies have examined the role of environmental factors in CRC incidence in large multiethnic populations. Associations with many established lifestyle risk factors vary across CRC subsites (Wang et al., 2020). In addition, distinct associations between the oral microbiome and risk of CRC by anatomic subsite have been reported (Vogtmann et al., 2025), highlighting the need for further investigations of environmental exposures by subsite.

The risk of CRC in association with exposure to PM2.5 has been investigated using data from the Prostate, Lung, Colon and Ovarian (PLCO) Cancer Screening trial, the Surveillance, Epidemiology, and End Results (SEER) Program, and the UK Biobank (Chu et al., 2021; Coleman et al., 2020; Jiang et al., 2024). A statistically significant positive association with CRC was observed in the PLCO study (Chu et al., 2021) and a positive association with CRC was suggested in the UK Biobank study (Jiang et al., 2024). In the SEER-based ecological study, county-level PM2.5 was not associated with colon cancer or rectal cancer after adjusting for multiple comparisons (Coleman et al., 2020). However, these previous studies assessed air pollutant exposure based solely on residential address at baseline (Chu et al., 2021; Jiang et al., 2024) or used county-level cross-sectional PM measures (Coleman et al., 2020), without accounting for longitudinal changes in residential history.

We used the California component of the Multiethnic Cohort Study (MEC) to conduct a prospective study to evaluate the associations of time-varying PM and gaseous air pollutants exposures, assessed from baseline (1993–1996) to December 31, 2018 with CRC incidence. To our knowledge, this is the first prospective study of time-varying measures of ambient air pollutants and CRC incidence that includes large numbers of racially and ethnically diverse individuals, and an assessment of sex-specific and subsite-specific associations.

2. Methods

2.1. Study subjects

The MEC is a large population-based prospective cohort study of more than 215,000 men and women from California (mostly Los Angeles County) and Hawai‘i, ages 45–74 years at enrollment in 1993–1996, and self-reported as African American, Japanese American, Latino, Native Hawaiian, and White (Kolonel et al., 2000). Participants completed a 26-page questionnaire (in English or Spanish) upon cohort entry and provided information on demographics, lifestyle, diet, medical history and family history of cancer. All participants were followed prospectively for diagnosis of incident invasive CRC (ICD-O-3 C18.0-C18.9, C19.9, C20.9, C26.0) through routine linkages with the Surveillance, Epidemiology and End Results (SEER) statewide registries in California and Hawai‘i, and for vital status through linkages to the National Death Index and state death certificate files. CRC subsite, histology, stage, grade, and other tumor characteristics were obtained from the cancer registries. For this study on air pollution, of the 107,024 eligible California MEC participants with baseline residential addresses, we excluded those who did not self-report as identifying as one of the five racial and ethnic groups (n=4,153), had CRC prior to cohort entry as determined by self-report at baseline questionnaire or cancer registry linkage (n=1,642), had a questionable address or invalid follow-up time (n=405), or required imputation for >50% of the air pollution data (n=2,149). The final study population included 98,675 California MEC participants. Study participants were followed from date of cohort entry (1993–1996) to the earliest date of invasive CRC diagnosis, death, study end date (December 31, 2018), or residential move outside of California. Institutional Review Boards at the University of Southern California, the University of Hawai‘i Cancer Center, and University of California, San Francisco approved the study protocol.

2.2. Address History and Geocoding

Residential addresses at baseline and changes during follow-up were obtained from several sources including participant self-report, follow-up questionnaires, annual newsletter mailings, and administrative and registry linkages. We geocoded MEC addresses from 1993 to 2018 using procedures described previously (Cheng et al., 2020; Conroy et al., 2018) in ArcGIS (ESRI, Redlands, CA), with the best available locator data from government agencies and commercial vendors. Addresses were assigned to land parcels (latitude and longitude) where possible, or to street segments. We then linked each geocoded residence to the pollutant surfaces by location: PM2.5 was assigned the value of the 1-km grid cell in which the coordinates fell, and PM10 (PM <10μm) and gaseous pollutants (nitrogen dioxide [NO2], nitrogen oxides [NOX], carbon monoxide [CO], and ozone [O3]) were assigned from the empirical Bayesian kriging predictions at those coordinates. Using residential histories, we built a monthly exposure series for each participant from cohort entry to end of follow-up, updating values at each move, and averaged these cumulatively as time-varying exposures. If participants moved outside of California, participation was censored at the last date of eligible residency.

2.3. Air pollution data

Ambient air pollutants were derived from various sources, including CO originating predominantly from traffic emissions and PM2.5, PM10, NO2, and NOx arising from both direct traffic emissions and secondary atmospheric formation. These pollutants typically reflect traffic-related pollution and reach peak concentrations near roadways. In contrast, O3 forms exclusively through secondary photochemical reactions, with peak concentrations occurring downwind in suburban areas. To reduce potential exposure misclassification due to factors affecting pollution levels (e.g., seasonal or policy changes), we used time-varying monthly averages of PM and gaseous pollutants for each participant based on their duration of residence at location(s) across follow-up. We generated individual-level, time-varying air pollutant exposure estimates for all California MEC participants’ residential addresses from cohort enrollment (1993–1996) through December 31, 2018, using previously validated spatiotemporal models (Cheng et al., 2022b; Wu et al., 2024).

For PM2.5, we used satellite-based estimates using a fine-resolution spatiotemporal model which provides validated, publicly available PM2.5 outputs at a 1-km resolution over North America (Cheng et al., 2022b; Meng et al., 2019). Cross-validation of that model against ground-based PM2.5 measurements yielded R2 of approximately 0.6 to 0.85 over the period relevant to our study (1988 onward) (Meng et al., 2019). For PM10 and gaseous pollutants (NO2, NOX, CO, O3), we applied empirical Bayesian kriging interpolation to spatially interpolate monthly concentrations of these pollutants from the US Environmental Protection Agency’s (EPA) routine monitoring station data (Cheng et al., 2022b; Wu et al., 2016). For these pollutants, leave-one-out cross-validation of the monthly estimates, conducted in a separate project that developed these surfaces, indicated reasonable performance: NO2 (R2 = 0.74; RMSE = 6.08 ppb) and O3 (R2 = 0.72; RMSE = 5.81 ppb) (Wu et al., 2016). PM10, NOx, and CO were interpolated using the identical empirical Bayesian kriging procedure applied to the same US EPA monitoring data; pollutant-specific cross-validation statistics were not separately derived for these three pollutants, which we note as a limitation given that monitoring density varies by pollutant. For participants with incomplete address records or incomplete air pollution data, exposure levels were imputed using last known estimates. In this analysis, 2% of participants had more than 50% imputed data and they were excluded from the analysis. Supplementary table S1 summarizes the demographic characteristics of the excluded participants and participants included in the analysis. Among those included in the analysis, 90.2% had 0% imputed, 4.0% had <10% imputed, 1.3% had 10 to <20% imputed, and 2.7% had 20 to <50% imputed exposure levels.

2.4. Participant characteristics and covariates

Risk factors assessed from the baseline study questionnaire included age at cohort entry, education (high school graduate or less, some college, college graduate, graduate and professional school, missing), body mass index (BMI; categorical variable calculated from height and weight measurements; underweight: <18.5 kg/m2, normal: 18.5 to >25 kg/m2, overweight: 25 to >30 kg/m2, obese: ≥30 kg/m2), alcohol intake (nondrinker, <12 grams/day, ≥12 grams/day), smoking status (no, formerly smoked with <20 pack-years, formerly smoked with ≥20 pack-years, formerly smoked with pack-years unknown, currently smoke with <20 pack-years, currently smoke with ≥20 pack-years, currently smoke with pack-years unknown, missing), first degree family history of CRC (no, yes, missing), multivitamin use at least once a week in the last year (no, yes, missing), NSAID use at least two times per week for one month or longer (no, yes, missing), moderate to vigorous physical activity in the past year (hours/week), energy intake in the past year (kcal/day), processed red meat consumption (grams/kcal/day), menopausal hormone use (never estrogen use with or without past or current progesterone use, past estrogen use with or without past progesterone use, current estrogen use alone, current estrogen use with past or current progesterone, missing or not applicable), and long-term oral contraceptive use (no, yes, missing). Examples of moderate physical activity provided on the study questionnaire included housework, brisk walking, golfing, bowling, bicycling on level ground, and gardening; examples of vigorous activities included jogging, bicycling on hills, tennis, racquetball, swimming laps, aerobics, moving heavy furniture, loading or unloading trucks, shoveling, and weightlifting. A composite measure of neighborhood socioeconomic status (nSES) was developed based on a principal component analysis of seven US Census-based socioeconomic indicators: education, proportion with a blue-collar job, unemployment, median income, median rent, median house value, and proportion below 200% of the poverty level (Yang et al., 2014; Yost et al., 2001) and appended to geocoded addresses at the level of the census block group for baseline and study end dates.

