Background
Idiopathic pulmonary fibrosis (IPF) is a severe and progressive fibrotic lung disease [
1]. Patients with IPF have a very poor prognosis, with a median survival of 3–5 years after diagnosis [
2,
3], and the survival rate is only 66% at 3 years after lung transplantation [
4]. The forced vital capacity (FVC), diffusing capacity of the lung for carbon monoxide (DLco), and transplant-free survival (TFS) were considered to be the key outcomes for assessing the prognosis of IPF [
5]. Evidence suggests that epithelial damage and abnormal wound repair contribute to the pathogenesis of IPF, and environmental exposure may be involved in this process, especially in patients with gastroesophageal reflux disease (GERD) [
1].
GERD encompasses a constellation of distressing symptoms and complications that arise due to the reflux of stomach contents into the esophagus [
6]. Although some studies have suggested that GERD-associated microaspiration may initiate or promote fibrosis and contribute to the disease progression of IPF [
7,
8], empirical acid suppression treatment did not slow the progression of IPF [
9,
10].
Smoking has been demonstrated to have a causal relationship with an increased susceptibility to both IPF and GERD [
11,
12]. Reynolds et al. [
13] found that GERD increased the risk of IPF using a bidirectional two-sample Mendelian randomization (MR) study, but they did not account for smoking as a confounding factor or adjust for it, potentially leading to false positive results. Therefore, when investigating the relationship between IPF and GERD, it is imperative to systematically exclude smoking-related SNPs and utilize a multivariate MR (MVMR) study to effectively adjust for smoking.
MR is a statistical technique employed to explore causal relationships between exposures and disease outcomes by utilizing genetic variants as instrumental variables (IVs) [
14]. It utilizes the principles of Mendelian inheritance to emulate the design of a randomized controlled trial, thus providing valuable insights into causality within observational studies [
14]. In comparison to conventional observational studies, MR can help mitigate bias resulting from confounding factors by utilizing genetic variants as IVs, as these variants are typically unaffected by confounders [
15].
Therefore, the objective of this study is to investigate the causal effect of GERD on the susceptibility of IPF while excluding confounding factors, such as smoking. Additionally, we also explored the causal relationship between GERD and the prognosis of IPF using the MR approach.
Data sources
The genome-wide association studies (GWAS) data for GERD were obtained from Ong et al. (129,080 cases and 473,524 control subjects) [
20]. The outcome datasets consisted of GWAS summary statistics for susceptibility (4125 cases and 20,464 control subjects) [
21], FVC (1048 cases and 4560 FVC measures) [
22], DLco (729 cases and 2795 DLco measures) [
22], and TFS (1481 cases, where endpoint events were defined as death or lung transplant) [
23] in the patients with IPF (Table
1). For replication, we extracted summary statistics for GERD (26,184 patients and 320,387 control subjects) and susceptibility of IPF (2018 patients and 373,064 control subjects) from the FinnGen biobank (Table
1) [
24]. Smoking initiation (1,232,091 patients) was obtained from the GSCAN study (Table
1) [
25]. For the FVC analysis the effect size is in terms of change in FVC in ml/year and for the DLco analysis the effect size is in terms of a change of mmol/min/kPa/year. Additionally, we collected the single nucleotide polymorphism (SNP) ID numbers (rs#) for TFS from dbSNP version 151, using version hg19 as the human reference genome [
26].
