The effects of SCFAs on glycemic control in humans: a systematic review and meta-analysis

ABSTRACT Background Noncommunicable disease development is related to impairments in glycemic and insulinemic responses, which can be modulated by fiber intake. Fiber's beneficial effects upon metabolic health can be partially attributed to the production of SCFAs via microbial fermentation of fiber in the gastrointestinal tract. Objectives We aimed to determine the effects of SCFAs, acetate, propionate, and butyrate on glycemic control in humans. Methods The CENTRAL, Embase, PubMed, Scopus, and Web of Science databases were searched from inception to 7 December 2021. Papers were included if they reported a randomized controlled trial measuring glucose and/or insulin compared to a placebo in adults. Studies were categorized by the type of SCFA and intervention duration. Random-effects meta-analyses were performed for glucose and insulin for those subject categories with ≥3 studies, or a narrative review was performed. Results We identified 43 eligible papers, with 46 studies within those records (n = 913), and 44 studies were included in the meta-analysis. Vinegar intake decreased the acute glucose response [standard mean difference (SMD), −0.53; 95% CI, −0.92 to −0.14; n = 67] in individuals with impaired glucose tolerance or type 2 diabetes and in healthy volunteers (SMD, −0.27; 95% CI, −0.54 to 0.00; n = 186). The meta-analyses for acute acetate, as well as acute and chronic propionate studies, showed no significant effect. Conclusions Vinegar decreased the glucose response acutely in healthy and metabolically unhealthy individuals. Acetate, propionate, butyrate, and mixed SCFAs had no effect on blood glucose and insulin in humans. Significant heterogeneity, risks of bias, and publication biases were identified in several study categories, including the acute vinegar glucose response. As evidence was very uncertain, caution is urged when interpreting these results. Further high-quality research is required to determine the effects of SCFAs on glycemic control.


Introduction
Noncommunicable diseases, such as type 2 diabetes (T2D) and cardiovascular disease, accounted for 44% of global deaths in 2019 (1). T2D diagnoses have quadrupled globally, from 108 million to 422 million, in the last 40 years (2). Elevations in blood glucose and insulin play a significant role in noncommunicable disease development, specifically of T2D (3)(4)(5)(6). Improving glycemic control can reduce the risk of complications associated with T2D (7).
Diet is a primary risk factor for the development of noncommunicable diseases. Western diets are often nutrient deficient, energy dense, and low in fiber (8), and populations following Western dietary patterns have high incidences of chronic disease (9,10). Fiber intake plays a determining role in the noncommunicable disease risk and is a strong indicator of the all-cause chronic disease mortality risk (11). Previous human nutrition studies have Quantifiable measures of glycemic control as the main or secondary outcome, such as fasting glucose or insulin, postprandial glucose or insulin (i.e., AUC), HbA1c, insulin sensitivity indexes (i.e., HOMA-IR, clamps) Studies which do not include a quantifiable measure of the outcomes of interest Study design Study designs that generate empirical data from interventional studies that are randomized controlled trials. Only results analyzed statistically will be included Reviews, conference abstracts, dissertation abstracts, lectures, information pieces, study registers and corrigendums were not included. Studies were limited to those in the English language published from 1980 onwards 1 Abbreviations: HbA1c, glycated hemoglobin A1c; PICOS, populations, intervention, comparison, outcomes, and study design.
shown that dietary fibers have a beneficial effect on glycemia (12). Dietary fiber passes through the upper gastrointestinal tract undigested and can act as a substrate for bacterial fermentation throughout the gut (13). After undergoing fermentation in the gastrointestinal tract, 10 g of fiber yields approximately 100 mmol/L of SCFAs. Acetate, butyrate, and propionate are produced in the largest quantities at a molar ratio of 3:1:1, respectively (14). SCFAs activate G-protein-coupled receptors, known as free-fatty-acid receptors (FFAR) 2 and 3, which are expressed in the gut and in metabolically active tissues, such as in the liver, adipocytes, myocytes, and pancreas (15). In vitro and animal studies have shown SCFAs influence the glucose metabolism in glucose-disposal tissues, such as hepatocytes (16), adipocytes (17), and myocytes (18). These SCFAs have been shown to directly stimulate the release of anorectic hormones, such as glucagon-like peptide 1 (GLP-1) and peptide YY, in colonic enteroendocrine cells (19)(20)(21). However, increasing fiber intake at the population level has proven challenging. Hence, providing a similar metabolic benefit via alternative methods is the aim of much current research.
Overall, evidence suggests that SCFAs may influence human glucose homeostasis. Compiling the studies exploring the impacts of SCFAs on glycemic control may help to elucidate the therapeutic potential of SCFAs within the systemic circulation and gut when administered at concentrations at or above that produced when the recommended fiber intake (30 g/d) is consumed. Here, we aim to investigate the effects of SCFA administration on glycemic control.

