Erythritol as a Potential Causal Contributor to Cardiometabolic Disease: A Mendelian Randomization Study Free

With the increase in diabetes and associated cardiometabolic diseases (CMD), low-calorie sweeteners are recommended by current diabetes and cardiac guidelines to improve glycemia and reduce caloric intake (1–3). Erythritol, a low-calorie sweetener, is increasingly being used as a sugar substitute (4).

Erythritol is a four-carbon polyol, found in low concentration in fruit, vegetables, mushrooms, and fermented foods (5,6). It can also be synthesized endogenously from glucose via the pentose-phosphate pathway: ∼5–10% can be metabolized to erythronate, and the remainder is excreted in urine (7). Exogenous erythritol consumption at a dose of 30 g increases circulating concentration ∼1,000-fold before returning to baseline after 2 days (8). Exogenous erythritol does not increase plasma glucose or insulin but does increase the secretion of glucagon-like peptide 1 and cholecystokinin and decreases gastric emptying (5). Daily erythritol supplementation has been reported to acutely improve endothelial function and reduce aortic stiffness over 4 weeks in people with type 2 diabetes (T2D) (9). Intriguingly, despite these potential beneficial effects, increased erythritol has been associated with adverse cardiometabolic effects in observational studies. Observational data from cohorts studied prior to the increased use of erythritol have shown increased risk of future centripetal adiposity and higher hemoglobin A1c (HbA_1c) with linear increase in T2D risk with increasing circulating erythritol (7,10). Recent data, which include cohorts studied both before and after the increased use of exogenous erythritol, show an increase in risk major cardiac events with increasing circulating erythritol (8,11). Specifically, a 16–31% increased risk of major cardiovascular events per 1-μmol increase in erythritol was reported across three cohorts recently· (8) In vitro and ex vivo studies indicate erythritol increases platelet aggregation at concentrations 4- to 10-fold higher than baseline fasting concentration, but this awaits confirmation in vivo in humans (8). Collectively, these data indicate that increased erythritol (likely from both endogenous production and exogenous consumption) is associated with adverse cardiometabolic traits, despite potential beneficial metabolic/endocrine effects after acute/short-term consumption.

Observational data can potentially be biased by confounding and reverse causality. T2D and dysglycemia are cardiovascular risk factors and can potentially increase endogenous erythritol production (7). Chronic kidney disease (CKD), a cardiovascular risk factor, can potentially impact excretion of endogenous and exogenous erythritol (7). Further, increased age and triglyceride associate with increased erythritol (8). Finally, exogenous erythritol consumption may be higher in people at risk for CMD.

Two-sample Mendelian randomization (MR) uses genetic instruments associated with an exposure in one population at genome-wide significance threshold (P < 5 × 10⁻⁸) to investigate potential causal association with an outcome of interest in another cohort. In comparison with observational studies, MR is less prone to confounding (12). Recent genome-wide association studies (GWAS) have identified variants strongly associated with circulating plasma/serum erythritol, including variants near TKT (encoding transketolase) and AKR1A1 (encoding Aldo-Keto Reductase Family 1 Member A1), which are plausible regulators of erythritol kinetics (13,14). We therefore undertook bidirectional MR to investigate the potential causal associations and direction of effect between erythritol, T2D, and associated CMD.

Using summary statistics from the largest published GWAS, univariable MR analyses were done in cohorts of European ancestry including NIHR (National Institute of Health Research), METSIM (Metabolic syndrome in men), Twins UK, and KORA (Cooperative Health Research in the Augsburg Region/Kooperative Gesundheitsforschung in der Region Augsburg) in whom erythritol GWAS was undertaken (Table 1, with further details in Supplementary Tables 1 and 2). Informed consent and research ethics approval were previously obtained by the individual cohort investigators. Primary MR analyses were undertaken separately using single nucleotide polymorphisms (SNPs) associated with erythritol (P < 5 × 10⁻⁸) in two of the largest GWAS studies in populations of European ancestry (NIHR and METSIM) (Table 1), which, based on linkage disequilibrium (LD), were partially overlapping (Supplementary Table 3) (13,14). Erythritol and threitol are isomers and cannot be distinguished by mass spectrometry unless separated by gas chromatography or liquid chromatography, with the latter dependent on the type of column used (15–18). In the METSIM and NIHR GWAS cohorts, liquid chromatography with hydrophilic interaction liquid chromatography column was undertaken to separate erythritol, but GWAS of threitol was not performed (14,15,19). We therefore undertook additional MR analyses using instruments from Twins UK/KORA in which erythritol and threitol were separated by gas chromatography prior to measurement with subsequent GWAS for both (18). Our instrument comprised SNPs on chromosome 1 associated with erythritol but not threitol; these were in the same loci as SNPs from the other instruments (Supplementary Table 3).