Table 1
GWAS summary statistics: source and description
GWAS of gastroesophageal reflux disease |
GERD (discovery cohort) | Ong et al. (2022) | 34,187,846 | 129,080 | 602,604 | European |
GERD (replication cohort) | FinnGen-R9 | 36,653,562 | 26,184 | 346,571 | European |
GWAS of idiopathic pulmonary fibrosis |
FVC | Allen et al. (2023) | 35,985,358 | – | 1048 | European |
DLCO | Allen et al. (2023) | 35,985,358 | – | 729 | European |
TFS | Oldham et al. (2023) | 36,780,644 | – | 1481 | European |
Susceptibility (discovery cohort) | Allen et al. (2022) | 35,688,625 | 4125 | 24,589 | European |
Susceptibility (replication cohort) | FinnGen-R9 | 36,653,562 | 2018 | 375,082 | European |
GWAS of smoking |
Smoking initiation | Liu M et al. (2019) | 30643251 | 1,232,091 | – | European |
Instrument selection
The genetic instruments were derived from a large genetic atlas of GERD [
20,
24]. SNPs that reached a significance level of
P < 5 × 10
−8 (
P < 1 × 10
−6 for replication cohort) were clumped for independence based on linkage disequilibrium (r
2 < 0.001 within 10,000 kb), using the European reference panel from the 1000 Genome Project [
27]. In cases where there were limited accessible SNPs for the outcomes, proxy SNPs with a high degree of linkage disequilibrium (r
2 > 0.8) were employed. The effects of the SNPs on exposures and outcomes were then harmonized to ensure that the beta values were assigned to the same alleles. Outliers were detected using the MR-PRESSO method (defined as those with
P > 0.05) [
28], but no outliers were found in the data. Subsequently, we manually screened and removed SNPs related to confounding factors and outcomes using the PhenoScanner database (
P-value < 1 × 10
−5, r
2 = 0.8, Proxies = EUR, Build = 37) [
29,
30]. The results of this screening are presented in Table S
1. For the susceptibility to IPF, potential confounders included pollution, occupation, animal antigens, and viral exposures [
1,
31]. The remaining SNPs were used to perform the MR study. For FVC, DLco, and TFS in patients with IPF, potential confounders included pulmonary function testing, pollution, occupation, and pulmonary infection (excluded previous exposures) [
1,
31].
Testing instrument strength
To assess the instrument strength for GERD, we employed two parameters: the proportion of variance (R
2) and the
F-statistic. The R
2 was calculated using the formula R
2 = 2 × EAF × (1–EAF) ×
β2 [
32], while the
F-statistic was computed as
F =
β2 / SE
2 [
33]. The
F-statistic is considered a measure of instrument strength, and a value greater than 10 indicates a sufficiently strong instrument [
34]. All
F-statistics of the SNPs in our study are ≥30, indicating a robust strength of genetic instruments (Table S
2).
Sensitivity analysis
The primary MR analysis was conducted using the inverse-variance weighted (IVW) method, which provides an unbiased estimate in the absence of horizontal pleiotropy and heterogeneity [
35]. Additionally, we performed other methods, including MR-Egger [
36], weighted median [
37], simple mode [
38], and weighted mode under different assumptions [
38]. However, the IVW method was preferentially applied when no heterogeneity and horizontal pleiotropy were present. To assess for heterogeneity and horizontal pleiotropy, we performed various tests, including the MR-Egger intercept test [
39], Cochran’s Q test [
40], and leave-one-out analyses [
41]. Lastly, we performed the MR-Steiger directionality test to evaluate the correct direction of causality in the presence of a causal association [
42].
Multivariable Mendelian randomization analysis
Since the effects of GERD on the susceptibility of IPF may also be influenced by smoking [
11], a MVMR analysis were conducted. In this analysis, GERD and smoking initiation was used as exposure variables to account for potential confounding by smoking. Two types of MVMR analyses were performed, namely multivariable IVW regression [
43] and MVMR-Egger regression [
44], as additional sensitivity analyses. In the MVMR approach, all genetic instruments, eliminated duplicate SNPs, and excluded correlated SNPs (r
2 ≥ 0.001) based on the minimum
P-value for genetic association with each trait were combined. Subsequently, the associations of the remaining SNPs with both the exposure and outcome variables and fitted multivariable models were extracted.
Statistical analysis
All statistical analyses were performed using the “TwoSampleMR” package [
45], “MRPRESSO” package [
28], and “MVMR” package [
46] in R (version 4.2.2) with RStudio (version 2022.07.2 Build 576). The threshold for statistical significance was set at
P-values below 0.05.
