Investigating the causal role of serum metabolites in substance use disorder risk: a study integrating Mendelian randomization and synthesis analysis

Abstract

Objective

This study aimed to explore how serum metabolites affect the risk of substance use disorders (SUD).

Methods

In the initial stage, Mendelian randomization was applied to assess the relationship between 1,400 serum metabolites and SUD. Inverse variance weighting, the Wald ratio odds ratio, and 95% confidence intervals were primarily used to evaluate causal relationships, and the false discovery rate was used for multiple comparison corrections. Sensitivity analysis was conducted via Cochran’s Q test and MR-PRESSO. The MR-Steiger test was used to examine reverse causality. In the validation stage, we sought additional GWAS data on SUD to verify the initial results. Furthermore, the pathway enrichment analysis was conducted for known metabolites that exhibited causal relationships with SUD in both phases.

Results

In the initial phase, we analysis suggests that these 77 metabolites may have potential causal associations with SUD, including 14 metabolite ratios and 63 metabolites (49 known and 14 unknown). In the validation phase, for 57 metabolites (38 known, 6 unknown, 13 ratios), confirmed associations may indicate causal effects on SUD incidence. The synthesis analysis results indicated that the overall effect of the combined metabolites was consistent with the primary analysis with two identified as risk factors and four as protective factors for SUD. Specifically, Erythronate levels, 1-(1-enyl-stearoyl)-2-oleoyl-GPE (P-18:0/18:1) levels, aspartate to citrulline ratios, and cis-4-decenoate (10:1n6) levels were negatively correlated with SUD, whereas gamma-glutamyl-alpha-lysine and ethyl alpha-glucopyranoside levels were positively correlated with disease incidence. The metabolites linked to the risk of SUD in both phases were primarily enriched in several metabolic pathways, including pantothenate and CoA biosynthesis; pyrimidine metabolism; biosynthesis of valine, leucine, and isoleucine; taurine and hypotaurine metabolism; histidine metabolism; and glycerolipid metabolism.

Conclusion

Circulating metabolites may have a causal relationship with the risk of SUD. “Specific metabolites may be potential biomarkers for SUD, contributing to risk prediction and the development of personalized treatment strategies”.

Supplementary Information

The online version contains supplementary material available at 10.1007/s44192-025-00275-6.

Introduction

Substance use disorders (SUD) are characterized by significant distress or suffering resulting from the use of one or more substances. Drug overdose deaths are at historic levels, with more than 106,000 drug overdose deaths in 2021 compared to just under 68,000 in 2018 [1]. According to the World Health Organization, an alarming 700,000 deaths occur globally each year due to SUD-related problems, suggesting that the dangers of SUD cannot be ignored [2]. The impact of SUD varies significantly across countries and regions, and the analysis of epidemiological data from developed versus developing countries can contribute to an in-depth understanding of the global burden of the disease and the factors that influence it. SUD are complex phenomena. The Diagnostic and Statistical Manual of Mental Disorders (DSM-5) criteria for diagnosing SUD are divided into impaired control over substance use (1–4), social impairment (5–7), risky use (8–9), and pharmacological criteria (10–11) [3]. Patients may be diagnosed with SUD from various substances, which the DSM-5 identifies as alcohol, caffeine, cannabis, hallucinogens, inhalants, opioids, sedatives, hypnotics, stimulants, and tobacco. The DSM-5 specifies severity modifiers, including mild, moderate, and severe. The development of substance-related problems is influenced by a range of factors, including broader social and economic conditions and individual genetic variations. Previous studies using multifactorial indices of clinical and environmental risk factors (e.g., childhood adversity, family history of SUD, early substance exposure, and frequent use during adolescence) have revealed that these indices help identify patients with persistent SUD. However, research on blood metabolic changes related to SUD is limited [4].

