Causal associations between human metabolites and functional dyspepsia: Evidence from two sample and multivariable Mendelian randomization

Abstract

Evidence for causal associations between functional dyspepsia (FD) and human metabolites remains limited. Two-sample and multivariable Mendelian randomizations to detect the causal relationships between FD and human metabolites. Single nucleotide polymorphisms significantly associated with human metabolites were selected as instrumental variables, the inverse variance weighting method was used as the primary analysis method, and the results were tested for heterogeneity and horizontal pleiotropy. Genetically predicted levels of 26 circulating metabolites and 5 cerebrospinal fluid metabolites were associated with FD. Among them, Total cholesterol in large low-density lipoprotein (odds ratio [OR] = 0.985, 95% confidence interval [CI] 0.975–0.995, P = .003) and cholesteryl esters to total lipids ratio in intermediate-density lipoprotein (OR = 0.983, 95% CI 0.971–0.996, P = .008) were negatively associated with FD. It was also positively associated with the level of phenylalanine (OR = 1.037, 95% CI 1.009–1.065, P = .009) and phospholipids to total lipids ratio in small high-density lipoprotein (OR = 1.015, 95% CI 1.003–1.027, P = .017). After multivariable adjustment, FD was causally associated with phospholipids to total lipids ratio in small high-density lipoprotein (OR = 1.332, 95% CI 1.060–1.674, P = .014). Sensitivity analyses showed no evidence of heterogeneity and horizontal pleiotropy in the above results. Human metabolites may be associated with FD.

1. Introduction

Functional dyspepsia (FD) is a prevalent gastrointestinal disorder characterized by persistent or recurrent upper abdominal discomfort, often associated with postprandial fullness, early satiation, epigastric pain, and nausea, without any identifiable organic cause.^[1] This condition affects a significant portion of the population and is increasingly recognized as a multifactorial syndrome, influenced by various physiological and psychological factors.^[2] Despite its high prevalence, the pathophysiology of FD remains poorly understood, necessitating further investigation into the underlying mechanisms and potential biomarkers for diagnosis and treatment.

Recent studies have highlighted the potential role of metabolites in the pathogenesis of FD.^[3] Metabolomics, the comprehensive study of metabolites in biological samples, offers a promising approach to identify metabolic signatures associated with functional gastrointestinal disorders.^[4] For instance, alterations in the levels of specific metabolites may reflect changes in gastrointestinal motility, gastric acid secretion, and mucosal integrity,^[5,6] which are often disrupted in FD patients. In this context, the identification of circulating metabolites and their relationship with clinical symptoms may provide valuable insights into the disease mechanism, thereby facilitating the development of targeted therapeutic strategies. Moreover, there is growing evidence suggesting a connection between cerebrospinal fluid (CSF) metabolites and FD. Metabolomic profiling of CSF has revealed distinct metabolic patterns that correlate with various neurological and gastrointestinal disorders.^[7] The interplay between the gut and brain, often referred to as the gut–brain axis, is thought to play a pivotal role in modulating gastrointestinal function and symptomatology.^[8,9] For example, metabolites such as short-chain fatty acids are produced during the fermentation of dietary fibers by gut microbiota, have been shown to influence intestinal motility and perhaps be implicated in the manifestation of dyspeptic symptoms.^[10] Furthermore, the identification of neuroactive metabolites in the CSF may elucidate the neurophysiological underpinnings of FD, potentially linking psychological factors such as anxiety and depression to gastrointestinal dysfunction.^[11]

Despite the advancements in understanding the relationship between metabolites and FD, significant gaps remain in the current literature. There is a need for large-scale, well-designed studies that employ comprehensive metabolomic approaches to identify specific biomarkers capable of reliably differentiating FD from other gastrointestinal disorders. Additionally, investigations into the effects of dietary interventions on metabolite profiles and symptom resolution could pave the way for personalized treatment strategies.^[12] The integration of metabolomic data with clinical assessments will not only enhance our understanding of FD but also facilitate the identification of novel therapeutic targets. This study employed Mendelian randomization (MR), a method for inferring causal relationships between exposure and outcome via the application of instrumental variable (IV), single nucleotide polymorphisms (SNPs), to address the existing research gaps.^[13] MR were performed with circulating metabolites and CSF metabolites as exposures, and FD as an outcome, supported by corresponding Genome-Wide Association Study (GWAS) and IV data. Our study was intended to fill the existing knowledge deficit by shedding light on the metabolic alterations associated with FD and their potential impact on diagnosis and management.

