Occupational asbestos exposure and lung cancer: contemporary research trends, exploratory genetic evidence, and global burden from 1990 to 2021

Abstract

Background

Asbestos exposure is a confirmed risk factor for lung cancer, but its causal relationship with different subtypes and its global disease burden remain poorly characterized. This study aims to comprehensively evaluate the association between occupational asbestos exposure and lung cancer through multidimensional analysis, integrating bibliometric analysis, mendelian randomization, and Global Burden of Disease data to identify research trends, explore potential subtype-specific associations, and characterize global disease burden patterns.

Methods

This study included a bibliometric analysis of publications from the Web of Science Core Collection on asbestos-related lung cancer. Mendelian randomization was performed using OpenGWAS data to investigate whether asbestos exposure is differentially associated with specific lung cancer subtypes. Global Burden of Disease 2021 data were analyzed to estimate the asbestos-related lung cancer burden by age and sex. Future burden was projected using the Bayesian age-period-cohort model.

Results

Bibliometrics revealed fluctuating publication trends, with a thematic shift toward molecular mechanisms. Mendelian randomization analysis identified a suggestive causal relationship between asbestos exposure and lung adenocarcinoma [odds ratio (OR) =4.901, 95% confidence interval (CI): 1.150–20.881, P=0.03], but no significant causal links were observed with overall lung cancer or other subtypes. None of the instrumental variables demonstrated horizontal pleiotropy or heterogeneity. Global Burden of Disease analysis showed a higher asbestos-related lung cancer burden in males and older populations, with the greatest attributable burden in high socio-demographic index regions. Projections indicate a continued decline in global burden by 2036.

Conclusions

The mendelian randomization results offer genetic evidence supporting a specific causal role of asbestos in lung adenocarcinoma development. Despite a decline in burden due to control policies, ongoing exposure in developing countries and legacy asbestos in buildings remain serious challenges. Future research should focus on larger genetic studies and analysis tailored to subtypes, while global efforts need to prioritize a comprehensive asbestos ban, improved screening for high-risk groups, and enhanced international cooperation to eliminate asbestos-related diseases.

Highlight box.

Key findings

• Mendelian randomization (MR) analysis identified a suggestive association between asbestos exposure and lung adenocarcinoma [odds ratio (OR) =4.901; P=0.03], while no causal links emerged for overall lung cancer or other subtypes.

• Research trends in the asbestos-related lung cancer have shifted from occupational exposure assessment toward molecular mechanisms and prevention strategies.

• The burden of asbestos-related lung cancer is higher in males, older populations, and in high-income regions, largely due to historical exposure and better diagnostic attribution.

• Projections suggest a 28.6% decline in male and 23.6% decline in female age-standardized death rates by 2036, consistent with the expected direction of post-ban cohort effects.

What is known and what is new?

• It is well established that exposure to asbestos is acknowledged as the primary cause of occupational lung cancer deaths.

• This study provides a novel and comprehensive insight into the research trends, causal relationships, and global disease burden patterns of asbestos-related lung cancer.

What is the implication, and what should change now?

• It is imperative to advocate for a complete asbestos ban.

• Policy should promote targeted lung cancer screening for people at high risk of asbestos exposure.

• International collaboration on asbestos exposure control and disease prevention should be strengthened.

Introduction

The initial research establishing a connection between asbestos and lung cancer was published in the 1930s (1). Over the subsequent decades, a series of landmark epidemiological investigations progressively solidified this association. These include the pioneering cohort study by Doll (2) among British textile workers, the prospective follow-up of insulation workers by Selikoff et al. (3), which first systematically quantified excess mortality from lung cancer and mesothelioma, the quantitative assessment of the synergistic effect between asbestos and smoking by Hammond et al. (4), and the dose-response meta-analyses conducted by Lash et al. (5). These foundational investigations, conducted before the widespread implementation of occupational exposure limits and national bans, remain essential for understanding the historical burden of asbestos-related lung cancer (ARLC).