2.5. Statistical analysis

CRC incidence was modeled using Cox proportional hazards regression with age as the time metric. To account for changes in air pollutant exposures over time, we estimated monthly air pollutant exposures in each calendar month/year from cohort entry through end of study (i.e., CRC diagnosis, death, or study end date). Specifically, we have information on month and year of each residential move that was appended to month and year of air pollutant concentration to assign a time-varying air pollution exposure. The month/year air pollution exposure was included in the cumulative average based on the residence in that month. Proportional values were assigned for partial months, such as in the start month/year, end month/year and the month/year of each individual moves. If the start date for the address is before or on the 15th day of the month, the calendar month for the date was used to compute month/year value, and if the start date is after the 15th day of the month, the preceding calendar month was used to compute month/year value. Similarly, if the end date for the address was before or on the 15th day of the month, the previous calendar month was used to compute year/month value, and if the date is after the 15th day of the month, the calendar month was used to compute month/year value.

We assessed pairwise correlations between air pollutant exposures at baseline and over follow-up. Averages of cumulative air pollutant exposures were entered into Cox proportional hazard models in standardized units as time-dependent variables. The interquartile ranges for the 12-month baseline measures were as follows: PM2.5 4.12 μg/m3, PM10 6.51 μg/m3, NO2 10.90 ppb, NOx 29.19 ppb, CO 443.43 ppb, O3 6.28 ppb. For each pollutant, we selected units that were frequently reported in prior studies (PM2.5: per 10 μg/m3, PM10: per 10 μg/m3, NO2: per 20 ppb, NOx: per 50 ppb, CO: per 1000 ppb, O3: per 10 ppb), allowing for an easy comparison of our results and published results (Bogumil et al., 2021; Cheng et al., 2020; Wu et al., 2024). Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated for minimally adjusted models with age at cohort entry and race and ethnicity as strata variables. Covariates were selected based on a priori knowledge and statistically significant associations with CRC in multivariable models among our study population. A stepwise model building approach was used to evaluate the changes in the HR estimates with sequential adjustment for confounders. Fully adjusted models included sex, education, nSES, BMI, alcohol intake, smoking status, family history of CRC, multivitamin use, NSAID use, physical activity, energy intake, processed red meat consumption, and menopausal hormone use. Cross-product interaction terms were used to assess the proportional hazards assumption for each covariate with time, and no serious violations were observed as there was less than a 10% change in the HR estimates when the covariates were included as strata variables in Cox regression models. We evaluated heterogeneity in the associations by sex, race and ethnicity, and potential effect modifiers using stratified analyses and the statistical significance of interactions evaluated by the Wald test of the cross-product interaction terms. Separate models by tumor subsite (right colon, ICD-O-3 C18.0-C18.5; left colon, C18.8-C18.7; rectum, C19.9, C20.9) and stage (local, regional + distant) were evaluated; differences by tumor subsite and stage were assessed by a competing risk analysis using the Lunn-McNeil approach (Lunn and McNeil, 1995). All hypotheses were two-sided and assessed at a significance level of P<0.05. Analyses were completed using SAS software version 9.4 (SAS Institute Inc., Cary, NC).

3. Results

3.1. Study population

We evaluated the association of ambient air pollutant exposures and CRC incidence among 98,675 California MEC participants with an average of 19.6 years of follow-up. The study population consisted of 42,088 (42.7%) men and 56,587 (57.3%) women who self-identified as African American (31.4%), Japanese American (11.7%), Latino (43.4%), Native Hawaiian (0.2%), or NH White (13.4%) (Table 1). At study enrollment, the average age of participants was 60.8 ± 8.4 years. Overall, 3,217 incident invasive CRC cases were diagnosed during the follow-up period. We observed racial and ethnic differences in the distribution of education, BMI, smoking status, alcohol intake, and other CRC risk factors. African American participants were the most likely to live in neighborhoods in the lowest quintile of baseline nSES (37.7%) and Japanese American participants were the most likely to live in neighborhoods in the highest quintile of baseline nSES (31.6%) compared to participants of the other racial and ethnic groups.

Table 1.

Summary of baseline study characteristics, overall and by race and ethnicity, in the California MEC Study, 1993–2018