2. Methods and materials

2.1. Study design

The study comprises 3 main components. First, we assessed the causal effects of circulating metabolites on FD. Second, we examined the causal effects of CSF metabolites on FD. Third, based on significant findings from the two-sample MR analysis, we conducted multivariable MR (MVMR) to further evaluate the independent causal effects of these metabolites. Before MR analysis, 3 fundamental assumptions must be addressed^[14]: the relevance assumption, which requires genetic variants to exhibit strong and significant associations with the exposure variables; the independence assumption, which mandates that IVs should be unrelated to any confounders; and the exclusion restriction assumption, which stipulates that genetic variants should affect the outcome solely through the exposure, without involvement in other pathways (Fig. 1).

Figure 1.

Mendelian randomization study design. This study initially employed two-sample Mendelian randomization to investigate the causal relationships between circulating metabolites, cerebrospinal fluid metabolites, and functional dyspepsia. To enhance the rigor of the conclusions, multivariable Mendelian randomization was applied to both sets of two-sample analyses. The robustness of the multivariable Mendelian randomization results was further evaluated using heterogeneity analysis, horizontal pleiotropy analysis, and the MR-PRESSO Global Test. CSF = cerebrospinal fluid, IVW = inverse-variance weighted, LD = linkage disequilibrium, MR = Mendelian randomization, MR-PRESSO = MR Pleiotropy RESidual Sum and Outlier, SNP = single nucleotide polymorphism, TSMR = two-sample Mendelian randomization.

2.2. Data source and outcome definition

Data on circulating metabolites were obtained from research by Karjalainen et al published in Nature.^[15] This study conducted a genome-wide association analysis of 233 circulating metabolic traits using nuclear magnetic resonance spectroscopy in over 136,000 participants. The research identified more than 400 independent genetic loci, with probable causal genes assigned to two-thirds of these loci. It emphasizes how participant characteristics can influence genetic associations and provides detailed metabolic profiling to better understand lipid metabolism. The study also explores genetic pleiotropy across metabolic pathways, uncovering a potential causal link between acetone and hypertension.

The CSF metabolite GWAS data was originally sourced from the study conducted by Panyard et al.^[16] In their research, participants were drawn from 2 cohorts: 532 from the Wisconsin Alzheimer’s Disease Research Center (WADRC) and 168 from the Wisconsin Registry for Alzheimer’s Prevention (WRAP). After quality control measures were applied, a total of 338 metabolites were analyzed across 155 samples from the WADRC cohort, with an average age of 64.7 years (standard deviation [SD] 6.4), and 136 samples from the WRAP cohort, with an average age of 62.0 years (SD 6.6). In the WADRC cohort, genome-wide significance was determined using a threshold of 5 × 10⁻⁸, with a Bonferroni correction applied for the 338 metabolites tested, resulting in a final significance threshold of 1.48 × 10⁻¹⁰. SNPs identified during the discovery phase were then compared with the WRAP cohort for replication, using a significance threshold of 0.05 and applying a Bonferroni correction for the number of significant SNPs tested in the WADRC cohort. The final significance threshold for replication was set at 8.25 × 10⁻⁵. A total of 213,718 SNPs were presented in the supplementary materials of their study.