Currently, exposure to asbestos is acknowledged as the primary cause of occupational lung cancer deaths, imposing a significant burden on the incidence and progression of lung cancer (6). Evidence indicates that the incidence of lung cancer attributable to asbestos is at least six times greater than that of mesothelioma (7). However, among asbestos-related cancers, mesothelioma tends to receive more attention than lung cancer because of its special association with asbestos (7,8). Moreover, as smoking is closely linked to lung cancer, many patients exposed to asbestos may be misclassified as smokers, seriously underestimating the role of asbestos (9). The use of asbestos has been limited for many years. In recent years, at least 67 countries have banned or heavily restricted the use of asbestos-containing products, thereby decreasing ongoing exposure to asbestos (10). However, certain countries still mine and utilize asbestos, and major importers and consumers of chrysotile asbestos remain (11).

Most studies on asbestos and lung cancer are retrospective cohort or case-control designs, which are prone to confounding factors such as smoking, an inherent limitation of observational studies (12-14). Furthermore, research on the carcinogenic mechanism of asbestos has mostly been conducted in cell or animal models, and differences in species limit the extrapolation of results to humans (15,16). Mendelian randomization (MR) uses genetic variation as an instrumental variable (IV) to simulate the lifetime effects of asbestos exposure, avoiding confounding factors and reverse causality, providing more robust evidence for causal inference (17). However, while MR strengthens causal inference, it does not directly capture the broader research landscape and global burden of ARLC.

Bibliometrics, a review method for evaluating the status and prospective developments of a research field, can systematically quantify research trends, identify knowledge gaps, and reveal multidisciplinary collaborative networks (18). Several bibliometric studies have been conducted on asbestos-related diseases (ARDs) (19-21). However, there has been no specific bibliometric analysis focused on ARLC so far. A comprehensive evaluation of the global burden and research trends of ARLC remains scarce (22-24). Conducting such assessments would equip policymakers and researchers with vital data to guide prevention strategies and future research efforts.

In this study, we systematically evaluated the association between occupational asbestos exposure and lung cancer from three complementary dimensions: research trends, causal evidence, and disease burden trajectories. To this end, we integrated bibliometric analysis, MR, and Global Burden of Disease (GBD) 2021 data to achieve the following objectives: (I) to characterize the evolution of research themes and identify contemporary hotspots related to ARLC through bibliometric analysis from 2008 to 2025, thereby providing contextual background for the subsequent MR-based causal analysis; (II) to examine potential subtype-specific associations between asbestos exposure and lung cancer subtypes using two-sample MR, with a particular focus on lung adenocarcinoma (LUAD), which was identified as the most frequent histological type among asbestos-exposed workers in early epidemiological studies (25); (III) to quantify ARLC burden by sex, age, and region during 1990–2021 using GBD 2021 data and project trends to 2036 via the Bayesian age-period-cohort (BAPC) model, thus translating causal findings into tangible indicators for policies such as asbestos bans, exposure limits, and high-risk screening. The overall study design is presented in Figure 1. This article is reported in accordance with the STROBE-MR reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-0975/rc).

Figure 1.

Flow diagram of the research. BAPC, Bayesian age-period-cohort; GBD, Global Burden of Disease Study; GWAS, genome-wide association study; IVW, inverse variance weighted; MR, mendelian randomization; MR-PRESSO, MR pleiotropy RESidual sum and outlier.

Methods

Bibliometric analysis

Data collection

All literature searches were performed on the Web of Science Core Collection from 2008 to 2025. Eligible publications included articles and review articles, excluding book chapters, corrections, editorial material, letters, meeting abstracts, etc. This search employed specific strings such as “asbestos” and “lung cancer”. The detailed search strategy is shown in Table S1. The search was conducted without language restrictions on August 13, 2025. After the initial search, two reviewers screened titles and abstracts to exclude literature unrelated to ARLC. Any disputes encountered during this process were resolved through consultation and discussion.

