Global impact of air pollution on cancer: causal evidence and health inequities across regions from 1990 to 2021.

Introduction
According to the Global Burden of Disease (GBD) 2021 report, cancer is the second leading cause of death worldwide, with over 20 million new cases reported in 2022, a number expected to rise in the coming years [1, 2]. Additionally, research highlights a strong link between airborne particulate matter, nitrogen oxides, and the development of respiratory, cardiovascular, and cancer-related diseases, with nearly 20% of non-communicable diseases attributed to air pollution [3, 4]. In 2013, the International Agency for Research on Cancer (IARC) classified outdoor air pollution and fine particulate matter (PM2.5) as human carcinogens, underscoring their role in cancer risk [5].
The association between air pollution and cancer varies by cancer type. A study of 17 European cohorts found statistically significant links between PM2.5 and PM10 exposure and lung cancer risk, with smaller particles posing greater harm, whereas no such association was observed for nitrogen oxides [6]. Additionally, each 10 μg/m3 increase in PM2.5 concentration correlated with a 15–27% rise in lung cancer mortality [7]. The harmful effects of PM are largely attributed to its high surface area and strong adsorption capacity, enabling it to carry toxic substances, bacteria, and viruses into the smallest airways and alveolar tissues, contributing to lung cancer [8].
Recent research further underscores the link between air pollution and urogenital cancers, showing that each 5 µg per cubic meter (μg/m3) increase in particulate matter with an aerodynamic diameter of less than 2.5 µm (PM2.5) raises the risk of urologic, bladder, and kidney cancers by 6%, 7%, and 9%, respectively, while each 10 μg/m3 rise in nitrogen dioxide (NO₂) correlates with a 3%, 4%, and 4% greater risk of urologic, bladder, and prostate cancers [9]. Additionally, nitric oxide (NO) acts as a cofactor in human papillomavirus (HPV) infection in cervical cancer, while HPV itself promotes NO release. This bidirectional interaction further suppresses early messenger RNA (mRNA) expression, reduces retinoblastoma protein (pRb) and tumor protein p53 (p53) levels, lowers p53 activity, and decreases apoptotic indices in HPV-infected cervical cells, ultimately enhancing the survival of mutant cells and driving carcinogenesis [10]. Furthermore, a systematic epidemiological review confirmed that PM2.5 exposure is associated with an increased risk of gastrointestinal cancer, with particularly strong associations with liver and colorectal cancer [11].
However, the relationship between specific types of particulate matter (PM), nitrogen oxides (NOx), and various cancers remains unclear, with conflicting findings across studies. For example, a study in California showed a positive association between air pollution and breast cancer risk, particularly among African Americans and Japanese Americans [12], while a cohort study covering 56 U.S. metropolitan areas found no such association in black women [13]. These contradictions highlight the need for further research to better understand air pollution’s role in cancer risk. Researchers should develop new approaches to more effectively link specific pollutants to cancer outcomes.
Despite extensive research on the carcinogenic effects of air pollution, significant gaps remain in linking population-level exposure patterns to causal biological pathways. While epidemiological studies quantify disease burden and genome-wide association studies (GWAS) identify risk loci, few studies have integrated these approaches to determine whether population-level trends reflect genetically validated causal relationships [14, 15]. Our dual-method design directly addresses this gap by correlating: (i) global and regional burden estimates derived from the Global Burden of Disease (GBD) data with (ii) Mendelian randomization (MR)-validated pollutant-cancer associations, thereby enabling a simultaneous evaluation of both public health impacts and biological plausibility.
This study investigates the impact of air pollution on cancer burden using GBD 2021 data, analyzing trends in mortality and disability-adjusted life years (DALYs) across age groups, sexes, and socio-demographic indices from 1990 to 2021, alongside projections of future disease burden. Additionally, we applied Mendelian randomization, leveraging GWAS data to assess the causal relationship between specific air pollutants (PM2.5, PM2.5–10, PM10, NOx, and NO2) and cancers at 20 sites, minimizing the effects of reverse causation and confounding factors. Ultimately, this study aims to inform strategies for early cancer prevention and burden reduction by characterizing air pollution, disease burden, and causality.

