1. Introduction
Gastric ulcer (GU) and gastric cancer (GC) are common diseases of the digestive system. The lifetime prevalence of peptic ulcer disease in the general population is approximately 5% to 10%. However, due to the high prevalence of rheumatic, cardiovascular, and cerebrovascular diseases globally, particularly among the elderly who widely use anti-inflammatory and antithrombotic medications, the diagnosis and treatment of GU are facing significant challenges again.[1,2] GC is the 5th most common cancer worldwide and the 4th leading cause of cancer death.
According to data from the International Agency for Research on Cancer in 2020, there are 1.09 million new cases of GC and 770,000 GC-related deaths globally each year.[3] GU and GC not only pose a serious threat to patients’ health and lives but also bring a heavy economic and disease burden to families and society, remaining significant public health issues worldwide.[4,5] The occurrence of GU and GC is the result of multiple factors. Their development is associated not only with Helicobacter pylori infection, age, and gender, but also with smoking, alcohol consumption, and dietary habits.[6,7] Gastrointestinal diseases like GU and GC account for a substantial portion of healthcare utilization and expenditures. An integrated approach of lifestyle interventions, such as controlling different food intakes, can not only bring more health benefits to patients but also reduce their economic burden.
Diet is considered one of the most important modifiable factors affecting human health. Specific foods can influence susceptibility to diseases, and the relationship between various food intakes and diseases is a key research focus in the health field. Several studies have confirmed causal relationships between the intake of non-oily fish, fresh fruits, grains, and the incidence of gastrointestinal malignancies such as colorectal cancer and esophageal cancer.[8,9] Many observational studies have found that the intake of garlic, citrus fruits,[10] dairy products,[11] cruciferous vegetables,[12] dietary fiber, carotenoids, and various diets is negatively correlated with the risk of GC.
On the other hand, coffee consumption,[13] high-fat diets, and red meat are potential risk factors for GC.[14] Excessive alcohol consumption is associated with an increased risk of several gastrointestinal diseases, including GU and GC.[15–17] However, the independent effects of different food intakes on GU and GC are not well understood. Clarifying the causal relationships between various food intakes and gastritis, GU, and GC is crucial, as this provides further evidence for the formulation and implementation of public health policies.
Mendelian randomization (MR) is an effective method for exploring the impact of exposures on disease risk.[18,19] In recent years, researchers have successfully employed MR to investigate the causal relationships between alcohol consumption, smoking, and 24 different gastrointestinal diseases.[20] In this study, we used bidirectional MR to evaluate both the effects of dietary intakes on GU and GC and the possibility of reverse causality. Additionally, we further applied mediation MR to test whether 1400 metabolites play a mediating role in these potential relationships. Together, these strategies help identify high-risk populations, clarify mechanistic pathways, and provide evidence to support targeted prevention and intervention strategies for GU and GC.
2. Materials and methods
2.1. Study design and data sources
The primary aim of this study is to systematically assess the causal associations between various dietary intakes and GU and GC. We conducted the analysis in several stages, as outlined in Figure 1. Initially, relevant datasets were obtained from genetic sequencing databases.
Next, instrumental variables (IVs) were selected and validated. Bidirectional MR was used to explore the causal relationships between different dietary intakes and GU and GC. Sensitivity analyses and heterogeneity tests were then performed.
Finally, a 2-step approach was employed to analyze whether 1400 metabolites mediate the potential relationship between dietary intakes and GC. MR is a research method that uses genetic variations as IVs for analysis, based on 3 core assumptions: relevance assumption: the IVs must be strongly associated with the exposure. Independence assumption: the IVs must not be directly related to confounders affecting both the exposure and the outcome.
Exclusivity assumption: the IVs should influence the outcome only through the exposure and not have a direct effect on the outcome. All procedures and reporting adhered strictly to the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization guidelines.
The exposure variables in this study are various dietary intakes, while the outcome variables are GU and GC. Data on 226 types of food intakes were sourced from the study by Li et al.[21] GU-related data were obtained from the summary statistics of the FinnGen R11 database (https://r11.finngen.fi/), which included 7012 GU cases and 385,082 controls. GC data included 1423 cases and 345,118 controls (ID: finngen_R11_C3_STOMACH_EXALLC), with genetic data also derived from the Finnish database.
