The ratio of omega-6 fatty acids to omega-3 fatty acids mediating the effect of gastroesophageal reflux disease on pancreatic cancer risk: a two-step mediation Mendelian randomisation study Authors: Jiqi Ouyang, Runshun Zhang, Wenliang Lv, Yanping Lu Published: 2025 ABSTRACT AbstractIntroductionPancreatic cancer (PC) is the third most common cause of cancer-related deaths worldwide, with the 5-year survival rate for patients with PC in developed countries being only 9–10%. Gastroesophageal reflux disease (GERD) is known to be a high-risk factor for several cancers. However, the relationship between GERD and PC remains controversial. Therefore, this study aimed to investigate the causal association between GERD and the risk of PC. Additionally, we sought to validate the role of omega-6/omega-3 polyunsaturated fatty acids (n-6/n-3 PUFA) as mediators in this pathway by calculating their mediating effect.MethodsFirst, a two-step, two-sample Mendelian randomisation (TSMR) framework was established to test the causal relationship between GERD and PC and to determine whether n-6/n-3 PUFA could act as a mediator of the causal pathway. Subsequently, using multivariate Mendelian randomisation (MVMR), the proportion of mediators mediated by n-6/n-3 PUFA was calculated using the product of coefficients approach.ResultsThe two-step TSMR results showed that GERD was positively associated with PC (n = 27 SNPs, OR 2.07 95% CI 1.27–3.39, P = 0.003 β = 0.72). Additionally, GERD was positively associated with n-6/n-3 PUFAs (n = 59 SNPs, OR 1.06, 95% CI 1.02–1.11 P = 0.001, β1 = 0.06), which are positively associated with PC (n = 28 SNPs, OR 1.23, 95% CI 1.04–1.46, P = 0.02, β = 0.20). The MRMV results demonstrated that after controlling for GERD, the n-6/n-3 PUFA had an impact on PC (n = 17 SNPs, se = 0.07, P = 0.004, β2 = 0.21). Furthermore, the causal effect of GERD on PC was partially mediated through the n-6/n-3 PUFA (n = 31 SNPs, se = 0.20, P = 0.03, β3 = 0.42) with a 2.9% mediating effect.ConclusionThis study provides insights into the pathological processes that link GERD to PC. In addition, it offers a novel approach to PC prevention and early detection. In clinical practice, early screening and drug usage control in patients with GERD must be prioritised. Furthermore, dietary supplementation with n-3 PUFAs and proper n-6 PUFA reduction may minimise the prevalence of PC.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12876-025-04445-2.Keywords: Gastroesophageal reflux disease, Polyunsaturated fatty acids, Pancreatic cancer, Mendelian randomization, Prevention Introduction Over the past two decades, the global incidence of pancreatic cancer (PC) has shown a continuous upward trend [1], largely driven by population aging, lifestyle factors such as obesity and smoking, and improvements in diagnostic techniques [2]. In 2022, approximately 510,000 new cases of PC were diagnosed, making it the 12th most common cancer worldwide. With a mortality rate of 4.8%, PC ranks as the sixth leading cause of cancer-related deaths globally [3], posing a serious threat to human health. Due to delayed diagnosis and limited therapeutic options, the five-year survival rate remains dismally low, ranging from 2% to 9% [4]. Consequently, identifying high-risk populations for targeted screening, reducing modifiable risk factors, and optimizing early prevention strategies are critical steps toward improving clinical outcomes [5]. Gastroesophageal reflux disease (GERD) is a highly prevalent disorder of gastrointestinal motility, with a worldwide prevalence of approximately 13% [6]. Many studies have confirmed that GERD is a high-risk factor for cancers such as those of the head and neck [7], lung [8], liver, and colorectal [9], largely due to cellular DNA damage from reflux (gastric acid, bile acids, pepsin, microbes, etc.) [10]. However, the relationship between GERD and PC is more complex and remains controversial. Most studies suggest that GERD may be associated with an increased risk of PC, potentially due to the presence of bile acids (BAs) in the refluxate and the long-term use of proton pump inhibitors (PPIs). Evidence indicates that BAs may contribute to pancreatic carcinogenesis by disrupting calcium and iron homeostasis [11], interacting with the gut microbiota [12], stimulating the secretion of glucagon-like peptide-1 (GLP-1) [13], and accelerating epithelial cell turnover. Kearns et al. conducted nested case-control and cohort studies demonstrating that prolonged PPI use increases the risk of PC [14]. This finding is supported by a meta-analysis showing a significant association between PPI use and PC, with a pooled odds ratio (OR) of 1.698 (95% CI: 1.200–2.402, P = 0.003) [15]. The underlying mechanisms may involve PPI-induced hypergastrinemia [16], hypercholesterolemia [17], alterations in the gut microbiome leading to increased nitrosamine production [18], and impaired vitamin B12 absorption [19]. However, conflicting data exist. A Korean cohort study reported a potential protective effect of GERD against PC (adjusted hazard ratio = 0.43, 95% CI: 0.24–0.76) [20], which may be attributed to the acid-dominant reflux pattern commonly seen in Asian populations and potential diagnostic bias. Acidic reflux may reduce duodenal pH, inhibit premature activation of pancreatic zymogens, and consequently lower the risk of pancreatic autodigestion and injury.These discrepancies may result from differences in study populations, designs, and residual confounding that is difficult to eliminate in observational studies. Therefore, further research is needed to clarify the potential role of GERD and PPI use in PC risk and to elucidate the underlying