Causal relationship between gastro-esophageal reflux disease and risk of lung cancer: insights from multivariable Mendelian randomization and mediation analysis

Abstract

Aim

A recent study has reported that anti-reflux surgery reduced the risk of lung cancer. However, the exact causal association between gastro-esophageal reflux disease (GORD) and lung cancer remains obscure. Therefore, we conducted a multivariable and network Mendelian randomization (MR) study to explore this potential association and mediation effect.

Methods

Independent single nucleotide polymorphisms (SNPs) strongly associated with GORD were selected as instrumental variables (IVs) from the corresponding genome-wide association studies (GWAS). The summary statistics were obtained from the largest GORD GWAS meta-analysis of 367 441 (78 707 cases) European individuals, and the summary statistics of lung cancer and pathological subtypes came from International Lung Cancer Consortium (ILCCO) and FinnGen databases. Univariable and multivariable MR analyses were performed to investigate and verify the causal relationship between genetically predicted GORD and lung cancer. Network MR analysis was conducted to reveal the mediating role of GORD between smoking initiation and lung cancer.

Results

The univariable MR analysis demonstrated that GORD was associated with an increased risk of total lung cancer in both ILCCO [inverse variance weighted (IVW): odds ratio (OR) = 1.37, 95% confidence interval (CI) 1.16–1.62, P = 1.70E-04] and FinnGen database (IVW: OR = 1.25, 95% confidence interval CI 1.03–1.52, P = 2.27E-02). The consistent results were observed after adjusting the potential confounders [smoking traits, body mass index (BMI) and type 2 diabetes] in multivariable MR analyses. In subtype analyses, GORD was associated with lung adenocarcinoma (IVW: OR = 1.27, 95% CI 1.02–1.59, P = 3.48E-02) and lung squamous cell carcinomas (IVW: OR = 1.50, 95% CI 1.22–1.86, P = 1.52E-04). Moreover, GORD mediated 32.43% (95% CI 14.18–49.82%) and 25.00% (95% CI 3.13–50.00%) of the smoking initiation effects on lung cancer risk in the ILCCO and FinnGen databases, respectively.

Conclusion

This study provides credible evidence that genetically predicted GORD was significantly associated with an increased risk of total lung cancer, lung adenocarcinoma and lung squamous cell carcinomas. Furthermore, our results suggest GORD is involved in the mechanism of smoking initiation-induced lung cancer.

Introduction

Lung cancer is one of the most prevalent malignancies, with a high incidence rate, and is a leading cause of cancer-related deaths worldwide.¹ The 5-year overall survival of lung cancer is nearly 19%, partly because more than one-half of cases are diagnosed at an advanced stage.¹ Therefore, early detection and diagnosis are of great importance to prevent cancer and improve prognosis. Recently, genetically predicted causal inference between complex risk factors and diseases has been popularized in epidemiological studies, which provide new insights into early screening and prevention of disease.^2–4 This study aims to investigate gastro-esophageal reflux disease (GORD) as the potential risk factor for lung cancer, using genetic instruments, and to explore the potential mediation effect, providing new insights for the early prevention and screening of lung cancer.

GORD is a gastrointestinal disorder characterized by frequent regurgitation of stomach acid and bile, which is known as the predominant risk factor of Barrett’s esophagus and esophageal adenocarcinoma.⁵ Airway aspiration is an important symptom of GORD, which could cause a series of pulmonary chemical irritation symptoms and inflammation.⁶ However, there are only a few studies, with insufficient evidence, which focus on the potential association between GORD and lung cancer.⁷ A recent multicentre study from Northern Europe demonstrated that anti-reflux surgery for patients with GORD decreased the risk of lung cancer, but the result might be biased due to the confounder of smoking.⁸ Additionally, an Asian retrospective study found that GORD was significantly associated with lung cancer.⁹ Due to the inevitable potential reverse causality and residual confounding in observational studies, whether there is a causal association between GORD and lung cancer remains obscure.