Data analysis

Subsequent bibliometric analysis utilized R (version 4.5.1) and VOSviewer (version 1.6.20). The bibliometric R-package was employed for overall publication trend and keyword analysis (26). The thematic map visualizes topics by dividing the coordinate axes into four quadrants based on a topic’s centrality and density (27). A position closer to the positive direction of the x-axis indicates higher centrality, while the positive direction of the y-axis signifies higher density (28). The upper right quadrant represents motor themes, while basic themes reside in the lower right quadrant. Niche themes occupy the upper left quadrant, and emerging or declining themes are positioned in the lower left quadrant. The trend topics enable visualization of thematic evolution over time within the research domain. National co-author analysis was conducted using VOSviewer (29). Within the network, each country is shown as a distinct node, with node size representing how often it appears (30). Links between nodes indicate co-occurrence between nations, where thicker links denote a higher frequency of joint appearances.

MR analysis

Data source

The single nucleotide polymorphisms (SNPs) used in this study were obtained from the OpenGWAS database (https://gwas.mrcieu.ac.uk/). Ultimately, one database was selected as the exposure data, and four databases were selected as the outcome data. The asbestos-related GWAS ID is ukb-d-22612_2, while the GWAS ID of overall lung cancer data is ieu-a-987 (Table 1). The GWAS IDs for small cell lung cancer (SCLC), lung squamous cell carcinoma (LUSC), and LUAD are ieu-a-988, ieu-a-989, and ieu-a-984, respectively.

Table 1. Characteristics of the GWAS datasets.

Trait	GWAS ID	Sample size (cases/controls)	Population	Build
Occupational asbestos exposure	ukb-d-22612_2	1,262/75,195	European	HG19/GRCh37
Lung cancer	ieu-a-987	29,863/55,586	European	HG19/GRCh37
Small cell lung carcinoma	ieu-a-988	2,791/20,580	European	HG19/GRCh37
Squamous cell lung cancer	ieu-a-989	7,704/54,763	European	HG19/GRCh37
Lung adenocarcinoma	ieu-a-984	11,245/54,619	European	HG19/GRCh37

GWAS, genome-wide association study.

SNPs selection

To ensure that SNPs are suitable for causal inference, the MR method must satisfy three core assumptions: SNPs are strongly associated with exposure, SNPs are independent of confounding factors, and SNPs influence outcomes solely through exposure (Figure 2A). To meet these requirements, we used three criteria for screening. First, SNPs linked to the exposure factor asbestos were selected using a P value threshold of 5×10⁻⁵. Thresholds were adjusted appropriately to increase the number of IVs, improving the robustness of causal inference (31). Second, linkage disequilibrium was removed by filtering for r²<0.001 and a window size of 10,000 kb to ensure SNP independence (32). Third, we selected SNPs with F-statistics greater than 10 to ensure sufficient statistical power and minimize bias (33). Detailed information on the SNPs can be found in Table S2.

Figure 2.

Mendelian randomization analysis results. (A) Key assumptions of the mendelian randomization analysis. (B) Visualization of mendelian randomization analysis results. CI, confidence interval; IVs, instrumental variables; IVW, inverse variance weighted; MR, mendelian randomization; OR, odds ratio; SNP, single nucleotide polymorphism; WM, weighted median.

Statistical analysis

All MR analyses were conducted using R (version 4.5.1). For two-sample MR, we employed the “TwoSampleMR”, “VariantAnnotation”, and “ieugwasr” packages. The primary method was inverse variance weighting (IVW) (34), with additional analyses using MR-Egger, weighted median, simple mode, and weighted mode techniques. The subsequent MR analysis results are presented in Figure 2B. To verify that the observed associations between exposure and outcome were not confounded, we carried out multiple sensitivity analyses. These included MR-Egger intercept tests (35) and MR-PRESSO tests (36) to assess the level of multihomogeneity, Cochran’s Q tests (37) to evaluate heterogeneity among IVs, and leave-one-out sensitivity analyses. LDtrait was also employed to identify potential confounders and exclude SNPs significantly associated with confounders (38).