pathophysiological mechanisms. Omega-3 and omega-6 polyunsaturated fatty acids (n-3/6 PUFAs) are common circulating polyunsaturated fatty acids in the human body that are absorbed in the digestive tract through food intake and metabolism [21]. Some researchers [22] have found that n-3 PUFAs protect against brain [23], breast [24], prostate [25], and colorectal cancers [26, 27]. Morales et al. [28] found that Kirsten rat sarcoma (KRAS) mutations in pancreatic cancer are associated with low intake of n-3 LC-PUFAs. Furthermore, Ding et al. [29] found that an ω3-rich diet prevented early PC by inhibiting proliferation associated with a reduction in phosphorylated AKT, while an ω6-rich diet accelerated tumour formation. Studies on the relationship between GERD and PUFAs are lacking, and some studies have found that Western dietary intake high in ω 6 can induce GERD symptoms [30]. Other studies have shown that an imbalance in n-3/n-6 PUFA intake leads to dysbiosis of the gastrointestinal microflora [31] and increases the risk of inflammation [32]. However, observational studies have several limitations [33], including their small sample sizes, confounding factors, and the presence of reverse causality. Consequently, the bias of standard epidemiological methodologies [34] hinders our understanding of the mediating effects of GERD as a risk factor in PC exposure and their mediating effects. The Mendelian randomisation (MR) approach, which uses genetic variants linked to exposures of interest as instrumental variables to investigate their causal influence on outcomes, is a potential alternative to observational studies [35]. As genetic variations are randomly assigned and detected prior to disease onset, the inadequacy of observational research can be compensated. Multivariate MR (MVMR) ensures a more accurate estimation of the causal association of numerous risk factors and various health outcomes [36]. When combined with two-step, two-sample MR (TSMR), it can facilitate the study of the mediation pathway while significantly decreasing the bias inherent in traditional multivariate approaches [37]. We searched for previous MR studies that confirmed that inflammatory bowel disease [38], type 2 diabetes mellitus [39], and obesity [40] can increase the risk of PC. However, no study has been conducted on the MR analysis of GERD in PC. Ghoneim et al. [41] performed MR analysis and found no significant causal relationship between n-6 LC-PUFA and PC risk. However, Haycock et al. [42] found that n-6 PUFA significantly increased the risk of colon and oesophageal cancer through MR analysis of European and East Asian ancestry. Based on the above observational and MR studies, we hypothesised that there is a causal relationship between GERD and PC and that n6/n-3 PUFAs act as potential mediators in the causal pathway. Our study aimed to validate the causal pathway using TSMR and MVMR methods. Materials and methods Study design This study employed a two-step, two-sample MVMR approach using pooled genome-wide association analysis (GWAS) data [43] sourced from publicly available datasets in the Integrative Epidemiology Unit database (https://gwas.mrcieu.ac.uk). These datasets encompassed the GWAS for GERD, n-6/n-3 PUFA, and PC. TSMR required the application of distinct datasets to establish connections between gene risk factors (GERD) and gene outcomes (PC). We divided the study into two parts: examination and the computation of the mediation effect. Initially, we employed a two-step TSMR analysis to assess the potential mediating effect of n-6/n-3 on the causal association between exposure and outcome. In the first step, we tested the causative relationship between GERD and n-6/n-3 PUFA and PC. Second, we tested the causal relationship between n-6/n-3 PUFA and PC. Subsequently, we used MVMR [44] to concurrently evaluate the effects of GERD and n-6/n-3 PUFA exposure on PC and to estimate the causal effect of exposure after correcting for others. To examine the extent to which the mediated effect contributed to the risk of GERD in PC, we used the product of the coefficients approach. We employed genetic variants such as IVs for MR analysis. Single nucleotide polymorphisms (SNPs) were considered valid IVs for GERD and n-6/n-3 PUFA if they meet the three key assumptions of MR. They must (1) be strongly related to the exposure of interest, (2) be linked to any confounding variables in the association between genetic variations and the desired outcome, and (3) only influence the outcome of interest through their relationship with the exposure of interest, without the influence of alternative pathways [45, 46]. (Fig. 1) Design of the Mendelian randomization study This MR study followed the recommendations of the STROBE-MR Guidelines, as detailed in the Supplementary Checklist. Data sources Instrumental variables for exposure Data on genetic association estimates for GERD were extracted from the most recent GERD GWAS meta-analysis and multi-trait GWAS (MTAG) analysis [47], which synthesized data from multiple large cohorts. This comprehensive analysis, building upon prior work [48], included 2,050,457 unique individuals—comprising 602,604 participants used in the primary analysis and 1,447,853 additional samples from 23andMe for replication and extension.The participating cohorts, as detailed in the supplementary material of Ong et al. encompassed a diverse population. The primary ancestry was European, stemming from cohorts such as the UK Biobank (UKB) (N = 450,210), FinnGen (N = 250,718), 23andMe (N = 1,027,515), QSkin (N = 38,206), among others. Additionally, East Asian ancestry populations were included from Biobank Japan (N = 188,432) and