Bidirectional univariable MR was undertaken with erythritol levels as the exposure and the following outcomes: BMI, waist-hip ratio adjusted and unadjusted for BMI (WHRadjBMI, WHRunadj), fasting glucose, HbA_1c, T2D, coronary artery disease (CAD), estimated glomerular filtration rate using creatinine (eGFRcreat), and CKD (defined as eGFR <60 mL/min/1.73 m²). We were unable to undertake MR analysis of the effects of erythritol on CAD using the NIHR UK bioresource, because the instrument or a suitable proxy was not identified. In the case of instruments with a single SNP, the Wald ratio was calculated. For multiple SNPs, inverse variance weighted (IVW) and additional sensitivity analyses, including MR-Egger, weighted median, weighted mode, and leave-one-out analyses, were reported (12).

There are three assumptions made in MR analyses. First, the instrument is robustly associated with the exposure; thus we used SNPs associated with the exposure at genome-wide significance (12). Second, there is no horizontal pleiotropy (i.e., the instrument does not influence the outcome via another pathway other than the outcome) (8). Lastly, the instrument is not influenced by any confounders (12).

By permitting the IVW to have a nonzero intercept, MR-Egger relaxes the assumption of no horizontal pleiotropy. This returns an unbiased causal estimate in the case of horizontal pleiotropy, providing that the horizontal pleiotropic effects are not correlated with the SNP-exposure effects (InSIDE assumption) (12,20). The median effect of all SNPs in the instrument was used, using weighted median MR, allowing SNPs with a greater effect to be evaluated by weighting the contribution of each SNP by the inverse variance of its association with the outcome (21). This is a robust method even if only 50% of the SNPs satisfy all three MR assumptions (21). Finally, SNPs were clustered into groups based on similarity of causal effects for weighted mode MR, with the cluster with the largest number of SNPs deriving the causal effect estimate (22). To assess heterogeneity, a Cochrane Q test was used, while leave-one-out analyses were conducted to assess whether any MR estimate was biased by a single SNP potentially with horizontal pleiotropic effect (12).

Each analysis was considered significant based on the Bonferroni corrected threshold of P < 0.0015 (0.05 divided by 34 tests) and nominally significant if P < 0.05. Odds ratios were calculated for binary outcomes, and F statistics were calculated. An F statistic of >10 was considered indicative of a strong instrument.

MR analyses were conducted using the TwoSampleMR package in R (RStudio version 1.3.1073 and R version 4.0.3). LD pruning was used to select a proxy (r² >0.8) if an SNP was not directly matched from the 1000 Genomes Project in the European superpopulation. Palindromes were not excluded. Plots were generated using the ggplot2 package in R. The Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) reporting guidelines were incorporated (Supplementary Table 5) (23).

All data generated or analyzed during this study are included in the published article (and its Supplementary Material).

Both studies in the primary analyses had SNPs in loci on chromosomes 1 and 3, with evidence that SNPs from chromosome 1 in each instrument (rs2229540 and rs72690839 near AKR1A1) were in LD (D′ = 0.935, r² = 0.86). (Supplementary Table 3) (24).

With the NIHR Bioresource instrument, MR suggests that increased erythritol might decrease BMI (b = −0.038, P = 0.035) at nominal significance. These results were concordant with MR using SNPs associated with erythritol in the METSIM study (Table 2). Increased erythritol significantly decreased BMI (b = −0.037, P = 1.23 × 10⁻⁵), with suggestive evidence that it might also increase WHR adjusted for BMI (b = 0.046, P = 0.035) at nominal significance. We did not find evidence that increased erythritol was causally associated with cardiometabolic outcomes (Table 2 and Supplementary Fig. 1).

Reverse MR did not find conclusive evidence that T2D, CAD, HbA_1c, fasting glucose, BMI, WHR, eGFRcreat, and CKD impact erythritol concentration (Supplementary Table 4).