Overall Race and Ethnicity
n=98675 African American n=30979 Japanese American n=11542 Latino n=42794 White n=13201
Sex
 Male 42088 (42.7) 11089 (35.8) 5600 (48.5) 20637 (48.2) 4674 (35.4)
 Female 56587 (57.3) 19890 (64.2) 5942 (51.5) 22157 (51.8) 8527 (64.6)
Age group, years
 ≤49 11906 (12.1) 4267 (13.8) 1424 (12.3) 4787 (11.2) 1396 (10.6)
 50–54 12937 (13.1) 4167 (13.5) 1429 (12.4) 5854 (13.7) 1457 (11.0)
 55–59 18098 (18.3) 4498 (14.5) 1607 (13.9) 9592 (22.4) 2363 (17.9)
 60–64 19131 (19.4) 4298 (13.9) 2014 (17.4) 10091 (23.6) 2690 (20.4)
 65–69 18544 (18.8) 6555 (21.2) 2221 (19.2) 7139 (16.7) 2615 (19.8)
 70–74 15434 (15.6) 6116 (19.7) 2299 (19.9) 4679 (10.9) 2334 (17.7)
 75–79 2625 (2.7) 1078 (3.5) 548 (4.7) 652 (1.5) 346 (2.6)
Education
 ≤High school graduate 50146 (50.8) 12772 (41.2) 3503 (30.4) 29071 (67.9) 4738 (35.9)
 Some college 28198 (28.6) 11122 (35.9) 4071 (35.3) 8785 (20.5) 4154 (31.5)
 College graduate 9909 (10.0) 3486 (11.3) 2433 (21.1) 2021 (4.7) 1952 (14.8)
 Graduate and professional school 8908 (9.0) 3136 (10.1) 1434 (12.4) 2110 (4.9) 2217 (16.8)
 Missing 1514 (1.5) 463 (1.5) 101 (0.9) 807 (1.9) 140 (1.1)
Baseline nSES
 Quintile 1 - Low 27764 (28.1) 11671 (37.7) 785 (6.8) 13961 (32.6) 1327 (10.1)
 Quintile 2 20598 (20.9) 7029 (22.7) 1162 (10.1) 10599 (24.8) 1790 (13.6)
 Quintile 3 18358 (18.6) 4979 (16.1) 2649 (23.0) 8114 (19.0) 2567 (19.4)
 Quintile 4 16384 (16.6) 4099 (13.2) 3152 (27.3) 5711 (13.3) 3385 (25.6)
 Quintile 5 - High 12982 (13.2) 2072 (6.7) 3645 (31.6) 3557 (8.3) 3674 (27.8)
 Missing 2589 (2.6) 1129 (3.6) 149 (1.3) 852 (2.0) 458 (3.5)
Body mass index, kg/m2
 <25 33197 (33.6) 8038 (25.9) 7496 (64.9) 11975 (28.0) 5631 (42.7)
 25–<30 40614 (41.2) 12282 (39.6) 3464 (30.0) 19896 (46.5) 4909 (37.2)
 ≥30 23507 (23.8) 9693 (31.3) 571 (4.9) 10574 (24.7) 2630 (19.9)
 Missing 1357 (1.4) 966 (3.1) 11 (0.1) 349 (0.8) 31 (0.2)
Alcohol intake, grams/day
 Nondrinker 52108 (52.8) 17544 (56.6) 7036 (61.0) 21930 (51.2) 5518 (41.8)
 <12 31490 (31.9) 9042 (29.2) 3066 (26.6) 14437 (33.7) 4893 (37.1)
 ≥12 15077 (15.3) 4393 (14.2) 1440 (12.5) 6427 (15.0) 2790 (21.1)
Smoking status and pack- years
 No 42672 (43.2) 11472 (37) 5509 (47.7) 20244 (47.3) 5381 (40.8)
 Formerly smoked, <20 27777 (28.1) 8970 (29) 3265 (28.3) 11990 (28.0) 3507 (26.6)
 Formerly smoked, ≥20 7305 (7.4) 2302 (7.4) 1217 (10.5) 2030 (4.7) 1746 (13.2)
 Formerly smoked, pack- years unknown 2541 (2.6) 900 (2.9) 139 (1.2) 1306 (3.1) 192 (1.5)
 Currently smoke, <20 9925 (10.1) 4385 (14.2) 662 (5.7) 4018 (9.4) 837 (6.3)
 Currently smoke, ≥20 5879 (6) 2284 (7.4) 634 (5.5) 1605 (3.8) 1346 (10.2)
 Currently smoke, pack- years unknown 391 (0.4) 202 (0.7) 11 (0.1) 149 (0.3) 29 (0.2)
 Missing 2185 (2.2) 464 (1.5) 105 (0.9) 1452 (3.4) 163 (1.2)
Family history of colon cancer in first degree relative
 No 75726 (76.7) 22952 (74.1) 9159 (79.4) 32979 (77.1) 10510 (79.6)
 Yes 6718 (6.8) 2446 (7.9) 1253 (10.9) 1910 (4.5) 1101 (8.3)
 Missing 16231 (16.4) 5581 (18.0) 1130 (9.8) 7905 (18.5) 1590 (12.0)
Multivitamin use
 No 45501 (46.1) 14586 (47.1) 4849 (42.0) 20267 (47.4) 5726 (43.4)
 Yes 50456 (51.1) 15318 (49.4) 6518 (56.5) 21293 (49.8) 7245 (54.9)
 Missing 2718 (2.8) 1075 (3.5) 175 (1.5) 1234 (2.9) 230 (1.7)
NSAID use
 No 38601 (39.1) 10398 (33.6) 6536 (56.6) 16634 (38.9) 4966 (37.6)
 Yes 56885 (57.6) 19408 (62.6) 4834 (41.9) 24563 (57.4) 7988 (60.5)
 Missing 3189 (3.2) 1173 (3.8) 172 (1.5) 1597 (3.7) 247 (1.9)
Physical activity (hours/week), quintiles
 Quintile 1: 0–1.50 (M); 0–0.75 (F) 21390 (21.7) 6601 (21.3) 1674 (14.5) 11153 (26.1) 1937 (14.7)
 Quintile 2: 1.50–3.25 (M); 0.75–2.50 (F) 20116 (20.4) 7272 (23.5) 2269 (19.7) 8105 (18.9) 2445 (18.5)
 Quintile 3: 3.25–5.75 (M); 2.50–5.00 (F) 19862 (20.1) 6633 (21.4) 2577 (22.3) 7869 (18.4) 2747 (20.8)
 Quintile 4: 5.75–11.00 (M); 5.00–8.50 (F) 16058 (16.3) 4657 (15.0) 2227 (19.3) 6576 (15.4) 2570 (19.5)
 Quintile 5: 11.00–93.00 (M); 8.50–93.00 (F) 18439 (18.7) 4803 (15.5) 2671 (23.1) 7600 (17.8) 3320 (25.1)
 Missing 2810 (2.8) 1013 (3.3) 124 (1.1) 1491 (3.5) 182 (1.4)
Energy intake (kcal/day), quintiles
 Quintile 1: 488.85–1,435.20 (M); 425.20–1,174.19 (F) 19734 (20.0) 7898 (25.5) 2097 (18.2) 7188 (16.8) 2531 (19.2)
 Quintile 2: 1,435.20–1,908.32 (M); 1,174.19–1,563.25 (F) 19736 (20.0) 6255 (20.2) 2901 (25.1) 7310 (17.1) 3231 (24.5)
 Quintile 3: 1,908.32–2,438.30 (M); 1,563.25–1,988.17 (F) 19735 (20.0) 5840 (18.9) 2929 (25.4) 7915 (18.5) 3024 (22.9)
 Quintile 4: 2,438.30–3,274.77 (M); 1,988.17–2,673.30 (F) 19736 (20.0) 5563 (18.0) 2390 (20.7) 9029 (21.1) 2723 (20.6)
 Quintile 5: 3,274.77–8,670.39 (M); 2,673.30–7,401.34 (F) 19734 (20.0) 5423 (17.5) 1225 (10.6) 11352 (26.5) 1692 (12.8)
Processed red meat (grams/kcal/day), quintiles
 Quintile 1: 0–9.94 (M); 0–7.15 (F) 19725 (20.0) 5061 (16.3) 2523 (21.9) 8831 (20.6) 3288 (24.9)
 Quintile 2: 9.94–16.40 (M); 7.15–12.76 (F) 19712 (20.0) 4916 (15.9) 2426 (21.0) 9466 (22.1) 2881 (21.8)
 Quintile 3: 16.40–22.77 (M); 12.76–18.60 (F) 19769 (20.0) 5389 (17.4) 2582 (22.4) 9161 (21.4) 2597 (19.7)
 Quintile 4: 22.77–31.32 (M); 18.60–26.64 (F) 19734 (20.0) 6492 (21.0) 2338 (20.3) 8492 (19.8) 2374 (18.0)
 Quintile 5: 31.32–215.89 (M); 26.64–184.98 (F) 19735 (20.0) 9121 (29.4) 1673 (14.5) 6844 (16.0) 2061 (15.6)
Hormone therapy use
 Never estrogen use, with or without past or current progesterone use 30454 (30.9) 11418 (36.9) 3242 (28.1) 12019 (28.1) 3739 (28.3)
 Past estrogen use, with or without past progesterone use 10177 (10.3) 3946 (12.7) 786 (6.8) 3810 (8.9) 1619 (12.3)
 Current estrogen use alone 6386 (6.5) 2082 (6.7) 742 (6.4) 2264 (5.3) 1288 (9.8)
 Current estrogen use with progesterone - past or current 6106 (6.2) 1248 (4.0) 984 (8.5) 2270 (5.3) 1597 (12.1)
 Missing or Not applicable (Male participant) 45552 (46.2) 12285 (39.7) 5788 (50.1) 22431 (52.4) 4958 (37.6)
Open in a new tab

All values are frequencies (column percentages).

Abbreviations: F, female; M, male; NSAID, non-steroidal anti-inflammatory drug; nSES, neighborhood socioeconomic status.

Overall, average pollutant concentrations at baseline were higher than the average cumulative concentrations across the follow-up period, except for O3 (Supplementary Table S2). African American participants had the highest baseline levels of PM2.5 and the highest average cumulative levels of NOx and CO. Japanese American participants had the highest baseline levels of NOx and CO. Latino participants had the highest levels of PM10 and NO2 at baseline and highest average cumulative levels of PM2.5, PM10 and NO2. White participants had the highest O3 levels at baseline and over follow-up. Across the follow-up period, positive correlations were observed among the average cumulative measures of PM2.5, PM10, NO2, NOx and CO (r = 0.62 to 0.90; Supplementary Table S3). O3 was inversely correlated with average cumulative measures of PM and the other gaseous pollutants (r = −0.71 to −0.15).

3.2. Ambient air pollution and CRC incidence overall and by sex

CRC incidence increased with increasing PM2.5 exposure (per 10 μg/m3; HR 1.13, 95% CI 0.96–1.33) and with NO2, NOx and CO exposures (respective HRs of 1.12, 1.07, 1.12 per standardized unit increase) but the 95% CIs included 1.0 (Table 2). Analyses by sex showed a statistically significant positive association between PM2.5 exposure and CRC incidence among female participants (HR 1.29, 95% CI 1.03–1.62) but not among male participants (HR 0.96, 95% CI 0.76–1.22) (Pheterogeneity = 0.08). Similarly, increasing NOx exposure was associated with a higher incidence of CRC among female participants (per 50 parts per billion [ppb]; HR 1.22, 95% CI 1.01–1.48) and a null association among male participants (HR 0.94; 95% CI 0.77–1.15) (Pheterogeneity = 0.07). We did not observe any statistically significant associations for PM10 and O3 with CRC incidence overall or by sex.