KOREF (N = 8,024). Most participants were aged 18–99 + years, with average ages typically between 40 and 65 years across cohorts. The sex distribution was consistent across major contributing cohorts, with females comprising approximately 53.0% to 56.0% of participants. GERD cases were defined based on a combination of self-reported symptoms, ICD-10 diagnostic codes of K21, and the use of GERD-related medications. Controls were individuals without any history of GERD diagnosis or related medication use. Participants with conflicting information between self-reported GERD status and medication records (n = 473,524) were excluded from the analysis. Detailed information on the GWAS can be found in the original publications. Instrumental variables for potential mediator Genetic association estimates for circulating PUFAs, considered potential mediators, were obtained from a genome-wide association study (GWAS) conducted by Borges et al. [49]. This study utilized data from the UK Biobank under projects 30,418 and 15,825 (REC reference: 11/NW/0382). The GWAS included 114,999 individuals of European ancestry, all recruited from the UKB. Circulating PUFA levels were quantified using targeted high-throughput nuclear magnetic resonance (NMR) spectroscopy. While the overall UKB cohort was recruited between ages 40–69, the PUFA GWAS subset had a median age of 58 years (IQR: 50–63), with approximately 54% female participants. Genetic analyses were conducted using the FUMA GWAS platform to identify independently significant SNPs. Summary-level genetic association data were subsequently deposited in the IEU GWAS database to support Mendelian randomization analyses. Full methodological details are available in the original publication. Instrumental variables for study outcome Genetic association estimates for PC were obtained from a large-scale cross-population genome-wide association meta-analysis conducted by Sakaue et al. [50], incorporating data from the BioBank Japan, UKB, and FinnGen cohorts. The study included 476,245 individuals, comprising 1,196 PC cases and 475,049 controls. PC was primarily defined using ICD-10 code C25, supplemented by electronic health and medication records. The study population was ancestrally diverse, with approximately 33.5% East Asian and 66.5% European participants. Across cohorts, participants were predominantly middle-aged to older adults, with average ages ranging from 52 to 63 years, and a balanced sex distribution, with 46% to 56% females. Detailed methodologies and phenotype definitions are available in the original publication and its supplementary materials. Table 1Characteristics (summary) of GWAS data for exposure, potential mediators, and outcomeCategoryTraitsSample sizeNo. of SNPsPMIDAuthor(Year)Exposure[47]GERD602,604(ncase:129,080;ncontrol:473,524)2,320,78134187846Ong JS et al. (2021)Potential mediators [49]n-6/n-3 PUFA114,99912,321,87535692035Borges CM et al. (2020)Outcome [50]PC476,245(ncase:1,196; ncontrol:475,049)24,195,22934594039Sakaue S et al. (2021)Abbreviations: GWAS genome-wide association studies, SNP single-nucleotide polymorphismThe genetic variants used as instrumental variables for GERD, PC, and n-6/n-3 PUFA risk factors are shown in Supplementary Information Characteristics (summary) of GWAS data for exposure, potential mediators, and outcome 602,604 (ncase:129,080; ncontrol:473,524) Abbreviations: GWAS genome-wide association studies, SNP single-nucleotide polymorphism The genetic variants used as instrumental variables for GERD, PC, and n-6/n-3 PUFA risk factors are shown in Supplementary Information Instrumental variables selection and evaluation SNPs related to GERD, n-6/n-3 PUFA, and PC in MR studies were acquired from GWAS datasets at a (P<5 × 10− 8) [51]. Next, separate SNPs for exposure were obtained by linkage disequilibrium clustering with an r2 < 0.001 threshold and allele distance >10,000 kb [52]. The screened SNP-associated phenotypes were then searched using the Human Genotype-Phenotype Association Database (http://www.phenoscanner.medschl.cam.ac.uk) [53] to eliminate confounding SNPs with strong correlations to outcome that have been shown to affect outcomes. The effects of SNPs on exposure and outcome were then coordinated to ensure that β-values were labelled as identical alleles. After data harmonisation, we removed the palindromic SNPs. Finally, the outlier pleiotropic SNPs were removed by a heterogeneity test using RadialMR [54] (p<0.05). Radial MR results are in the Supplementary Material Fig. 1–1, 1–2, and 1–3. Therefore, these carefully selected SNPs were used as IVs in subsequent MR analyses. We used F-statistics to evaluate the strength of the association between genetic tools and phenotypes. The proportion of phenotypic variety their genetic variant tool, R2, was calculated using the formula [55]: where SD and β are the standard deviation and β coefficient for effect size, respectively and MAF is the minor allele frequency for each SNP. The F-statistic was used to evaluate the strength of the effect of SNPs on GERD and n-6/n-3 PUFA in a single TSMR while calculating the conditional F-statistic in the MVMR using the formula: where N is the sample size and the total number of SNPs selected for the MR analysis. An F-statistic > 10 suggests a sufficient (adequate) instrument strength, while ≤ 10 implies a weak instrument strength. In this study, we retained only strong instrumental variables. MR analysis TSMR We aligned and merged alleles based on the information extracted from the screened GERD-related instrumental variables for n-6/n-3 PUFA and PC outcomes. We used five different analytical approaches for MR analysis: Inverse Variance Weighted (IVW), MR-Egger, weighted median (WM), Simple Mode, and Weighted Mode. Among these, the IVW is the predominant method of analysis [52]. This is because it meta-analyses the Wald ratios of the genetic causal effects of each SNP, assuming that all SNPs are valid IVs and that there is no indication of directional pleiotropy. Therefore, this method is considered to provide estimates with the highest power and precision [56, 57]. Weighted according to the precision of the variant-outcome genetic association estimates, MR-Egger regresses the variant-exposure genetic association estimates against the variant-outcome genetic association estimates [58]. The MR estimate is denoted by the slope of the regression line, and the directional pleiotropy can be determined by examining whether the intercept deviates from zero. The WM method selects the median value by ranking the MR estimates of individual SNPs in order of precision-weighted magnitude, which complements the MR-Egger method [59]. Comparing the results from all three methods increased our confidence and the reliability of the findings. MVMR To estimate the mediating effect of n-6/n-3 PUFA in the genetically predicted effect of GERD on PC risk, we performed MVMR analyses. In these analyses, the total effect of each exposure is decomposed into direct and indirect effects to estimate mediation and the proportion of the effect of the main exposure of interest on the mediated outcomes [36]. Genetic association estimates for GERD and n-6/n-3 PUFA were regressed and weighted according to the precision of the genetic association estimates, with the intercept fixed at zero. When performing MVMR analyses, we assessed the proportion of mediating effects for all individuals and combined the mediating effects by dividing the indirect effect by the overall effect. Due to the inherent nature of MVMR to adjust for each other, multiple tests were not used [60]. Sensitivity analysis We conducted heterogeneity, multidimensionality, and sensitivity analyses to assess potential biases from horizontal pleiotropy in the study’s MR analyses. First, we calculated Cochran’s Q statistic to quantitatively evaluate the level of heterogeneity of individual SNPs [61]. Using the IVW, MR-Egger Intercept, WM, and Pleiotropy RESidual Sum and Outlier (MR-PRESSO) methods [62], we performed a series of sensitivity analyses. By deleting one or more pleiotropic outlying SNPs and conducting MR analyses, we generated outlier-adjusted estimates of causal correlations using the MR-PRESSO approach [63]. Subsequently, we ran a ‘Leave-one-out’ sensitivity analysis, eliminating a different SNP in each iteration to see if the total MR estimations were altered by the removed SNPs [64]. If any SNPs significantly altered MR estimates, they were eliminated, and MR analyses were re-run. Statistical tools All statistical analyses and visualisations were performed using R statistical software (version 4.3.3) with the ‘TwoSampleMR’, ‘mr_pleiotropy_test’, ‘mv_lasso_feature_selection’ and ‘mv_multiple’ functions in TwoSampleMR package, ‘RadialMR’ and ‘forestplot’ Packages. Results Selection and validation of instrumental variables for MR analysis We selected 80 independent genome-wide significant (P < 5 × 10− 8) SNPs that were not shared with variants instrumenting n-6/n-3 PUFA, which independently contributed to GERD as genetic instruments, from the most recent GWAS. The PC outcome extracted 80 SNPs from GERD. Following Penoscanner searches, 42 SNPs (rs1011407, rs10133111, rs11762636, and rs12357321 et.) associated with body mass Index [65] and C-reactive protein levels [40] were deleted. Subsequently, 38 SNPs were left after deletion. GWAS was used to retrieve PC information. The association between the 38 SNPs and the outcome was determined by analysing the study findings and combining the exposure data and the outcome to show the relationship between the 38 instrumental factors and the outcome and exposure. Nine palindromic sequences (rs10837002, rs12598916, rs1592757, and rs2016933, among others) were identified; 2 outlier SNPs(rs1431196, rs9373363) were examined by the IVW method of RadialMR. 27 instrumental variables were considered the final variables for associating GERD with PC. Based on previous studies, no significant confounders associated with n-6/n-3 PUFA were observed among the 80 GERD-related SNPs recovered by the GWAS for exposure and mediator instrumental variable screening. 79 SNPs were extracted with n-6/n-3 PUFA with coordinated. Fourteen palindromic sequences (rs10837002, rs12598916, rs1592757, and rs2016933) were excluded. 6 outlier SNPs(rs12357321, rs12997558, rs1716171, rs2782641, rs4382592, rs9372625) were examined by the IVW method of RadialMR. Eventually, 59 SNPs were identified as instrumental factors in GERD related to n-6/n-3 PUFA. Using the same steps, we integrated the mediator and outcome instrumental variable datasets. The GWAS database was searched for 41 n6/n-3 PUFA-associated genetic variant SNPs that matched all the relevant assumptions. The PC ending extracted 40 SNPs from n-6/n-3 PUFA because one unidentified SNP was eliminated. 40 SNPs were extracted with PC with coordinated. Four SNPs (rs10733306, rs112875651, rs7916868, and rs1260326) associated with BMI and CRP levels were eliminated. Of the 36 SNPs identified, 4 palindromic SNPs (rs12976395, rs4860987, rs72789541, and rs964184) were revealed. From the 36 SNPs revealed, 4 palindromic SNPs (rs12976395, rs4860987, rs72789541, rs964184), and 1 duplicate SNPs are removed. 