GBD analysis

Data extraction

Data for this study were obtained from GBD 2021, which provided disease burden estimates of 371 diseases and injuries, as well as 88 risk factors from 1990 to 2021 (39,40). In GBD 2021, ARLC is not modelled as a standalone diagnostic category. In this study, ARLC was therefore operationally defined as the burden of tracheal, bronchus, and lung (TBL) cancer attributable to occupational exposure to asbestos. The attributable burden was estimated using the GBD comparative risk assessment framework, whereby population attributable fractions for the risk-outcome pairing were derived from exposure distributions, relative risk estimates, and the theoretical minimum risk exposure level, and then applied to the total TBL cancer burden to obtain asbestos-attributable TBL cancer burden (40). Since GBD 2021 excludes ARLC burden data for individuals under 20 years of age, our analysis primarily focuses on those aged 20 and older, grouped into 16 five-year age cohorts ranging from 20–24 years to 95+ years. We extracted global ARLC deaths and disability-adjusted life years (DALYs) from 1990 to 2021, along with their 95% uncertainty intervals (UIs) for subsequent analysis and prediction. All data used in this study are publicly available via the GBD results tool (http://ghdx.healthdata.org/gbd-results-tool). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Data analysis

We conducted cross-sectional analyses to assess the burden of ARLC. The mortality and DALYs for ARLC across different genders and age groups in 1990 and 2021 were analyzed. Concurrently, the lung cancer burden attributable to occupational asbestos exposure was quantified and visualized across 21 GBD regions and 5 socio-demographic index (SDI) regions (40). Finally, the disease burden for ARLC from 2022 to 2036 was projected using the BAPC model (41). All data analyses and graphical plotting were performed using the statistical software R (version 4.5.1).

Results

Bibliometric analysis

Literature search

A total of 3,026 relevant documents were retrieved from the Web of Science Core Collection. First, 308 documents were excluded based on document type, retaining only articles and article reviews. Additionally, 2,240 documents addressing only mesothelioma or ARDs other than lung cancer were excluded through title and abstract screening. Ultimately, 478 documents were included in this analysis, comprising 395 articles and 83 reviews.

Analysis of publications

Figure 3A illustrates the annual publication trend from 2008 to 2025. During this period, the annual number of publications fluctuated within a relatively narrow range of 18–35 articles.

Figure 3.

Bibliometric analysis results. (A) Annual publication trend. (B) Co-authorship analysis of countries. (C) Thematic map. (D) Trend topics.

Analysis of countries

The co-authorship analysis of countries is shown in Figure 3B. The largest nodes represent the United States (documents =140), followed by Italy (documents =90) and China (documents =50). Based on total link strength, the top five countries were the United States (n=123), Italy (n=100), the Netherlands (n=88), France (n=87), and the United Kingdom (n=83), with the United States exhibiting particularly strong collaborative ties with other nations.

Analysis of keywords

The top twenty most frequent keywords in the literature are shown in Table 2. In addition to the search keywords used, other primary keywords reflecting research trends in the field include “mesothelioma” (occurrences =88), “occupational exposure” (occurrences =31), and “cancer” (occurrences =26). The thematic map displays the development and relevance of eighteen identified thematic clusters (Figure 3C). Figure 3D illustrates the trend topics from 2008 to 2025, displaying the duration and frequency of occurrence for each keyword. Recent research hotspots are primarily focused on asbestos prevention and research into asbestos-induced carcinogenesis at the cellular level.

Table 2. List of the top 20 most frequent author’s keywords.