The development of SUD may be closely linked to abnormal changes in serum metabolite levels. Metabolomics studies have revealed that serum biomarkers associated with neuroinflammation, energy metabolism, and lipid metabolism, which may contribute to SUD pathogenesis by affecting central nervous system function [5]. Additionally, lipid metabolism disorders observed in conditions such as hyperuricemia, including perturbations in glycerophospholipid and sphingolipid metabolism, have been associated with chronic inflammatory states. Given that chronic inflammation is established to influence neuroplasticity in addiction, we hypothesize these metabolic alterations may represent a potential biological link to SUD pathogenesis. However, direct causal validation in SUD cohorts remains to be established [6]. Additionally, perturbations in glycerophospholipid and sphingolipid metabolism—observed in conditions like hyperuricemia—are established inducers of chronic inflammation. While chronic inflammation is mechanistically linked to impaired neuroplasticity in addiction, the direct involvement of this lipid-inflammation axis in SUD requires validation through causal frameworks [7, 8]. Notably, dysregulation of amino acid metabolism (e.g., glutamate and alanine) and energy metabolism (e.g., tricarboxylic acid cycle) can exacerbate substance dependence by altering neurotransmitter balance and mitochondrial function [9]. Although direct metabolomics studies on SUD remain limited, insights from metabolic pathway analyses of other neuropsychiatric disorders suggest that systematic screening of serum lipids, amino acids, and inflammation-related metabolites could provide new directions for understanding SUD metabolomics and identifying biomarkers.

Mendelian randomization (MR) is an epidemiological method that uses genetic variation, primarily single nucleotide polymorphisms (SNPs), as instrumental variables (IVs) to replace the target exposure variable, with the aim of assessing the causal relationship between exposure and specific outcomes [10–12]. Genome-wide association studies have linked circulating metabolites to SNPs. Mendelian randomization was used to investigate the causal relationship between circulating metabolites and SUD. Our study focused on identifying metabolites that may contribute to the onset of SUD and discovering potential targets for assessing disease activity, diagnosis and treatment.

Subjects and methods

Study design

To clarify the causal relationship between serum metabolites and the risk of SUD, a two-sample MR approach was performed to screen for causal relationships between 1,400 metabolites and SUD. Additionally, several sensitivity analyses were conducted to confirm the robustness of the results. Subsequently, SUD GWAS data from other sources were used to validate the initial phase results. Finally, a synthesis analysis was performed to combine the effect sizes of metabolites with the causal effects of the two SUD outcomes, leading to reliable causal inferences (Fig. 1).

Fig. 1.

A illustrates the three key hypotheses proposed in the study and their interrelationships; B shows in detail the flow of the analysis from the initial phase to the replication phase and the subsequent metabolic pathway enrichment analysis and synthesis analysis. Abbreviations: IVW, inverse-variance weighted; WM, weighted median

Data sources

The summary statistics of the serum metabolites used in this study were derived from the Canadian Longitudinal Study of Aging (CLSA) [13], which conducted genome-wide genotyping and assessment of circulating plasma metabolites in 8,299 unrelated European individuals. Assertions of rigorous pre-GWAS genotype quality control must be accompanied by a detailed description of the quality control procedures employed [14]. Following stringent pre-GWAS genotype quality control, approximately 15.4 million SNPs were retained for GWAS assessment. The levels of 1,458 metabolitesin plasma samples were quantified usingthe Ultrahigh Performance Liquid Chromatography-Tandem Mass Spectroscopy (UPLC-MS/MS) platform. After standardization, including the removal of artifacts representing system noise, mismatches, and background noise, a strict quality control process identified 1,091 metabolites (850 known substances and 241 unknown entities) and 309 metabolite ratios for genome-wide association analysis. The 850 metabolites withestablished identities were categorized into eight super-pathways: carbohydrates, amino acids, nucleotides, vitamins, lipids, peptides, energy metabolites, and xenobiotics. Given that many metabolites serve as substrates and products of enzymatic reactions, studying the genetic determinants of substrate‒product ratios can increase the understanding of broader biological processes; thus, 309 metabolite ratios were also included in this study [15].

The initial phase outcome data came from the FinnGen biobank analysis consortium, which included 7,232 SUD patients and 385,885 healthy controls [16]. Validation data were obtained from a study published by Jiang L in 2021, which included 236 SUD patients and 49,112 healthy controls [17]. The genetic data used in this study complied with ethical standards and were approved by the local ethics review committee, which represents legally publicly available research data. As this study was based on publicly available GWAS summary data, missing data processing was completed in the original study. No additional missing data processing was performed in this study.