All study participants were of European ancestry to minimize bias due to ethnic differences. The data used in this study are publicly available, so no additional ethical approval was required.
2.2. Selection and exclusion of IVs
In this study, different dietary intakes were used as exposure factors, and single nucleotide polymorphisms (SNPs) associated with these dietary intakes were selected as IVs. The outcome variables were GU and GC. In the reverse MR study, GC and GU were used as exposure factors, and different dietary intakes were treated as outcome variables.
The SNPs used met the commonly applied significance threshold in MR studies of P < 5 × 10⁻8. To remove linkage disequilibrium and ensure independent genetic variants, an r² threshold of 0.001 and a 10,000 kilobase (kb) window were set.
The F-statistic was used to assess the strength of the IVs, with F > 10 indicating that the IVs are unlikely to suffer from weak instrument bias. Additionally, palindromic SNPs were removed during the harmonization of genetic variants. The Phenoscanner v2 database – a platform that integrates genotype and phenotype associations: was utilized to determine whether the SNPs were associated with potential confounding factors.[22] SNPs related to confounding factors were further excluded based on literature review.
2.3. Bidirectional MR analysis
In this study, the MR analysis adhered to 3 fundamental assumptions. Four statistical validation methods were employed to identify and address potential pleiotropy in the genetic IVs. MR analyses were conducted using the inverse variance weighted (IVW) fixed-effect and random-effect models, MR-Egger method, weighted median method, and weighted mode method.
The IVW method served as the primary MR analysis approach, which involves a meta-analysis of all individual Wald ratios for the genetic variants related to the exposure and outcome. In other words, the IVW method provides a weighted average estimate of the effect of genetically predicted exposure on the risk of the outcome. The MR-Egger method was used to detect violations of the standard IV assumptions and to provide effect estimates that are not biased by such violations.
To reduce the probability of Type I errors, false discovery rate correction was applied.
2.4. Sensitivity analysis
Heterogeneity was assessed using the Cochran Q test, where a P-value < .05 indicates the presence of heterogeneity among the included IVs, in which case the results from the IVW random-effects model were primarily considered. Pleiotropy, which occurs when a genetic variant influences the outcome through pathways independent of the risk factor, was evaluated using the MR-Egger intercept method and the MR-pleiotropy residual sum and outlier global test. If horizontal pleiotropy was detected (P < .05), the results from the MR-Egger method were taken as the main reference.
The leave-one-out sensitivity analysis was also conducted to assess the robustness of the results. Steiger directionality test was used to refute biases due to reverse causality, ensuring the reliability of the bidirectional MR analysis results.
2.5. Mediation MR analysis
A 2-step MR analysis was conducted to test the exclusion restriction assumption and to determine whether any of the 1400 metabolites act as mediators in the relationship between various dietary intakes and GC. Dietary intakes that showed a causal relationship with GC were selected as exposures for the mediation analysis, with GC as the outcome. Positive mediator variables were identified among the 1400 metabolites, the genetic data of which were obtained from the study conducted by Chen et al.[23] Mediation was considered present if the following conditions were met: the exposure factor is associated with the outcome (β0); the exposure factor is associated with the mediator (β1); the mediator is associated with the outcome (β2).
The mediation ratio was calculated as (β1×β2/β0).
2.6. Enrichment analysis
We selected SNPs that were significantly associated with exposure and outcome, utilizing the Single Nucleotide Polymorphism Database for their conversion. Subsequently, enrichment analysis was performed on the transformed genes.[24] This study employed 2 distinct enrichment analysis methodologies: Kyoto Encyclopedia of Genes and Genomes and Gene Ontology (GO). Kyoto Encyclopedia of Genes and Genomes focuses on determining whether genes or metabolites within the gene set are enriched in specific biological pathways.
GO assesses whether the genes in the gene set are enriched at various levels of the GO hierarchy, thereby enhancing our understanding of these genes’ functions concerning cellular composition, biological processes, and molecular functions. The integration of both approaches provides a more comprehensive and nuanced interpretation of gene functions and disease mechanisms.