Table 2.

Association of ambient air pollutants and CRC, overall and by sex, in the California MEC Study, 1993–2018

Overall Male Female Pheterogeneity by sex
Cases HR (95% CI)a Cases HR (95% CI)a Cases HR (95% CI)a
PM2.5 3,217 1.13 (0.96–1.33) 1,533 0.96 (0.76–1.22) 1,684 1.29 (1.03–1.62) 0.08
PM10 3,187 1.02 (0.93–1.11) 1,521 1.00 (0.87–1.14) 1,666 1.04 (0.92–1.17) 0.65
NO2 3,187 1.12 (0.96–1.31) 1,521 1.12 (0.89–1.41) 1,666 1.14 (0.92–1.41) 0.95
NOx 3,129 1.07 (0.93–1.23) 1,501 0.94 (0.77–1.15) 1,628 1.22 (1.01–1.48) 0.07
CO 3,187 1.12 (0.91–1.39) 1,521 0.98 (0.72–1.34) 1,666 1.29 (0.97–1.72) 0.21
O3 3,187 1.02 (0.91–1.15) 1,521 1.08 (0.91–1.28) 1,666 0.97 (0.82–1.14) 0.38
Open in a new tab

Multivariable models adjusting for sex (except in sex-stratified models), education, nSES, body mass index, alcohol use, smoking status, family history of colon cancer, multivitamin use, NSAID use, physical activity, energy intake, processed red meat consumption, and menopausal hormone use (among female participants). All models include age at baseline and race and ethnicity as strata variables. Abbreviations: CI, confidence interval; CO, carbon monoxide; CRC, colorectal cancer; HR, hazard ratio; MEC, Multiethnic Cohort; NO2, nitrogen dioxide; NOx, nitrogen oxide; NSAID, non-steroidal anti-inflammatory drug; nSES, neighborhood socioeconomic status; O3, ozone; PM, particulate matter; ppb, parts per billion.

a

Colorectal cancer risk per 10 μg/m3 satellite-based PM2.5, per 10 μg/m3 krigged PM10, per 20 ppb of krigged NO2, per 50 ppb of krigged NOx, per 1000 ppb of krigged CO, and per 10 ppb of krigged O3.

3.3. Ambient air pollution and CRC incidence by race and ethnicity

In analyses by race and ethnicity (Table 3), all pollutants (except for O3) were associated with increased HRs for CRC incidence among Japanese American and Latino participants, whereas the HR estimates were near null for African American and White participants. Among Latino participants, a statistically significant association was found for PM2.5 (HR 1.47, 95% CI 1.03–2.08) and NO2 (per 20 ppb; HR 1.42, 95% CI 1.04–1.93) with CRC incidence, but there was no evidence of statistically significant heterogeneity in these associations by race and ethnicity (Pheterogeneity ranged from 0.17 to 0.73). A summary of our findings using a stepwise model building approach from the minimally adjusted to the fully adjusted model are presented in Supplementary Table S4.

Table 3.

Association of ambient air pollutants and CRC, by race, in the California MEC Study, 1993–2018

Black Japanese American Latino White Pheterogeneity by race and ethnicity
Cases HR (95% CI)a Cases HR (95% CI)a Cases HR (95% CI)a Cases HR (95% CI)a
PM2.5 1,102 0.96 (0.77–1.19) 487 1.19 (0.69–2.06) 1,201 1.47 (1.03–2.08) 422 1.31 (0.85–2.03) 0.31
PM10 1,087 0.98 (0.87–1.10) 483 1.16 (0.85–1.58) 1,196 1.15 (0.96–1.39) 416 0.87 (0.67–1.13) 0.38
NO2 1,087 0.98 (0.78–1.24) 483 1.52 (0.94–2.44) 1,196 1.42 (1.04–1.93) 416 0.90 (0.61–1.32) 0.17
NOx 1,031 1.03 (0.85–1.25) 483 1.41 (0.90–2.22) 1,196 1.10 (0.84–1.43) 414 0.94 (0.65–1.38) 0.73
CO 1,087 0.99 (0.75–1.32) 483 1.83 (0.89–3.77) 1,196 1.42 (0.92–2.20) 416 0.84 (0.46–1.53) 0.34
O3 1,087 1.12 (0.94–1.32) 483 0.73 (0.48–1.12) 1,196 1.01 (0.82–1.25) 416 0.96 (0.70–1.32) 0.48
Open in a new tab

Multivariable models adjusting for sex, education, nSES, body mass index, alcohol use, smoking status, family history of colon cancer, multivitamin use, NSAID use, physical activity, energy intake, processed red meat consumption, and menopausal hormone use (among female participants). All models include age at baseline as a strata variable. Models among the overall population additionally include race and ethnicity as a strata variable.

Abbreviations: CI, confidence interval; CO, carbon monoxide; CRC, colorectal cancer; HR, hazard ratio; MEC, Multiethnic Cohort; NO2, nitrogen dioxide; NOx, nitrogen oxide; NSAID, non-steroidal anti-inflammatory drug; nSES, neighborhood socioeconomic status; O3, ozone; PM, particulate matter; ppb, parts per billion.

a

Colorectal cancer risk per 10 μg/m3 satellite-based PM2.5, per 10 μg/m3 krigged PM10, per 20 ppb of krigged NO2, per 50 ppb of krigged NOx, per 1000 ppb of krigged CO, and per 10 ppb of krigged O3.

3.4. Associations by tumor characteristics

We evaluated the associations of ambient air pollutant exposure by CRC stage at diagnosis and did not observe heterogeneity in associations by stage (Supplementary Table S5).

In subgroup analyses by tumor anatomical subsite (Table 4), there was no evidence of associations between air pollutant exposures and right colon cancer (HRs ranged from 0.94 to 1.03; all 95% CIs included the null). Increased HRs for the associations of PM and gaseous pollutants (except for O3) with left colon cancer and rectal cancer were observed although none of the 95% CI excluded the null. However, exposure to PM2.5 was associated with a statistically significantly increased risk of left colon and rectal cancers combined (HR 1.36, 95% 1.05–1.76), NO2 (HR 1.32, 95% 1.05–1.68) and CO (per 1000 ppb; HR 1.36; 95% CI 1.01–1.84), but the findings did not statistically significantly differ from estimates for right colon cancers (respective Pheterogeneity were 0.08, 0.06, and 0.13). Among female participants, risk of left colon and rectal cancers combined increased in association with exposure to PM2.5 (HR 1.63, 95% CI 1.11–2.39), NOx (HR 1.41, 95% CI 1.04–1.93), and CO (HR 1.73 95% CI 1.10–2.73) but there was no evidence of statistically significant differences by sex (Pheterogeneity=0.18, 0.11, and 0.16, respectively).

Table 4.