Rank	Author’s keywords	Occurrences
1	Asbestos	187
2	Lung cancer	154
3	Mesothelioma	88
4	Asbestosis	35
5	Chrysotile	33
6	Occupational exposure	31
7	Cancer	26
8	Mortality	24
9	Epidemiology	22
10	Asbestos exposure	19
11	Occupational health	14
12	Pleural plaques	14
13	Smoking	13
14	Exposure	12
15	Lung neoplasms	12
16	Malignant mesothelioma	12
17	Screening	12
18	Cohort study	11
19	Lung	11
20	Carcinogenicity	10

MR analysis

Asbestos and overall lung cancer

Through multistage genetic variant screening, we selected 85 SNPs associated with asbestos exposure and overall lung cancer as IVs. Since no evidence of heterogeneity was detected in Cochran’s Q test (Q =93.458, P=0.23), a fixed-effects model was employed in the primary analysis of MR (Table 3). The IVW analysis revealed no significant causal association between asbestos and overall lung cancer [odds ratio (OR) =2.110, 95% confidence interval (CI): 0.738–6.031, P=0.16], and none of the four additional methods demonstrated a significant causal relationship. Furthermore, no horizontal pleiotropy was identified in the MR-PRESSO test (P=0.22) or the MR-Egger intercept test (intercept =0.002, P=0.73). The supplementary materials include scatter plots (Figure S1A), funnel plots (Figure S1B), forest plots (Figure S1C), and leave-one-out plots (Figure S1D), illustrating the MR sensitivity analyses for overall lung cancer.

Table 3. Results of multiple sensitivity analyses in the mendelian randomization (MR) analysis.

Exposure	Outcome	MR-Egger intercept test		MR-PRESSO	Cochran’s Q test
		Egger intercept	P value	P value	Q	Q_pval
Worked with materials containing asbestos: often	Lung cancer	0.002	0.73	0.22	93.458	0.225
Worked with materials containing asbestos: often	Small cell lung carcinoma	−0.001	0.94	0.44	63.804	0.448
Worked with materials containing asbestos: often	Squamous cell lung cancer	0.004	0.65	0.12	98.517	0.118
Worked with materials containing asbestos: often	Lung adenocarcinoma	−0.010	0.16	0.23	91.747	0.239

Asbestos and SCLC

After rigorous screening, we included 64 SNPs associated with asbestos exposure and SCLC. Heterogeneity testing (Cochran’s Q =63.804, P=0.45) supported the use of a fixed-effects model for MR causal estimation (Table 3). Primary MR analysis results indicated no significant causal association between asbestos and SCLC (OR =8.910, 95% CI: 0.549–144.543, P=0.12), and supplementary methods also failed to demonstrate causality. No horizontal pleiotropy was detected in the MR-Egger intercept test (intercept =−0.001, P=0.94) or the MR-PRESSO test (P=0.44). Figure S2A-S2D display the relevant​ scatter plots, funnel plots, forest plots, and leave-one-out plots, respectively.

Asbestos and LUSC

Building on the fundamental assumptions of IVs, we chose 84 SNPs linked to asbestos exposure and LUSC. Cochran’s Q test indicated no significant heterogeneity (Q =98.517, P=0.12), leading us to use a fixed-effects model for the main MR analysis (Table 3). IVW results suggested no significant causal association between asbestos exposure and LUSC, consistent with results from supplementary methods. The MR-Egger intercept test (intercept =0.004, P=0.65) and the MR-PRESSO test (P=0.12) both indicated no significant horizontal pleiotropy in the IVs. The corresponding scatter plots, funnel plots, forest plots, and leave-one-out plots are presented in Figure S3A-S3D, respectively.

Asbestos and LUAD

Initially, this study screened and included 86 causal inference SNPs for asbestos and LUAD. However, Cochran’s Q test indicated heterogeneity among these IVs (Q =108.251, P=0.045). Additionally, two outliers identified in the MR-PRESSO test—IV rs146479864 and IV rs62095961—were excluded from further analysis. Ultimately, 84 relevant SNPs were included (Cochran’s Q =91.747, P=0.24). The IVW analysis demonstrated a suggestive subtype-specific association between asbestos exposure and LUAD (OR =4.901, 95% CI: 1.150–20.881, P=0.03), consistent with the MR-Egger findings (OR =30.902, 95% CI: 1.634–584.331, P=0.03). Scatter plots supported this suggestive association, exhibiting positive slopes in both the IVW and MR-Egger regression lines (Figure S4A). As shown in Table 3, no evidence of horizontal pleiotropy was observed in the MR-Egger intercept test (intercept =−0.010, P=0.16) or the MR-PRESSO test (P=0.23). SNPs displayed a symmetric distribution in the funnel plot (Figure S4B). In addition, the overall effect's CI in Figure S4C excluded zero, indicating a statistically significant association within the population. Finally, leave-one-out analysis (Figure S4D) revealed no variants exerting disproportionate influence on the aggregate estimate.