Method

Study data of GBD
Our study utilized data from the Global Burden of Disease Study (GBD) 2021, which is available at http://ghdx.healthdata.org/gbd-results-tool. GBD 2021 provides a comprehensive assessment of 369 diseases and injuries and 87 risk factors across 204 countries, covering the period from 1990 to 2021.
The data for total cancer mortality was sourced from vital registration systems, surveillance data, and verbal autopsies. To enhance accuracy and address data gaps and misclassifications, this information was adjusted. The updated data were processed through the Cause of Death Ensemble model (CODEm), which generated estimates of total cancer mortality by location, year, age, and gender [16]. A comparative risk assessment was also conducted to identify major risk factors for total cancers. The population attributable fraction (PAF) was used to quantify the contribution of air pollution to the total cancer burden. The estimates for cancer mortality and disability-adjusted life years (DALYs) related to air pollution were derived by applying the PAFs to the respective mortality and DALYs data by location, year, age, and gender [17]. DALYs are a comprehensive measure of disease burden, combining years lost due to premature death (YLLs) and years lived with disability (YLDs). YLLs were calculated by multiplying cancer-related deaths in each age group by the expected residual life expectancy. YLDs were computed by multiplying the prevalence of low respiratory infections (LRIs) by the disability weights (DWs) reflecting their severity. The Socio-demographic Index (SDI), a composite index ranging from 0 (lowest) to 100 (highest), was derived based on the total fertility rate for individuals under 25 years old (TFU25), average educational attainment for those over 15 years old (EDU15 +), and adjusted per capita income. Using SDI, the 204 countries and regions were grouped into five categories: low SDI, low-middle SDI, middle SDI, high-middle SDI, and high SDI [18].

Statical analysis of GBD
Age-standardized rates (ASR) were applied to adjust mortality and DALY rates for differences in age distribution and demographic characteristics across countries. A linear model was fitted to the natural logarithm of the rate over time, expressed as y = α + βx + ε, where x represents the year, and y is the natural log of the rate. The estimated annual percentage change (EAPC) was calculated using the formula 100 * (e^β − 1), with a corresponding 95% confidence interval (95% CI). An increase in ASR was determined if both the EAPC and the lower bound of the 95% CI were positive, while a decrease was noted if the EAPC and the upper bound of the 95% CI were negative. No change in ASR was concluded if neither condition was met [19]. The correlation between ASR and the Socio-demographic Index (SDI) was assessed using Gaussian process regression with a Loess smoothing function and confirmed through Spearman rank order correlation tests.
The ARIMA model was further applied to examine the impact of air pollution on cancer trends and to project trends at the global, regional, and national levels from 2020 to 2050. ARIMA, which stands for autoregressive integrated moving average, combines autoregressive, moving average, and differencing components. In the ARIMA model (p, d, q), ‘AR’ refers to autoregression, with p representing the number of autoregressive terms; ‘MA’ indicates the moving average component, with q as the number of terms; and d represents the number of differencing steps to achieve stationarity [20]. Model selection was carried out using the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC).
A 95% uncertainty interval (UI) was computed for all estimates, and rates were expressed per 100,000 people. Data analysis, management, and visualization were performed using R software version 4.3.2.

Air pollution GWAS Summary data sources of MR
The air pollution exposure GWAS data were derived from the UK Biobank, a large-scale prospective cohort study consisting of 500,000 participants from the UK. Researchers used participants' geographic location data, integrating it with estimated annual air pollution levels obtained from the European Cohort Study on the Effects of Air Pollution (ESCAPE) (https://www.escapeproject.eu/) [21]. This integration enabled them to estimate air pollution concentrations using land-use regression (LUR) modeling. LUR models are a valuable tool for estimating air pollution exposure, offering advantages such as high spatial resolution and cost-effectiveness [22]. However, it is important to also acknowledge their limitations, particularly with regard to temporal resolution, reliance on proxy data, and challenges in generalizing the models to other regions or populations. The study focused on five primary air pollutants: PM2.5, PM10, PM2.5–10, NO2, and NOx. Corresponding data IDs for these pollutants were ukb-b-10817, ukb-b-18469, ukb-b-12936, ukb-b-9942, and ukb-b-12417 (Table 2). The GWAS data on smoking status were obtained from the Genomic Sequencing Consortium for Alcohol and Nicotine Use (GSCAN), which includes data from 607,291 participants [23].
The GWAS data on cancer outcomes used in this study were obtained from the FinnGen database (Version R12), a European-origin cohort specifically designed to minimize the risk of sample overlap [24]. The analysis included 17 cancer types across various sites, including: cancer, skin cancer, breast cancer, colorectal cancer, bronchus and lung cancer, bladder cancer, rectal cancer, urinary organ cancer, kidney cancer (excluding renal pelvis), ovarian cancer, cancers of the liver, bile duct, and gallbladder, lymphoid leukemia, brain cancer, small intestine cancer, liver cancer (hepatocellular carcinoma), stomach cancer, and esophageal cancer. Detailed information on ovarian cancer and other malignancies is presented in Table 2.