GWAS summary data of FD were derived from the study by Garcia-Etxebarria et al,^[17] which investigated the genetic and comorbid associations of FD, a common gastrointestinal disorder with unclear pathophysiology, using data from 3 large biobanks (UK Biobank, Estonian Genome Center of the University of Tartu, and Michigan Genomics Initiative). The analysis of 10,078 FD cases and 351,282 controls revealed significant associations between FD and other conditions, particularly gastrointestinal disorders, anxiety, ischemic heart disease, and infectious diseases. A genome-wide association meta-analysis estimated FD heritability at around 5%, with genetic correlations indicating shared predispositions with several other diseases. Thirteen loci were identified as potentially linked to FD risk, with 2 loci replicated in an independent cohort. The findings suggest that while FD has a weak genetic component, it shares common pathogenetic mechanisms with multiple conditions, enhancing the understanding of etiology and potential treatment approaches towards FD.

2.3. IV selection

P-value screening for both circulating and CSF metabolites was set at P < 5 × 10⁻⁸ to determine rigorous correlation between IVs and exposures. To minimize the impact of linkage disequilibrium on our analysis, we set threshold parameters at r² = 0.001 and a distance criterion of 10,000 kb. To ensure a robust association between IVs and exposures, we calculated the explained variance (R²) and assessed the strength of each IV using the F-statistic, excluding any SNPs with an F-statistic below 10.^[18] These calculations were based on several parameters: minor allele frequency (MAF), effect size on exposure (β), SD, sample size (N), and the number of IVs (k), R² = 2 × (1 − MAF) × MAF×β², F = R² (n − 2)/1 − R². To enhance the reliability of our analysis by minimizing biases and confounding, we excluded palindromic SNPs with intermediate allele frequencies that could interfere with genotype–phenotype associations and compromise result accuracy.^[19]

2.4. Statistics analysis

The inverse-variance weighted (IVW) method was employed as the primary analysis approach, and the results obtained through IVW were considered the main findings.^[20] In addition, 3 other analytical approaches, MR-Egger, weighted median, and weighted mode, were applied.^[21,22] The results were evaluated from 2 key perspectives: if the findings from the additional 3 methods were statistically significant and if the odds ratios were consistent in direction with the IVW results, this would further strengthen the conclusions. Analyses using all methods were only conducted when the number of SNPs exceeded 2; if there were exactly 2 SNPs, only the IVW method was used, and for a single SNP, only the Wald ratio method was applicable.

Multivariate MR was adopted in further analysis of positive results for circulating metabolites and spinal metabolites with FD, respectively, with IVW to be the primary outcome.^[23] Sensitivity analyses included tests for heterogeneity tests and horizontal pleiotropy. For the positive results of multivariate MR, MRpresso analyses were subsequently performed to further ascertain that there was no horizontal pleiotropy in the positive results by global test. The R software packages “devtools,” “TwoSampleMR,” “LDlinkR,” and “MRPRESSO” were utilized for statistical analyses.

