Associations between artificial sweeteners and cardiovascular disease, stroke, and diabetes: A Mendelian randomization study

Abstract

Background

Erythritol is a widely used artificial sweetener, yet its long-term impact on cardiometabolic health remains debated. This study aimed to investigate the genetic associations of erythritol with cardiovascular disease (CVD), stroke, and diabetes using a two-sample Mendelian randomization (TSMR) approach.

Methods

We utilized single-nucleotide polymorphisms (SNPs) associated with erythritol levels from genome-wide association studies (GWAS) as instrumental variables (IVs). The primary analysis employed the inverse-variance weighted (IVW) method. Robustness was assessed using multiple sensitivity analyses (including MR-Egger, weighted median, weighted multitude, and simple mode). Heterogeneity test, pleiotropy test, and sensitivity analysis were also conducted to further ensure the accuracy and stability of the research results.

Results

Erythritol showed positive associations with coronary heart disease (CHD) (OR = 1.0020, 95% CI: 1.0007–1.0034, P = 0.0034), myocardial infarction (MI) (OR = 1.0015, 95% CI: 1.0004–1.0026, P = 0.0090), and stroke (OR = 1.0463, 95% CI: 1.0010–1.0937, P = 0.0449) according to the IVW method. There was suggestive evidence of a positive association between erythritol and CHD, MI, and stroke. No significant causal association was observed between erythritol with heart failure (HF) and diabetes.

Conclusions

This TSMR study provides genetic evidence suggesting erythritol is associated with an increased risk of CHD, MI, and stroke, but not with HF or diabetes. Our findings could further clarify the effect of erythritol on CVD, stroke and diabetes, and thus be more beneficial in reducing the risk of disease. Clinical trial number: not applicable.

Graphical abstract

1. Introduction

Erythritol, classified as a polyhydroxy compound, exists as a natural sweetening agent present in select botanical sources and microbially processed edibles. This low-calorie carbohydrate derivative demonstrates unique metabolic characteristics in the human body, with a sweetness of approximately 60–70 % that of sucrose and nearly zero calories. Erythritol, a frequently used artificial sweetener, is often incorporated into a range of food and drink items such as baked goods, dairy products, and low-calorie beverages [1]. In recent years, the demand for low-calorie sweeteners has surged, leading to an increased prevalence of erythritol in consumer products. A comprehensive assessment of major low-calorie and non-caloric sweeteners conducted between 2008 and 2018 indicated that current intake levels remain within the acceptable daily intake for the general population. However, the long-term health impacts of artificial sweeteners, including erythritol, remain contentious. While a limited number of clinical trials suggest potential minor benefits of artificial sweeteners, meta-analyses and observational studies have revealed that low-calorie and non-nutritive sweeteners often fail to deliver expected health benefits and are associated with adverse cardiovascular metabolic risk factors [2], including weight gain, increased body fat, type 2 diabetes [3], and unfavorable cardiac events [4].

Currently, the global burden of cardiovascular disease (CVD), stroke, and diabetes has reached unprecedented levels. Demographic health assessments reveal that CVD maintain their position as the primary determinant of worldwide fatality patterns. Surveillance metrics demonstrate a twofold escalation in disease burden, with documented instances surging from 271 million to 523 million cases between the last decade of the 20th century and the concluding year of the 2010s. Concurrently, mortality trajectories exhibited a 53.2 % increment, ascending from 12.1 million to 18.6 million fatalities during this thirty-year observation window [5]. Epidemiological studies position stroke as the world’s second most prevalent mortality determinant and third predominant contributor to functional impairment across international populations, with 12.2 million new cases and a prevalence of 101 million in 2019, resulting in 6.55 million deaths [6]. Diabetes affects over 463 million adults worldwide, with projections suggesting this figure will reach 700 million by 2045 [7]. There exists a close interrelationship among these chronic diseases, as diabetes is a significant risk factor for CVD and stroke, primarily mediated through hyperglycemia, insulin resistance, and metabolic dysregulation, which accelerate atherosclerosis and substantially increase the risk of cardiovascular events and stroke [8,9]. Some studies have established a link between the consumption of artificially sweetened soft drinks and an increased risk of metabolic syndrome [10], stroke [11], and myocardial infarction (MI) [12]. The elevated risk of stroke may be attributed to erythritol's potential to impair ischemic brain endothelial progenitor cells and reduce angiogenesis, thereby exacerbating cerebral ischemia [13].