Two-sample Mendelian randomization (MR) analysis is a popular method using single nucleotide polymorphisms (SNPs) as instrumental variables (IVs) to infer credible causal relationship when randomized controlled trials (RCTs) are not feasible.¹⁰ As genetic variations were randomly assigned at meiosis and were independent of environmental and other acquired factors, MR design provides a randomization procedure and minimizes the interference of residual confounding and reverse causality.³ To achieve a credible causal inference, IVs of MR analysis must satisfy three assumptions. (i) The instrumental variables must be strongly related to exposure. (ii) The instrumental variables are not associated with any confounders. (iii) The instrumental variables are not directly related to outcomes. Recently, MR studies have revealed a causal association between several risk factors [e.g. smoking, body mass index (BMI), diabetes] and lung cancer.^11–14 Nonetheless, there is as yet no MR evidence to prove the causal association and mediation effect between GORD and lung cancer.

Considering that tobacco smoking is a common risk factor for both GORD and lung cancer,^15,16 we used mediation analysis to investigate the mediation role of GORD between smoking initiation and lung cancer. Mediation analysis is a statistical approach that aims to identify intermediate factors as potential intervention targets and to improve the cognition of aetiological composition.¹⁷ Traditional non-IV mediation analysis is limited by the lack of strong causal assumptions. MR design overcomes these limitations by using the instrumental variable technique. Therefore, network MR analysis is widely used to investigate potential mediation effects in a causal pathway.^2,4

In this study, we conducted two-sample Mendelian randomization (MR) analyses utilizing genome-wide association study (GWAS) summary data to investigate the causal effect of GORD on lung cancer susceptibility, based on two large databases. Moreover, we further conducted network MR analysis to explore the potential mediator role of GORD between smoking initiation and lung cancer. These analyses provide causal evidence for the role of GORD in lung cancer genesis, which provides a new perspective on the prediction and prevention of lung cancer.

Methods

We conducted two-sample MR analyses based on the largest GORD dataset and two large lung cancer datasets, to infer the causal effects of GORD on overall lung cancer and its pathological subtypes. Multivariable MR analysis was performed to reveal the direct causal effect of GORD on lung cancer by adjusting potential confounders. Furthermore, we performed network MR analysis to investigate the potential mediation role of GORD between smoking initiation and lung cancer. A detailed flowchart summarizes the study design (Figure 1). This study is designed and reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline, specific for MR (STROBE-MR) (Checklist).¹⁸

GWAS data sources

GWAS for GORD

We identified genetic instruments for GORD from the largest GORD GWAS meta-analysis in Europe, which included a total of 367 441 (78 707 cases) European individuals from the UK Biobank study (354 285 individuals) and QSKIN study (13 156 individuals).¹⁹ UK Biobank is a population-based cohort study of more than 500 000 individuals and biological samples from across Great Britain. GORD cases from UK Biobank were defined as a mixture of self-reported GORD symptoms, ICD-9/10 diagnosis and GORD-related medication use.²⁰ QSKIN study is conducted by the QIMR Berghofer Medical Research Institute and aims to investigate risk factors for skin cancers and other complex diseases, which includes 43 794 participants between 40 and 69 years old from Queensland, Australia. GORD in the QSKIN cohort was defined as individuals who self-reported heartburn and took one or more types of GORD-related medications.²¹ More detail on the meta-analysis method may be obtained from the supplemental method and the original research (Supplementary Table S1 and Supplementary Method, available as Supplementary data at IJE online).