Subgroup analysis for the association of ambient air pollutants and CRC by tumor subsite (right, left, rectum), overall and by sex, in the California MEC Study, 1993–2018

Total population Male Participants Female Participants Pheterogeneity by sex
Cases HR (95% CI)a Cases HR (95% CI)a Cases HR (95% CI)a
PM2.5
 Right colon 1493 0.99 (0.80–1.25) 624 0.81 (0.58–1.15) 869 1.16 (0.87–1.56) 0.13
 Left colon 847 1.36 (0.96–1.91) 428 1.19 (0.76–1.89) 419 1.53 (0.92–2.56) 0.47
 Rectum 751 1.37 (0.93–2.02) 428 1.12 (0.67–1.89) 323 1.76 (0.98–3.14) 0.25
 Pheterogeneityb 0.20 0.36 0.38
 Left colon + Rectum 1598 1.36 (1.05–1.76) 856 1.15 (0.82–1.63) 742 1.63 (1.11–2.39) 0.18
 Pheterogeneityc 0.08 0.16 0.18
PM10
 Right colon 1482 0.94 (0.84–1.06) 621 0.87 (0.72–1.05) 861 1.01 (0.86–1.18) 0.23
 Left colon 837 1.10 (0.92–1.30) 423 1.10 (0.86–1.42) 414 1.08 (0.86–1.35) 0.89
 Rectum 744 1.16 (0.95–1.41) 424 1.15 (0.87–1.53) 320 1.16 (0.89–1.52) 0.95
 Pheterogeneityb 0.14 0.16 0.65
 Left colon + Rectum 1581 1.12 (0.99–1.28) 847 1.13 (0.93–1.36) 734 1.11 (0.94–1.33) 0.94
 Pheterogeneityc 0.05 0.06 0.41
NO2
 Right colon 1482 0.97 (0.78–1.22) 621 0.90 (0.64–1.28) 861 1.06 (0.78–1.43) 0.49
 Left colon 837 1.28 (0.93–1.76) 423 1.23 (0.81–1.89) 414 1.35 (0.84–2.18) 0.78
 Rectum 744 1.39 (0.98–1.99) 424 1.46 (0.88–2.41) 320 1.30 (0.79–2.15) 0.75
 Pheterogeneityb 0.17 0.26 0.62
 Left colon + Rectum 1581 1.32 (1.05–1.68) 847 1.32 (0.96–1.83) 734 1.32 (0.94–1.87) 0.99
 Pheterogeneityc 0.06 0.11 0.33
NOx
 Right colon 1444 0.99 (0.82–1.21) 608 0.86 (0.63–1.16) 836 1.13 (0.87–1.47) 0.17
 Left colon 826 1.22 (0.93–1.61) 417 1.10 (0.77–1.57) 409 1.41 (0.91–2.18) 0.38
 Rectum 737 1.13 (0.85–1.52) 423 0.93 (0.62–1.38) 314 1.43 (0.93–2.21) 0.16
 Pheterogeneityb 0.46 0.57 0.54
 Left colon + Rectum 1563 1.18 (0.97–1.44) 840 1.01 (0.78–1.32) 723 1.41 (1.04–1.93) 0.11
 Pheterogeneityc 0.23 0.42 0.28
CO
 Right colon 1482 0.98 (0.73–1.32) 621 0.87 (0.54–1.40) 861 1.09 (0.74–1.61) 0.47
 Left colon 837 1.31 (0.86–1.99) 423 1.08 (0.62–1.88) 414 1.67 (0.87–3.20) 0.30
 Rectum 744 1.44 (0.94–2.21) 424 1.17 (0.64–2.11) 320 1.84 (0.99–3.40) 0.34
 Pheterogeneityb 0.28 0.72 0.28
 Left colon + Rectum 1581 1.36 (1.01–1.84) 847 1.11 (0.74–1.66) 734 1.73 (1.10–2.73) 0.16
 Pheterogeneityc 0.13 0.44 0.13
O3
 Right colon 1482 1.03 (0.86–1.22) 621 1.00 (0.77–1.32) 861 1.03 (0.83–1.29) 0.87
 Left colon 837 1.01 (0.79–1.29) 423 1.10 (0.80–1.53) 414 0.89 (0.60–1.31) 0.36
 Rectum 744 0.91 (0.71–1.17) 424 1.01 (0.73–1.41) 320 0.80 (0.54–1.18) 0.37
 Pheterogeneityb 0.74 0.89 0.49
 Left colon + Rectum 1581 0.96 (0.81–1.15) 847 1.07 (0.85–1.35) 734 0.84 (0.64–1.11) 0.19
 Pheterogeneityc 0.61 0.73 0.27
Open in a new tab

Multivariable models adjusting for sex (except in sex-stratified models), education, nSES, body mass index, alcohol use, smoking status, family history of colon cancer, multivitamin use, NSAID use, physical activity, energy intake, processed red meat consumption, and menopausal hormone use (among female participants only). All models include age at baseline and race and ethnicity as strata variables.

Abbreviations: CI, confidence interval; CO, carbon monoxide; CRC, colorectal cancer; HR, hazard ratio; MEC, Multiethnic Cohort; NO2, nitrogen dioxide; NOx, nitrogen oxide; NSAID, non-steroidal anti-inflammatory drug; nSES, neighborhood socioeconomic status; O3, ozone; PM, particulate matter; ppb, parts per billion.

a

Colorectal cancer risk per 10 μg/m3 satellite-based PM2.5, per 10 μg/m3 krigged PM10, per 20 ppb of krigged NO2, per 50 ppb of krigged NOx, per 1000 ppb of krigged CO, and per 10 ppb of krigged O3.

b

Test for differences in association by colorectal subsites right colon, left colon, and rectum.

c

Test for differences in association by colorectal subsites right colon and left colon + rectum (combined).

Subgroup analysis by tumor subsite and race and ethnicity revealed some notable differences (Supplementary Table S6). Among Latino participants, incidence of rectal cancer was statistically significantly associated with PM2.5 (HR 2.83, 95% CI 1.33–6.04), PM10 (HR 1.60, 95% CI 1.17–2.20), NO2 (HR 3.04, 95% CI 1.52–6.09), CO (HR 2.86, 95% CI 1.33–6.18) and O3 (HR 0.65, 95% CI 0.42–0.998); there was statistical evidence of heterogeneity by race and ethnicity for PM2.5 and NO2 (Pheterogeneity=0.04 and 0.03).

3.5. Evaluating potential effect modifiers of the PM2.5-CRC association

We observed that increasing PM2.5 exposure was associated with a statistically significantly higher incidence of CRC among never smokers (HR 1.33; 95% CI 1.01–1.73) (Table 5), nondrinkers of alcohol (HR 1.28; 95% CI 1.02–1.62), and current users of menopausal hormone therapy (HR 1.76; 95% CI 1.003–3.08). However, statistically significant heterogeneity was not observed across levels of these factors (Pheterogeneity>0.05). The association of PM2.5 exposure and CRC risk did not differ by history of a prior colonoscopy or sigmoidoscopy (P=0.96).

Table 5.

Minimally and fully adjusted models evaluating the association of PM2.5 with CRC by strata of potential effect modifiers in the California MEC Study, 1993–2018

Cases Model 1
HR (95% CI)a 		Model 4
HR (95% CI)a
Neighborhood SES
 Low Q1–Q3 2,162 1.20 (0.98–1.47) 1.07 (0.87–1.32)
 High Q4–Q5 1,028 1.32 (1.02–1.72) 1.20 (0.92–1.58)
 Pheterogeneity 0.58 0.51
Body Mass Index (kg/m3)
 <25 1,047 1.34 (1.02–1.75) 1.16 (0.87–1.56)
 25–<30 1,343 1.35 (1.06–1.71) 1.07 (0.82–1.38)
 ≥30 782 1.34 (0.99–1.81) 1.07 (0.77–1.49)
 Pheterogeneity 0.99 0.89
Smoking pack-years
 Never 1,306 1.57 (1.23–2.02) 1.33 (1.01–1.73)
 <20 pack years 1,292 1.24 (0.99–1.57) 1.02 (0.80–1.32)
 ≥20 pack-years 477 1.43 (0.98–2.10) 1.12 (0.74–1.69)
 Pheterogeneity 0.39 0.38
Alcohol intake (g/day)
 Nondrinker 1,663 1.49 (1.20–1.85) 1.28 (1.02–1.62)
 <12 958 1.25 (0.95–1.64) 0.99 (0.74–1.33)
 ≥12 596 1.32 (0.93–1.86) 1.00 (0.68–1.46)
 Pheterogeneity 0.58 0.32
Reported previous colonoscopy or sigmoidoscopy through follow-up
 No 1,172 1.39 (1.06–1.83) 1.13 (0.85–1.52)
 Yes 1,388 1.40 (1.11–1.75) 1.15 (0.90–1.46)
 Pheterogeneity 0.99 0.96
Menopausal hormone therapy use (among females)
 Never 911 1.43 (1.06–1.91) 1.27 (0.93–1.75)
 Past use 365 1.42 (0.92–2.17) 1.29 (0.82–2.04)
 Current use 317 2.12 (1.26–3.57) 1.76 (1.003–3.08)
 Pheterogeneity 0.40 0.60
Long-term oral contraceptive use (among females)
 No 1,320 1.41 (1.12–1.78) 1.22 (0.95–1.56)
 Yes 267 2.20 (1.15–4.22) 1.91 (0.94–3.89)
 Pheterogeneity 0.21 0.24
Open in a new tab

Model 1 is adjusting for sex. Model 4 is adjusting for sex, nSES, body mass index, alcohol use, smoking status, family history of colon cancer, multivitamin use, NSAID use, physical activity, energy intake, processed red meat consumption, and menopausal hormone use (among female participants). All models include age at baseline and race and ethnicity as strata variables. Abbreviations: CI, confidence interval; CRC, colorectal cancer; HR, hazard ratio; MEC, Multiethnic Cohort; NSAID, non-steroidal anti-inflammatory drug; nSES, neighborhood socioeconomic status; PM, particulate matter.

a

Colorectal cancer risk per 10 μg/m3 satellite-based PM2.5.