Furthermore, recent research synthesizing in vitro, ex vivo, and in vivo findings suggests that erythritol may promote platelet activation and thrombosis [14]. However, other studies contend that current evidence is insufficient to definitively establish a relationship between dietary erythritol intake and these disease risks, indicating that the relationship between erythritol and conditions such as stroke is complex and warrants further investigation [11,[14], [15], [16]]. Presently, research on the associations between erythritol and adverse cardiac events, stroke, and diabetes has not yielded consistent conclusions, necessitating further exploration.

Given the inconsistent findings in the current literature, this study employs Mendelian randomization (MR) analysis using genetic instruments to systematically assess the long-term metabolic effects of erythritol on CVD, stroke, and diabetes. By leveraging genetic variants as proxies for exposure, MR helps mitigate unmeasured confounding and reverse causation—common limitations in observational studies—thus providing more robust evidence for causal inference regarding erythritol intake and health outcomes. The results of this research will support public health policy and nutrition guidance by contributing to the development of safer and more effective dietary strategies aimed at improving health. Furthermore, this work will establish a foundation for future investigations into the biological mechanisms of artificial sweeteners, guiding subsequent basic research in this area.

2. Materials and methods

2.1. Study design

This study employs a Two-Sample Mendelian Randomization (TSMR) framework to investigate potential causal relationships between erythritol levels and the risks of CVD, stroke, and diabetes. TSMR is especially suitable for this purpose, as it helps minimize confounding biases and reverse causation—common limitations in observational studies of dietary components such as erythritol, where self-reported intake is often unreliable and long-term health effects are challenging to assess through conventional trial designs. The approach utilizes summary-level genetic data, with erythritol concentration serving as the exposure. Its potential associations with five targeted health endpoints were investigated: coronary heart disease (CHD), MI, heart failure (HF), stroke, and diabetes. This analytical framework operates under the three core assumptions of Mendelian Randomization: (A) single nucleotide polymorphisms (SNPs) strongly associate with erythritol levels; (B) SNPs are independent of known confounders; and (C) SNPs influence CVD, stroke, and diabetes solely through their impact on erythritol levels [17,18]. (Fig. 1)

Fig. 1.

Three key assumptions of the Mendelian randomization study. (A) SNPs are strongly associated with erythritol; (B) SNPs are independent of confounders; (C) SNPs must only affect cardiovascular diseases, stroke, and diabetes via erythritol. SNP: single-nucleotide polymorphism.

2.2. Data sources

The investigation leveraged publicly available summary-level data primarily from European populations, sourced via the Integrative Epidemiology Unit OpenGWAS Database Project [19]. The focus on European ancestry populations was due to the greater availability of large-scale genome-wide association studies (GWAS), which maximizes statistical power and helps minimize potential bias from population stratification. However, this focus may limit the generalizability of our findings to other ethnic groups.

GWAS summary statistics for circulating erythritol levels were obtained from a study available via the European Bioinformatics Institute (EBI) database [20]. To mitigate potential bias from sample overlap [21], outcome datasets for CHD (n = 361,194), MI (n = 361,194) and HF (n = 361,194) were derived from the UK Biobank study, while datasets for stroke and diabetes were sourced from the FinnGen biobank. Specific details for each GWAS dataset, including sample sizes, the number of SNPs, unique GWAS identifiers, and data release years, are provided in Table 1. All GWAS data used in this secondary analysis are publicly available and had received prior ethical approval from their respective institutional review boards; thus, no additional ethical approval was required for this study.

Table 1.

Summary of genetic data information.