3 outlier SNPs (rs2232143, rs73109460, rs8074191) were examined by the IVW method of RadialMR. The n6/n-3 PUFA to PC ratio was used as the final instrumental variable, and 28 SNPs were retained. The F-statistics for all regressions of GERD and the potential mediators on their genetic variant instrumental variables were all greater than 10, indicating that the instrumental variables we used were all strong (Supplementary material raw data). 3.2 Results of the Two-Step Mendelian Randomization Analysis. Using a TSMR approach, we systematically evaluated three causal pathways: (1) GERD → PC, (2) GERD → n-6/n-3 PUFA ratio, and (3) n-6/n-3 PUFA ratio → PC. Across all three steps, the IVW method consistently revealed statistically significant positive associations, with direction and magnitude of effect supported by multiple complementary MR methods, reinforcing the robustness of the findings. In the first step, genetically predicted GERD was significantly associated with increased risk of PC (OR = 2.07, 95% CI: 1.27–3.39, P = 0.003; β = 0.72). This finding was strongly corroborated by the Weighted Median method (OR = 2.01, 95% CI: 1.06–3.79, P = 0.03; β = 0.70), and supported in direction and effect size by the MR-Egger, Weighted Mode, and Simple Mode methods, though the latter three showed wider confidence intervals and non-significant p-values. Importantly, the MR-Egger intercept test indicated no evidence of directional pleiotropy. In the second step, a significant positive causal effect of GERD on the n-6/n-3 PUFA ratio was observed (IVW: OR = 1.06, 95% CI: 1.02–1.10, P = 0.001; β1 = 0.06). This result was reinforced by the Weighted Median (OR = 1.07, 95% CI: 1.01–1.13, P = 0.01; β = 0.07), while the MR-Egger, Weighted Mode, and Simple Mode approaches yielded concordant effect directions, albeit with wider intervals and non-significant results. Again, no evidence of horizontal pleiotropy was detected. In the third step, the n-6/n-3 PUFA ratio was significantly associated with higher PC risk (IVW: OR = 1.23, 95% CI: 1.04–1.46, P = 0.02; β = 0.20). This was further validated by both the Weighted Median (OR = 1.26, 95% CI: 1.05–1.51, P = 0.01; β = 0.23) and Weighted Mode (OR = 1.27, 95% CI: 1.06–1.53, P = 0.02; β = 0.24) methods. While MR-Egger and Simple Mode produced non-significant results, both showed effect directions consistent with the IVW estimate, and no significant pleiotropy was detected via MR-Egger intercept. Collectively, these findings demonstrate consistent and directionally concordant causal associations across five MR methods, particularly supporting the primary IVW results. The robustness of our conclusions is further strengthened by the lack of significant heterogeneity or directional pleiotropy across all analyses. The results of the two-step TSMR are summarised in Table 2. The related forest and scatter plots are shown in Figs. 2 and 3, respectively. Table 2The results of Two-step TSMRMethodNumber of SNPsbetaOR (95% CI)PGERD on PCIVW270.722.07(1.27–3.39)0.003MR Egger271.464.32(0.06–336.20)0.52Weighted median270.702.01(1.06–3.79)0.03Weighted mode271.052.84(0.85–9.56)0.10Simple mode270.982.66(0.81–8.77)0.12GERD on n-6/n-3 PUFAIVW590.061.06(1.02–1.10)0.001MR Egger590.101.10(0.90–1.35)0.34Weighted median590.071.07(1.01–1.13)0.01Weighted mode590.061.07(0.95–1.19)0.28Simple mode590.061.06(0.94–1.20)0.33n-6/n-3 PUFA on PCIVW280.201.23(1.04–1.46)0.02MR Egger280.141.15(0.92–1.43)0.24Weighted median280.231.26(1.05–1.51)0.01Weighted mode280.241.27(1.06–1.53)0.02Simple mode280.571.78(0.82–3.82)0.15Abbreviations: TSMR = Two sample mendelian randomization, IVW = Inverse Variance Weighted The results of Two-step TSMR Abbreviations: TSMR = Two sample mendelian randomization, IVW = Inverse Variance Weighted Two-step two-sample Mendelian randomized forest plot A Scatter plot of SNP effects on relative GERD versus PC. B Scatter plot of SNP effects on relative GERD versus n-6/n-3 PUFA. C Scatter plot of SNP effects on relative n-6/n-3 PUFA versus PC. The slope of each line corresponding to the estimated MR effect per method Sensitivity analyses and assessment of Pleiotropy and heterogeneity Sensitivity analysis of GERD in PC showed no significant heterogeneity in the estimates calculated using the IVW method (Q = 6.79, P = 1.00). The MR-Egger intercept test did not demonstrate significant multiplicity in the estimates (Egger intercept = −0.02, SE = 6.73%, P = 0.74). Our “Leave-one-out” sensitivity analyses showed in supplementary Fig. 2 − 1, suggesting robustness in our results. The MR-PRESSO analyses did not reveal any significant outliers in the exposure instrument (P = 0.98). Sensitivity analyses of GERD on n-6/n-3 PUFA showed no heterogeneity in the estimates calculated using the IVW method (Q = 63.6, P = 0.29). The results of the MR-Egger intercept (Egger intercept = −1.0 × 10 − 3, SE = 0.34%, P = 0.71) suggest that the instrumental variables did not significantly affect the mediator in ways other than GERD. Our “Leave-one-out” sensitivity analyses showed in supplementary Fig. 2–2, suggesting robustness in our results. MR-PRESSO analyses (P = 0.14) did not find significant outliers for the exposure instrument. No significant variation in the estimates produced by the IVW approach was noted, as seen by the n-6/n-3 PUFA in the PC sensitivity analysis (Q = 29.04, P = 0.35). The MR-Egger intercept test results (Egger intercept = 0.01, SE = 1.32%, P = 0.31) showed that the outcome PC was not affected by any confounders. Our “Leave-one-out” sensitivity analyses showed in supplementary Fig. 2–3, suggesting robustness in our results. The findings of the MR-PRESSO studies (P = 0.41) did not find significant outliers for the exposure instrument. The heterogeneity and horizontal pleiotropy results are summarised in Table 3. Table 3The results of the heterogeneity and horizontal