3. Results

3.1. Two sample MR results

The two-sample IVW analysis revealed significant associations between 26 circulating metabolites and FD (Fig. 2). Specifically, total cholesterol levels in small high-density lipoprotein (HDL; β = −0.021, 95% confidence interval [CI]: −0.039 to −0.003, P = .020), the cholesteryl ester-to-total lipid ratio in intermediate-density lipoprotein (β = −0.017, 95% CI: −0.029 to −0.004, P = .008), cholesterol esters in small HDL (β = −0.016, 95% CI: −0.031 to −0.001, P = .032), and the phospholipid-to-total lipid ratio in large very-low-density lipoprotein (β = −0.016, 95% CI: −0.033 to 0.0000, P = .047) were inversely associated with FD. In addition, total cholesterol in large low-density lipoprotein (LDL; β = −0.015, 95% CI: −0.025 to −0.005, P = .003), phospholipids in large LDL (β = −0.015, 95% CI: −0.025 to −0.005, P = .005), and cholesterol esters in large LDL (β = −0.014, 95% CI: −0.024 to −0.004, P = .005) also showed significant negative associations. Other metabolites, such as total cholesterol levels in LDL (β = −0.013, 95% CI: −0.023 to −0.004, P = .008), cholesterol esters in medium LDL (β = −0.013, 95% CI: −0.023 to −0.003, P = .009), and the total cholesterol-to-total lipid ratio in small HDL (β = −0.013, 95% CI: −0.025 to −0.001, P = .035), were similarly associated. Moreover, the cholesteryl ester-to-total lipid ratio in small LDL (β = −0.013, 95% CI: −0.023 to −0.003, P = .015), total cholesterol in medium LDL (β = −0.012, 95% CI: −0.022 to −0.002, P = .016), and the total cholesterol-to-total lipid ratio in small LDL (β = −0.012, 95% CI: −0.023 to −0.002, P = .020) exhibited consistent trends. Additional findings included significant associations for total cholesterol in small LDL (β = −0.012, 95% CI: −0.022 to −0.002, P = .017), total lipids in large LDL (β = −0.012, 95% CI: −0.021 to −0.002, P = .023), free cholesterol in medium LDL (β = −0.011, 95% CI: −0.022 to −0.001, P = .038), and the concentration of large LDL particles (β = −0.011, 95% CI: −0.021 to −0.001, P = .030). In contrast, the phospholipid-to-total lipid ratio in small LDL (β = 0.012, 95% CI: 0.001–0.022, P = .028) and the phospholipid-to-total lipid ratio in small HDL (β = 0.015, 95% CI: 0.003–0.027, P = .017) were positively associated with FD. Summary of positive results was presented in Table 1. Full details of the additional metabolite associations were provided in Table S1, Supplemental Digital Content 1.

Figure 2.

Forest plot displaying the results of the two-sample Mendelian randomization analysis for circulating metabolites and functional dyspepsia. CI = confidence interval, HDL = high-density lipoprotein, IDL = intermediate-density lipoprotein, LDL = low-density lipoprotein, TSMR = two-sample Mendelian randomization, VLDL = very-low-density lipoprotein.

Table 1.

Causal relationships between functional dyspepsia and circulating metabolites in TSMR.

Exposures	Method	SNP numbers	β (95% CI)	OR (95% CI)	P value
Total cholesterol levels in small HDL	IVW	34	−0.021 (−0.039 to −0.003)	0.979 (0.962 to 0.997)	.020
Cholesteryl esters to total lipids ratio in IDL	IVW	46	−0.017 (−0.029 to −0.004)	0.983 (0.971 to 0.996)	.008
Cholesterol esters in small HDL	IVW	38	−0.016 (−0.031 to −0.001)	0.984 (0.969 to 0.999)	.032
Phospholipids to total lipids ratio in large VLDL	IVW	32	−0.016 (−0.033 to 0.000)	0.984 (0.968 to 1.000)	.047
Total cholesterol in large LDL	IVW	70	−0.015 (−0.025 to −0.005)	0.985 (0.975 to 0.995)	.003
Phospholipids in large LDL	IVW	71	−0.015 (−0.025 to −0.005)	0.985 (0.975 to 0.995)	.005
Cholesterol esters in large LDL	IVW	70	−0.014 (−0.024 to −0.004)	0.986 (0.976 to 0.996)	.005
Total cholesterol levels in LDL	IVW	69	−0.013 (−0.023 to −0.004)	0.987 (0.977 to 0.996)	.008
Cholesterol esters in medium LDL	IVW	61	−0.013 (−0.023 to −0.003)	0.987 (0.977 to 0.997)	.009
Total cholesterol to total lipids ratio in small HDL	IVW	56	−0.013 (−0.025 to −0.001)	0.987 (0.975 to 0.999)	.035
Cholesteryl esters to total lipids ratio in small LDL	IVW	55	−0.013 (−0.023 to −0.003)	0.987 (0.977 to 0.997)	.015
Cholesteryl esters to total lipids ratio in medium LDL	IVW	48	−0.012 (−0.023 to −0.002)	0.988 (0.977 to 0.998)	.025
Total cholesterol in medium LDL	IVW	63	−0.012 (−0.022 to −0.002)	0.988 (0.978 to 0.998)	.016
Total cholesterol to total lipids ratio in small LDL	IVW	58	−0.012 (−0.023 to −0.002)	0.988 (0.977 to 0.998)	.020
Total cholesterol in small LDL	IVW	68	−0.012 (−0.022 to −0.002)	0.988 (0.978 to 0.998)	.017
Total lipids in large LDL	IVW	72	−0.012 (−0.021 to −0.002)	0.988 (0.979 to 0.998)	.023
Free cholesterol in medium LDL	IVW	68	−0.011 (−0.022 to −0.001)	0.989 (0.978 to 0.999)	.038
Concentration of large LDL particles	IVW	70	−0.011 (−0.021 to −0.001)	0.989 (0.979 to 0.999)	.030
Concentration of medium LDL particles	IVW	68	−0.011 (−0.021 to −0.001)	0.989 (0.979 to 0.999)	.032
Cholesteryl esters to total lipids ratio in large LDL	IVW	54	−0.011 (−0.021 to 0.000)	0.989 (0.979 to 1.000)	.047
Phospholipids in IDL	IVW	71	−0.010 (−0.020 to 0.000)	0.990 (0.980 to 1.000)	.041
Free cholesterol in large LDL	IVW	71	−0.010 (−0.020 to 0.000)	0.990 (0.980 to 1.000)	.046
Phospholipids to total lipids ratio in small LDL	IVW	63	0.012 (0.001 to 0.022)	1.012 (1.001 to 1.022)	.028
Phospholipids to total lipids ratio in small HDL	IVW	55	0.015 (0.003 to 0.027)	1.015 (1.003 to 1.027)	.017
Histidine levels	IVW	20	0.031 (0.004 to 0.058)	1.032 (1.004 to 1.060)	.023
Phenylalanine levels	IVW	21	0.036 (0.009 to 0.063)	1.037 (1.009 to 1.065)	.009