Selection of instrumental variables

Instrumental variables were identified via the following steps: (1) Genetic variations linked to metabolic traits were identified at a study-wide significance threshold (P < 5 × 10⁻⁷) for metabolomic traits [18]. (2) Independent SNPs were chosen as instrumental variables on the basis of a linkage disequilibrium assessment. When screening tool variables, we excluded SNPs with linkage disequilibrium (LD) effects. Specifically, within a genomic distance of 5,000 kb, only one SNP was retained if the LD coefficient r² between SNPs was greater than 0. 01 [19]. (3) The strength of each SNP was assessed by calculating the F statistic, retaining only those with values greater than 10. The F statistic was calculated via the formula F = R²(N − K−1)/K(1 − R²). Here, R² represents the proportion of variance in the exposure explained by the instrumental variable and is calculated as R² = [2EAF(1 − EAF)β²]/[2EAF(1 − EAF)β²+ EAF(1 − EAF)NSE²]. Here, N is the sample size of the exposure, K is the number of instrumental variables, EAF is the effect allele frequency, β is the effect size, and SE is the standard error of the effect size [20, 21].

Two-sample Mendelian randomization

This study primarily employed inverse variance weighting (IVW), weighted median (WM), MR‒Egger, weighted mode, simple mode, and Wald ratio methods for the analysis. The IVW method assumes that all SNPs are valid and independent, constraining the regression intercept to zero and using the inverse of the variance of the outcome as the fitting weight. When each genetic variation meets the assumption of instrumental variables, IVW is assumed to provide the most accurate results; therefore, IVW results are typically regarded as the primary standard for assessing causal effects [18, 19]. In this study, the P-value from IVW was used as the main indicator to evaluate the causal relationship between exposure and outcome. If only one instrumental variable was obtained, the P-value from the Wald ratio was used for evaluation. Other methods were used to supplement the assessment of the MR results, and the consistent direction of effect sizes (β values) across different methods indicated robust findings. The false discovery rate (FDR) correction was applied to adjust for multiple tests of all P-values obtained from the IVW and Wald ratio methods. An FDR-adjusted P-value (q-value) < 0.05 provides statistically significant evidence supporting a potential causal association. Conversely, a q-value ≥ 0.05 indicates insufficient evidence for causality; such results require cautious interpretation as they may represent false negatives or underpowered associations. It is important to note that a single P-value or FDR value alone does not provide definitive causal evidence. The reliability of Mendelian randomization analysis also depends on the validity of the instrumental variables, sample size, and the results of sensitivity analyses.

Sensitivity analyses

Various methods were used to assess the heterogeneity and pleiotropy of the results. The P-value of Cochran’s Q test was greater than 0.05, indicating that no significant evidence of heterogeneity was found; [22]. The MR-Egger regression intercept term and MR-PRESSO test P-value were greater than 0.05, indicating that no significant evidence of horizontal pleiotropy was found [23, 24]. Notably, when the number of SNPs is insufficient (3 for MR-PRESSO and 2 for MR‒Egger), Pleiotropy testing was not conducted.

The Steiger test was utilized to evaluate the directionality of the instrumental variable’s impact on the outcome variable, The MR-Steiger directionality test is used to detect evidence of a reverse causal relationship between exposure and outcome, to ensure that the results support the initial hypothesis. This step is crucial for verifying the alignment of the findings with the initial hypothesis. Specifically, a P-value exceeding 0.05 for the instrumental variable is indicative of reverse causation [25, 26].

Replication stage and synthesis analysis

We sought additional sources of outcome data for SUD and conducted MR validation on the initial stage results under the same conditions, retaining metabolites and metabolite ratios that still exhibited a causal effect with SUD. Subsequently, we merged the two MR results (odds ratios, OR; 95% confidence intervals, CI) through a synthesis analysis. If the merged results showed heterogeneity, indicated by an I² value greater than 50%, a random-effects model was utilized; otherwise, a fixed-effects model was applied. The statistical threshold for the synthesis analysis was set at 0.05. The aforementioned study was primarily conducted using the Two Sample MR and meta packages within the R statistical software, version 4.3.1, with a significance level (alpha) of 0.05.

Enrichment analysis of metabolic pathways

The HMDB IDs of known metabolites were retrieved from the Human Metabolome Database (https://hmdb.ca/), and enrichment analysis was conducted on the metabolic pathways associated with these metabolites. The analysis was performed via MetaboAnalyst 5.0 (https://www.metaboanalyst.ca/). The pathway libraries selected for this analysis were the Small Molecule Pathway Database (SMPDB) and the Kyoto Encyclopedia of Genes and Genomes (KEGG). The enrichment method used was the hypergeometric test, and the significance level for metabolic pathway analysis was set at 0.01.