Phenotypes	GWAS ID	Sample size	SNPs	Year
Erythritol levels	ebi-a-GCST90026156	291	6864,181	2021
Coronary heart disease	ukb-D-I9_CHD	361,194	13,295,130	2018
Myocardial infarction	ukb-D-I9_MI	361,194	12,640,541	2018
Heart failure	ukb-D-I50	361,194	9806,537	2018
Stroke	finn-b-C_STROKE	180,862	16,380,350	2021
Diabetes	finn-b-E4_DIABETES	218,792	16,380,466	2021

2.3. Selection and validation of SNPs

Three criteria were applied to select suitable SNPs for MR analysis. First, SNPs associated with erythritol levels were chosen based on genome-wide significance (P < 5 × 10⁻⁸). Second, the independence among selected SNPs was evaluated using pairwise linkage disequilibrium, with SNPs having an r² > 0.001 (within a clumping window of 10,000 kb) being assessed to ensure independence [22]. SNPs that were highly correlated or had lower significance (higher P-values) were excluded [18]. Third, the F-statistic was calculated to validate the strength of individual SNPs as instrumental variables (IV) [23]. An F-statistic greater than 10 indicates a strong enough correlation between the IV and exposure, thus protecting the MR analysis from weak instrumental bias [[23], [24], [25]].

Before MR analysis, data harmonization steps were performed to ensure consistency, requiring that the effect of an SNP on both the exposure (erythritol levels) and the outcome (CVD, stroke, and diabetes) corresponded to the same allele.

2.4. Two-sample Mendelian randomization analysis

As the principal analytical framework, the study implemented the inverse-variance weighted (IVW) technique functioning as the core statistical procedure. In this approach, estimates were aggregated using the IVW fixed-effect method and presented as odds ratios (ORs) with corresponding 95 % confidence intervals (CIs), assuming all SNPs were valid instruments [26]. To assess the robustness of our findings and account for potential violations of MR assumptions (particularly pleiotropy), several sensitivity analyses were systematically implemented, each relying on different assumptions about pleiotropy. These included the MR-Egger regression, which can detect and provide an estimate adjusted for directional pleiotropy, albeit with potentially less statistical power [28]; the weighted median method, providing a consistent estimate if at least 50 % of the weight in the analysis derives from valid instrumental variables [27]; and additionally, the weighted mode and simple mode methods were employed. To evaluate variability within the chosen IV, Cochrane's Q-value was applied in the analytical process [26]. Furthermore, a "leave-one-out" analysis was performed to identify potentially heterogeneous IVs by sequentially excluding each instrumental SNP and assessing its impact via forest plots. SNPs that introduced bias were excluded, and the aforementioned analyses were repeated. To assess heterogeneity visually, scatter plots were employed. Funnel plots and MR-Egger regression intercept tests were conducted to further examine directional pleiotropy [27,28]. All data processing was performed through the TwoSampleMR toolkit implemented in R software (version 4.3.3), developed by the R Foundation for Statistical Computing based in Shanghai.

3. Results

3.1. SNP selection and validation

Based on established screening procedures, fifty-one SNPs were ultimately selected as IVs for erythritol levels after rigorous quality control measures were applied to the combined genomic data (detailed list provided in Supplementary Table 1). All IVs demonstrated statistically robust metrics exceeding the critical threshold (F > 10), substantially reducing potential distortions caused by weak instruments and thus reinforcing the validity of analytical outcomes.

3.2. Cardiovascular diseases

Following inclusion/exclusion parameters, genomic datasets revealed 51 erythritol-linked SNPs for CHD, 51 for MI, and 36 for HF following quality control procedures that removed abnormal SNPs from aggregated analyses, as detailed in Fig. 2.

Fig. 2.

Associations between erythritol levels and cardiovascular disease based on Mendelian randomization estimates.

Heterogeneity tests between erythritol and different CVDs indicated no significant heterogeneity (all P > 0.05), allowing the primary use of IVW analysis. This consistency was further evidenced in scatter plots depicting the correlations with CHD and MI (Figs. 3A and 4A). The MR-Egger intercept method demonstrated non-significant P-values (all >0.05) in horizontal pleiotropy assessments, consistently confirming the lack of directional bias and reinforcing the reliability of MR outcomes (Table 2). Symmetrical funnel plots for CHD and MI (Figs. 2 and 3B) and non-significant MR-Egger intercept (P = 0.8503 and 0.8895) reinforced the robustness of results (Table 1).

Fig. 3.