CI = confidence interval, HDL = high-density lipoprotein, IDL = intermediate-density lipoprotein, IVW = inverse-variance weighted, LDL = low-density lipoprotein, OR = odds ratio, SNPs = single nucleotide polymorphisms, TSMR = two sample Mendelian randomization, VLDL = very-low-density lipoprotein.

Five CSF metabolites, tryptophan betaine (β = −0.018, 95% CI: −0.033 to −0.003, P = .016), Acisoga (β = 0.009, 95% CI: 0.001 to 0.017, P = .023), trimethylamine N-oxide (β = 0.017, 95% CI: 0.000 to 0.034, P = .050), X-15245 (β = 0.020, 95% CI: 0.008 to 0.033, P = .002), and alpha-tocopherol (β = 0.034, 95% CI: 0.003 to 0.064, P = .033) were also found to be associated with FD (Table 2; Fig. 3). Full details of the additional metabolite associations were provided in Table S2, Supplemental Digital Content 2.

Table 2.

Causal relationships between functional dyspepsia and CSF metabolites in TSMR.

Exposure	Method	SNP numbers	β (95% CI)	OR (95% CI)	P value
Tryptophan betaine levels	IVW	2	−0.018 (−0.033 to −0.003)	0.982 (0.967 to 0.997)	.016
Acisoga levels	IVW	15	0.009 (0.001 to 0.017)	1.009 (1.001 to 1.017)	.023
Trimethylamine n-oxide levels	IVW	2	0.017 (0.000 to 0.034)	1.017 (1.000 to 1.035)	.050
X-15245 levels	IVW	4	0.020 (0.008 to 0.033)	1.020 (1.008 to 1.033)	.002
Alpha-tocopherol levels	IVW	4	0.034 (0.003 to 0.064)	1.034 (1.003 to 1.067)	.033

CI = confidence interval, CSF = cerebrospinal fluid, IVW = inverse-variance weighted, OR = odds ratio, SNPs = single nucleotide polymorphisms, TSMR = two sample Mendelian randomization.

Figure 3.