4. Discussion

In this prospective study of ambient air pollutant exposure and CRC incidence, increasing PM2.5 and NOx exposures were associated with an increased incidence of CRC among female participants. We observed suggestive differences in associations by tumor subsite as increasing PM2.5, NO2, and CO exposures were associated with an increased incidence of left colon and rectal cancers combined, but not for right colon cancers. The increased CRC incidence with increasing PM2.5 and NO2 exposure were most prominent among Latino participants in the MEC.

Increasing PM2.5 exposure was associated with elevated CRC incidence in published studies with some differences (Chu et al., 2021; Coleman et al., 2020; Jiang et al., 2024). Although air pollution is a complex mixture including PM, gaseous pollutants, and metals, the previous SEER-based and PLCO studies only investigated associations with PM2.5 (Chu et al., 2021; Coleman et al., 2020) while the UK Biobank study additionally investigated PM10, NO2, NOx, and a combined air pollution exposure score (Jiang et al., 2024). The PM2.5–CRC associations by sex or tumor anatomical location were investigated only in the UK Biobank. Their results showed that PM2.5 exposure was associated with CRC risk overall (per 5-μg/m3; HR 1.09; 95% CI 0.95–1.24), and a stronger association among female (HR 1.35; 95% CI 0.91–2.01) than male participants (HR;1.04, 95% CI 0.73–1.47). However, in their analyses by tumor subsite, no association was observed with rectal cancer (HR 0.83; 95% CI 0.48–1.42) (Jiang et al., 2024). In the PLCO study, equally strong and statistically significant associations were found between PM2.5–CRC incidence in men and women (Chu et al., 2021). All three studies were conducted in predominantly NH White populations (Chu et al., 2021; Coleman et al., 2020; Jiang et al., 2024) and were not able to evaluate racial and ethnic differences in CRC–air pollution associations. It is well documented that low SES neighborhoods often include a larger proportion of racial and ethnic minoritized groups (Hajat et al., 2015; Miranda et al., 2011; Woo et al., 2019) and these communities tend to have high levels of air pollution due to decades of racial and ethnic segregation (Nardone et al., 2020; Woo et al., 2019). As in the UK Biobank study, we observed a statistically significant association of PM2.5 exposure with CRC risk among never smokers (UK Biobank HR 1.49; 95% CI 1.00–2.20; MEC HR 1.33; 95% CI 1.01–1.73)(Jiang et al., 2024). Although the HRs in never smokers were not statistically significantly different from the HR estimates among smokers in both studies, our finding adds to the evidence that PM2.5 exposure is associated with an increased risk of CRC when residual confounding due to smoking exposure is eliminated or lessened. Our findings of differences in CRC incidence according to air pollutants by tumor subsite and race and ethnicity highlight the importance of evaluating the role of air pollutants within these subgroups.

Air pollution may affect cancer risk through the development of PM-induced DNA adducts, leading to oxidative stress and inflammation that promote tumor growth (Pritchett et al., 2022; Turner et al., 2020). Also, there is supportive evidence for the biological plausibility of exposure to PM2.5 and NO2 and associations with CRC incidence as PM mixtures of metals and polycyclic aromatic hydrocarbons can lead to increases in inflammation and oxidative stress, causing DNA mutations and methylation changes (Cheng et al., 2024). In the UK Biobank Mendelian randomization analysis, genetically predicted methylation at PM2.5-related and NO2-related CpG sites were associated with increased risk of CRC (Jiang et al., 2024). The methylation markers at protein-coding genes TMBIM1/PNKD, CXCR5, and TMEM110, have been associated with increasing oxidative stress, increasing inflammation, and may promote tumor growth and proliferation via the PI3K/AKT pathway (Jiang et al., 2024). Another Mendelian randomization study of genome-wide data of ambient air pollution and CRC incidence in the UK Biobank found a statistically significant association between methylation-marker predicted NO2 exposure and CRC risk (Xu et al., 2025).

Air pollution-CRC associations were stronger among female than male participants in the MEC, and among current menopausal hormone users and long-term oral contraceptive users, suggesting that a hormonal pathway may be implicated. Previous studies of air pollution exposure and risk of pancreatic (Bogumil et al., 2021) and lung cancer (Cheng et al., 2022a) among California MEC participants also tended to find stronger HR estimates in female participants. Decreased estrogen receptor β (ERβ) expression has been cited as a potential driver of sex-related differences in CRC as reduced expression has been associated with increased odds of tumor progression, advanced stage at diagnosis and poor survival (Liang et al., 2026; Niv, 2015). In a lung cancer mouse model study, exposure to PM2.5 organic extracts was associated with an increase in ERβ and interleukin-6 expression through the activation of MAPK/ERK and PI3K/AKT pathways in bronchial epithelial cells, promoting inflammation and carcinogenesis (Luo et al., 2021). In another rodent experiment, there were sex-dependent responses (e.g., expression of ERβ) to concentrated ambient particles in the nasal epithelium, in support of higher risks of females exposed to urban PM exposures (Yoshizaki et al., 2016). In addition, the associations of PM2.5 organic extract exposure with inflammatory cytokines in female mice compared to male mice exhibited higher levels of tumor necrosis factor-α, interleukin-5, and growth-related oncogene/keratinocyte chemoattractant in the bronchoalveolar lavage fluid via ERβ activation (Guo et al., 2022). While ERβ activation is typically associated with anti-inflammatory and anti-carcinogenic processes in the colon (Das et al., 2023), these findings suggest that pollutant-induced activation of ERβ may increase tissue inflammation and carcinogenesis to a greater extent in females.

Our findings of stronger air pollution-CRC associations among Latino participants in the MEC are intriguing. In two other studies of air pollution and cancer incidence in the MEC, risk associations were more pronounced among Latino participants (Bogumil et al., 2021; Cheng et al., 2022a). Specifically, exposures to PM2.5, PM10, and NOx were statistically significantly associated with and increased incidence of pancreatic cancer among Latino MEC participants (respective HRs were 3.59, 1.45, 1.72), while these increased HRs were lower (ranged from 1.00 to 1.49) in African American and White participants (Bogumil et al., 2021). In our study of PM and gaseous air pollutants (except for O3) and lung cancer risk, the HR estimates were also higher in Latino than African American and White participants (Cheng et al., 2022a). Although Latino participants represent the largest ethnic group among California MEC participants, the larger size is unlikely the only explanation as the incidence of all three cancers were higher among African American participants. In areas with higher proportions of racially and ethnically minoritized populations, characteristics of the built environment may increase co-exposure to other environmental stressors (e.g., unmeasured air toxics) that could potentially amplify air pollution-related health effects. However, we may not have been able to detect a significant association of air pollution and CRC incidence among African American participants due, in part, to residual confounding, such as the proportion of ever and current smokers was highest among this racial and ethnic group.