Scatter plot and funnel plot of the association between erythritol levels and coronary heart disease. A, Scatter plot of instrument associations with coronary heart disease (y-axis) and erythritol levels (x-axis) with the slopes of Mendelian randomization (MR) estimates. B, Funnel plot of instrument precision (y-axis) and MR associations between erythritol levels and coronary heart disease (x-axis).

Fig. 4.

Scatter plot and funnel plot of the association between erythritol levels and myocardial infarction. A, Scatter plot of instrument associations with myocardial infarction (y axis) and erythritol levels (x axis) with the slopes of Mendelian randomization (MR) estimates. B, Funnel plot of instrument precision (y axis) and MR associations between erythritol levels and myocardial infarction (x axis).

Table 2.

Heterogeneity and pleiotropy test results of erythritol levels with cardiovascular disease, stroke and diabetes.

Outcomes	methods	Q	df	P for Heterogeneity test	Egger regression intercept	P for MR Egger intercept
Coronary heart disease
	IVW	49.3460	50	0.4996
	MR Egger	49.3098	49	0.4607	9.6389e-05	0.8503
Myocardial infarction
	IVW	44.3545	50	0.6984
	MR Egger	44.3350	49	0.6624	5.9197e-05	0.8895
Heart failure
	IVW	31.1954	35	0.6524
	MR Egger	28.1611	34	0.7489	0.0004	0.0906
Stroke
	IVW	31.8488	47	0.9555
	MR Egger	31.7898	46	0.9450	0.0004	0.8090
Diabetes
	IVW	79.7396	48	0.0027
	MR Egger	79.4406	47	0.0022	0.0079	0.6759

Multiple MR techniques such as IVW, MR-Egger, weighted median, weighted mode, and simple mode were utilized to comprehensively evaluate causal associations. Erythritol demonstrated a positive association with CHD (IVW method: OR = 1.0020, 95 % CI: 1.0007–1.0034, P = 0.0034). Although MR Egger, weighted median, weighted mode, and simple mode yielded P > 0.05, the direction of effect was consistent with IVW (Fig. 2), indicating stable findings. Erythritol also correlated positively with MI (IVW method: OR = 1.0015, 95 % CI: 1.0004–1.0026, P = 0.0090). Additional sensitivity assessments reinforced the validity of these relationships through comprehensive evaluations, with weighted median providing a more precise estimate (OR = 1.0017, 95 % CI: 1.0001–1.0034, P = 0.0417) compared to MR- Egger (OR = 1.0013, 95 % CI: 0.9987–1.0040), simple mode (OR = 1.0015, 95 % CI: 0.9979–1.0051), and weighted mode (OR = 1.0020, 95 % CI: 0.9993–1.0047). Notably, rigorous analysis of the data revealed no statistically significant causal link connecting erythritol consumption to HF, as evidenced by comprehensive tabular and graphical presentations (Table 2, Fig. 5, and Supplementary Fig. 1). Additionally, the stability of our findings was confirmed by a "leave-one-out" analysis, as excluding individual SNPs did not significantly alter the results of TSMR correlation analyses for CHD and MI (Supplementary Fig. 2A and 2B).

Fig. 5.

Associations between erythritol levels with stroke and diabetes based on Mendelian randomization estimates.

3.3. Stroke and diabetes

Using established screening parameters, 48 and 49 erythritol-linked SNPs were respectively detected in diabetes and stroke, after excluding abnormal SNPs from the integrated dataset (Fig. 5). In this study, all IVs demonstrated an F value > 10, effectively mitigating bias from weak instruments and confirming the reliability of our results.

The heterogeneity test for erythritol-stroke interactions resulted in a P-value exceeding 0.05, suggesting negligible heterogeneity, which prompted predominant utilization of the IVW methodology. Consistent results were observed in scatter plots depicting the correlation between erythritol levels and stroke (Fig. 6A). MR-Egger's test for horizontal pleiotropy consistently yielded P > 0.05 across all analyses, suggesting no directional pleiotropy (Table 2). The symmetrical funnel plot for stroke (Fig. 6B) and non-significant MR-Egger intercept (P = 0.8090) further supported the robustness of our findings (Table 1).