Results

Initial-phase Mendelian randomization analysis results

Among the 1,400 metabolites, 77 metabolites exhibited statistically supported evidence for a potential causal relationship with SUD risk (IVW-FDR < 0.05), consisting of 14 metabolite ratios and 63 individual metabolites (including 49 identified and 14 unidentified metabolites) (Fig. 2). After adjusting for the FDR, six metabolites or metabolite ratios remained significantly associated with SUD. In addition, among the 139 SNPs that remained significantly associated with six phenotypes after FDR correction, some SNPs were also significantly associated with SUD, but not all (see Supplementary Table 1).

Fig. 2.

Ahows the classification of the 77 metabolites identified in the initial analysis as potentially causally associated with SUD, including categories such as lipids, amino acids, and exogenous substances

The 49 identified metabolites primarily consisted of lipids, xenobiotics, and amino acids. Among the 38 analyzed amino acid metabolites, erythronate had the strongest protective effect (OR: 0.85, 95% CI 0.76–0.95, P_{Wald ratio}= 0.006, P_FDR = 0.035). In contrast, valine levels had the strongest effect on risk (OR: 1.21,95% CI 1.05–1.39; P_{Wald ratio}= 0.008, P_FDR = 0.035).

Among the 14 xenobiotics, elevated X-17,351 was associated with the most significant positive causal relationship with SUD incidence (OR: 1.21, 95% CI 1.08–1.38, P_{Wald ratio}= 0.001, P_FDR= 0.035). Conversely, elevated X-24,295 exhibited the most significant negative causal relationship (OR: 0.83, 95% CI 0.73–0.94, P_{Wald ratio}= 0.005, P_FDR= 0.035). Among the 11 lipid metabolites, elevated sphingomyelin (d18:1/17:0, d17:1/18:0, d19:1/16:0) levels had the most significant positive causal relationship with SUD incidence (OR: 1.11, 95% CI 1.02–1.22, P_{Wald ratio}= 0.023, P_FDR= 0.039), whereas elevated glycerol levels were associated with the most significant negative causal relationship (OR: 0.88, 95% CI 0.79–0.98, P_{Wald ratio}= 0.016, P_FDR= 0.036). Can be found in Supplementary Table 2.

Among the various analytical methods, the direction of effect estimates for most metabolites was consistent, enhancing the reliability of the results. The specific MR analysis results for the different methods are shown in Fig. 3 and Supplementary Table 2.

Fig. 3.

Presents the results of five Mendelian randomization analysis methods (IVW, MR-Egger, weighted median, weighted pattern, and simple pattern) for analyzing the causal association between metabolites and SUD. The depth of color indicates the consistency of effect size and direction

Sensitivity and reverse causal analysis

Cochran’s Q test (via TwoSampleMR) and the MR-PRESSO global test were used to evaluate heterogeneity and detect potential pleiotropy among the selected instrumental variables. The MR-Steiger directional test confirmed the accuracy of the exposure direction of the results, with all instrumental variables showing P Steiger < 0.05. P-values for testing the reverse relationship between each metabolite and SUD were all less than 0.05, suggesting no evidence of potential reverse causation. Can be found in Supplementary Table 3.

Replication stage Mendelian randomization results and synthesis analysis

Among the 57 metabolites identified as having a causal relationship with SUD, six were successfully validated, including one metabolite ratio and five known metabolites. Figure 5A shows the sensitivity analysis results for the six metabolites. Detailed information on the main MR results for the six metabolites is shown in Fig. 4 and Supplementary Table 4.

Fig. 5.

A Displays the heterogeneity and pleiotropy analysis results of the 6 metabolites that were successfully validated. B Enriched pathways of the metabolites in the KEGG database and C Enriched pathways of the metabolites in the SMPDE database

Fig. 4.