Forest plot displaying the results of the two-sample Mendelian randomization analysis for cerebrospinal fluid metabolites and functional dyspepsia. CSF = cerebrospinal fluid, CI = confidence interval, TSMR = two-sample Mendelian randomization.

3.2. Multivariable MR

After performing MVMR analysis on the positive results, only the phospholipids to total lipids ratio in small HDL maintained a statistically significant causal relationship with FD (β = 0.287, 95% CI: 0.058 to 0.515, P = .014) among circulating metabolites (Table 3; Fig. 4). For CSF metabolites, 3 remained causally associated with FD: Acisoga levels (β = 0.012, 95% CI: 0.003 to 0.020, P = .006), X-15245 levels (β = 0.019, 95% CI: 0.007 to 0.031, P = .003), and alpha-tocopherol levels (β = 0.037, 95% CI: 0.004 to 0.070, P = .027; Table 4; Fig. 5).

Table 3.

Causal relationships between functional dyspepsia and circulating metabolites in MVMR.

Exposure	SNP numbers	β (95% CI)	OR (95% CI)	P value
Cholesterol esters in medium LDL	127	−2.283 (−5.757 to 1.192)	0.102 (0.003 to 3.294)	.198
Total cholesterol levels in LDL	127	−1.745 (−4.360 to 0.869)	0.175 (0.013 to 2.385)	.191
Concentration of large LDL particles	127	−1.016 (−4.132 to 2.100)	0.362 (0.016 to 8.166)	.523
Phospholipids in large LDL	127	−0.938 (−2.225 to 0.348)	0.391 (0.108 to 1.417)	.153
Free cholesterol in large LDL	127	−0.409 (−1.802 to 0.984)	0.664 (0.165 to 2.676)	.565
Cholesterol esters in large LDL	127	−0.409 (−3.797 to 2.980)	0.665 (0.022 to 19.686)	.813
Total cholesterol to total lipids ratio in small LDL	127	−0.296 (−0.831 to 0.239)	0.743 (0.435 to 1.270)	.278
Phospholipids to total lipids ratio in small LDL	127	−0.253 (−0.570 to 0.063)	0.776 (0.566 to 1.065)	.117
Free cholesterol in medium LDL	127	−0.167 (−1.119 to 0.785)	0.846 (0.327 to 2.192)	.731
Phospholipids to total lipids ratio in large VLDL	127	−0.127 (−0.258 to 0.003)	0.880 (0.773 to 1.003)	.055
Total cholesterol levels in small HDL	127	−0.111 (−0.448 to 0.227)	0.895 (0.639 to 1.254)	.520
Total cholesterol in small LDL	127	−0.071 (−0.949 to 0.807)	0.932 (0.387 to 2.241)	.874
Histidine levels	127	0.006 (−0.027 to 0.039)	1.006 (0.974 to 1.039)	.717
Cholesteryl esters to total lipids ratio in large LDL	127	0.008 (−0.426 to 0.442)	1.008 (0.653 to 1.556)	.971
Phenylalanine levels	127	0.031 (−0.004 to 0.065)	1.031 (0.996 to 1.067)	.081
Cholesteryl esters to total lipids ratio in medium LDL	127	0.033 (−0.503 to 0.570)	1.034 (0.605 to 1.768)	.904
Cholesteryl esters to total lipids ratio in IDL	127	0.045 (−0.100 to 0.190)	1.046 (0.905 to 1.209)	.545
Cholesterol esters in small HDL	127	0.203 (−0.179 to 0.586)	1.225 (0.836 to 1.796)	.298
Total cholesterol to total lipids ratio in small HDL	127	0.206 (−0.085 to 0.497)	1.229 (0.919 to 1.644)	.165
Cholesteryl esters to total lipids ratio in small LDL	127	0.218 (−0.348 to 0.783)	1.243 (0.706 to 2.188)	.451
Phospholipids in IDL	127	0.260 (−0.578 to 1.099)	1.297 (0.561 to 3.000)	.543
Phospholipids to total lipids ratio in small HDL	127	0.287 (0.058 to 0.515)	1.332 (1.060 to 1.674)	.014
Concentration of medium LDL particles	127	0.606 (−1.364 to 2.577)	1.834 (0.256 to 13.151)	.546
Total lipids in large LDL	127	1.429 (−1.995 to 4.853)	4.174 (0.136 to 128.122)	.413
Total cholesterol in large LDL	127	2.156 (−2.360 to 6.672)	8.637 (0.094 to 789.901)	.349
Total cholesterol in medium LDL	127	2.422 (−1.851 to 6.696)	11.271 (0.157 to 809.141)	.267