Our study has several strengths. First, we examined a large prospective study with extensive residential histories over the follow-up period, allowing for the evaluation of long-term air pollution exposure. We evaluated six common PM and gaseous pollutants in urban environments. Given our large racially and ethnically diverse population, we were able to assess racial and ethnic-specific effect estimates. The study was also strengthened by the availability of detailed, individual-level risk factor questionnaire data and cancer incidence ascertained through linkage with population-based cancer registries with virtually complete case-ascertainment. We evaluated potential differences in cancer etiology with analyses by sex and tumor subsite. However, we note some limitations of the study. We adjusted for risk factors that were assessed at baseline but some behaviors may have changed over time. Air pollution is a heterogenous mixture with highly correlated components and common sources, making it difficult to identify the effects of individual pollutants or components. As such, the inverse association between O3 and CRC risk may be due to the negative correlation between O3 and NOx concentrations caused by the photochemical reaction between O3 and nitric oxide (Sillman et al., 1997). Empirical Bayesian kriging interpolation may not fully capture fine-scale spatial variation, particularly for NO2, which declines steeply near roadways. Publicly available modeled surfaces can better resolve these gradients but were not available as monthly, time-varying products covering all study pollutants across our full follow-up period. Any resulting misclassification would likely be non-differential with respect to CRC status and would therefore attenuate associations toward the null, indicating our estimates are conservative. In addition, we do not have information on pollutant exposures at work, in transit, indoor exposures, or across lifetime residential histories (before study enrollment). Given our reliance on residential exposures, this may have contributed to an information bias that may underestimate the true magnitude of the association due to the absence of non-residential exposure data. We did not formally adjust for multiple comparisons and recognize that there is a possibility of chance findings; however, we apply the Goldberg and Silbergerld framework for multiple comparisons (Goldberg and Silbergeld, 2011) as these analyses were hypothesis-driven and many of the pollutant exposures were correlated. These findings may not be generalizable to populations outside of California. Examining the role of ambient air pollutants using mixture modeling would be a natural next step.

5. Conclusion

Our study adds to the growing evidence of the deleterious effects of ambient air pollution on CRC. Ambient air pollution may be a modifiable environmental risk factor for CRC incidence that can be addressed with regional policies and health education to promote CRC prevention. Our comprehensive assessment of differences by race and ethnicity, sex, and CRC subsite add to the growing recognition that further understanding of the etiology of CRC will require better understanding of exposures and risk patterns by location within the large bowel. Additional studies are needed to evaluate the biological mechanisms by which ambient air pollutant exposure may contribute to colorectal carcinogenesis through inflammatory pathways, especially among females.

Supplementary Material

1

Highlights.

  • PM2.5 and NOx were linked to higher colorectal cancer risk in female participants

  • PM2.5 and NO2 were linked to higher colorectal cancer risk in Latino participants

  • PM2.5, NO2 and CO were associated with left colon and rectal cancers combined

  • Associations of PM2.5 and NO2 with rectal cancer differed by race and ethnicity

Acknowledgments

We are grateful to all the Multiethnic Cohort Study participants and the study teams at the University of Hawai‘i, University of Southern California, and University of California, San Francisco. We thank Dr. David Bogumil for his technical support.

Funding

This work was supported by the National Cancer Institute of the National Institutes of Health grants U01 CA164973, U01 CA164973-11S1, T32CA229110, 5-R21-CA094723-03, and P01CA138338; National Institute Environmental Health Science grant R01ES026171; National Institute of Environmental Health Sciences Environmental Exposures, Host Factors, and Human Disease grant P30 ES0070480; and California Air Resource Board contract 04-323.

Footnotes

Conflicts of Interest: The authors declare they have no conflicts of interest related to this work to disclose.

Declaration of interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

CRediT authorship contribution statement

Ugonna Ihenacho: Formal analysis, Funding acquisition, Investigation, Visualization, Writing – original draft, Writing – review & editing. Chiuchen Tseng: Data curation, Formal analysis, Investigation, Writing – review & editing. Jun Wu: Data curation, Investigation, Writing – review & editing. Scott Fruin: Investigation, Writing – review & editing. Timothy V. Larson: Investigation, Writing – review & editing. Salma Shariff-Marco: Investigation, Writing – review & editing. Loïc Le Marchand: Funding acquisition, Investigation, Project administration, Resources, Supervision, Writing – review & editing. Daniel O. Stram: Investigation, Writing – review & editing. Brian Z. Huang: Investigation, Writing – review & editing. Lynne R. Wilkens: Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review & editing. Christopher A. Haiman: Funding acquisition, Investigation, Project administration, Resources, Supervision, Writing – review & editing. Beate Ritz: Investigation, Writing – review & editing. Iona Cheng: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Visualization, Writing – review & editing. Anna H. Wu: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Visualization, Writing – review & editing.

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.

Data Statement

The Multiethnic Cohort investigators and institutions affirm their intention to share the research data consistent with all relevant NIH resource/data sharing policies. Data requests should be submitted through Multiethnic Cohort online data request system at https://www.uhcancercenter.org/for-researchers/mec-data-sharing.