PleiotropyCausal pathCochran’s Q heterogeneity testMR Egger pleiotropy test P valueOutliers (Yes/No)P Value before MRPRESSO correctionAfter correction and eliminationGERD on PC6.790.74No0.98-GERD on n-6/n-3 PUFA63.600.71No0.14-n6/n3 PUFA s on PC29.040.31No0.10- The results of the heterogeneity and horizontal Pleiotropy MVMR analysis and mediation effect of n-6/n-3 PUFA The results of the MVMR analysis indicate that n-6/n-3 PUFA have a negative impact on PC, even after accounting for the influence of GERD (n = 17 SNPs, se = 0.07, P = 4.0 × 10− 3, β2 = 0.21). After accounting for mediators, we observed that the positive causal association between GERD and PC remained significant (n = 31 SNPs, se = 0.20, P = 0.03, β3 = 0.42). The effect of a 2.9% change was modulated by n-6/n-3 PUFA (Table 4; Fig. 4). Table 4The mediation effect of GERD on PC via n-6/n-3 PUFAMediatorTotal effectDirect effect ADirect effect BMediation effect P Mediatedproportion (%)ββ1β2β3n6/n3 PUFA0.730.060.210.420.0042.9 The mediation effect of GERD on PC via n-6/n-3 PUFA Mediation analysis of the causal effect of GERD on PC via n6/n3 PUFA potential mediators ‘Total effect’ indicates the effect of relative GERD on PC, ‘direct effect A’ indicates the effect of GERD on n6/n3 PUFA, ‘Direct effect B’ indicates the effect control of GERD of n6/n3 PUFA on PC, ‘Mediation effect’ indicates the effect of relative GERD on PC through n6/n3 PUFA.β, β1, were derived by IVW of TSMR; β2, β3 were derived by using the delta method of MVMR. P < 0.05 was considered significant. Discussion To the best of our knowledge, this is the first study to use MR to explore the extent to which it influences the causal relationship between GERD and PC. Our study, which used a mediated MR design with a larger sample size and greater genetic variation, showed that GERD increased risk of PC, with 2.9% of this effect being mediated by n-6/n-3 PUFA. Even after adjusting for mediators, GERD increased the risk of PC. These results imply that PC may occur more frequently in patients with GERD and that reducing n-6/n-3 PUFA may help to prevent the early development of PC. All sensitivity analyses demonstrate the resilience of the results. We combined the conclusions drawn from this study with those of previous studies to elucidate the pathological mechanisms by which GERD causes PC. The primary mechanism by which GERD increases the risk of cancers such as esophageal and lung cancer involves the reflux of gastric acid, bile acids, pepsin, and other substances, which can damage cellular DNA in affected organs. This process alters cell proliferation and induces resistance to apoptosis, which activates oncogenic pathways [66]. BAs dysregulation due to acid reflux in GERD may be one of the mechanisms for PC [67]. The imbalance in bile secretion leads to elevated biliary pressure sufficient to cause reflux into the pancreatic ducts, thereby exacerbating cellular injury and inducing recurrent inflammatory responses. This chronic insult promotes the dedifferentiation of pancreatic acinar cells into ductal progenitor-like cells, ultimately contributing to malignant transformation [68]. BAs are protonated in highly acidic environments [69], which facilitates their transmembrane passage and enhances cytotoxicity. They induce cell cycle arrest and apoptosis through a protein kinase C-dependent pathway [70]. Studies have found that BAs induce the expression of EGFR and HER2 and activate downstream pathways that promote cancer progression [71]. l Overexpression of HER2 in pancreatic cancer is often associated with poorer survival outcomes [72]. On the other hand, the increased risk of PC associated with GERD may be partly attributed to the use of PPIs, a standard treatment for GERD. Several studies have reported a positive association between PPI use and PC risk. A meta-analysis of 25 studies involving over 2.7 million individuals found that long-term PPI use was linked to an increased risk of hepatobiliary and pancreatic cancers (OR: 1.61; 95% CI: 1.23–2.11) [73]. Another meta-analysis including nearly 950,000 participants reported a pooled OR of 1.70 (95% CI: 1.20–2.40) for PPI-associated PC, regardless of study region, design, or PPI type [74]. Case-control and cohort studies from France [75] and South Korea [76] further confirmed that higher cumulative PPI doses were associated with increased PC risk. Mechanistically, PPIs disrupt gut microbiota by altering gastrointestinal pH [77], promoting chronic inflammation. Dysregulation of the gastrointestinal flora leads to an increase in lipopolysaccharide-producing bacteria, which are involved in PC progression by triggering an immune response via Toll-like receptors, thus implicating n-6/n-3 PUFA. They also irreversibly inhibit gastric H⁺/K⁺-ATPase, leading to compensatory hypergastrinemia. Elevated gastrin levels may stimulate PC cell growth via pathways such as MAPK/ERK. Experimental and clinical studies suggest that gastrin receptor antagonists can inhibit PC cell proliferation, underscoring gastrin’s role in tumor progression [78]. We suggest that a positive causal effect of GERD on n-6/n-3 PUFA exists, which correlates with the dietary tendencies of people with GERD who are likely to be exposed to high n-6/n-3 PUFAs, such as fried foods, desserts, and beverages [79]. A high n-6/n-3 PUFA ratio increases the risk of pancreatic cancer in patients with GERD, which means that more omega-3 fatty acids and less omega-6 fatty acids should be consumed. Although this mediating role was only 2.9% in the causal pathway, it was still significant. According to Cholewski et al. [21], omega-6 fatty acids compete for enzymes that convert omega-3 fatty acids to EPA, DPA, and