Presents the synthesis analysis results for six metabolites significantly associated with SUD, including effect size, 95% confidence interval, and p-value

In the MR analysis, heterogeneity tests for the six metabolites yielded Cochran’s Q P > 0. 1, suggesting that any between-instrument heterogeneity, if present, was not detected at conventional power levels. MR analysis confirmed a stable causal relationship between these metabolites and SUD (P < 0.05). The synthesis analysis indicated that among the six metabolites, increased ethyl alpha-glucopyranoside levels were a significant risk factor for SUD (OR: 1.87, 95% CI 1.11–3.14; P < 0.001). Increased erythronate levels were a strong protective factor against SUD (OR: 0.47, 95% CI 0.26–0. 88; P < 0.001). Additionally, 1-(1-enyl-stearoyl)-2-oleoyl-GPE (P-18:0/18:1) levels (OR: 0.52, 95% CI 0.25–0.96, P < 0.001) and Cis-4-decenoate (10:1n6) levels (OR: 0.47, 95% CI 0.26–0. 88, P < 0.001) were negatively correlated with an increased risk of SUD. Further details are provided in Fig. 4.

Results of metabolic pathway enrichment analysis

We queried the HMDB IDs of 49 known metabolites associated with SUD risk in the initial phase, including five from the replication phase, and performed metabolic pathway enrichment analysis on 21 identifiable compounds, with one from the replication phase. The enrichment results from both phases highlighted the major metabolic pathways, such as glutathione metabolism, biosynthesis of unsaturated fatty acids, α-linolenic acid and linoleic acid metabolism (initial phase), glycerophospholipid metabolism, and arachidonic acid metabolism (replication phase). The limited number of metabolites prevented the SMPDB from conducting an enrichment analysis during the replication phase. The specific enrichment results are shown in Fig. 5B and C.

Discussion

To our knowledge, no study has investigated the causal relationship between serum metabolites and SUD using MR analysis. Our study pioneers the systematic exploration of the association between serum metabolites and the risk of SUD through MR analysis, complemented by pathway enrichment analysis. We strengthened our preliminary findings with validation analyses and meta-analyses, identifying six metabolites causally linked to the risk of developing SUD. The known metabolites are predominantly involved in lipid metabolism. Pathway analysis highlighted the importance of glutathione metabolism, alpha-linolenic and linoleic acid metabolism, and the biosynthesis of unsaturated fatty acids. We reviewed the literature on 16 known metabolites and summarized the reported roles of nine searchable metabolites in different diseases (Supplementary Table 5).

In this study, we revealed the causal relationship between multiple metabolites and SUD through a Mendelian randomization approach combined with data from a large-scale genome-wide association study (GWAS) of serum metabolites and SUD(substance use disorder). Erythronate is a metabolite associated with glucose metabolism, and its metabolic process in the body may be related to energy metabolism and maintenance of cellular functions. In glucose metabolism, erythronate levels may reflect the efficiency of cellular glucose utilization and the ability to produce energy. Lower erythronate levels may be associated with disruptions in cellular energy metabolism, which can affect nervous system function and cognitive or behavioral regulation. The reduction of this metabolite may be related to the occurrence of SUD, as it may impact the brain’s reward system and self-control abilities, thereby increasing susceptibility to SUD. In a metabolomics study on renal function by Sekula, 493 small molecules in human serum were quantified using an untargeted metabolomics approach using gas and liquid chromatography coupled with mass spectrometry [27]. Erythronate was found to be significantly correlated with the glomerular filtration rate, estimated based on creatinine and cystatin C levels [27]. This suggests that erythronate may be a potential marker of renal function that could help improve estimates of glomerular filtration rate. Furthermore, in the field of lung cancer research, erythronate has also been found to have potential diagnostic and therapeutic applications, indicating its broader relevance beyond metabolic and renal studies. A study using an untargeted metabolomics approach to analyze A549 lung cancer cells identified erythronate as one of the metabolites labeled from [U-13C6] glucose. Erythronate levels are significantly elevated in lung cancer tissues and cancer cell lines, suggesting that erythronate may serve as a biomarker for cancer [28]. In addition, erythronate was listed as one of the top 10 blood metabolites based on a multiplicative increase in erythronate levels in a study by Mindikoglu, which was associated with the metabolomic profiles of hepatic and renal dysfunction [29]. Erythronate showed a negative correlation with substance use disorder (SUD) in this study, suggesting that it may act as a protective factor against SUD. In clinical applications, erythronate level detection can be used as a biomarker for SUD risk assessment. By monitoring changes in erythronate levels, high-risk individuals can be identified in advance, and timely interventions can be implemented. In addition, the metabolic pathways of erythronate involve glycolysis and the pentose phosphate pathway, providing a theoretical basis for the development of drugs targeting these metabolic pathways.