References

  1. Bogumil D, et al. 2021. The association between ambient air pollutants and pancreatic cancer in the Multiethnic Cohort Study. Environ Res. 202, 111608. 10.1016/j.envres.2021.111608. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Carethers JM, 2021. Racial and ethnic disparities in colorectal cancer incidence and mortality. Adv Cancer Res. 151, 197–229. 10.1016/bs.acr.2021.02.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Cheng HJ, et al. 2024. Involvement of autophagy and gut dysbiosis in ambient particulate matter-induced colonic inflammation. Ecotoxicol Environ Saf. 286, 117171. 10.1016/j.ecoenv.2024.117171. [DOI] [PubMed] [Google Scholar]
  4. Cheng I, et al. 2020. Association between ambient air pollution and breast cancer risk: The multiethnic cohort study. Int J Cancer. 146, 699–711. 10.1002/ijc.32308. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Cheng I, et al. 2022a. Traffic-Related Air Pollution and Lung Cancer Incidence: The California Multiethnic Cohort Study. Am J Respir Crit Care Med. 10.1164/rccm.202107-1770OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Cheng I, et al. 2022b. Traffic-related Air Pollution and Lung Cancer Incidence: The California Multiethnic Cohort Study. Am J Respir Crit Care Med. 206, 1008–1018. 10.1164/rccm.202107-1770OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Chu H, et al. 2021. A prospective study of the associations among fine particulate matter, genetic variants, and the risk of colorectal cancer. Environ Int. 147, 106309. 10.1016/j.envint.2020.106309. [DOI] [PubMed] [Google Scholar]
  8. Coleman NC, et al. 2020. Fine Particulate Matter Exposure and Cancer Incidence: Analysis of SEER Cancer Registry Data from 1992–2016. Environ Health Perspect. 128, 107004. 10.1289/EHP7246. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Conroy SM, et al. 2018. Characterizing the neighborhood obesogenic environment in the Multiethnic Cohort: a multi-level infrastructure for cancer health disparities research. Cancer Causes Control. 29, 167–183. 10.1007/s10552-017-0980-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Das PK, et al. 2023. Implications of estrogen and its receptors in colorectal carcinoma. Cancer Med. 12, 4367–4379. 10.1002/cam4.5242. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Edwards BK, et al. 2010. Annual report to the nation on the status of cancer, 1975–2006, featuring colorectal cancer trends and impact of interventions (risk factors, screening, and treatment) to reduce future rates. Cancer. 116, 544–73. 10.1002/cncr.24760. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Godoy-Vitorino F, et al. 2026. Environmental exposures, microbiome dynamics and chronic disease risk in climate-vulnerable regions: interdisciplinary perspectives from Puerto Rico. Front Public Health. 14, 1807779. 10.3389/fpubh.2026.1807779. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Goldberg M, Silbergeld E, 2011. On multiple comparisons and on the design and interpretation of epidemiological studies of many associations. Environ Res. 111, 1007–9. 10.1016/j.envres.2011.08.010. [DOI] [PubMed] [Google Scholar]
  14. Guo H, et al. 2022. The role of estrogen receptor beta in fine particulate matter (PM(2.5)) organic extract-induced pulmonary inflammation in female and male mice. Environ Sci Pollut Res Int. 29, 60922–60932. 10.1007/s11356-022-20055-x. [DOI] [PubMed] [Google Scholar]
  15. Hajat A, et al. 2015. Socioeconomic Disparities and Air Pollution Exposure: a Global Review. Curr Environ Health Rep. 2, 440–50. 10.1007/s40572-015-0069-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Hamra GB, et al. 2014. Outdoor particulate matter exposure and lung cancer: a systematic review and meta-analysis. Environ Health Perspect. 122, 906–11. 10.1289/ehp/1408092. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Harmon BE, et al. 2014. Ethnic differences and predictors of colonoscopy, prostate-specific antigen, and mammography screening participation in the multiethnic cohort. Cancer Epidemiol. 38, 162–7. 10.1016/j.canep.2014.02.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. He X, et al. 2026. Short-term exposure to ambient air pollution, gut microbiota, and pro-atherosclerotic responses in healthy adults: A prospective panel study. Environ Res. 301, 124532. 10.1016/j.envres.2026.124532. [DOI] [PubMed] [Google Scholar]
  19. Islami F, et al. 2018. Proportion and number of cancer cases and deaths attributable to potentially modifiable risk factors in the United States. CA Cancer J Clin. 68, 31–54. 10.3322/caac.21440. [DOI] [PubMed] [Google Scholar]
  20. Jiang F, et al. 2024. Impact of ambient air pollution on colorectal cancer risk and survival: insights from a prospective cohort and epigenetic Mendelian randomization study. EBioMedicine. 103, 105126. 10.1016/j.ebiom.2024.105126. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Kish L, et al. 2013. Environmental particulate matter induces murine intestinal inflammatory responses and alters the gut microbiome. PLoS One. 8, e62220. 10.1371/journal.pone.0062220. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Kolonel LN, et al. 2000. A multiethnic cohort in Hawaii and Los Angeles: baseline characteristics. Am J Epidemiol. 151, 346–57. 10.1093/oxfordjournals.aje.a010213. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Le Marchand L, et al. 1997. Associations of sedentary lifestyle, obesity, smoking, alcohol use, and diabetes with the risk of colorectal cancer. Cancer Res. 57, 4787–94. [PubMed] [Google Scholar]
  24. Liang B, et al. 2026. Sex as a Contextual Modifier in Colorectal Cancer: Integrating Tumor Sidedness, Molecular Subtype, Immune Ecology, and Early-Onset Disease. Cancers (Basel). 18. 10.3390/cancers18081309. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Lunn M, McNeil D, 1995. Applying Cox regression to competing risks. Biometrics. 51, 524–32. [PubMed] [Google Scholar]
  26. Luo F, et al. 2021. PM2.5 organic extract mediates inflammation through the ERbeta pathway to contribute to lung carcinogenesis in vitro and vivo. Chemosphere. 263, 127867. 10.1016/j.chemosphere.2020.127867. [DOI] [PubMed] [Google Scholar]
  27. Meng J, et al. 2019. Estimated Long-Term (1981–2016) Concentrations of Ambient Fine Particulate Matter across North America from Chemical Transport Modeling, Satellite Remote Sensing, and Ground-Based Measurements. Environ Sci Technol. 53, 5071–5079. 10.1021/acs.est.8b06875. [DOI] [PubMed] [Google Scholar]
  28. Miranda ML, et al. 2011. Making the environmental justice grade: the relative burden of air pollution exposure in the United States. Int J Environ Res Public Health. 8, 1755–71. 10.3390/ijerph8061755. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Mutlu EA, et al. 2018. Inhalational exposure to particulate matter air pollution alters the composition of the gut microbiome. Environ Pollut. 240, 817–830. 10.1016/j.envpol.2018.04.130. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Nardone A, et al. 2020. Associations between historical residential redlining and current age-adjusted rates of emergency department visits due to asthma across eight cities in California: an ecological study. Lancet Planet Health. 4, e24–e31. 10.1016/S2542-5196(19)30241-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Niv Y, 2015. Estrogen receptor beta expression and colorectal cancer: a systematic review and meta-analysis. Eur J Gastroenterol Hepatol. 27, 1438–42. 10.1097/MEG.0000000000000471. [DOI] [PubMed] [Google Scholar]
  32. Ollberding NJ, et al. 2011. Racial/ethnic differences in colorectal cancer risk: the multiethnic cohort study. Int J Cancer. 129, 1899–906. 10.1002/ijc.25822. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Polite BN, et al. 2017. Charting the Future of Cancer Health Disparities Research: A Position Statement from the American Association for Cancer Research, the American Cancer Society, the American Society of Clinical Oncology, and the National Cancer Institute. Cancer Res. 77, 4548–4555. 10.1158/0008-5472.CAN-17-0623. [DOI] [PubMed] [Google Scholar]
  34. Pritchett N, et al. 2022. Exposure to Outdoor Particulate Matter Air Pollution and Risk of Gastrointestinal Cancers in Adults: A Systematic Review and Meta-Analysis of Epidemiologic Evidence. Environ Health Perspect. 130, 36001. 10.1289/EHP9620. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Siegel RL, et al. 2025. Cancer statistics, 2025. CA Cancer J Clin. 75, 10–45. 10.3322/caac.21871. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Sillman S, et al. 1997. The Use of Photochemical Indicators to Evaluate Ozone-NO(x)-Hydrocarbon Sensitivity: Case Studies from Atlanta, New York, and Los Angeles. J Air Waste Manag Assoc. 47, 1030–1040. 10.1080/10962247.1997.11877500. [DOI] [PubMed] [Google Scholar]
  37. Sninsky JA, et al. 2022. Risk Factors for Colorectal Polyps and Cancer. Gastrointest Endosc Clin N Am. 32, 195–213. 10.1016/j.giec.2021.12.008. [DOI] [PubMed] [Google Scholar]
  38. Speizer FE, 1983. Assessment of the epidemiological data relating lung cancer to air pollution. Environ Health Perspect. 47, 33–42. 10.1289/ehp.834733. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Turner MC, et al. 2020. Outdoor air pollution and cancer: An overview of the current evidence and public health recommendations. CA Cancer J Clin. 10.3322/caac.21632. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Vogtmann E, et al. 2025. The human oral microbiome and risk of colorectal cancer within three prospective cohort studies in the United States. Cancer. 131, e35802. 10.1002/cncr.35802. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Wang L, et al. 2020. Risk Factor Profiles Differ for Cancers of Different Regions of the Colorectum. Gastroenterology. 159, 241–256 e13. 10.1053/j.gastro.2020.03.054. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Woo B, et al. 2019. Residential Segregation and Racial/Ethnic Disparities in Ambient Air Pollution. Race Soc Probl. 11, 60–67. 10.1007/s12552-018-9254-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Wu AH, et al. 2024. Air Pollution and Breast Cancer Incidence in the Multiethnic Cohort Study. J Clin Oncol. JCO2400418. 10.1200/JCO.24.00418. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Wu J, et al. 2016. Adverse Reproductive Health Outcomes and Exposure to Gaseous and Particulate-Matter Air Pollution in Pregnant Women. Res Rep Health Eff Inst. 2016, 1–58. [PMC free article] [PubMed] [Google Scholar]
  45. Xu X, et al. 2025. The association between ambient air pollution and colorectal cancer: a Mendelian randomization study. Int J Environ Health Res. 35, 495–505. 10.1080/09603123.2024.2361453. [DOI] [PubMed] [Google Scholar]
  46. Yang J, et al. 2014. Developing an area-based socioeconomic measure from American Community Survey data. Fremont, CA: Cancer Prevention Institute of California. [Google Scholar]
  47. Yoshizaki K, et al. 2016. The effects of urban particulate matter on the nasal epithelium by gender: An experimental study in mice. Environ Pollut. 213, 359–369. 10.1016/j.envpol.2016.02.044. [DOI] [PubMed] [Google Scholar]
  48. Yost K, et al. 2001. Socioeconomic status and breast cancer incidence in California for different race/ethnic groups. Cancer Causes Control. 12, 703–11. 10.1023/a:1011240019516. [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
NIHMS2200316-supplement-1.pdf (303.2KB, pdf)

Data Availability Statement

The Multiethnic Cohort investigators and institutions affirm their intention to share the research data consistent with all relevant NIH resource/data sharing policies. Data requests should be submitted through Multiethnic Cohort online data request system at https://www.uhcancercenter.org/for-researchers/mec-data-sharing.

RESOURCES