Fig. 6.

Scatter plot and funnel plot of the association between erythritol levels and stroke. A, Scatter plot of instrument associations with stroke (y axis) and erythritol levels (x axis) with the slopes of Mendelian randomization (MR) estimates. B, Funnel plot of instrument precision (y axis) and MR associations between erythritol levels and stroke (x axis).

This research applied five analytical approaches—IVW, MR-Egger, weighted median, weighted mode, and simple mode—to evaluate causal relationships. Erythritol showed a positive correlation with stroke (IVW method: OR = 1.0463, 95 % CI: 1.0010–1.0937, P = 0.0449). Although multiple pleiotropy-robust approaches (MR-Egger, weighted median, weighted mode, and simple mode) showed non-significant associations (all P > 0.05), the effect orientation derived from IVW exhibited directional coherence with supplementary methodologies in the multivariable framework (Fig. 5). Sensitivity analyses using the additional 4 MR methods demonstrated similar associations with stroke as observed with the IVW method. The analysis revealed that erythritol intake failed to establish a causal relationship with diabetes risk (refer to Table 2, Fig. 5, and Supplementary Fig. 3). To assess parameter stability, a "leave-one-out" analysis was systematically performed, temporarily omitting each genetic variant during iterations. This methodological examination demonstrated that no single SNP exerted a substantial influence on outcome consistency across experimental cycles. Hence, the findings of the TSMR correlation analyses of erythritol with stroke were deemed stable and reliable (Supplementary Fig. 2 C).

4. Discussion

To the best of our knowledge, the present study is the first to systematically assess the association of erythritol with cardiometabolic risk using a two-sample Mendelian randomization (TSMR) approach with an innovative analytical framework, which focused on three outcomes: cardiovascular disease (CVD), stroke, and diabetes. Accordingly, our findings provide evidence of a positive association between erythritol intake and CHD, MI, and stroke. However, our analysis found no evidence supporting a causative link between erythritol consumption and the development of HF or diabetes.

This research demonstrates multiple methodological advantages, particularly through the integration of MR analysis, which provides a robust framework to address potential confounding factors and bidirectional causality issues. In contrast to such capabilities of MR, it is important to note that epidemiological and cohort studies suggest that individuals with poor metabolic status tend to prefer sugar-free diets even before diagnosis of cardiovascular diseases, aiming to avoid weight gain or hyperglycemia [29]. To address limitations inherent in conventional epidemiological investigations, the MR framework, by utilizing genetic polymorphisms as analytical instruments, effectively circumvents residual confounding bias characteristic of conventional epidemiological investigations [17,30]. This genetic epidemiology approach therefore enables deductive causal attribution that transcends the correlational limitations inherent in retrospective cohort analyses [31,32], particularly through its pleiotropy-robust framework. Furthermore, MR is particularly advantageous for establishing causal relationships in disease research, circumventing ethical challenges inherent in randomized controlled trials (RCTs) that involve leaving patients untreated.

Many observational epidemiological studies have explored the association between artificial sweetener use and various adverse health outcomes, including cardiovascular disease mortality, with varying results [29,33,34]. Indeed, many studies report that artificial sweetener use is associated with various adverse health outcomes, including CVD mortality [[35], [36], [37]], while others do not [38,39]. In relation to erythritol and CVD, our study's findings are consistent with some previous observational studies, showing an association between erythritol and CVD. For example, Witkowski et al. conducted a non-targeted metabolomics study involving 1157 high-CVD-risk individuals, including 22 % with diabetes, to explore circulating metabolites associated with increased cardiovascular-risk [4]; they found that plasma erythritol levels were significantly associated with major cardiovascular events (such as death, MI, and stroke) over a 3-year follow-up. Additionally, further independent validation cohorts in the US (n = 2149, NCT00590200) and Europe (n = 833, DRKS00020915) also confirmed this association, with adjusted hazard ratios (95 % CI) of 1.80 (1.18–2.77) and 2.21 (1.20–4.07), respectively [4]. These studies highlight erythritol's potential to increase the risk of CHD and MI, aligning with our study's conclusions. However, compared to our study, Witkowski et al.'s research suggests a possibly more pronounced effect of erythritol on CVD risk, likely due to their inclusion of patients recruited from tertiary referral centers with high cardiovascular disease and traditional risk factors. With respect to stroke, our findings provide evidence of a positive association between erythritol and stroke. Nevertheless, existing studies provide limited evidence regarding the association between erythritol consumption and stroke risk with current understanding remaining incomplete.