2.7. Statistical analysis
MR results were expressed as odds ratios (OR) with corresponding 95% confidence intervals (CI), and results were considered statistically significant when P < .05. All MR analyses were conducted using R software (version 4.2.1; R Foundation for Statistical Computing, Vienna, Austria), the “TwoSampleMR” package (version 0.5.7), and the “MRPRESSO” package (version 1.0).
3. Results
3.1. Bidirectional MR analysis of 226 dietary intakes and GU
In this study, all genetic variants were aggregated, and SNPs exhibiting linkage disequilibrium or palindromic structures were removed. Additionally, SNPs associated with confounding factors were filtered out using the Phenoscanner v2 database. The results of the forward MR analysis indicated a positive causal relationship between GU and average weekly beer plus cider intake (OR = 1.429, 95% CI: 1.015–2.009, P = .039), as well as hard cheese intake (OR = 1.322, 95% CI: 1.009–1.731, P = .042).
Conversely, a negative causal relationship was observed between GU and average monthly red wine intake (OR = 0.701, 95% CI: 0.501–0.964, P = .028), cereal intake (OR = 0.542, 95% CI: 0.418–0.704, P < .01), courgette intake (OR = 0.597, 95% CI: 0.389–0.916, P = .018), dried fruit intake (OR = 0.560, 95% CI: 0.409–0.768, P < .01), other egg intake (OR = 0.425, 95% CI: 0.182–0.995, P = .048), seeds intake (OR = 0.381, 95% CI: 0.175–0.825, P = .014), and spinach intake (OR = 0.633, 95% CI: 0.416–0.963, P = .032) (Fig. 2).
The IVW, MR-Egger, and MR-pleiotropy residual sum and outlier analyses revealed no evidence of heterogeneity or pleiotropy between the exposures and outcomes (P > .05). The leave-one-out sensitivity analysis confirmed that the exclusion of individual SNPs did not substantially affect the results, further validating the robustness of our findings. The results are illustrated in a scatter plot (Fig.
3), a funnel plot (Figure S1, Supplemental Digital Content, https://links.lww.com/MD/R53), and a leave-one-out plot (Figure S2, Supplemental Digital Content, https://links.lww.com/MD/R53).
3.2. Reverse MR analysis results
The reverse MR analysis indicated that GU could decrease the intake of cheesy biscuits (beta = −0.008, 95% CI: −0.015 to 0.000, P = .035), oily fish (beta = −0.014, 95% CI: −0.024 to −0.004, P = .006), other drinks (beta = −0.014, 95% CI: −0.026 to −0.001, P = .029), raw vegetables (beta = −0.008, 95% CI: −0.016 to 0.000, P = .027), and yogurt (beta = −0.036, 95% CI: −0.071 to −0.001, P = .039). Conversely, GU could increase the intake of coleslaw (beta = 0.022, 95% CI: 0.001–0.044, P = .038), courgette (beta = 0.031, 95% CI: 0.012–0.051, P < .01), fortified wine (beta = 0.011, 95% CI: 0.002–0.020, P = .014), herbal tea (beta = 0.714, 95% CI: 0.015–1.413, P = .045), melon (beta = 0.013, 95% CI: 0.000–0.029, P = .029), mixed vegetables (beta = 0.034, 95% CI: 0.002–0.066, P = .032), poultry (beta = 0.013, 95% CI: 0.001–0.024, P = .022), pulses (beta = 0.033, 95% CI: 0.009–0.057, P = .006), and sweet potatoes (beta = 0.015, 95% CI: 0.004–0.025, P = .004) (Fig. 4).
The Cochran Q test did not reveal significant heterogeneity among the SNPs (P > .05), and the MR-Egger intercept was not significantly different from 0 (P > .05). The scatter and funnel plots demonstrated a symmetrical distribution of SNPs, suggesting the absence of horizontal pleiotropy in the MR analysis (see Fig. 5 and Figure S3, Supplemental Digital Content, https://links.lww.com/MD/R53).
Additionally, the leave-one-out sensitivity analysis revealed that no individual SNP had a significant impact on the overall results (Figure S4, Supplemental Digital Content, https://links.lww.com/MD/R53).
