Investigating the causal association between serum uric acid levels and gastric cancer risk: a Mendelian randomization study.

Introduction
Cancer represents a major public health challenge globally. Gastric cancer, a common malignancy, is recognized as the fourth most frequent cause of oncologic mortality worldwide1,2. In 2020, the global incidence of gastric cancer was estimated at approximately 1.089 million new cases, while the disease claimed approximately 769,000 lives3. The insidious onset of gastric cancer frequently results in a low early diagnosis rate. Approximately 70% of patients are diagnosed with the disease at an advanced stage, missing out on optimal treatment opportunities4. Therefore, it is crucial to enhance early screening, diagnosis, and intervention for gastric cancer. The occurrence and progression of gastric cancer are not solely attributed to a single factor but rather to a variety of factors including genetic predisposition, environmental influences, and nutritional factors5. High-nitrogen, high-salt, and high-fat diets, along with excessive consumption of processed meats, and red meat, coupled with inadequate intake of fresh vegetables and fruits, have been recognized as significant risk determinants for gastric cancer. Additionally, excessive alcohol consumption and smoking are also recognized as significant contributors to the development of this malignancy. Furthermore, conditions such as diabetes6 and obesity7; a history of Epstein-Barr virus (EBV) infection; Helicobacter pylori infection; human papillomavirus (HPV) infection; a family history of cancer; and the presence of precancerous lesions can further elevate the risk for gastric cancer. Lastly, exposure to ionizing radiation and genetic predispositions are also considered potential risk factors associated with gastric cancer. By controlling these risk factors associated with cancer, the risk can be mitigated. Investigating these risk factors that promote tumor development and implementing interventions will yield benefits for both the country and society8. Hence, exploring potential risk factors related to gastric cancer is essential.
Uric acid (UA), a pivotal metabolic byproduct in humans, is synthesized through the action of xanthine oxidoreductase (XOR), which catalyzes the oxidation of xanthine and hypoxanthine9. There are numerous factors that can influence SUA levels in the human body. Research indicates that multiple genetic loci, such as ABCG2 and SLC2A9, are significantly correlated with variations in SUA concentrations10. Additionally, individuals exhibiting components of metabolic syndrome—such as hypertension, diabetes mellitus, and dyslipidemia—typically present with elevated UA levels compared to those without these conditions11. Dietary intake plays a crucial role in modulating UA levels; notably high fructose consumption and substantial alcohol intake (particularly from beer) have been shown to increase SUA concentrations11. Obese individuals typically exhibit significantly elevated UA levels compared to their normal-weight counterparts11. The kidneys are essential for the excretion of UA; thus, patients suffering from chronic kidney disease often demonstrate elevated SUA levels due to impaired renal function12. Moreover, the long-term use of certain pharmacological agents known to adversely affect renal function, including corticosteroids, diuretics, and antibiotics, can also contribute to an increase in SUA concentrations. The interplay among these various factors ultimately determines individual SUA levels and may significantly influence the onset and progression of related diseases.
SUA stands out as one of the predominant antioxidants in human blood, functioning as both a chelator of transition metal ions and a free radical scavenger13. Consequently, elevated levels of UA are believed to inhibit the development of malignant tumors14. However, numerous large-scale epidemiological and prospective studies have yielded inconsistent and even contradictory conclusions regarding the link of heightened SUA levels with the incidence and mortality of malignant tumors. Increasing evidence now suggests that high SUA levels may also function as a pro-oxidant15, promoting tumor cell proliferation, migration, as well as survival through mechanisms involving reactive oxygen species (ROS) and inflammatory stress, thereby facilitating tumor formation13. Recent studies have also proved that high levels of SUA are associated with the occurrence and progression of various malignant tumors16–18. However, research exploring the specific relationship between SUA levels and gastric cancer risk remains scarce. The sole pertinent observational study identified followed 7,889 men of Japanese descent for over 20 years, concluding that increased UA levels did not correlate with a heightened risk of most malignancies, including gastric cancer19. However, this study only examined males of Japanese descent. Furthermore, given that observational researches often cannot effectively eliminate confounding biases and reverse causation, the causal conclusions drawn from this study may be subject to uncertainty20–22. Consequently, the causal association between SUA levels and gastric cancer risk has yet to be determined.
The randomized controlled trial (RCT) is heralded as the paramount methodology for establishing causality, but it comes with ethical and cost-related issues. An alternative methodological approach, known as Mendelian randomization (MR), utilizes genetic variants to investigate potential causality between exposures and outcomes20–22. Because genetic variations are randomly assigned at conception, MR is capable of circumventing some of intrinsic limitations associated with observational studies and RCTs. The growing accessibility of genome-wide association studies (GWAS) has facilitated the adoption of MR analysis as a suitable approach to uncover causal relationships. Our research aims to examine potential causal relationships of SUA to gastric cancer risk, as well as explore potential ethnic disparities by employing MR analyses within European and East Asian cohorts.