As a phospholipid, 1-(1-enyl-stearoyl)-2-oleoyl-GPE (P-18:0/18:1) is negatively correlated with substance use disorder (SUD), indicating that it may play a protective role in SUD development. This metabolite belongs to the phospholipid class and is an important component of cell membranes. A study found that 1-(1-enyl-stearoyl)-2-oleoyl-GPE (P-18:0/18:1) may play a mediating role in the causal relationship between the gut microbiota and Alzheimer’s disease (AD), as revealed by a bidirectional two-sample mediated Mendelian randomization analysis [30]. Specifically, this metabolite may be involved in the association between Lactobacillaceae and AD [27]. GPE is an alcohol-derived phospholipid formed from phosphatidylcholine and can be used to assess alcohol consumption in the past few weeks [31]. One study measured PEth by tandem mass spectrometry. subject work characteristics (ROC) and column tabulation analyses were used to assess PEth in detecting the presence of alcohol consumption and an average of 4 or more drinks per day in the past 30 days. PEth can be used as a biomarker of alcohol use in alcohol-dependent SUD patients for response to treatment and monitoring of abstinence effects, alcohol withdrawal, and ongoing alcohol use Assessment.

Cis-4-Decenoate (10:1n6) is an unsaturated fatty acid belonging to the class of long-chain polyunsaturated fatty acids (PUFAs). In metabolic studies, cis-4-decenoate (10:1n6) was negatively correlated with substance use disorder (SUD), indicating that it may play a protective role in SUD. Lower levels of cis-4-decenoate (10:1n6) may be associated with disorders of fat metabolism, which can affect the fluidity of cell membranes and signal transduction functions, potentially contributing to increased susceptibility to substance use. It is also been found to be associated with various physiological processes. In a study of neoagarotetraose in a mouse model of dextran sodium sulfate-induced colitis, cis-4-decenoate (10:1n6) was found to induce significant metabolic changes in this model, suggesting that it may play an important role in the metabolic processes associated with intestinal inflammation [32, 33]. Cis-4-decenoate (10:1n6), a long-chain monounsaturated fatty acid, in studies on lipid metabolism, this compound has been identified as one of the metabolites with significant changes, indicating that it may be involved in energy metabolism and cellular signaling [29]. In the future, it may be possible to improve neuroprotective effects and energy metabolism by modulating metabolic pathways (e.g., supplementation with medium-chain fatty acids) for SUD treatment to alleviate anxiety and seizure risk during withdrawal.

The positive correlation between gamma-glutamyl-alpha-lysine and ethyl-alpha-glucopyranoside with SUD suggests that they may play a contributory role in the development of SUD. Gamma-glutamyl-alpha-lysine (GGL) is a heteropeptide-linked compound formed from glutamine and lysine catalyzed by transglutaminase (TGase). This compound plays an important role in various biological processes, particularly in protein cross-linking and extracellular matrix stabilization [30]. Gamma-glutamyl-alpha-lysine also shows potential as a biomarker for SUD, suggesting its relevance in the pathophysiology and diagnosis of the disorder [34, 35].

We used Mendelian Randomization (MR) analysis to explore the causal links between metabolite levels and SUD. Though our analysis suggested that some metabolites might influence SUD, caution is needed in interpreting these results due to the complex, bidirectional relationship between them. Substance use can directly alter metabolic pathways. For example, chronic alcohol use affects liver metabolism, changing lipid and amino acid profiles, while stimulant use influences neurotransmitter levels and energy metabolism. Conversely, certain metabolic profiles may predispose individuals to SUD. Variations in neurotransmitter metabolites could affect the brain’s reward system, influencing susceptibility to substance use and addiction. Moreover, genetic variants can influence both metabolic pathways and the risk of developing SUD, creating a complex web of interactions. Environmental factors, such as diet, physical activity, and stress, can also modulate metabolic profiles and contribute to the risk of SUD. Given these complexities, our MR analysis has limitations in fully capturing the bidirectional nature of these interactions. Future research should include longitudinal studies to track changes in both metabolite levels and SUD status over time, multi-omics approaches to offer a more comprehensive understanding of the underlying biological mechanisms, experimental models to elucidate the direct effects of substances on metabolic pathways, and large-scale meta-analyses to provide more robust estimates of the relationships [36].