Selection of instrumental variables of MR
When applying the standard significance threshold (P < 5E-08), some air pollution exposures lacked a sufficient number of single-nucleotide polymorphisms (SNPs) to serve as instrumental variables (IVs) for MR analysis. To address this, thresholds of 5E-06 and 1E-05 were adopted in this study for SNP selection, ensuring a strong association between IVs and air pollution exposure. These two groups served as the discovery and replication datasets, respectively. To minimize the effects of linkage disequilibrium, IVs were revalidated using a 10,000 kb window and an R2 threshold of < 0.001 [25]. Furthermore, SNPs with an F-statistic < 10 or a palindromic structure were excluded to mitigate the influence of weak IVs and potential biases.

Mendelian randomisation and sensitivity analysis
In the MR analyses, the (IVW) method was employed as the primary approach for effect estimation, combining the effect sizes of individual SNP, representing the effect of exposure on outcome, weighted by the inverse of their standard errors, to provide a robust representation of the overall IV effect [26]. The Weighted Median (WM) and MR-Egger methods were used for validation and to adjust for horizontal pleiotropy, while the Maximum Likelihood Estimation method served as a complementary analytical approach. Furthermore, previous studies have established smoking as a common risk factor for cancer. Accordingly, this study applied multivariable Mendelian randomization (MVMR) to assess whether the association between air pollution and cancer incidence remains significant after adjusting for smoking [27].
The study employed Cullen's Q-test to evaluate heterogeneity and used MR-Egger intercept regression to identify potential horizontal pleiotropy [28]. Additionally, the MR-PRESSO method was applied to further assess pleiotropy and to test whether individual SNPs acted as outliers that significantly influenced the results. A threshold of P < 0.05 was used to determine statistical significance in this study.

Result

Spatiotemporal patterns of total cancers attributable to air pollution
Air pollution was responsible for approximately 0.37 million deaths and 8.93 million DALYs due to total cancers, resulting in an age-standardized mortality rate (ASMR) of 4.3407 (95% UI, 2.7394, 6.0404) and an age-standardized DALY rate (ASDR) of 102.0777 (95% UI, 64.8861, 141.6184) per 100,000 population. Over the past 30 years, the total burden from air pollution decreased significantly until 2020, followed by an increase in 2020–2021 (Table 1, Fig. 1).
In terms of socio-demographic index (SDI) regions, areas with high-middle SDI exhibited the highest total cancer burden linked to air pollution. In contrast, regions with high SDI showed substantial reductions in cancer burdens from air pollution. Across the SDI gradient, the burden declines markedly when referenced to the high-middle SDI stratum. In 2021, ASMR decreased from 6.22 (95% UI 3.95–8.83) in high-middle SDI to 5.66 (3.49–7.96; − 9%), 2.84 (1.82–3.83; − 54%), 2.36 (1.54–3.21; − 62%), and 2.23 (1.27–3.23; − 64%) in middle, low-middle, low, and high SDI, respectively. The corresponding ASDR decreased from 145.62 (92.82–205.83) to 129.71 (80.65–181.50; − 11%), 72.66 (46.67–98.15; − 50%), 58.84 (37.73–80.23; − 60%), and 49.61 (28.27–70.76; − 66%). Put differently, compared with high-middle SDI, ASMR and ASDR are lower by roughly 2.8–2.9-fold in high SDI and ~ 2.0–2.6-fold in low/low-middle SDI.
Regionally, China and countries in Africa had the highest cancer burdens attributable to air pollution, with the greatest ASMR and ASDR values. Similar patterns were evident in the EAPC of both ASDR and ASMR (Table 1, Table S1, Fig. 2).

Age and sex pattern
Figure 3 illustrates the age-specific global mortality and DALY rates for total cancers in 2021 and their changes from 1990 to 2021. The data revealed an inverted U-shaped distribution, with mortality rates rising for individuals under 45, peaking between 45 and 85 years, and then declining in those over 85. DALY rates peaked in the 70–74 age group. Throughout all age groups, males showed consistently higher rates than females due to air pollution, although the EAPC for mortality and DALY rates was relatively lower for males. Across different SDI regions, the gap in total cancer burden due to air pollution narrowed (Fig. 4).