Materials and methods

Study design
To assess the causal relationship of SUA to gastric cancer risk, a two-sample MR analysis was conducted. This involved analyzing genome-wide associations between exposure and outcome23,24. In our MR analysis, SUA and gastric cancer were served as exposure and outcome variables, respectively, with single nucleotide polymorphisms (SNPs) serving as instrumental variables (IVs). The validity of two-sample MR approach rested on three critical assumptions: (a) IVs manifest a robust association with SUA; (b) IVs are not influenced by confounding variables; and (c) IVs affect gastric cancer risk solely through their effect on SUA25 (Fig. 1). To further investigate potential racial differences, we conducted the aforementioned MR analysis within both European and East Asian populations. Datasets utilized in the present investigation were derived from publicly accessible sources, with ethical clearances specifically documented within the original GWAS. Therefore, patient informed consent and ethical approval were deemed unnecessary for this study.

Data source
For the European population, the genome-wide association summary statistics data of UA and gastric cancer utilized in this study were obtained from the IEU OpenGWAS project website (https://gwas.mrcieu.ac.uk/)26. The UA GWAS data (ID: ukb-d-30880_irnt27), obtained from the analysis of UK Biobank phenotypes by Neale lab, comprised 343,836 participants with 13,585,994 SNPs. The gastric cancer GWAS data (ID: finn-b-C3_STOMACH_EXALLC) was obtained from FinnGen Biobank and included 174,639 participants, consisting of 633 gastric cancer patients and 174,006 control samples (where individuals with other types of cancer were excluded). The total number of SNPs analyzed amounts to 16,380,305. For the East Asian population, the UA GWAS data was obtained from Taiwan Biobank (TWB) in China and included 92,615 participants with 8,239,189 SNPs28. The ages of participants ranged between 30 and 70 years, averaging 50 years, with a standard deviation (SD) of 10.9 years. UA concentrations were determined through blood sample assays, with an average UA concentration of 5.46 mg/dl and a SD of 1.43 mg/dl. Data download is accessible via the GWAS Catalog website (https://www.ebi.ac.uk/gwas/, ID: GCST90278646)29. The gastric cancer GWAS data (ID: bbj-a-119) was obtained from Biobank Japan (BBJ) and comprised 202,308 participants, consisting of 6,563 gastric cancer patients (including 4,885 males and 1,678 females) and 195,745 control samples (including 97,655 males and 98,090 females). The total SNP count for this dataset was 8,885,324. All cases of gastric cancer were diagnosed by physicians, and other types of cancer were systematically excluded from the control group30. Detailed information about the included GWAS can be found in Table 1. Exposure and outcome data were sourced from different cohorts to avoid any issues related to sample overlap.
To strengthen the robustness of our results and mitigate potential bias arising from genetic differences, we incorporated an additional dataset on SUA levels of 109,029 participants from Biobank Japan in our analysis of the East Asian population31. This dataset is also accessible for download via the IEU OpenGWAS website (ID: bbj-a-57). In our study, unless otherwise indicated, the MR analysis of the East Asian population refers to the analysis conducted using IVs selected based on UA data from Taiwan Biobank.

Instrumental variables selection
We identified significant SNPs as IVs from the SUA data using P < 5E-8 as the screening criteria in order to meet the first MR assumption. Independent SNPs that were not affected by linkage disequilibrium (LD) were screened with thresholds of r2 > 0.001 and clump distance > 10,000 kb. In this process and beyond, the population chosen for LD reference panel was consistent with the population for exposure and outcome. Subsequently, the proportion of variance (R2) explained for the association between a specified SNP and exposure was ascertained, and thereafter, the F-statistic corresponding to this SNP was computed32–34. SNPs with an F-statistic below 10 were excluded in order to mitigate the potential bias from weak IVs in MR analysis35. Next, the selected SNPs were queried against the gastric cancer GWAS database. If SUA-related index SNPs were absent from the gastric cancer dataset, proxy SNPs demonstrating LD (r2 > 0.8) was used. Subsequently, during the process of variant harmonization, we eliminated specific SNPs associated with incompatible alleles and removed palindromic SNPs with intermediate allele frequencies. Then, we utilized a threshold of P > 5E-8 to select SNPs demonstrating no association with the outcome in order to meet the third MR assumption. To fulfill the second MR assumption, we analyzed potential confounders associated with the selected SNPs utilizing LDtrait module of LD Link (https://ldlink.nih.gov/?tab=ldtrait in April, 2024). SNPs that exhibited significant associations with established risk factors for gastric cancer, such as smoking, Helicobacter pylori infection, alcohol abuse, and BMI, were removed. Finally, we applied Steiger filtering to evaluate the directionality of causality for each SNP. A P-value of ≥ 0.05 indicates that the SNP has a reverse causal relationship and should be removed36. Figure 2 illustrates the flowchart for selecting IVs. The final set of SNPs was then employed for the MR analysis.