Methods
We performed a systematic review of peer-reviewed literature published since inception. The systematic review was conducted in adherence with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (22). The formal screening of papers began on 15 November 2020, and the registration of the protocol for this review to PROS-PERO was submitted on 14 January 2021 with the reference CRD42021231115.

Eligibility criteria
The PICOS (patients, intervention, comparator, outcomes, and study design) criteria were used to establish study eligibility ( Table 1).

Search strategy
The online databases PubMed (Medline), Cochrane CEN-TRAL, EMBASE, Web of Science, and Scopus were used to identify records published from inception to 7 December 2021. The search algorithm used for each database is described in Supplemental Table 1. In addition, a manual search of reference lists of reviews on the topic was performed, to identify additional relevant articles. When necessary, the authors were contacted to obtain data of interest. Studies were excluded if authors did not respond.

Study selection
The study selection was performed using the online software Covidence systematic review software (Veritas Health Innovation; www.covidence.org). All articles identified by the search strategy were screened by title and abstract by 2 reviewers independently (SA-A and JEP). After screening, full texts deemed to be potentially relevant were assessed for eligibility against the defined inclusion and exclusion criteria independently (SA-A and JEP; Table 1). During the study screening and assessment, any discrepancies in the eligibility of papers were resolved by consulting a third party (AC-M). Excluded studies and reasons for exclusion can be found in Supplemental Table 2.

Data extraction and quantification
Data were extracted by 4 reviewers independently (SA-A, AC-M, DH, and JEP). Articles deemed eligible for inclusion were assigned to subject categories according to the nature of the study intervention (acute or chronic) and type of SCFA [acetate, butyrate, propionate, mixed or vinegar (acetic acid)]. Study characteristics were extracted for each category, including the authors' names, publication year, study design, length of intervention [acute (<24 hours) or chronic], sample size, participants' demographic characteristics (gender, age, BMI, and any health conditions), SCFA concentration, route of administration (oral, intravenous, or gastrointestinal), measurement period, energy and macronutrient matching, and outcomes analyzed. Some identified records contained multiple studies, which were extracted individually. Whilst all comparisons within the same record were captured in the tables summarizing the studies, not all comparisons were meta-analyzed. Selection of the comparisons against the control was based on the highest dose or on the format used in real life (e.g., liquid vinegar over pill). Table 2 summarizes the eligible study characteristics.

Data synthesis and statistical analysis
Descriptive data were reported as the mean ± SD unless otherwise stated. Glycemic control measurements included the blood glucose and insulin (raw or change from baseline) AUC or incremental AUC (iAUC), fasting blood glucose (FBG) or insulin, glycated hemoglobin (Hb1Ac; as a percentage), or insulin sensitivity indexes (e.g., HOMA-IR). For the subject categories that included <3 studies, a narrative review was conducted in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (23). For categories that had ≥3 studies, a meta-analysis was performed. For the meta-analysis, raw data or changes from the baseline iAUC were extracted, and the variance was transformed to SD. When data were available as individual time points, means and variances were extracted using the online tool WebPlotDigitizer 4.4 (Ankit Rohatgi) (https://apps.automer is.io/). Then, the iAUC of the mean and variance was calculated by the trapezoidal rule (24).

Meta-analysis
A meta-analysis to estimate the pooled effects of the SCFAs on the different glycemic outcomes was performed for each subject category that included ≥3 studies reporting the same glycemic outcome. These were: acute acetate, acute vinegar, and acute and chronic propionate administration. For acute studies, meta-analyses of postprandial blood glucose (PBG) and insulin iAUC were performed. For chronic studies, meta-analyses of PBG and insulin iAUC, fasting glucose and insulin, and glycated hemoglobin (HbA1c) were performed. For these outcomes for each subject category, the weighted effect estimates, reported as the standard mean differences (SMDs) and corresponding 95% CIs, were calculated as using a Sidik-Jonkman random-effects model to allow a wide 95% CI in order to reflect uncertainty in the estimation of between-study heterogeneity. For crossover studies, the SMD was calculated by assuming a parallel design for a more conservative analysis. Heterogeneity was assessed using the I 2 statistic and visual inspection of the Galbraith plot (Supplemental Figures 1.2

Risk-of-bias assessment
Studies were assessed for the risk of bias by 3 independent reviewers (SA-A, JEP, DH) following the revised Cochrane riskof-bias tool for randomized trials (26). The 6 methodological features assessed were randomization, assignment of the intervention, adherence to the intervention, missing outcome data, measuring the outcome, and selection of the reported result. Studies were classified as having a high risk of bias if they contained methodological flaws that may have influenced the results, having a low risk if the flaw was not deemed to have affected the results, and having some concerns if not enough information was provided to pass a judgement. Disagreements in the classification were resolved by consulting a third party (AC-M).