CI = confidence interval, HDL = high-density lipoprotein, IDL = intermediate-density lipoprotein, LDL = low-density lipoprotein, OR = odds ratio, MVMR = multivariable Mendelian randomization, SNPs = single nucleotide polymorphisms, VLDL = very-low-density lipoprotein.

Figure 4.

Forest plot displaying the results of the multivariable Mendelian randomization analysis for circulating metabolites and functional dyspepsia. CI = confidence interval, HDL = high-density lipoprotein, IDL = intermediate-density lipoprotein, LDL = low-density lipoprotein, MVMR = multivariable Mendelian randomization, VLDL = very-low-density lipoprotein.

Table 4.

Causal relationships between functional dyspepsia and CSF metabolites in MVMR.

Exposures	SNP number	β (95% CI)	OR (95% CI)	P value
Tryptophan betaine levels	26	−0.009 (−0.023 to 0.004)	0.991 (0.977 to 1.004)	.176
Acisoga levels	26	0.012 (0.003 to 0.020)	1.012 (1.003 to 1.020)	.006
Trimethylamine n-oxide levels	26	0.015 (−0.002 to 0.033)	1.016 (0.998 to 1.033)	.083
X-15245 levels	26	0.019 (0.007 to 0.031)	1.019 (1.007 to 1.032)	.003
Alpha-tocopherol levels	26	0.037 (0.004 to 0.070)	1.038 (1.004 to 1.073)	.027

CI = confidence interval, CSF = cerebrospinal fluid, MVMR = Multivariable Mendelian randomization, OR = odds ratio, SNPs = single nucleotide polymorphisms.

Figure 5.

Forest plot displaying the results of the multivariable Mendelian randomization analysis for cerebrospinal fluid metabolites and functional dyspepsia. CSF = cerebrospinal fluid, CI = confidence interval, MVMR = multivariable Mendelian randomization.

3.3. Robustness of the results

Sensitivity analyses of two-sample MR showed no heterogeneity or horizontal pleiotropy in the main positive results. Phospholipids to total lipids ratio in small HDL (Q test P [IVW] = .449, Egger test P = .179, Global test P = .438), Acisoga levels (Q test P [IVW] = .963, Egger test P = .883, Global test P = .962), X-15245 levels (Q test P [IVW] = .721, Egger test P = .969, Global test P = .726), and Alpha-tocopherol levels (Q test P [IVW] = .842, Egger test P = .468, Global test P = .841). Likewise, multivariate sensitivity analyses also revealed the absence of heterogeneity and horizontal pleiotropy. Phospholipids to total lipids ratio in small HDL (Q test P [IVW] = .486, Egger test P = .516), Acisoga levels (Q test P [IVW] = .813, Egger test = 0.844), X-15245 levels (Q test P [IVW] = .813, Egger test P = .844), and Alpha-tocopherol levels (Q test P [IVW] = .813, Egger test P = .844).

4. Discussion

Our study revealed statistically significant causal associations between 26 circulating metabolites, 5 CSF metabolites, and FD. Through MVMR, we further confirmed the causal relationships between the phospholipid-to-total lipid ratio in HDL, Acisoga levels, X-15245 levels, alpha-tocopherol (α-tocopherol) levels, and metabolic FD.