GWAS for smoking traits

To adjust for potential confounders and investigate the mediation effect of GORD between smoking and lung cancer, we obtained genetic instruments for smoking traits including smoking initiation, smoking duration (age of smoking initiation) and smoking frequency (cigarettes smoked per day) from the Sequencing Consortium of Alcohol and Nicotine use (GSCAN) project. GSCAN provided the largest smoking-related GWAS meta-analysis and identified 566 genetic variants associated with four smoking phenotypes in 1 232 091 individuals of European ancestry.²² More information on the study design and GWASs in this meta-analysis study can be found elsewhere.²²

GWAS for other traits

To validate the effectiveness of the GORD instruments, we obtained the GWAS meta-analysis data of Barrett's esophagus, which was derived from four large European ancestry cohorts.²³ Besides, to avoid the interference of potential confounders and explore the direct causal effect of GORD on lung cancer, we obtained the genetic instruments for BMI and type 2 diabetes from the most recent and largest studies. GWAS summary statistics for BMI were obtained from the Genetic Investigation of Anthropometric Traits (GIANT) consortium, which included 152 893 participants.²⁴ Summary statistics for type 2 diabetes were obtained from Diabetes Genetics Replication and Meta-analysis (DIAGRAM), which included 149 821 (34 840 cases) participants.²⁵ The details of traits involved in this study are presented in Supplementary Table S1).

GWAS for outcomes

The GWAS data for lung cancer were obtained from two large databases, International Lung Cancer Consortium (ILCCO) and FinnGen. ILCCO is a multicentre project dedicated to the study of genetic variants associated with lung cancer.²⁶ FinnGen database is a combined project that integrates digital health records from the Finnish Health Registry with genetic data from the Finnish biobank (https://www.finngen.fi/en). In this study, we obtained two GWAS summary statistics for total lung cancer from ILCCO (11 348 cases and 15 861 controls) and FinnGen (4030 cases and 238 678 controls) databases as the primary outcomes. We further used GWAS for lung adenocarcinoma (3442 cases and 14 894 controls) and lung squamous cell carcinoma (3275 cases and 15 038 controls) from ILCCO, and the GWAS for small cell lung carcinoma (461 cases and 238 678 controls) from FinnGen, as secondary outcomes to reveal the association between GORD and lung cancer pathological subtypes (Supplementary Table S2, available as Supplementary data at IJE online). All summary statistics were filtered at the minimum variant allele (MAF) frequency >0.01. UK Biobank database is excluded in the selection of outcome data due to the consideration of sample overlap between exposure and outcome.

Genetic instruments selection criteria

Genetic instruments (SNPs) for GORD, Barrett's esophagus, smoking traits, BMI and type 2 diabetes were all selected at genome-wide significant threshold (P <5E-8). We identified independent SNPs by performing linkage disequilibrium (LD) pruning for each trait based on the 1000 Genomes LD reference panel in European ancestry. The threshold of LD was set to r² >0.01 and clump window <10 kb. In order to infer correct causal estimates from the MR analysis, the effect of an SNP on an outcome and exposure must be harmonized to be relative to the same allele. Thus, we used the ‘TwoSampleMR’ R package and performed variant harmonization to identify and exclude palindromic SNPs for which the correct orientation of the allele could not be determined. To avoid the confounding effect, SNPs significantly associated with outcome (P < 5E-8) were excluded. Also we excluded SNPs pleiotropically associated with second traits which might confound the causal estimate, through PhenoScanner v2 [PhenoScanner (cam.ac.uk), P <1E-5]²⁷ (Supplementary Table S3, available as Supplementary data at IJE online). Besides, we calculated F statistics to evaluate the validity of instrument variables. A value less than 10 suggests a weak instrumental variable.

Statistical analyses

Main Mendelian randomization analyses

We used the random effect inverse variance weighted (IVW) method as the main method for all MR analyses. IVW model provided a weighted regression of the IV-specific causal estimation and could provide a stable causal inference even in the presence of horizontal pleiotropy.²⁸ Additionally, we performed analyses in both ILCCO and FinnGen databases to ensure the consistency of the inference. Considering smoking traits (smoking initiation, smoking duration and smoking frequency), BMI and type 2 diabetes were potential risk factors for lung cancer, we adjusted these confounders by using multivariate MR analyses and explored the direct effect of GORD on lung cancer. Moreover, we further explored the causal relationship between GORD and lung cancer pathological subtypes to refine the risk of GORD for lung cancer.