GBD analysis

Differences in burden across age and sex

In 1990, the age-sex pattern of ARLC deaths is shown in Figure 4A. By 2021, as illustrated in Figure 4B, the death rate for ARLC increased with age and was higher in males than in females within the same age group. Male death rate peaked at 90–94 years, while female deaths rate peaked at 95+ years. Deaths cases were highest in the 75–79 age group for both sexes. For DALYs, the age-sex pattern in 1990 is presented in Figure 4C. Both male and female DALYs cases peaked at 70–74 years in 2021, and the same gender-age pattern observed in deaths rate was also evident when considering DALYs (Figure 4D). Disease burden trends and gender disparities in 1990 were largely similar to those observed in 2021, though the disease burden in 2021 had increased compared to 1990.

Figure 4.

Differences in burden across age and sex. (A) Deaths cases and rate in 1990. (B) Deaths cases and rate in 2021. (C) DALYs cases and rate in 1990. (D) DALYs cases and rate in 2021. DALYs, disability-adjusted life years.

Attributable burden of risk factors

Figure 5 shows DALYs from occupational asbestos exposure for males and females in 1990 and 2021. Asbestos had a greater impact on male DALYs. The proportion of DALYs from asbestos remained mostly stable, with slight improvement in 2021. Across the five SDI regions, asbestos’s impact grew with SDI, mainly affecting Western Europe and Australasia for both genders.

Figure 5.

Attributable burden of risk factors. (A) Proportion of lung cancer DALYs attributable to asbestos risk factor in 1990. (B) Proportion of lung cancer DALYs attributable to the asbestos risk factor in 2021. GBD, Global Burden of Disease Study; DALYs, disability-adjusted life years; SDI, socio-demographic index.

BAPC prediction of ARLC

From 2021 to 2036, the global male age-standardized deaths rate (ASDR) is projected to decline from 7 per 100,000 population in 2021 to 5 per 100,000 by 2036, representing a 15-year decline rate of approximately 28.6% (Figure 6A). In comparison, the ASDR for females is projected to decrease from 0.89 per 100,000 in 2021 to 0.68 per 100,000 in 2036, for a 15-year decline of nearly 23.6% (Figure 6B). The trends in age-standardized DALYs rate for males and females correspond to their respective ASDR patterns (Figure 6C,6D, respectively), showing a year-on-year decline, with the trends broadly aligning.

Figure 6.

BAPC prediction of ARLC. (A) Trends of age-standardized deaths rate for ARLC among males. (B) Trends of age-standardized deaths rate for ARLC among females. (C) Trends of age-standardized DALYs rate for ARLC among males. (D) Trends in age-standardized DALYs rate for ARLC among females. Agestd, age standardized; ARLC, asbestos-related lung cancer; BAPC, Bayesian age-period-cohort; DALYs, disability-adjusted life years.

Discussion

This study presents a systematic evaluation of the association between occupational asbestos exposure and trends in lung cancer burden, integrating bibliometric analysis, MR, and GBD 2021 data. Our integrated findings reveal a complex picture of ARLC. Bibliometric analysis indicates a sustained yet stable level of research output, accompanied by a gradual shift in focus toward molecular mechanisms and prevention strategies. MR analysis suggests a potential genetic association between asbestos exposure and LUAD, although null results for other subtypes warrant cautious interpretation. Meanwhile, GBD data highlight a persistent and unequal global ARLC burden, with the BAPC model projecting a decline by 2036. The projected decline is directionally consistent with historical asbestos restrictions and the gradual phase-out of asbestos. However, other factors such as long exposure latency and variation in cancer registration completeness warrant further attention.