Statistical analysis
In our investigation, R software (version 4.3.2) along with “TwoSampleMR” package were employed to perform both MR analysis and sensitivity analyses22. The principal analytical strategy used to derive an assessment of the causal impact was the inverse variance weighted (IVW) model37. Heterogeneity within the IVW estimates was assessed using Cochran’s Q test38. The detection of notable heterogeneity (P < 0.05) necessitated the adoption of a random-effects IVW model. Conversely, when heterogeneity was not present, a fixed-effects model was deemed suitable. Additionally, other MR analysis models including MR-Egger39, weighted mode40 and weighted median41 were employed to validate the causal relationship of SUA to gastric cancer. The MR-derived odds ratio (OR) and the corresponding 95% confidence interval (95% CI) for gastric cancer per standard deviation increase in SUA levels were calculated. Subsequently, a funnel plot was employed to evaluate the heterogeneity in the causal effect estimates. The stability of this estimate was determined by examining the plot’s symmetry. Asymmetry within the funnel plot suggests the potential presence of directional pleiotropy, which could influence the results of the MR analysis. Potential horizontal pleiotropy was evaluated using MR-Egger regression analysis, and a significant non-zero intercept would suggest its presence. Additionally, we conducted an MR-PRESSO global test to assess the presence of pleiotropy. The MR-PRESSO outlier test was performed to identify and exclude any SNPs exhibiting potential pleiotropic effects, followed by an estimation of the adjusted results42. Should outliers be detected, the MR-PRESSO distortion test was used to evaluate if significant differences exist between the results before and after adjustment. Subsequently, the robustness of the MR findings was assessed through a leave-one-out sensitivity test.

Results

MR analysis of SUA and gastric cancer in European population
In the MR-PRESSO test, no outlier SNPs were detected. 206 SNPs were ultimately identified for MR analysis in the European population, collectively explaining 10.93% of the variance in SUA levels. The minimum value of F-statistics was 29.90, suggesting a reduced likelihood of bias due to weak IVs in the causal inference. Steiger filtering revealed that all SNPs exhibited unidirectional causality (all P < 0.05). In addition, the minimum P-value observed for the correlation between these SNPs and gastric cancer was 0.0024. Further details regarding these SNPs can be found in Supplementary Table 2. In addition, Cochran’s Q test revealed no significant heterogeneity across the included SNPs (PIVW=0.102, PMR−Egger=0.095; Table 2). Employing the fixed-effects IVW method, SUA levels did not significantly correlate with gastric cancer risk (OR: 1.04; 95% CI: 0.77 ~ 1.41; P = 0.778). This result was further confirmed by the findings of three other MR analysis methods. Causal relationships of SUA to gastric cancer risk are visually represented in scatter and forest plots (Fig. 3A and Supplementary Fig. 1). Funnel plot (Supplementary Fig. 2) demonstrated that all included SNPs were symmetrically distributed, suggesting no directional pleiotropy and reinforcing the robustness of the assessment results. Detailed causal estimates of SUA on gastric cancer in different models are presented in Fig. 4. Additionally, the absence of pleiotropy was confirmed by MR-Egger regression test (intercept = -0.00282, P = 0.682) and MR-PRESSO global test (P = 0.099; as indicated in Table 2). The outcomes of leave-one-out sensitivity analysis are presented in Supplementary Fig. 3, which further substantiates the robustness of the findings.