Network Mendelian randomization analyses

To investigate the potential mediator role of GORD between smoking initiation and lung cancer, we conducted network MR analysis. Three estimates were performed for network MR analysis: (i) the total effect of smoking initiation on lung cancer; (ii) the direct effect α of smoking initiation on GORD; and (iii) the direct effect β of GORD on lung cancer. Mediation effect can be calculated by the follwing equation: $Mediation effect=α*β.$The calculation process is described in detail in Supplementary Method (available as Supplementary data at IJE online).²⁹ The proportion of the mediation effect was estimated as the total causal effect of smoking on lung cancer divided by the mediation effect. To ensure the reliability of the results, we performed network MR analyses in both ILCCO and FinnGen databases and compared them for consistency.

Sensitivity analyses

Sensitivity analyses were performed by weighted median, MR-Egger regression and MR-Pleiotropy Residual Sum and Outlier (MR-PRESSO) methods to verify the basic assumptions for univariable MR analyses. Multivariable MR-Egger regression was used to detect the pleiotropy for multivariable MR analyses.³⁰ The weighted median model calculates a weighted value of the median of the IV-specific causal estimation. When more than half of the analytical weights are derived from valid IVs, the weighted median model can provide a consistent causal estimation.³¹ MR-Egger regression allows free estimation of the horizontal pleiotropic effects, whereas it compromises statistical power.³² MR-PRESSO is developed to detect and correct horizontal pleiotropic outliers. The MR-PRESSO distortion test demonstrates the difference in estimation before and after outlier elimination.³³ Cochran’s Q test is used to estimate heterogeneity among SNPs for exposure and checks the consistency between MR assumption and analyses. Based on the methods of MR estimate and sensitivity analysis mentioned above, we considered a credible causal inference must meet the following criteria. (i) The results of MR estimate and sensitivity analysis presented a consistent direction among the three methods. (ii) Intercept test of MR-Egger suggested no horizontal pleiotropic effect. We calculated F statistics to evaluate the strength of IVs for exposures in univariable MR analysis (Supplementary Method).³⁴ Two‐sample conditional F statistic (F_TS) was calculated by the ‘MVMR’ R package to evaluate the strength of IVs for exposures in multivariable MR analyses.³⁵ F statistic >10 suggests a strong instrumental variable. Statistical power of each MR analysis was calculated by the mRnd power calculator for binary outcomes: mRnd: power calculations for Mendelian Randomization [cnsgenomics.com].³⁶

All the above processes were carried out in R 4.1.0 [https://www.R-project.org/]. R package ‘TwoSampleMR’,³⁷ ‘MRPRESSO’³³ and ‘MendelianRandomization’³⁸ were used to perform MR and sensitivity analyses.

Results

Genetic instruments

The process of IV selection for this study is presented in Supplementary Figure S1 (available as Supplementary data at IJE online). According to the criteria of genetic instrument selection, 77 and 76 independent SNPs were finally selected as IVs for GORD in the analyses of ILCCO and FinnGen databases, respectively, and 74 and 75 SNPs were selected as IVs for smoking initiation, respectively. The detail for IVs of each exposure is reported in Supplementary Tables S4 and S5 (available as Supplementary data at IJE online). The selected IVs overall explained 3.58% of the GORD variance and 2.15% of the smoking initiation variance. All F statistics for IVs used in this univariable MR analyses = >10, which suggested strong instrument variables (Supplementary Table S6, available as Supplementary data at IJE online). However in multivariable MR analysis, the F_TS statistic of the GORD instrument variable was less than 10 after adjusting for smoking duration and BMI, indicating weak IVs (Supplementary Table S7, available as Supplementary data at IJE online).