It should be noted that pre-2008 publications were not included in the bibliometric analysis. Instead, this study deliberately focused on publications from 2008 onward to capture contemporary research trajectories following the adoption of WHA60.26 in 2007 and the launch of the World Health Organization (WHO) Global Plan of Action on Workers’ Health 2008–2017. The bibliometric analysis indicates that the annual publication output in the ARLC field remained relatively consistent from 2008 to 2025. Research hotspots evolved over time, initially emphasizing occupational exposure tracking in high-risk industries such as textiles (42), mining, and milling (43), consistent with the keywords in trend topics. However, the heavy reliance on retrospective designs weakened the evidence quality. Recently, new topics have gained attention, including studies on multi-walled carbon nanotubes (44), asbestos exposure prevention (45), and efforts to reduce the global burden of ARDs (46). Several research papers have examined how asbestos influences the expression of p53 and p16 genes (16,47). International collaboration networks reveal disparities in ARLC research, with most partnerships occurring among developed nations like the United States and Italy from 2008 to 2025. Wealthier countries typically have more resources, fostering stronger cross-border research collaborations and knowledge sharing.

Despite the predominance of retrospective designs identified by our bibliometric analysis, causal evidence at the subtype level remains limited. We therefore performed a two-sample MR analysis to explore potential subtype-specific associations, which revealed that asbestos exposure was suggestively associated only with LUAD (OR =4.901, P=0.03), with no association found for overall lung cancer or other subtypes. Although asbestos exposure increases the risk of all histological types of lung cancer (48,49), numerous studies indicate that adenocarcinoma is the most common subtype in ARLC (50,51). Concurrently, the driver genes of asbestos-related LUAD have been extensively studied. Wright et al. identified ADAM28 as a specific oncogene for asbestos-related LUAD, validated across multiple independent cohorts (52). MRPL1, SDK1, SEMA5B, and INPP4A were previously demonstrated by Mäki-Nevala et al. as candidate genes for asbestos-related LUAD (53). The MR results linking asbestos to adenocarcinoma further corroborate this association.

Discrepancies between the MR findings and those of traditional observational studies for different subtypes and overall lung cancer can be explained by several factors. First, IVs used for asbestos exposure might not fully capture all biological effects of asbestos, possibly affecting specific lung cancer pathways and resulting in different outcomes. Second, the smaller sample sizes for SCLC and LUSC, compared to LUAD, may limit the statistical power of these analyses. Moreover, asbestos exposure might be more strongly associated with certain subtypes (54), which could be diluted when considering the full lung cancer population. Lastly, asbestos-exposed workers tend to have higher smoking rates (55), and the influence of asbestos on lung cancer risk may be obscured by the effects of smoking. Although MR analysis provides genetic-level insights into the asbestos-lung cancer association, the public health magnitude of this relationship across demographic and regional strata warrants quantification via population-level burden data.

Our GBD 2021 analysis reveals diverse patterns of ARLC burden variation both horizontally and vertically. The latency period between initial asbestos exposure and the clinical onset of ARDs typically spans several decades (56). Moreover, following the publication of the first study linking asbestos to lung cancer in 1935, most developed nations took three to four decades to implement preventive measures (57). This lag may explain why the disease burden of ARLC in 2021 exceeded that of 1990. Since major measures to reduce occupational and general population exposure to asbestos were only implemented in the 1980s, individuals born prior to this period had greater opportunities for substantial asbestos exposure (58-60). Consequently, mortality rates and DALY rates are higher among older age cohorts. Additionally, the absolute number of deaths is influenced by the population size of each age cohort, which can shift the peak of death burden away from the oldest old toward younger elderly groups. The greater ARLC burden among males is primarily driven by occupational exposure patterns, as men are more likely to be employed in industries with high asbestos exposure levels, resulting in higher asbestos exposure among men than women (13,61). The asbestos attribution risk findings show a seemingly paradoxical pattern: areas with strict regulations and high income tend to have a higher burden. This is primarily due to the legacy effects of past exposure (57). These countries industrialized early and used asbestos extensively, resulting in a large population base of individuals with occupational asbestos exposure (62,63). Moreover, the well-established occupational health systems in these regions enable more accurate attribution of causes of death, whereas many cases in low- and middle-income areas are categorized as ordinary lung cancer (64,65).