Future trends
Predictions for ASMR and ASDR due to air pollution-related cancers in 2030 and 2050 are shown in Fig. 5. At the national level, the distribution is expected to remain similar in both 2030 and 2050. However, the anticipated burden of total cancers from air pollution will be notably higher in China and African countries than in other regions for both years.

MR analysis results of the discovery group
At a threshold of 5E-06, we identified 79, 83, 57, 23, and 29 SNPs associated with NOx, NO2, PM2.5, PM2.5–10, and PM10, respectively, as instrumental variables (IVs), all with F-statistics greater than 10 (Table S1). Using the Inverse-Variance Weighted (IVW) method, five significant causal relationships were identified: elevated levels of NOx (OR = 1.8492, 95% CI = 1.3491–2.5347, P = 0.0001), PM2.5 (OR = 1.5492, 95% CI = 1.0884–2.2062, P = 0.0151), and NO2 (OR = 3.8573, 95% CI = 1.2845–11.5834, P = 0.0161) were found to be associated with an increased risk of bronchus and lung cancer. Similarly, higher levels of NO2 (OR = 2.5158, 95% CI = 1.4526–4.3570, P = 0.0010) and PM10 (OR = 1.4458, 95% CI = 1.0306–2.0282, P = 0.0328) were linked to an increased risk of stomach cancer (Fig. 6 and Table S2). Furthermore, the directionality of results obtained using the WM, MR-Egger, and Maximum Likelihood Estimation methods was consistent with those from the IVW method, reinforcing the robustness of these findings.
In sensitivity analyses, the MR-Egger intercept test did not detect evidence of pleiotropy, and partial Coulomb's Q test produced P-values below 0.05, prompting the use of a random-effects model within the IVW framework. For the NO2–bronchus and lung cancer pair, the MR-PRESSO test identified rs10116277 as an outlier, and the causal relationship between the two became more significant after excluding this SNP (Table S3).

MR analysis results of the repeated group
At the more relaxed threshold of 1E-5, we identified 127, 118, 104, 41, and 62 IVs for NOx, NO2, PM2.5, PM2.5–10, and PM10, respectively (Table S4). The study results indicated that NOx (OR = 1.6106, 95% CI = 1.2435–2.0861, P = 0.0003), NO2 (OR = 1.3733, 95% CI = 1.0498–1.7965, P = 0.0207), and PM2.5 (OR = 1.6344, 95% CI = 1.2515–2.1344, P = 0.0003) maintained significant causal associations with bronchus and lung cancer. Similarly, a positive causal association was observed between NO2 (OR = 2.1295, 95% CI = 1.3152–3.4480, P = 0.0021) and stomach cancer; however, the causal relationship between PM10 and stomach cancer was not revalidated. The results of heterogeneity and pleiotropy tests, as shown in Table 2 and Table S5, align with previous findings, confirming that, after excluding the influence of rs10116277, the association between NO2 and bronchus and lung cancer became even more significant (Table S6).
Scatter plots (Figure S1) illustrating the causal effects of air pollutant exposure on cancer outcomes are presented for four methods: IVW, WM, MR-Egger, and Maximum Likelihood Estimation, all of which demonstrate consistent directionality. In the funnel plots (Figure S2), the SNPs appear approximately symmetrical around the IVW axis, and no outlier instrumental variables were detected in the leave-one-out (LLO) plots (Figure S3).
Subsequently, we performed MVMR analysis on the primary positive findings. At the 5E-06 threshold, a causal association was observed between NO2 and gastric cancer (OR = 1.8756, 95% CI = 1.1643–3.0214, P = 0.0097) as well as lung cancer (OR = 1.6802, 95% CI = 1.2250–2.2495, P = 0.0005). Consistent results were obtained at the 1E-05 threshold, where NO2 remained significantly associated with gastric cancer (OR = 2.1295, 95% CI = 1.3260–3.4197, P = 0.0018), and NOx showed a significant association with lung cancer (OR = 1.4156, 95% CI = 1.0417–1.9237, P = 0.0264) (Table S7).