In order to validate the effectiveness of the GORD IVs, we performed a test MR analysis on GORD and Barrett's esophagus. The results demonstrated that genetically predicted GORD was associated with an increased risk of Barrett's esophagus [IVW: odds ratio (OR) = 1.26, 95% confidence interval (CI) 1.05–1.52, P = 1.28E-02), and sensitivity analysis indicated consistent findings. This suggests that the instrumental variables for GORD used in this study are valid (Supplementary Table S8, available as Supplementary data at IJE online). Additionally, the statistical power of all MR analyses was 1.00 (>0.80; Supplementary Table S9, available as Supplementary data at IJE online).

Causal effects of GORD on lung cancer and pathological subtypes

By using the genetic IVs for GORD, univariate MR analysis provided credible evidence that GORD was associated with an increased risk of total lung cancer in the ILCCO database (IVW: OR = 1.37, 95% CI 1.16–1.62, P = 1.70E-04), and the estimate was consistent in the FinnGen database (IVW: odds ratio (OR) = 1.25, 95% confidence interval (CI) 1.03–1.52, P = 2.27E-02). In the subgroup analysis of lung cancer pathological subtypes, GORD increased the risk of lung adenocarcinoma (IVW: OR = 1.27, 95% CI 1.02–1.59, P = 3.48E-02) and lung squamous cell carcinomas (IVW: OR = 1.50, 95% CI 1.22–1.86, P = 1.52E-04). However, only a weak causal association was observed between GORD and small cell lung cancer (IVW: OR = 2.15, 95% CI 0.87–5.31, P = 0.10) (Figure 2).

Consistent directions of the causal estimates were observed in most sensitivity analyses, but the direction of ORs in MR-Egger was inconsistent in the analyses of adenocarcinoma and small cell lung cancer (0.84 and 0.03, respectively) (Figure 2). One outlier in the analysis of total lung cancer in the ILCCO database was detected by MR-PRESSO, but the result was still consistent after eliminating the outlier variant (OR = 1.33, 95% CI 1.31–1.35, P = 6.90E-04). Moderate heterogeneity was observed in the analysis of total lung cancer in the ILCCO database. Furthermore, no horizontal pleiotropy was detected in MR-Egger intercept tests for any of the MR analyses conducted (Table 1).

Multivariable MR analyses adjusting potential confounders

In most multivariable MR analyses, the causal effect of GORD on lung cancer and pathological subtypes remained consistent after adjusting the five potential confounders (the three smoking traits, BMI and type 2 diabetes). However, the causal effect of GORD on lung adenocarcinoma was attenuated after adjusting smoking initiation (IVW: OR = 1.17, 95% CI 0.98–1.36, P = 0.11) and smoking frequency (IVW: OR = 1.02, 95% CI 0.81–1.28, P = 0.88) (Figure 3).

In sensitivity analyses, no horizontal pleiotropy was detected by MR-Egger regression of the multivariable MR analyses, and mild to moderate heterogeneity was observed in the analysis of total lung cancer and lung adenocarcinoma (Table 2).

GORD mediates the causal effect of smoking initiation on lung cancer

We performed network MR analyses to investigate the potential mediating role of GORD in the association between smoking initiation and lung cancer, in both ILCCO and FinnGen databases. In the ILCCO analysis, we observed that smoking initiation (IVW: OR = 1.44, 95% CI 1.18–1.76, P = 2.66E-04) was associated with an increased total risk of lung cancer. For direct effects, smoking initiation (IVW: OR = 1.45, 95% CI 1.33–1.59, P = 2.74E-16) was associated with increased risk of GORD (direct effect α). In addition, GORD has exhibited a causal relationship with increased risk of lung cancer (IVW: OR = 1.37, 95% CI 1.16–1.62, P = 1.70E-04) (direct effect β). The proportion of the causal effect of smoking initiation on lung cancer mediated by GORD was 32.43% (95% CI 14.18%–49.82%). We repeated the network analysis in the FinnGen database and obtained consistent causal estimates of the mediating proportion of GORD in smoking initiation to lung cancer risk (25.00%, 95% CI 3.13%–50.00%). Moreover, no horizontal pleiotropy was detected in all network MR analyses in both ILCCO and FinnGen datasets (Table 3; Supplementary Table S10, available as Supplementary data at IJE online).