The BAPC model reveals that the worldwide burden of lung cancer caused by occupational asbestos exposure is expected to decrease. This decrease aligns with the declining trend in global asbestos production and use over recent decades, resulting from national bans and increased health awareness (66). There is an increasing body of research advocating for stronger international cooperation, greater public awareness, and policy to reduce ARDs globally (45,67). Although this research did not directly assess the impact of asbestos policies, the implementation of asbestos regulations and the annual reduction in ARLC cases offer indirect support for stricter laws (68).

In the future, we should focus on promoting a thorough understanding and awareness of asbestos risks in countries where its use is still unregulated, aiming for a complete ban on asbestos (69). However, simply banning asbestos is not enough, as we will continue to inhabit buildings and workplaces constructed with this material, which can degrade at any time (70). This poses a risk of long-term exposure, requiring remediation efforts. Additionally, regular lung cancer screenings should be provided for people at high risk of asbestos exposure (65,71,72). Research institutions from both developed and developing nations need to work together to improve global prevention efforts, reduce environmental health disparities, and ultimately eliminate ARDs (66).

This study has certain limitations. First, the bibliometric analysis was deliberately restricted to publications from 2008 onward. Earlier studies were therefore not included, as the analysis aimed to capture contemporary research trajectories following WHA60.26 [2007] and the launch of the WHO Global Plan of Action on Workers’ Health [2008–2017]. Second, MR primarily utilized GWAS data from European populations, limiting its generalizability. Enrolling genetically diverse study cohorts is crucial for enhancing the applicability of findings across populations. Third, our MR analysis detected suggestive causal evidence only for LUAD, likely due to limited statistical power from smaller subtype-specific GWAS and reliance on proxy genetic instruments with modest strength. These are common challenges in asbestos-related MR that future studies using direct exposure instruments and larger samples should address. Fourth, the accuracy of ARLC burden estimates is influenced by the quality of cancer registry data and the GBD’s inability to distinguish between lung cancer and TBL cancers, or to analyze histological subtypes separately. Further efforts should focus on refining disease classification and improving data collection in resource-limited regions. Fifth, GBD burden estimates are model-derived approximations rather than direct registry-observed counts, and uncertainty is higher in regions with less robust surveillance infrastructure. Therefore, caution is warranted when interpreting stratified results by age, sex, and region. Sixth, this study exclusively examined asbestos exposure without considering potential interactions with other risk factors. Subsequent research should adopt a multi-exposure framework to better understand compound effects. Finally, the MR analysis presented here lacks independent external validation in an alternative cohort or dataset. Replication in populations with diverse ancestries and more precise exposure assessment would strengthen confidence in the observed associations.

Conclusions

This study used an MR approach to suggest a potential causal relationship between occupational asbestos exposure and LUAD, partially addressing common limitations in observational studies such as confounding factors and reverse causality. While no causal link was found between overall lung cancer or other subtypes and asbestos exposure, this genetic evidence emphasizes the unique carcinogenic mechanisms across different subtypes. Meanwhile, bibliometric analysis shows a research shift from exposure assessment to molecular mechanisms, although international collaboration remains mainly in developed regions. ARLC burden varies significantly by gender, age, and geography, with higher socioeconomic areas experiencing a greater burden due to historical exposure and better diagnostics. Projections suggest a continued decline in burden through 2036, consistent with the expected impact of asbestos restrictions.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.