Multiple biological hypotheses exist regarding the mechanisms linking high erythritol levels with CHD, MI, and stroke (enhanced platelet reactivity and thrombosis, macrophage dysfunction and inflammatory responses, impaired vascular repair and reduced angiogenesis). Research by the Witkowski team demonstrated that erythritol directly increases platelet reactivity and thrombosis risk by inducing elevated intracellular calcium release and enhancing platelet aggregation responses to various agonists [4]. A recent study by the Alamri team found that erythritol may trigger necroptosis in THP-1-derived macrophages, thereby disrupting normal macrophage function and promoting inflammatory responses commonly associated with cardiometabolic diseases [40]. Animal studies by Dong et al. suggested that long-term consumption of erythritol and other artificial sweeteners may exacerbate cerebral ischemic injury in mice, partly due to impaired endothelial progenitor cell (EPC) function and reduced angiogenesis in ischemic brain regions [13]. Studies on lipid metabolism have yielded conflicting results regarding the effects of erythritol. Some indicate that erythritol activates the Nrf2 signaling pathway, exerting antioxidant effects that suppress endoplasmic reticulum stress and reduce lipid accumulation [41]. In contrast, other studies report that erythritol promotes hepatic lipid droplet accumulation [42]. Meanwhile, several investigations have found no significant changes in blood lipid profiles following erythritol consumption [43]. Additionally, erythritol has been shown to enhance platelet activation and thrombus formation, which could contribute to atherosclerotic plaque destabilization and potentially exacerbate atherosclerotic cardiovascular disease (ASCVD). Thus, the overall impact of erythritol on lipid metabolism remains unclear, warranting further investigation to clarify these mechanisms. Although no other studies directly indicate that erythritol increases stroke risk, some suggest it may indirectly contribute by promoting risk factors such as CHD, hypertension [44], and thrombosis [4,45]. These experimental findings on erythritol's effects on platelet function, macrophage activity, and vascular biological processes provide multiple biologically plausible explanations for the positive associations revealed in our Mendelian randomization study between circulating erythritol levels and the risks of CHD, MI, and stroke. However, consistent with our MR study's null finding for an association between erythritol and heart failure, there is a notable paucity of mechanistic research specifically investigating direct links between erythritol and heart failure pathophysiology. While our evidence does not suggest that erythritol is a primary causal factor for overall heart failure risk, future studies could explore potential indirect effects or specific roles in particular heart failure subtypes or etiologies, especially should relevant leads emerge from other research areas.

Regarding diabetes, erythritol, as a low-calorie sugar substitute, has garnered considerable interest regarding its impact on the risk of developing diabetes. While our study did not find an association between erythritol intake and the risk of diabetes, current literature exploring this specific relationship remains limited. In contrast, preliminary evidence suggests that erythritol may potentially aid in improving glucose metabolism. For example, erythritol is considered particularly suitable for individuals with impaired glucose control or obesity due to its minimal short-term effects on insulin and blood glucose levels [46]; furthermore, animal studies indicate that erythritol may reduce postprandial blood glucose levels and improve glucose metabolism—its potential mechanism likely involves stimulating the gut-brain-pancreas axis [e.g., through Glucagon-Like Peptide-1 (GLP-1), cholecystokinin, and Peptide YY], or by inhibiting α-glucosidase, with GLP-1 playing a crucial role as an incretin. These conclusions were typically drawn under conditions of single-dose or short-term erythritol administration [15,47,48]. Diabetes, as a chronic disease, develops gradually under lifelong nutritional and environmental exposures. Therefore, it is plausible that these short-term or acute glucoregulatory effects of erythritol are insufficient to translate into a decisive modification of long-term diabetes risk at the population level. The MR methodology, by assessing the effects of lifelong genetic predisposition, is well-suited to detect such chronic effects; thus, our null finding for diabetes further suggests that the acute mechanisms mentioned may not reflect a significant long-term impact on diabetes incidence.