MR analysis of SUA and gastric cancer in East Asian population
Similar to the European population analysis, no outlier SNPs were detected in the East Asian population. 45 SNPs were ultimately identified for MR analysis, collectively explaining 4.97% of the variance in SUA levels. The minimum value of F-statistics was 31.57, suggesting a reduced likelihood of bias due to weak IVs in the causal inference. Steiger filtering revealed that all SNPs exhibited unidirectional causality (all P < 0.05). In addition, the minimum P-value observed for the correlation between these SNPs and gastric cancer was 0.0090. Further SNP details are available in Supplementary Table 3. In addition, Cochran’s Q test revealed no significant heterogeneity across the included SNPs (PIVW=0.106, PMR−Egger=0.106; Table 2). Employing the fixed-effects IVW method, SUA levels did not significantly correlate with gastric cancer risk (OR: 0.93; 95% CI: 0.81 ~ 1.05; P = 0.245). This result was further confirmed by the findings of three other MR analysis methods. Causal relationships of SUA to gastric cancer risk are visually represented in scatter, forest and funnel plots (Fig. 3B, Supplementary Fig. 4, and Supplementary Fig. 5). Detailed causal estimates of SUA on gastric cancer in different models were presented in Fig. 4. Sensitivity analysis confirmed the robustness of the findings (Supplementary Fig. 6). Additionally, the absence of pleiotropy was confirmed by MR-Egger regression test (intercept=-0.00608, P = 0.349) and MR-PRESSO global test (P = 0.124; Table 2). Employing SNPs selected from UA data provided by Biobank Japan, we repeated the entire analysis process and achieved consistent results (IVW estimate, OR: 0.92; 95%CI: 0.82 ~ 1.03; P = 0.130), further confirming the robustness of our conclusions.