3.3. Bidirectional MR analysis of 226 food intakes and GC
The forward MR analysis revealed that instant soup intake (OR = 4.636, 95% CI: 1.095–19.617, P = .037) and other milk-based pudding intake (OR = 4.787, 95% CI: 1.044–21.944, P = .043) were positively associated with GC. Conversely, oily fish intake (OR = 0.221, 95% CI: 0.097–0.505, P < .001), hot soup intake (OR = 0.263, 95% CI: 0.081–0.856, P = .027), average weekly red wine intake (OR = 0.522, 95% CI: 0.286–0.951, P = .034), pineapple intake (OR = 0.288, 95% CI: 0.088–0.946, P = .040), intake of artificial sweetener added to cereal (OR = 0.274, 95% CI: 0.077–0.976, P = .046), and water intake (OR = 0.596, 95% CI: 0.358–0.992, P = .047) were negatively associated with GC (Fig. 6).
Cochran Q test did not detect any significant heterogeneity among the SNPs (P > .05), and the MR-Egger intercept did not significantly differ from 0 (P > .05), indicating no substantial horizontal pleiotropy. We present the results using scatter plots (Fig. 7).
The funnel plot demonstrated symmetry among the SNPs, supporting the absence of horizontal pleiotropy in the MR analysis (Figure S5, Supplemental Digital Content, https://links.lww.com/MD/R53). Leave-one-out sensitivity analysis further confirmed that no individual SNP had a significant impact on the overall results (Figure S6, Supplemental Digital Content, https://links.lww.com/MD/R53).
The reverse MR analysis indicated that GC did not have a significant effect on the intake of different foods.
3.4. Mediation MR results
The mediation analysis results indicate that the causal relationship between other meat intake and GC is mediated by salicylate levels, while the causal relationship between average weekly red wine intake and GC is mediated by carotene diol (1) levels. The mediation effect proportions are 15.09% and 13.10%, respectively (P < .05), as shown in Table 1.
3.5. Enrichment analysis
3.5.1. GU
Following the conversion of SNPs into genes, enrichment analysis was conducted (Figures 8 and 9, Figures S7 and S8, Supplemental Digital Content, https://links.lww.com/MD/R53). As illustrated in Figure 8, average weekly beer plus cider intake may elevate the risk of GU through multiple mechanisms, including oxidative stress, diminished repair capacity of the gastric mucosa, and dysregulation of gastric acid secretion. These mechanisms collectively exert an adverse effect on gastric health.
Cereal intake has the potential to modulate the stability and reparative capacity of the gastric environment by influencing various processes such as protection of the gastric mucosa, metabolic regulation, DNA repair, and inflammatory responses; thus affecting the likelihood of developing a GU. Dried fruit consumption may impact both the protection and repair mechanisms of the gastric mucosa via Hseveral metabolic pathways (including cholesterol and steroid metabolism) as well as cellular responses (such as mRNA metabolic regulation, chromatin organization, and protein interactions), thereby reducing the risk associated with GU. However, specific mechanisms warrant further experimental investigation for validation.
3.5.2. GC
Average weekly red wine intake may influence the incidence and progression of GC through various mechanisms, particularly in areas such as immune modulation, metabolic changes, and gene expression regulation. The consumption of red wine could impact the transcriptional dysregulation associated with GC by modulating chromatin structure and gene expression. By affecting nucleosome assembly and chromatin modifications, the bioactive components in red wine may interfere with oncogene expression, thereby inhibiting the advancement of GC.
Conversely, water intake may mitigate the risk of developing GC through multiple biological pathways, including enhancement of immune functions, regulation of inflammatory responses, and maintenance of normal metabolic states. It can modulate the immune response to Helicobacter pylori infection while reducing inflammation caused by this infection; it also influences cytokine networks to sustain normal immune and inflammatory responses and prevent chronic inflammation. Additionally, it regulates lipid metabolism and cellular processes to avert abnormal cell proliferation resulting from metabolic disorders (Fig.
9, Figure S8, Supplemental Digital Content, https://links.lww.com/MD/R53).
4. Discussion
Observational studies have confirmed that many chronic diseases, such as gallstones, gout, diabetes, and fatty liver disease, are influenced by a combination of genetic factors, lifestyle choices, and dietary habits. With rapid societal development, there is increasing attention on how healthy diets can reduce the incidence of these diseases. Numerous studies have demonstrated that diet plays a critical role in the development, progression, and treatment of cancer, with specific foods either increasing or decreasing cancer susceptibility in populations.[25,26] Both domestic and international research has extensively explored the relationship between diet and gastrointestinal diseases, but most studies have focused on individual nutrients or food components, and the causal relationship between different food intakes and gastrointestinal diseases remains unclear.[27] MR provides a powerful approach to explore causal relationships between exposures and outcomes that are difficult or impossible to observe in clinical research, thus paving the way for identifying disease risk factors, conducting risk stratification, and developing intervention strategies.