From a genetic perspective, the precise functional roles of the 51 SNPs significantly associated with circulating erythritol levels in our study remain to be fully elucidated. Currently, these SNPs primarily serve as genetic instruments for overall erythritol levels. However, future functional genomics research should prioritize exploring whether any of these genetic loci reside within or near genes known to be involved in erythritol metabolism (e.g., key enzymes in the pentose phosphate pathway, sugar transport systems) or broader cardiometabolic processes. The results of our MR study should stimulate such investigations to uncover the specific molecular links between these genetic variants and erythritol homeostasis, which, in turn, could reveal more direct therapeutic or preventative targets.

Regarding the strengths of this research, leveraging dual-sample MR methods enhances statistical power, crucial for detecting causal links, especially in studies involving binary disease outcomes [49]. Moreover, the reliability of our findings is bolstered by utilizing genetic instruments derived from multiple robust GWAS, thereby minimizing selection bias through a large sample size. First, the findings rely on three core assumptions of MR analysis. Although we selected SNPs significantly associated with circulating erythritol levels (P < 5 × 10⁻⁸) and employed MR-Egger regression to assess pleiotropy, completely eliminating confounding factors and bidirectional pleiotropic effects remains challenging [50], particularly considering that erythritol metabolism involves polyol pathways and other pleiotropic biological mechanisms. Second, current genetic instruments cannot distinguish between endogenous and exogenous erythritol, nor can they reflect metabolic differences across tissues or microbial communities, potentially leading to underestimation of true causal associations. Third, the study primarily utilized data from European ancestry populations, and genetic functional variants combined with ethnic differences in dietary habits may limit the global generalizability of the conclusions. Additionally, although we endeavored to ensure independence between exposure and outcome datasets, potential sample overlap in GWAS studies may still affect result precision. Finally, it is important to note that MR analysis reflects the cumulative effects of lifelong genetic predisposition—the effect sizes cannot be directly equated to short-term dietary intake impacts, nor can they capture dose-response relationships or critical exposure time windows. Future studies should employ trans-ethnic research designs, integrate multi-omics data, and incorporate clinical intervention trials to further validate and refine these findings. We acknowledge that while the reported ORs between erythritol levels and risks of CHD, MI, and stroke in our study were statistically significant, their absolute magnitudes were modest. Several factors may contribute to these effect sizes, including inherent challenges in quantifying the impact of ubiquitous dietary exposures, as well as study limitations such as the predominant inclusion of European ancestry populations, which may influence both the generalizability of the associations and the observed effect sizes. Notably, despite ORs approximating unity, the effect direction remained consistently positive (OR > 1) across various sensitivity analyses within our Mendelian randomization framework, suggesting a small yet robust and genetically supported adverse association. Importantly, the public health and clinical relevance of these findings should be interpreted in the context of both exposure prevalence and outcome severity. Although current data on population exposure proportions of erythritol remain limited, its widespread use as a common sugar substitute in daily diets suggests potentially extensive population-level exposure. Given that CHD, MI, and stroke represent conditions with substantial morbidity, mortality, and adverse outcomes, even marginal increases in individual relative risk attributable to highly prevalent exposures like erythritol could translate to non-negligible population attributable fractions. Thus, when these modest effect sizes are applied to large populations with frequent erythritol exposure, they may indeed constitute meaningful real-world risks that warrant careful consideration from a public health perspective.