Discussion
This study was the first to assess the impact of air pollution on global cancer burden from 1990 to 2021. It was found that air pollution contributed to approximately 0.37 million cancer deaths and 8.93 million DALYs annually. Among all SDI regions, the highest burden was observed in medium- and high-SDI regions, while nationally, China and several African countries suffered most. Males witnessed higher cancer burden, while mortality followed an inverted U-shaped distribution among all age groups, peaking at age 85. MR analyses assessed causal relationships between specific air pollutants and cancer at the genetic susceptibility level. The results confirmed a positive causal link between NOx, PM2.5, and NO2 with bronchial and lung cancers in both discovery and validation datasets. Additionally, PM10 and NO2 were identified as risk factors for stomach cancer, with the causal relationship between NO2 and gastric cancer remaining significant in the validation set. Notably, the association between NO2 and gastric cancer remained significant even after adjustment for smoking.
Lung cancer is a leading cause of morbidity and mortality, with increasing evidence implicating air pollution as a significant risk factor. This study established a causal link between genetic susceptibility to bronchial and lung cancers and exposure to PM₂.₅, NO₂, and NOx. Previous researches have indicated that the combined effects of rising air pollution and smoking had led to an approximately 30% increase in lung cancer mortality since 2007. In 2017, PM₂. ₅ pollution accounted for an estimated 265,267 lung cancer deaths (14.1%, UI: 9.8–18.7%) [29]. Another cohort study, which analyzed 352,053 lung cancer patients diagnosed in California between 1988 and 2009, found that exposure to PM₂. ₅ (HR = 1.38, 95% CI: 1.35–1.41) and NO₂ (HR = 1.30, 95% CI: 1.28–1.32) was associated with reduced survival, particularly among early-stage non-small-cell lung cancer patients, especially those with adenocarcinomas [30]. Besides, a meta-analysis of 14 studies from North America and Europe has demonstrated that every 10 μg/m3 increase in PM₂. ₅ concentration could lead to a 9% higher risk of lung cancer incidence or mortality (95% CI: 4–14%) [31]. Furthermore, extensive research highlighted the role of nitrogen oxides in lung cancer development. A meta-analysis encompassing 28 cohorts found that each 10 μg/m3 increase in NO₂ exposure raised the risk of lung cancer death by 1.04- to 1.05-fold [32]. Another meta-analysis including 20 studies reported a 4% increase in lung cancer risk per 10 μg/m3 NO₂ exposure (95% CI: 1%–8%), while a 10 μg/m3 rise in NOx levels was associated with a 3% increase in lung cancer risk (95% CI: 1%–5%) [33].
Inhalation of air pollutants primarily impacted the respiratory tract, airways, and alveoli, thus contributing to lung carcinogenesis through mechanisms such as genetic mutations, oxidative stress, and chronic inflammation. Studies have shown that mice exposed to air pollution would exhibit greater genetic variation at tandem repeat DNA loci, while the incidence of lung cancer-associated mutations in high-exposure populations was three times higher than in low-exposure areas [34, 35]. Specifically, low-dose PM2.5 exposure induced epigenetic silencing of TP53 in human alveolar epithelial cells, disrupting cell proliferation, apoptosis, and damage repair, thereby facilitating lung cancer development [36]. Additionally, miR-182, miR-185, and miR-802 were significantly downregulated in PM2.5-exposed human bronchial and A549 cells, potentially weakening oncogene suppression (SLC30A1, SERPINB2, AKR1C1) and promoting tumorigenesis [37, 38]. Nitrogen oxides were closely linked to the generation of reactive oxygen species (ROS) and reactive nitrogen species (RNS), with ROS-exposed mouse fibroblasts exhibiting increased proliferation, invasion, angiogenesis, and metastasis while resisting apoptosis [39, 40]. Finally, air pollutants would stimulate the release of pro-inflammatory cytokines (IL-6, TNF-α, GM-CSF), leading to sustained low-grade inflammation in the airways, which fosters the tumor microenvironment [41]. These findings were supported by extensive evidence from human, animal, cellular, and mechanistic studies. Consequently, in 2013, the International Agency for Research on Cancer (IARC) classified outdoor air pollution and particulate matter as Group 1 human carcinogens for lung cancer [5].
Epidemiologic data on the association between air pollution and cancers other than lung cancer remain limited. This study was the first to establish a positive causal relationship between PM10 and NO₂ and stomach cancer. Previous researches have primarily focused on PM2.5 in relation to gastric cancer, with no observed association between NO₂ and gastric cancer. The NHIS study reported a significant adverse association between PM2.5 and gastric cancer mortality (per 10 μg/m3: HR = 1.87, 95% CI 1.20–2.92, n = 525) [42]. In addition, a study analyzing long-term ambient air pollution exposure and gastric cancer incidence in 11 European cohorts found that the estimated risk associated with PM2.5 was higher in men (HR = 1.98,95% CI 1.30–3.01) compared with women (HR = 0.85,95% CI 0.5–1.45), possibly because men were more likely to be exposed to smoking, occupational exposures, and Helicobacter pylori infection [43].