We undertook additional analyses using an instrument from a GWAS that identified variants associated with erythritol but not its structural isomer threitol (after separation with gas chromatography) (18). The variant in chromosome 1 (rs7542172) is in the same loci as SNPs from the NIHR Bioresource and METSIM study (Supplementary Table 3). Our MR analyses were concordant with prior analyses, with nominal evidence that increased erythritol may decrease BMI (b = −0.08 ± 0.09, P = 0.04) but with no evidence that erythritol impacts other cardiometabolic parameters (Table 3 and Fig. 1). Reverse MR analyses did not find evidence that cardiometabolic traits influence erythritol concentration (Supplementary Table 4).

Recent observational studies suggest an association between increased erythritol and CMD (8,10,11). These findings include studies undertaken both before and after the widespread use of erythritol and thus suggest that both endogenous and exogenous erythritol may have deleterious effects, with ex vivo and in vitro evidence that erythritol may increase platelet aggregation (8). Our MR analyses suggest increased erythritol might decrease BMI and consequently increase BMI adjusted WHR. However, we did not find evidence that erythritol increases CAD, T2D, fasting glucose, eGFR, or CKD, nor did we find evidence that evaluated cardiometabolic and anthropometric traits significantly influence erythritol concentration. Collectively, we did not find evidence of causal associations between erythritol and CMD. A previous in vitro study reported that increased oxidative stress, which is implicated in cardiovascular disease, increased endogenous erythritol (25,26). Endogenous erythritol may thus be a marker of oxidative stress, but this awaits in vivo confirmation. Reverse causality with increased erythritol consumption in people at higher cardiometabolic risk could potentially explain the association of increased erythritol with CMD in more recent studies. Whether the observed ex vivo and in vitro effects of erythritol on platelet function are attenuated by its documented beneficial effects on gut peptides such as GLP-1 and endothelial function is not known (5,9). Adequately designed prospective studies are needed to confirm these findings.

The strengths of this study include using the largest available GWAS in populations of European ancestry. Primary analyses with the two instruments, which were strong based on F statistic > 10, were generally concordant, with further confirmation using a strong instrument associated with erythritol but not its isomer threitol.

It is not known how much of the measured erythritol levels is exogenous versus endogenous. GWAS were undertaken prior to the widespread use of erythritol and likely represent endogenous concentration. The MR analyses assessed linear relationships between erythritol and CMD, based on reported linear increase in risk for T2D and major cardiac disease in studies conducted both before and after increased adoption of erythritol in the food industry (8,10,11). Whether further increases in cardiometabolic risk are seen with erythritol concentration above a certain threshold is not established. For each 1 μmol/L increase in erythritol, an increase in risk of major cardiovascular events of 31%, 21%, and 16% across three cohorts was recently reported (8). However, formal statistical analyses of whether a nonlinear relationship exists between circulating erythritol and cardiovascular disease was not undertaken: the data do not exclude this possibility, with highest risk seen in the fourth quartile. Analyses were adjusted for age, sex, BMI, presence of diabetes, systolic blood pressure, current smoking status, LDL, HDL, and triglyceride. However, participants in the fourth quartile had a higher prevalence of baseline cardiovascular disease across all three cohorts, which may potentially confound the results (8). Further studies assessing the potential nonlinear relationship between erythritol concentration and cardiometabolic traits will be informative, as exogenous erythritol can increase circulating concentration >1,000-fold and can impact platelet function at concentrations 4- to 10-fold higher than baseline, ex vivo and in vitro (8). Finally, the number of GWAS analyzing erythritol are fewer compared with studies of cardiometabolic traits (Supplementary Table 1) and are limited to participants of European ancestry; thus our findings may not be generalizable to other ethnicities. Finally, individual-level data including erythritol consumption and medications were not available.

We did not find supportive evidence from MR analyses that erythritol has deleterious cardiometabolic effects. Well-designed prospective studies will be needed to confirm these findings.

Acknowledgments. The authors thank Dr. Andreas Schulze, Sick Kids Hospital, Toronto, Canada, for his invaluable advice in preparing the revised manuscript.

Funding. S.D. is funded by Canadian Institute of Health Research, Heart & Stroke Foundation of Canada, and Banting & Best Diabetes Centre.

Duality of Interest. S.D. has received consulting and speaker fees from Novo Nordisk, Eli Lilly, and Medison. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. R.K., A.D.P. and S.D. designed the study. All authors analyzed the data. R.K. and S.D. wrote the manuscript, and all authors read and edited the manuscript. S.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.