Discussion
The present investigation sought to elucidate the potential causal relationship of SUA levels to gastric cancer risk by employing a two-sample MR analysis to provide initial insights. Since the exposure and outcome datasets utilized in this MR analysis were from different cohorts, any potential bias resulting from sample overlap issues was eliminated. The findings showed that, in both European and East Asian populations, there was no causal link between SUA levels and gastric cancer risk. This conclusion was consistently supported across different MR methods, including fixed-effects IVW, MR Egger, weighted mode, and weighted median, each reinforcing the credibility of the causal conclusion. Particularly in East Asian population, we employed two sets of genetic instruments sourced from different regions as exposures, and obtained consistent results. Steiger filtering corroborated that our findings are not affected by reverse causation. Furthermore, the Cochran’s Q test revealed no evidence of heterogeneity in the data. Additionally, both the MR-Egger regression test and MR-PRESSO test did not identify any horizontal pleiotropy. The robustness of our research findings was further substantiated by the leave-one-out sensitivity analysis.
Cancer constitutes a significant and escalating challenge in public health. UA plays a multifaceted role in cancer development. On one hand, it is believed that high UA levels can inhibit the formation of malignant tumors by clearing oxygen free radicals and inhibiting lipid peroxidation14. A study conducted in the Netherlands revealed an inverse relationship between SUA levels and overall cancer mortality in men over a 38-year follow-up period43. Furthermore, Horsfall et al. discovered that lower SUA levels correlated with a heightened incidence of lung cancer44. Additionally, a study utilizing the EPIC-Heidelberg cohort demonstrated an inverse association between higher SUA levels and the risk of breast cancer17. Contrarily, mounting evidence suggests that elevated UA levels may exert a pro-oxidant effect15, promoting the proliferation, migration, and survival of tumor cells via mechanisms involving ROS and inflammatory stress, thereby facilitating tumor formation13. A study revealed a significant relationship of higher SUA levels to the recurrence and metastasis of breast cancer16. Furthermore, another study indicated that higher SUA levels are significantly correlated with an increased risk of colorectal cancer45. Additionally, in hypertensive Chinese individuals, Yang et al. observed that increased SUA levels are linked to a higher risk of overall and digestive cancer incidences, as well as cancer mortality46. Another investigation highlighted gender-specific associations between SUA levels and the incidence of hepatobiliary and pancreatic cancers; in females, higher SUA levels were linked to an increased risk of pancreatic cancer, whereas in males, an elevated risk was observed for gallbladder cancer47. Due to its dual role, the predictive role of SUA in tumors has been a topic of controversy in academic research. Numerous findings have even yielded inconsistent or contradictory conclusions concerning the relationship of elevated UA levels to the risk of developing cancer. For instance, a cohort investigation of Japanese descent Hawaiian men indicated that higher SUA levels were not linked to an increased risk of most malignancies, such as colon, rectal, and bladder cancers19. However, it was identified as posing a heightened risk for prostate cancer19. In addition, a MR study involving an East Asian population has identified high SUA levels as a possible risk factor for prostate cancer18. In contrast, a MR study conducted within a European population found insufficient evidence to substantiate a causal relationship of SUA levels to prostate cancer risk48. These divergent outcomes underscore the complexity of UA’s role in cancer development and the need for further detailed investigations.
Limited previous research has been conducted on the relationship of SUA levels to gastric cancer risk. Only one relevant observational study was identified in the literature, which included 7,889 men of Japanese descent in Hawaii and followed them for over 20 years19. The study concluded that heightened UA levels did not correlate with an increased incidence of gastric cancer19. Nonetheless, it is critical to acknowledge that this research only focused on males, thereby leaving the potential relationship of UA levels to gastric cancer risk in females undetermined. Additionally, the inclusion solely of individuals of Japanese descent raises questions regarding the generalizability of these findings to other ethnic groups. Furthermore, considering the potential confounding factors in observational studies and the possible influence of follow-up time, these conclusions are still subject to question. The MR approach has the potential to address the inherent limitations of observational studies20–22. However, to date, no MR study has yet explored the potential causal relationship of SUA levels to gastric cancer risk.
In the present investigation, a two-sample MR study was designed to assess the potential causality between SUA levels and gastric cancer risk. For the validity of an MR study, three fundamental assumptions must be met20–22. Initially, in our investigation, the selected SNPs demonstrated significant associations with exposure, as evidenced by P-values less than 5E-8. Additionally, the F-statistics for IVs exceeded 10, indicating that all instruments were sufficiently effective. Secondly, SNPs that showed significant associations with established risk factors for gastric cancer, such as smoking, Helicobacter pylori infection, alcohol abuse, and BMI, were systematically excluded. Furthermore, we employed Steiger filtering to remove all SNPs exhibiting reverse causality, thereby ensuring that our study was not influenced by reverse confounding factors. Moreover, only SNPs exhibiting no association with the outcome (P > 5E-8) were retained for analysis. In addition, potential horizontal pleiotropy was assessed and ruled out using both the MR-Egger regression test and the MR-PRESSO test, indicating no detectable pleiotropic effects that could bias the causal estimates. These stringent analytical measures suggest that the MR study adhered to the necessary assumptions, thereby minimizing the likelihood of biases and enhancing the credibility of the findings.
When presenting the results, several limitations must be acknowledged and carefully considered. Firstly, the sample size for gastric cancer cases within European population was notably small, comprising only 633 cases. Despite this constraint, it is noteworthy that this is currently the largest database of gastric cancer cases in Europe available for analysis. Therefore, there is an urgent need for larger case sample sizes in GWAS data for MR analysis to improve statistical accuracy49. Secondly, the explanatory variance of the IVs used in our MR analysis was relatively low, which could potentially limit the power of our causal inferences. Therefore, additional GWAS incorporating a greater number of valid SUA-associated SNPs are imperative to substantiate the present findings. Thirdly, due to the aggregate nature of the GWAS data, it was not possible to conduct stratified analyses incorporating key covariates such as gender, tumor staging, and age. Extensive research has shown that males globally experience approximately double the incidence rates of gastric cancer compared to females, highlighting significant gender disparities5. Furthermore, research indicates that SUA levels are associated with the risk of various cancers—including colorectal, pancreatic, and gallbladder cancers—in a gender-specific manner47,50. Additionally, differences in SUA levels between genders have been observed51. Discrepancies in the gender ratio of SUA data compared to that of gastric cancer data may potentially influence the results of our study. Therefore, analyzing the gender differences in the association between SUA levels and gastric cancer risk is of paramount importance. The absence of gender-specific analysis represents a significant limitation in our study. Fourthly, another inherent limitation of this two-sample MR study is the assumption of linearity. Although a clear linear relationship exists between SUA levels and the risk of certain cancers, studies have also revealed non-linear associations. For instance, a prospective cohort study demonstrated a U-shaped correlation between SUA levels and colorectal cancer risk in both genders50. Additionally, another study found that in males, SUA levels are similarly U-shaped in relation to liver cancer risk, indicating an increased risk at both extremely low and high SUA levels47. A further study utilized penalized splines to analyze the relationship between SUA levels and the overall cancer incidence in males, revealing a J-shaped association52. However, in our research, we were unable to assess the potential existence of non-linear associations between exposures and outcomes. Fifthly, while our results were consistent across European and East Asian cohorts, the generalizability of these findings to other ethnic groups remains uncertain. Despite these limitations, it is pertinent to recognize that this investigation offers the closest approximation currently available to a RCT concerning the impact of SUA levels on gastric cancer risk. This highlights the value of our findings while also acknowledging the areas in need of further research to validate and expand upon these initial results.
In conclusion, this MR study did not find any evidence of a causal relationship between genetically predicted SUA levels and gastric cancer risk. Consistent results were observed in both European and East Asian populations. It is strongly recommended to conduct internal and external validation of these findings. Future MR investigations should prioritize gender-specific analyses, employ more effective SNPs, and incorporate a broader spectrum of cancer cases and types to enhance the analytical scope and depth.

Electronic Supplementary Material
Below is the link to the electronic supplementary material.