Air pollution may contribute to gastric cancer through multiple mechanisms. Studies indicate that ultrafine carbon particles can translocate from the lungs to extrapulmonary organs, including the gastrointestinal tract, after whole-body inhalation exposure in rats [44]. Additionally, the gastrointestinal tract and lungs are interconnected via circulation and the lymphatic system, allowing mutual influence [45]. Chronic air pollution also alters the composition and diversity of the gut microbiota, impacting immune function and increasing cancer risk [46, 47]. These findings suggest that air pollution may promote gastric cancer through particulate matter translocation, systemic interactions, and gut microbiota disruption.
These findings highlight the significant impact of air pollution on cancer risk and survival, emphasizing the urgency of implementing stringent environmental policies and targeted public health interventions. The strong association between air pollution and bronchial and lung cancers reinforces the need for effective mitigation strategies to reduce exposure to PM₂.₅, NO₂, and NOx. Beyond lung cancer, this study provides novel evidence linking PM₁₀ and NO₂ to stomach cancer, underscoring the importance of further research into the effects of air pollution on other cancer types.
The inflection observed in 2020–2021 is best interpreted as a joint effect of the COVID-19 pandemic and ambient air pollution rather than a purely pollution-driven signal [48]. Long-term exposure to PM₂.₅/NO₂ can exacerbate systemic inflammation and cardiopulmonary comorbidity, plausibly heightening susceptibility to SARS-CoV-2 and worsening cancer care trajectories (treatment delays, stage migration, reduced therapy tolerance) [49–51].
A key strength of this study lies in its comprehensive and systematic approach, integrating GBD estimates with MR analyses to assess the relationship between specific air pollutants and cancer. The GBD approach quantifies the global cancer burden attributable to air pollution, providing robust evidence to support stronger pollution control measures. Meanwhile, MR analysis, leveraging genetic variation as an instrumental variable, establishes causal relationships while minimizing confounding and reverse causation, offering a more reliable assessment of air pollution’s impact on cancer risk. Given the disparities in air pollution exposure across different SDI regions, policy measures should be tailored accordingly. In high- and medium-SDI regions, where air pollution is a major contributor to cancer burden, stricter emissions regulations and pollution reduction initiatives are essential. In low-SDI regions, addressing environmental risk factors alongside broader health infrastructure improvements may be a more effective strategy for reducing cancer incidence. This study reinforces the need for region-specific, evidence-based policies to mitigate outdoor air pollution and improve public health outcomes worldwide.
This study has certain limitations. While it provides a global analysis of the cancer burden associated with air pollution, the causal assessment was restricted to individuals of European ancestry, necessitating further validation in Asian, African, and other populations to ensure broader applicability. A key constraint is the uneven quality and limited availability of regional data, especially from less developed regions (e.g., sub-Saharan Africa). Consequently, estimates that depend on mathematical models for disease burden in these settings carry considerable uncertainty. Besides, GBD’s comparative risk assessment uses relative-risk functions derived from long-term exposure studies, which implicitly account for latency in cancer development. Nonetheless, our annual attributable estimates align current-year exposure distributions with these long-term RRs rather than individual exposure histories; in settings with rapidly changing pollution, some timing mismatch may remain. Additionally, although the GBD framework and MR analysis accounted for confounding factors, air pollution comprises not only major pollutants such as PM and nitrogen oxides but also various organic and chemical substances that may have influenced the study's conclusions. Furthermore, the generalizability of the MR results may be limited due to regional differences in air pollution composition and pollutant toxicity. While this study provides valuable insights, further validation in regions with varying air pollution profiles and toxicity is necessary to ensure the broader applicability of the findings. Finally, while GBD and MR confirm the association and causal link between specific air pollutants and cancer burden, further cellular, animal, and clinical research is needed to elucidate the underlying biological mechanisms.

Conclusion
This study underscores the significant impact of air pollution on cancer risk, confirming the causal association of PM₂.₅, NO₂, and NOx with bronchial and lung cancers and identifying a new link between PM₁₀, NO₂, and gastric cancer. These findings highlight the necessity of stringent air pollution control to reduce cancer incidence and mortality, reinforcing the urgent need for targeted policies and interventions to protect public health.

Supplementary Information