Our study had several limitations. First, the strict screening criteria we employed (P < 5 × 10⁻⁸, r² < 0.001 within a 10,000 kb distance) hindered the extraction of SNPs for many metabolites. After broadening the criteria (P < 5 × 10 ⁻⁷, r < 20.01 within a 5,000 kb distance), some metabolites still presented few SNPs, potentially leading to biased results [37]. Second, many metabolites screened initially failed to achieve statistical significance after strict multiple testing adjustments. Third, the population data came from individuals of European ancestry, and the results may differ when the findings are generalized to other ethnicities. More individuals with different genetic backgrounds should be analyzed to improve the generalizability of the results. In addition, the sample size of the cyclic metabolite GWAS data used in this study was relatively small (N = 8,299), which was significantly insufficient compared to large-scale GWAS studies of other complex traits (such as height and body mass index). This may have led to reduced statistical power, and some true associations between metabolites may not have been effectively identified. Therefore, future studies require larger sample sizes of metabolome GWAS data to further validate the findings of this study. Although IVW indicated significant associations for six metabolites after multiple-testing correction, effect directions were not entirely consistent across MR-Egger,

weighted median and weighted mode, and some P-values did not survive correction.These limitations should be borne in mind before considering these metabolites as ‘validated’ causal factors for SUD. Future studies require larger sample sizes and stricter validation criteria to further confirm these findings. Finally, the use of genetic variations related to metabolites as instrumental variables reflects long-term exposure, whereas short-term dietary supplementation may have different effects on outcomes.

Conclusion

Overall, our study identified six metabolites linked to the risk of SUD. This includes one metabolite ratio, two lipids, and several amino acid metabolites. These metabolites could act as biomarkers for predicting the onset and progression of SUD. This study also suggests dietary adjustments and nutritional interventions for clinical patients with SUD.Specifically, increasing the intake of medium-chain fatty acids (MCFAs) and phosphatidylethanolamine may positively impact the management of these disorders.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Weixiong Xu and Dandan Xie wrote the main text of the manuscript, and ZhenZhu Zhang, Pibo Du, Yongjian Ye and Xubo Dai contributed to the diagrams. All authors reviewed the manuscript.Each of the named authors made significant contributions to the conduct of the basic research and the drafting of this manuscript. In addition, to the best of our knowledge, none of the authors has a conflict of interest, financial or otherwise.

Funding

The authors declare that no funds, grants, or other support was received for this study. The research was conducted independently, and all associated costs were covered by the authors personally.

Data availability

No datasets were generated or analysed during the current study. The experimental data generated/analyzed in this study have been stored in the [Finnish database and the GAWA Catalog. https://www.finngen.fi/en/access_results(finngen_R10_F5_SUBSNOALCO.gz). https://www.ebi.ac.uk/gwas/home(GCST90043693_buildGRCh37.tsv.gz.)

Declarations

Ethical approval and consent to participate

This study involves secondary analysis of aggregated-level data from published studies or publicly available databases. All original studies were approved by the respective institutional review boards at the time of publication and comply with the principles of the Declaration of Helsinki. This secondary analysis study does not involve direct contact with or collection of new human subject data or biological samples. Therefore, in accordance with the regulations of the Ethics Review Committee of the 910th Hospital of the Chinese People’s Liberation Army Joint Logistics Support Force, no additional ethical approval is required. The conduct of this study strictly adhered to all relevant data usage agreements and ethical guidelines. This study uses publicly available, de-identified summary statistics that do not contain any personally identifiable information. Therefore, it is not necessary to obtain new informed consent specifically for this study. The informed consent process for participants in the original studies was completed during the data generation phase (i.e., in the original GWAS and metabolite studies), which were approved by the respective ethics committees and obtained written informed consent from all participants to use their data for future genetic and health-related research.

Consent for publication

All data used in this study are publicly available summary statistics from published literature or public databases. The results reported in this study are based solely on summary-level statistical analyses and do not disclose any information that could identify individuals. Therefore, no additional consent is required for the publication of the results of this study.

Competing interests

The authors declare no competing interests.