Inflammation is an underlying pathophysiological mechanism in many chronic diseases, and diet serves as a modulator of chronic inflammation,[28] making it a key modifiable factor influencing the risk of gastrointestinal diseases. GU is a chronic disease characterized by gastric mucosal inflammation accompanied by varying degrees of epithelial damage.[29] High-fat diets, such as those including desserts and hard cheeses, can produce excess fatty acids that damage gastric wall cells, leading to the excessive production of pro-inflammatory cytokines that activate innate immunity and cause chronic gastric inflammation.[30] Moreover, the interaction between dietary patterns and food components also influences the body’s inflammatory processes. On the one hand, diet directly affects the secretion of inflammatory cytokines by modulating immune responses.
Pro-inflammatory dietary components, such as saturated fats, cholesterol, and added sugars, can directly interact with immune cells and receptors, promoting the proliferation of M1 macrophages (pro-inflammatory cells), inhibiting regulatory T cells, and activating mechanisms like Toll-like receptor 4, which cause chronic immune system activation and increase the levels of pro-inflammatory cytokines such as interleukin (IL)-1, IL-6, IL-12, and tumor necrosis factor alpha. These cytokines are closely linked to the tumor immune microenvironment and tumorigenesis.[31,32] The role of diet and nutrition in cancer prevention cannot be overlooked. Therefore, diet-related inflammatory indices, which serve as mediators linking diet and inflammation, have been developed and rapidly applied in epidemiological studies investigating the relationship between diet and cancer.[33] On the other hand, diet indirectly influences inflammation by regulating the gut microbiota and its metabolism.
Anti-inflammatory diets can increase gut microbiota diversity, enhance the activity and abundance of beneficial gut bacteria, increase short-chain fatty acid levels, repair mucosal oxidative damage, improve gut barrier function, and reduce endotoxins, thereby decreasing inflammation.[31,34] A large prospective cohort study in Korea found that individuals with pro-inflammatory diets had a 1.47-fold higher risk of developing GU compared to those with anti-inflammatory diets, highlighting the importance of anti-inflammatory diets in reducing the risk of GU in the general population.[35] Increasing anti-inflammatory components and reducing pro-inflammatory components in the diet, and tailoring dietary plans according to individual characteristics, play a crucial role in the prevention and treatment of GU and GC.
Alcohol consumption is a significant part of contemporary culture and is also a recognized risk factor for numerous diseases. There is a well-established association between alcohol intake and increased risk of gastrointestinal diseases, including alcoholic liver disease, cirrhosis, and chronic pancreatitis. Recent research by Monika Laszkowska et al identified a strong correlation between heavy alcohol consumption and GC risk (OR 3.13, 95% CI: 1.15–8.64).[36] In this study, average monthly red wine intake was found to be inversely associated with GU.
Protein, an essential component of the human body, is considered crucial for patients with GC. Lower preoperative serum albumin levels correlate with shorter survival in GC patients.[37,38] This study suggests that higher protein intake may reduce the incidence of GC. Additionally, high salt intake, particularly from foods like salted nuts, has been strongly linked to GC,[39,40] which is consistent with the findings of this research.
Salt plays a pivotal role in the early carcinogenesis stages, including gastritis and atrophic gastritis, and continues to be influential during the later stages of cancer development by impacting intestinal metaplasia and dysplasia.[41] However, to date, no cohort studies have investigated the relationship between dietary salt intake and GC, indicating a need for further research to clarify the impact of salt intake on GC risk.