DHA, blocking their conversion to anti-inflammatory compounds. Taha et al. [80] observed that lowering dietary omega-6 fatty acid intake for 12 weeks increase in plasma omega-3 fatty acids. Hence, an appropriate ratio of n-6/n-3 PUFA [81] can limit tumour progression by decreasing mitochondrial function [82], triggering tumour cell apoptosis [23], reducing the number of pro-tumour macrophages [83], and preventing metastasis by anti-angiogenesis [84]. Wang J et al. [85] conducted a meta-analysis of available studies on the effect of omega-3 intake on PC (RR = 0.71; 95% CI:0.45–1.13; I2 = 70.2%) and found no statistical significance. Animal studies demonstrate that diets rich in omega-3 PUFAs reduce the incidence and progression of precancerous pancreatic lesions compared to diets high in omega-6 PUFAs or standard diets [29]. Further research indicates that omega-3 PUFAs inhibit pancreatic tumorigenesis by suppressing cell proliferation and promoting apoptosis [86]. Our data demonstrate that high n-6/n-3 PUFA levels are associated with an increased risk of PC, which is not consistent with the findings of previous observational studies, thereby adding to the genetic evidence. A review of prior MR studies revealed that Chen et al. identified a genetic predisposition to GERD as increasing PC risk [87]. However, no studies have yet explored potential mediators of the causal relationship between GERD and PC or quantified their mediating effects. Haycock et al. [42] employed MR to evaluate the GWAS data of PUFAs in 67 malignancies, although the analysis excluded PC-associated SNPs. Furthermore, Ghoneim et al. [41] conducted an MR analysis and discovered that n-6 PUFAs have no genetic effect on PC risk; however, the study did not include SNPs for n-3 PUFAs. Another MR study found no significant association between omega-3 fatty acid levels and the risk of GERD, esophageal ulcers, reflux esophagitis, or Barrett’s esophagus [88], suggesting no reverse causality between PUFAs and GERD. Unlike the above studies, our study incorporated the most recent GWAS data for PC and n-6/n-3 PUFA-related GWAS by combining TSMR and MVMR methods. The causal and mediating effects of GERD and PC were obtained, complementing the results of the above studies. This study explored the causal association between GERD and pancreatic cancer PC, as well as the mediating role of the n-6/n-3 PUFA ratio. The findings provide new insights into potential population characteristics of PC and offer novel directions for identifying high-risk groups, informing preventive strategies, and improving future cancer risk prediction—contributing important clinical value.First, Early detection and treatment of pancreatic cancer are crucial for improving long-term survival. However, excessive screening in the general population may lead to unnecessary harm [89].Therefore, screening and surveillance should focus on high-risk individuals. Patients with GERD may represent a genetically susceptible population and should be considered for more frequent screening to reduce the high mortality associated with late-stage PC [90]. Clinicians could assess PC risk based on the duration and severity of GERD symptoms and use this information to develop individualized screening protocols. The use PPIs in GERD management also requires monitoring [14]. Evidence suggests that the efficacy of PPIs is influenced by CYP2C19 genotypes [91]. Clinicians might consider CYP2C19 testing for GERD patients who exhibit poor response or adverse effects to PPIs to optimize treatment and avoid potential risks associated with long-term PPI use, including the increased risk of pancreatic cancer. From a dietary intervention perspective, this study highlights the importance of modulating the n-6/n-3 PUFA ratio to reduce PC risk. Using SNPs reflecting genetically predicted PUFA levels, our results suggest that dietary adjustments may alleviate GERD symptoms and prevent PC progression. Clinicians may advise patients to increase intake of n-3-rich foods such as fatty fish (e.g., salmon, mackerel, sardines) [92],, algae, and flaxseeds. In clinical practice, patients should be guided to adjust their n-6/n-3 PUFA consumption based on the dietary recommendations published by primary healthcare organisations, such as the Dietary Guidelines for Americans [93]. Future research should focus on the genetic heterogeneity of GERD populations and factors influencing progression to PC, to better understand the underlying mechanisms and identify novel biomarkers and therapeutic targets. Large-scale observational studies on n-6/n-3 PUFA intake in PC patients, particularly in high-risk populations, are warranted to evaluate the benefits and optimal intake ratios. Randomized controlled trials assessing the effect of n-3 supplementation on PC risk and GERD symptomatology would also be valuable. In addition, novel treatment strategies for GERD should be explored to reduce PC risk. For instance, endoscopic radiofrequency ablation has shown promise in improving GERD symptoms with sustained efficacy. Further studies should investigate its potential protective role in high-risk populations, along with the effectiveness and safety of other endoscopic and surgical approaches in both symptom control and cancer prevention. To our knowledge, this is the first study to examine the causal relationship between GERD and PC risk and evaluate the role of n-6/n-3 PUFA as a moderator. The advantage of using a GWAS database to obtain a large number of samples and SNPs is that it avoids the limitations of traditional multivariate regression methods in terms of mediating effects. In addition, we used different methods to assess consistency among