Discussion

Our Mendelian randomization study revealed that genetically predicted GORD was associated with an increased risk of total lung cancer, lung adenocarcinoma and lung squamous cell carcinomas, and this causal association was independent of smoking duration, BMI and type 2 diabetes. Besides, we provided causal evidence that GORD mediated a considerable proportion of smoking initiation effect on lung cancer risk. However, the study was analysed at the genetic level, and thus these results should be interpreted with caution.

The risk of GORD for lung cancer may be mediated by ongoing pulmonary aspiration and the resulting chronic inflammation.⁸ Previous study has proved that chronic inflammation induced by refluxate (i.e. stomach acid and bile salts) increases the risk of esophagus cancer and laryngeal/pharyngeal cancer,³⁹ and here we revealed the genetically predicted association between GORD and lung cancer. Long-term exposure to stomach acid and bile salts, which aspirate into the airway, cause chronic inflammation and increase the release of reactive oxygen species (ROS).⁶ Accumulated ROS could cause ongoing oxidative DNA damage or mutation, whereas the mutant cells proliferating in an inflammatory environment potentiates tumorigenesis.⁴⁰

Epidemiological evidence for the association between GORD and lung cancer is conflicting. Yanes et al.⁸ reported that anti-reflux surgery significantly reduced the risk of lung small cell carcinoma and squamous cell carcinoma, but not of adenocarcinoma. Another single preliminary report demonstrated the significant association between GORD and lung cancer irrespective of histology.⁴¹ Moreover, the credibility of epidemiological evidence is still under challenge due to the interference of smoking and other uncontrollable environmental variables.

Recently, Li et al. ⁴² reported that GORD was associated with an increased risk of lung cancer, which is consistent with some of our results. Compared with prior research, our research elucidated the causal impact of GORD on overall lung cancer, adenocarcinoma and squamous cell carcinoma, based on the ILCCO database, which was further validated using the FinnGen database. We also conducted multivariable MR analysis to adjust for potential confounders, and investigated the direct effect of this causal relationship. Moreover, we further explored the mediating effect of GORD in the causal pathway between smoking initiation and lung cancer. In our results, there is no causal association between GORD and small cell lung carcinoma. As a high-grade neuroendocrine carcinoma, the pathogenesis of small cell lung cancer is different from that of non-small cell lung cancer, which might be one of the reasons for the negative result.

Previous studies have found that smoking is a common risk factor for both GORD and lung cancer.^15,43 Here we revealed a direct causal association between GORD and total lung cancer which is independent of smoking traits, BMI and type 2 diabetes. However, adjusting for smoking (especially smoking frequency) greatly attenuated the causal association between GORD and lung cancer risk. This suggests that when considering smoking status, the independent impact of GORD on lung risk is relatively small. In order to investigate the relationship between smoking, GORD and lung cancer risk, we further conducted and reported the proportion of the causal effect of smoking initiation on lung cancer mediated by GORD. This finding highlights that smoking cessation intervention is a vital measure for lung cancer prevention. Moreover, GORD was elucidated as part of the smoking initiation-induced lung cancer mechanism.

Our findings provided new perspectives on the early screening and prevention of lung cancer. Patients with severe GORD may need to consider the potential risk of developing lung cancer and take timely measures to intervene. Medical therapy for GORD might be a potential strategy for lung cancer prevention, although a study demonstrated that medical therapy could not reverse pulmonary aspiration.⁴⁴ Anti-reflux surgery might be considered as the potential option to reduce the risk of lung cancer in patients with severe GORD.⁸ Furthermore, the results of the mediation analysis highlight the importance of smoking cessation in the prevention of both GORD and lung cancer.