It is widely accepted today that increased consumption of vegetables and fruits is beneficial in reducing the incidence of various diseases.[10,42] Anti-inflammatory dietary components such as fiber, vitamins, and polyphenols provide protective effects on the gastrointestinal tract.[31] It has been demonstrated that a diet rich in vegetables, fruits, saturated fatty acids from olive oil and nuts, legumes, and dairy products can reduce gastrointestinal inflammation, thereby lowering the incidence of gastrointestinal diseases.[43] This aligns with the results of this study. Anti-inflammatory components in various fruits and vegetables help prevent ulcer formation by reducing oxidative stress through mechanisms such as the activation of glutathione, superoxide dismutase, CAT, and nitric oxide, inhibition of pepsin activity, suppression of gastrin and histamine release, cyclooxygenase-2 and prostaglandin E2-mediated cytoprotection, and inhibition of H pylori growth.[44]
The complex interactions between the gut microbiota and dietary components can disrupt or integrate gut barrier and lipid metabolism.[45] High-fiber diets enhance the mucous layer and promote immunoglobulin A production, thereby strengthening the gut mucosal barrier, inhibiting the development, maturation, and function of inflammatory cells, and reducing the production of inflammatory cytokines. The short-chain fatty acids produced by dietary fiber regulate the expression of tight junction proteins, thereby enhancing gut barrier function. The relationship between vegetables and fruits such as eggplant, spinach, and oranges with GU and GC needs to be confirmed through large-scale, aggregated prospective data analysis.
A genetically predicted intake of artificial sweeteners added to cereal demonstrated a statistically significant inverse association with GC risk (OR = 0.274, 95% CI: 0.077–0.976, P = .046). This counterintuitive observation may be attributed to specific dietary patterns among European populations, in which artificial sweeteners are frequently consumed in combination with whole-grain cereals known for their protective gastrointestinal effects. Furthermore, low-to-moderate consumption levels may fall below the threshold for inducing harmful effects and may serve as a substitute for added sugars, which are recognized as a risk factor for GC.
To explore whether dietary intake has bidirectional causal links with GU and GC, we used bidirectional and mediation MR. For these analyses, we applied strict criteria to screen IVs, used multiple analytical methods (IVW, MR-Egger, weighted median), and implemented rigorous quality control steps (e.g., removing SNPs with F-statistics < 10 to avoid weak instrument bias). The results demonstrated stability and reliability, providing a higher level of evidence for the causal relationship between diet and these gastrointestinal conditions.
Despite the robust statistical analyses, there are some limitations to acknowledge. Firstly, the genome-wide association study data relied upon were primarily derived from European populations, which may limit the applicability of the identified genetic variations to other ethnic groups. Future studies should replicate these bidirectional and mediation MR analyses using genome-wide association study data from diverse ancestral groups.
Such cross-ancestry replication will help clarify whether the dietary–GU/GC causal associations identified here are consistent across populations or context-dependent on genetic and environmental factors. Secondly, while multiple foods were identified as having causal relationships with these conditions, the mechanisms and functions of some remain unknown, necessitating further research. Lastly, to improve the accuracy and reliability of analyses, future studies should aim to increase sample sizes and further investigate the clinical applicability of different dietary intakes on GU and GC.
5. Conclusion
Exploring risk factors for GU and GC is crucial for their prevention and treatment. This study employed genetic variation to infer causal relationships, utilizing bidirectional and mediation MR analysis with data from the Finnish database of European descent. The study offers 2 key conclusions based on the European population.
First, varying dietary intakes may influence the risk of GU and GC, and conversely, the occurrence of GU and GC may impact dietary habits. These findings reinforce the theoretical basis for considering different food intake levels as potential biomarkers in the pathogenesis of GU and GC, providing new directions for developing dietary treatment strategies for these conditions.
Acknowledgments
We sincerely thank the investigators and participants of the original genome-wide association studies (GWAS) whose publicly available data were used in this study. Dietary intake data were obtained from the study by Li et al GU and GC genetic summary statistics were accessed from the FinnGen R11 database (https://r11.finngen.fi/).
Author contributions
Conceptualization: Lirong Yang.
Data curation: Lirong Yang, Linhui Li, Xinlei Luo.
Formal analysis: Zhijing Liu, Feng Li, Xinlei Luo.
Project administration: Mei Yang.
Resources: Liulin Chen.
Supervision: Zongting Meng.
Validation: Wanqi Wang, Yuan He.
Writing – original draft: Jinyi Cai, Xinlei Luo.
Writing – review & editing: Yufeng Wang, Xinlei Luo.
Supplementary Material