The findings from this MR study, suggesting potential causal links between the widely used food additive erythritol and increased risks of CHD, MI and stroke, carry significant public health and nutritional implications that warrant careful consideration and potential regulatory review. Given erythritol's extensive incorporation into diverse food products and its growing global consumption, even a modest individual effect size for these serious outcomes could translate into a considerable public health burden, particularly for high-risk populations susceptible to thrombosis [4]. Although the European Food Safety Authority (EFSA) established an acceptable daily intake (ADI) for erythritol following a reassessment in December 2023, this was primarily based on its laxative effects. If the causal relationship between erythritol and coronary heart disease, myocardial infarction, and stroke—as suggested by this Mendelian randomization study—is further confirmed, relevant regulatory agencies may need to consider whether additional investigations and research are warranted to more clearly delineate erythritol's potential risks, particularly for susceptible populations and individuals at high cardiovascular risk. To mitigate the potential health risks associated with artificial sweetener use, a multidimensional intervention strategy integrating telemedicine and mobile health technologies can be implemented. First, personalized dietary guidance should be provided to diabetic patients and high-risk populations through telemedicine platforms [51]. Second, physiological parameters such as heart rate variability should be monitored via wearable devices and mobile applications to generate valuable data for subsequent research [52]. Additionally, high cardiovascular risk patients should be encouraged to adopt healthier dietary habits through real-time health data feedback [53]. This comprehensive approach combining telemedicine, smart monitoring, and behavioral intervention could more effectively reduce the potential health risks of artificial sweeteners. Notably, this field currently lacks clear international guidelines. Therefore, our findings highlight two urgent needs: further research initiatives and proactive reevaluation of the widespread use of erythritol, which will better inform regulatory policy formulation, optimize dietary recommendations, and ultimately safeguard public health.

The findings of this Mendelian randomization study, which provide novel insights into the potential causal associations of erythritol with cardiometabolic diseases, naturally pave the way for future investigations. Crucially, replication studies in larger and more ancestrally diverse populations are warranted to confirm the robustness and generalizability of our observed genetic associations with cardiovascular and cerebrovascular outcomes. Furthermore, in-depth exploration of the underlying biological and metabolic pathways is essential to elucidate precisely how erythritol levels might contribute to the identified risks for conditions such as coronary heart disease, myocardial infarction, and stroke, or, conversely, why no significant causal associations were found for other conditions like diabetes and heart failure; this will likely necessitate a combination of experimental models and advanced human studies. To gain deeper molecular insights, future research should also prioritize the functional characterization of the genetic variants identified in this study as being associated with erythritol levels. Leveraging the integration of multi-omics data (e.g., genomics, metabolomics, proteomics) will also be pivotal in constructing a more comprehensive understanding of the molecular sequelae of altered erythritol levels. Such continued research, potentially incorporating advanced genetic epidemiological approaches like multivariable Mendelian randomization to dissect specific metabolic effects, is vital for translating these initial genetic findings into a more complete understanding of erythritol's cardiometabolic impact and for ultimately refining public health guidance regarding this widely used sweetener.

5. Conclusions

This TSMR analysis provided genetic evidence of a positive association between high erythritol levels and an increased risk of CHD, MI, and stroke, while found no genetic evidence supporting a causal link between erythritol levels and the development of HF or diabetes. These findings provide a scientific basis for balancing the potential benefits of erythritol as an artificial sweetener and the cardiovascular disease risks explained by genetic evidence, thereby informing strategies to reduce the population's risk of disease.

Funding

This research received no external funding.

Institutional review board statement

Not applicable.

Informed consent statement

Ethical approval was not required for this study as all data were publicly available and did not involve individual-level information. Clinical trial number: not applicable.

Data availability

All custom code and data supporting this study will be made available upon request to the corresponding author.

CRediT authorship contribution statement

Jinming Fan: Conceptualization, Software, Validation, Writing – original draft. Yifei Hu: Conceptualization, Validation, Writing – original draft. Junzhu Zhang: Resources, Data curation. Jiawen Chen: Methodology, Formal analysis, Investigation. Yajun Yuan: Writing – review & editing, Visualization, Supervision. Benshuai Yu: Writing – review & editing, Project administration, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix. Supplementary materials

Supplementary Materials: Table S1. Single nucleotide polymorphisms used as instrumental variables in the Mendelian randomization analyses of erythritol levels. Figure S1. Scatter plot, funnel plot, inverse-variance weighted, and leave-one-out plot of the association between erythritol levels and heart failure. Figure S2. Leave-one-out plot of the association between erythritol levels and coronary heart disease, myocardial infarction, and stroke. Figure S3. Scatter plot, funnel plot, inverse-variance weighted, and leave-one-out plot of the association between erythritol levels and diabetes.