This study has several highlights. The major merit of MR design is to minimize the biases caused by confounding and reverse causality. Second, we performed the causal estimate in two large databases to ensure consistency and thus obtained a credible causal inference. The population of our study was limited to European ancestry, which minimized the bias due to population stratification. The MR-Egger and MR-PRESSO analyses detected no evidence of pleiotropic effect, indicating that the observed causal estimates were not induced by confounding factors. Moreover, the appropriate sample size of exposures provided sufficient statistical power (>0.80) for our study.

There are also limitations in this work. We used outcomes from FinnGen and ILCCO databases to control the bias caused by sample overlap, but there is still potential overlap due to the large sample size and limiting the area to Europe.⁴⁵ Some cases of GORD were not fully identified based on physician diagnosis, and case identification through self-report and medication use may lead to potential exposure misclassification and bias in the results. In the sensitivity analysis, the inconsistent direction of ORs was observed in the analyses of adenocarcinoma and small cell lung cancer. The possible reason is that using the MR-Egger method to estimate causality could be biased by inflating the type 1 error rates.⁴⁶ Similar situations have been reported elsewhere.⁴⁷ The exaggeration of type 1 error rates may lead to the bias of ORs. Besides, using the product method to calculate the mediation effect for binary outcome is challenging, due to the non-collapsibility of the odds ratio.^48,49 A recent work simulates mediation analysis of rare binary outcomes on the log odds ratio scale, using network MR analysis. The results show the bias is small and typically would not affect the conclusions made.¹⁷ The direct causal effect of GORD adjusted by smoking duration and BMI should be interpreted with caution because of the weak instrument bias (F_TS<10). The different definitions of lung cancer might cause heterogeneity in the analysis of the ILCCO database. Nonetheless, the use of the IVW random effects method and the absence of horizontal pleiotropy suggest that our results are unlikely to be disturbed by heterogeneity. Finally, due to this study restricting the population to Europe, generalization of our results to other populations should be done prudently.

Conclusion

The present study provides credible evidence that genetically predicted GORD was associated with an increased risk of total lung cancer, lung adenocarcinoma and lung squamous cell carcinoma, based on both ILCCO and FinnGen databases, which indicated GORD might be a novel screening signal for early-stage lung cancer. The mediation effect of GORD in the association between smoking initiation and lung cancer revealed that GORD might be a part of the mechanism for smoking-induced lung cancer. However, future randomized clinical trials will be still needed to confirm this genetic inference.

Ethics approval

This study used summary-level statistics from published studies and publicly available GWASs. No ethical approval was required for this study.

Data availability

GWAS summary statistic data for GORD was obtained from the previous study.¹⁹ GWAS data for total lung cancer and pathological subtypes may be obtained from the ILCCO at ILCCO: Research projects [ILCCO: Research projects (who.int)] and FinnGen consortia at [https://www.finngen.fi/en]. GWAS data for smoking traits,²² Barrett's esophagus,²³ BMI²⁴ and type 2 diabetes²⁵ were all obtained from publicly available sources.

Supplementary data

Supplementary data are available at IJE online.

Author contributions

All authors contributed to the design of this study. Material preparation and data collection were performed by Y.L., R.Z. and L.X. Yi.L. and H.L. performed the analyses and wrote the manuscript. L.L. supervised the project.

Funding

This study was supported by the 1.3.5 Project for Disciplines of Excellence, West China Hospital, Sichuan University [ZYGD18021] (to L.L.), Key Projects of Sichuan Province (No. 2022YFS0208 for L.X.).

Acknowledgements

We thank the ILCCO at ILCCO: Home [iarc.fr], FinnGen [https://www.finngen.fi/en], GSCAN,²² GIANT²⁴ and DIAGRAM²⁵ consortia, James et al.,²⁶ Ong et al. ¹⁹ and Gharahkhani et al. ²³ for making the summary statistics involved in this study publicly available. Figure 1 was drawn partly with modified Servier Medical Art templates [http://smart.servier.com/], licensed under a Creative Common Attribution 3.0 Generic License [https://creativecommons.org/licenses/by/3.0/].

Conflict of interest

The authors declare no competing interests.

References

Author notes