Studies investigating the relationship between metabolites and gastrointestinal diseases have been increasingly conducted.^[24,25] To date, we were the first to reveal a significant association between human metabolites and the severity of FD. Previous research on HDL has predominantly focused on its role in cardiovascular health, particularly the lipid metabolism of larger HDL particles.^[26,27] Our study broke new ground by focusing on smaller HDL subfractions, demonstrating that an elevated phospholipid ratio in small HDL is significantly correlated with the occurrence of FD. These findings suggest that lipid metabolism may not only influence the cardiovascular system but also play a critical role in gastrointestinal function. Although the specific role of phospholipids in small HDL remains unclear, our findings suggest that this lipid-related pathway may be involved in the pathophysiology of FD. Further studies are warranted to clarify the biological mechanisms underlying this association and to determine whether alterations in HDL lipid composition are relevant to dyspeptic symptoms.

Additionally, Acisoga was identified as another metabolite significantly associated with FD. Through untargeted metabolomics analysis that Acisoga levels were significantly elevated in patients with FD. Although previous studies have suggested a link between Acisoga and cardiovascular disease,^[28] we were the first to confirm its association with FD in MR study. This finding provided compelling evidence for a new role of Acisoga in gastrointestinal health, indicating that it may be involved in metabolic pathways that contribute to dyspepsia. The potential mechanisms underlying Acisoga influence could open up new avenues for treatment, particularly in exploring how targeted interventions on this metabolite may improve gastrointestinal symptoms. Further mechanistic studies will help gain a deeper understanding of Acisoga specific role in the digestive system and pave the way for the development of more precise therapeutic strategies.

Moreover, α-tocopherol, a known antioxidant,^[29] has been found to be closely associated with the occurrence of FD, with lower serum α-tocopherol levels observed in patients compared to healthy controls. This suggests that α-tocopherol may exert a protective effect on gastrointestinal function by reducing oxidative stress and inflammation. Although previous research has primarily focused on α-tocopherol role in alleviating exercise-induced oxidative stress, its potential benefits for gastrointestinal health warrant further investigation,^[24,30] our study was the first to validate its protective potential in the context of human gastrointestinal diseases. The role of α-tocopherol in gut health extended beyond its basic antioxidant function, highlighting its potential as a therapeutic option for managing FD. Future research should explore how α-tocopherol can be leveraged to reduce oxidative stress and inflammation, thereby alleviating dyspeptic symptoms and enhancing patient outcomes.

The strength of this study lies in the fact that the possible potential causal relationships between circulating metabolites, CSF metabolites, and FD were revealed via MR, and possible confounding between metabolites was eliminated by multifactorial MR analysis. However, the limitations of this study should be noted. Firstly, only the GWAS data source of European group was used for analysis, which implied that the conclusions of the study may have some application constraints in other populations.^[31] Otherwise, it was acknowledged that the development of diseases was basically determined by the combination of environmental and genetic variations,^[32] while the results of our study were mainly derived from the latter. The study does not take into account the fact that, in the case of diseases, the development of disease is determined by a combination of both the environment and genetic variation. No possible effects of environmental factors were considered to contribute to the metabolic reprogramming in humans, which may result in limitations in our conclusions to some extent.

5. Conclusion

Our results pointed out the causal associations between the alteration of 26 circulating metabolites, 5 CSF metabolites and FD. Further MVMR confirmed the statistically significant causality between FD and the ratio of phospholipids to total lipids in small HDL, Acisoga, X-15245, and α-tocopherol. Our study linked metabolites and FD together, which was expected to expand our cognition of FD and other digestive diseases with metabolic levels of the organism.

Acknowledgments

The authors wish to acknowledge the participants and investigators of the Original GWAS.

Author contributions

Conceptualization: Liqiong Liu, Lijuan Zhang.

Data curation: Lijuan Long.

Formal analysis: Lijuan Long, Lijuan Zhang.

Methodology: Mengyu Li.

Software: Mengyu Li.

Visualization: Liqiong Liu.

Writing – original draft: Lijuan Long, Mengyu Li, Liqiong Liu, Lijuan Zhang.

Writing – review & editing: Lijuan Long, Lijuan Zhang.