Discussion
We conducted an investigation to explore the causal associations between GERD and the susceptibility and prognosis of IPF. In our study, after adjusting for smoking, we found no evidence that GERD increases susceptibility to IPF. Furthermore, our genetic evidence demonstrates no causal impact of GERD on FVC, DLco, and TFS in patients with IPF.
Similar to previous studies, initially we did not exclude SNPs associated with confounding factors, primarily those related to smoking (e.g., rs215614, rs12357321, rs329122, rs324769, and rs12967855). Consequently, we initially arrived at a similar conclusion to previous studies, suggesting a causal effect of GERD on susceptibility to IPF [
13,
47]. However, Zhu, J et al. performed a multivariable MR and demonstrated that there is no causal effect of GERD on susceptibility to IPF after adjusting for smoking in a replicate cohort [
47]. Smoking has been shown to have a causal relationship with an increased susceptibility to both IPF and GERD [
11,
12]. Therefore, in the present study, smoking was included as potential confounder, and the results indicated no causal effect of GERD on susceptibility to IPF after adjusting for smoking.
Previous studies have suggested that GERD is an important risk factor for IPF, as gastroesophageal reflux has been reported in 76–94% of patients with IPF [
48,
49]. Therefore, recent studies have explored the potential role of antacid medication in halting the progression of IPF [
7,
8,
10,
50‐
54]. However, the majority of research did not yield the expected results [
10,
50,
51,
53,
54], and two meta-analyses demonstrated that antacid medication had no statistically significant effect on arresting the disease progression of IPF [
9,
10]. Therefore, guidelines do not recommend antacid medication and other interventions for improving respiratory outcomes in IPF [
5]. Our study revealed no causal effect of GERD on FVC, DLco, and TFS in patients with IPF, providing some support for the recommendations outlined in the guidelines.
Anti-reflux surgery is designed to prevent both acid and non-acid refluxate. A prospective, randomized controlled trial was conducted to compare the decline in FVC between patients with IPF who underwent the surgery and those who did not. The study included 58 patients, and it was observed that the surgical group experienced a slower decline in FVC over a 48-week period (0.05 L) compared to the non-surgical group (0.13 L), but the difference did not reach statistical significance (
P = 0.28) [
55]. After conducting a recent meta-analysis of case-control studies, it has been suggested that the association between IPF and GERD may not stem from a direct causal relationship. Instead, it could be influenced by confounding factors, particularly smoking [
56]. Combining our results, the recommendation to universally treat GERD in patients with IPF is further called into question.
The greatest strength of this study is its consideration of smoking and smoking-related SNPs in the MR analysis to examine the causal relationship between GERD and the susceptibility of IPF. The demonstrated absence of a causal relationship is attributed to the adjustment for smoking. Additionally, our findings suggest that there is no causal effect of GERD on FVC, DLco, or TFS in IPF. These results provide insights into the treatment options for IPF, indicating that the administration of universally recommended GERD therapy in the patients with IPF may not be supported.
This study has several limitations that should be acknowledged. First, the relatively small sample size in both the IPF GWAS (discovery and replication cohorts) and GERD GWAS (replication cohort) limited the precision of population parameter estimates, leading to larger standard errors. However, it’s noteworthy that these sample sizes were the largest ever used for these specific research questions. Second, we observed a relatively small number of significantly associated genetic loci for GERD within the replication cohort. The limited number of patients in the GERD replication cohort could have played a role in this restriction of significant loci, possibly leading to potential false-negative findings. Third, further investigations among populations with diverse racial and ethnic backgrounds are necessary, as the GWAS predominantly includes individuals of European ancestry. Therefore, caution must be exercised when generalizing the results to other populations.
Conclusions
This study employed large GWAS data for an MR investigation into the relationship between GERD and susceptibility, FVC, DLco, and TFS of IPF. Our findings suggest that the association of GERD with susceptibility to IPF may not be directly causal and could be explained by confounding factors, particularly smoking. Furthermore, no observed causal effect of GERD on FVC, DLco, and TFS of IPF was found.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.