the SNPs. However, this study has a few limitations. First, population stratification or correlations originating from subpopulations with distinct distributions of genetic variation and exposure/outcomes are potential confounders that may emerge at the sample level. Second, there is partial sample overlap among the GWAS datasets for GERD, n-6/n-3 PUFA, and pancreatic cancer due to shared use of the UK Biobank. Although this could theoretically bias two-sample MR estimates, previous studies [94] have shown that such bias is minimal when instruments are strong (F-statistics >10), as in our analysis. Simulation using an online tool (https://sb452.shinyapps.io/overlap/) further confirmed that the expected bias and type I error inflation are negligible across realistic levels of sample overlap (Detailed results are shown in the Supplementary Figures). Third, GERD-associated SNPs were partly based on self-reported data [47] which may introduce recall bias. Finally, because MR estimates reflect lifelong genetic effects, they may not fully capture time-specific exposure effect [95]. Conclusion In conclusion, our study used GWAS data in the MR analysis. Our findings support the evidence that GERD has causally detrimental effects on endpoint PC risk. A small fraction of this effect is mediated by n-6/n-3 PUFA, which may be a public health goal for PC prevention, early identification, and mortality reduction. Early screening and long-term monitoring of PC in patients with GERD needs attention. Focusing on the treatment of GERD, controlling disease progression, and dietary adjustments could be an important strategy for PC prevention. However, a majority of the consequences of GERD in PC remain unknown. These findings suggest the need for large-scale cohort studies to assess PC risk in patients with GERD. The correlation between GERD and PC at the pathophysiological level could be examined to identify additional risk factors. Additionally, interventional trials could be conducted to investigate whether lowering n-6/n-3 PUFAs or consuming more n-3 PUFAs could help reduce the risk of GERD in high-risk populations. Supplementary Information Supplementary material 1. Supplementary material 2. Supplementary material 3. Supplementary material 4. Acknowledgements The authors would like to acknowledge the UKB, MRC-IEU, NBDC for providing original data for our Mendelian randomization analyses. We would like to thank Editage (www.editage.cn) for English language editing. Abbreviations Genome-wide association studies Pancreatic cancer Gastroesophageal reflux disease The ratio of omega-6 fatty acids to omega-3 fatty acids Two-sample Mendelian randomisation Multivariate Mendelian randomisation Inverse Variance Weighted Single-nucleotide polymorphism Authors’ contributions Jiqi Ouyang and Wenliang Lv conceptualized the study and developed the research design. Jiqi Ouyang and Runshun Zhang collected and curated the data. Yanping Lu extracted appropriate genetic instruments from published GWAS and ensured data consistency and quality. Jiqi Ouyang and Yanping Lu performed the statistical analyses. Jiqi Ouyang drafted the manuscript, including the methodology, results, and discussion sections. Wenliang Lv and Runshun Zhang provided revisions and final editing to ensure clarity and accuracy. All authors read and approved the final manuscript. Funding This work was supported by the Supported by Sanming Project of Medicine in Shenzhen, Grant No.SZZYSM202311014. Central High-level Chinese Medicine Hospital Clinical Research and Achievement Translation Capacity Enhancement Project - Special Project for Inheriting Academic Experiences of Famous Elderly Chinese Medicine Experts : Research on the Inheritance of Academic Thought and Clinical Experiences of Qihuang Scholar Lu Wenliang, HLCMHPP2023013. Data availability The datasets analysed in this study are publicly available summary statistics. Summary statistics for the GWASs concerning the exposures, and outcome are available from IEU database (https://gwas.mrcieu.ac.uk/). All data and results generated in this work are presented in the main manuscript and the related supplementary files. Investigators who have made their genome-wide data available may not necessarily agree with comments made in this article and the authors take full responsibility for the content of this article. For the datasets used or analysed, and the codes used during the current study, please contact the first author at ouyangjiqi@126.com on reasonable request. Declarations Ethics approval and consent to participate Not Applicable. Consent for publication Not Applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Wenliang Lv and Yanping Lu contributed equally to this work. Contributor Information Wenliang Lv, Email: lvwenliang@sohu.com. Yanping Lu, Email: 2725850656@qq.com. References Associated Data Supplementary Materials Supplementary material 1. Supplementary material 2. Supplementary material 3. Supplementary material 4. Data Availability Statement The datasets analysed in this study are publicly available summary statistics. Summary statistics for the GWASs concerning the exposures, and outcome are available from IEU database (https://gwas.mrcieu.ac.uk/). All data and results generated in this work are presented in the main manuscript and the related supplementary files. Investigators who have made their genome-wide data available may not necessarily agree with comments made in this article and the authors take full responsibility for the content of this article. For the datasets used or analysed, and the codes used during the current study, please contact the first author at ouyangjiqi@126.com on reasonable request.