Certainty of evidence: Grading of Recommendations, Assessment, Development, and Evaluation
The table summarizing the findings was constructed using GRADEpro software (McMaster University and Evidence Prime Inc.) (http://gradepro.org, accessed 10 December 2021). Certainty of evidence was assessed using Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) recommendations. The certainty of evidence was graded as high, moderate, low, and very low (27).

Assessment of confounders of glycemic control
Studies that controlled for factors influencing the glycemic response may produce more accurate results. Acute confounders of glycemic control are physical activity, the length of the fasting period, and fiber or alcohol intake. Chronic confounders include changes in the body weight and body fat percentage over the course of the study. Included studies were assessed for how they controlled for elements known to confound glycemic control (e.g., body weight or body fat change, physical activity, or alcohol intake prior to the intervention or an overnight fast), and the results are summarized in Supplemental Table 3.

Description of studies
The systematic literature search produced a total of 5932 references following the database and manual search (Figure 1). Specifically, 3514 publications were identified from PubMed, 289 from Cochrane CENTRAL, 1018 from EMBASE, 325 from Web of Science, 786 from Scopus, and 1 from the manual search. Duplicates (n = 1862) were removed.
After screening, 4014 records were excluded. A total of 56 full-text articles were assessed for eligibility against the PICOS criteria, 14 of which were excluded. This yielded 43 records to be included in this review. Some records had more than 2 intervention arms per test intervention or more than 2 studies within the same record. This was the case for acetate studies, such as that of Scheppach et al. (28) (2 studies within the same record), and vinegar studies, such as those of Brighenti et al. (29) (2 interventions compared with the same control), Johnston and Buller (30)  Nevertheless, not all comparisons were meta-analyzed. The comparisons chosen for meta-analysis are described in the footnote of each forest plot. Therefore, from the 43 identified records, there were 52 studies within those records, 44 of which were included in the meta-analysis. Five investigated acetate (all acute interventions), 2 investigated butyrate (both reporting the same chronic study), 14 investigated propionate (8 acute and 6 chronic interventions), 31 studies investigated vinegar (25 acute and 6 chronic interventions), and 5 investigated mixed SCFAs.
Tables 2-6 describe the study design and participant characteristics of all eligible studies. In the interest of clarity, SCFA interventions will be referred to in the text in the simple forms of acetate, butyrate, or propionate. However, some studies have used different compounds of the SCFAs, such as sodium propionate.

Risk of bias in included studies
The risks of bias for the included studies are described in Figure 2. Out of the studies, 47% were determined to have a high risk of bias, 44% to have a moderate risk of bias, and 9% to have a low risk of bias. The domains of greatest concern were risks of bias arising from deviations from intended interventions (D2.2), in measurements of the outcome (D4), and in the selection of reported results (D5).

Acetate.
The characteristics of the eligible acetate studies are summarized in Table 2.

Records identified from
Reports not retrieved (n = 0)

Vinegar
The characteristics of the eligible vinegar studies are summarized in Table 3.
Homogeneity via a Galbraith plot, publication bias via a funnel plot, and a sensitivity analysis via a leave-1-out plot were assessed and are reported in Supplemental Figures 3-7.
Homogeneity via a Galbraith plot, publication bias via a funnel plot, and a sensitivity analysis via a leave-1-out plot were assessed and are reported in Supplemental Figure 8.
Two eligible studies (49,50) investigated the effects of chronic vinegar supplementation on HbA1c in individuals with T2D. Patients were supplemented with 15 mL and 20 mL of vinegar for a month or 10 weeks, respectively (49,50). The authors in both the 15-mL and 20-mL studies reported a significant decrease in HbA1c (of 7% and 9%, respectively), whereas in the placebo group the HbA1c levels decreased by 1% and increased by 2%, respectively (49,50).
Three studies investigated the degree of insulin resistance, using HOMA-IR, following chronic vinegar intake (46)(47)(48). Two of these were investigated in a healthy cohort, showing that while 4 weeks of supplementation of 21 g of vinegar did not result in a significant change compared to the control group (48), 8 weeks of supplementation led to a significant decrease of 8% in HOMA-IR compared to the control groupn (47). One study provided vinegar supplementation in people with T2D for 8 weeks and reported that HOMA-IR, Quantitative Insulin-Sensitivity Check Index (QUICKI), and HOMA-β values were not significantly different compared to those in the control group (46).

Butyrate
The characteristics of the eligible butyrate studies are summarized in Table 4.

Chronic interventions
Two interventions reported glycemic outcomes following a chronic intervention with butyrate (52, 53). These 2 records described the same study, so a meta-analysis was not possible. In this study, 60 participants with type 2 diabetes were randomized in a parallel design to 4 groups (n = 15 in each), in which they had to consume 6 oral capsules (100 mg) and 10 g of powder a day for 45 days. Two of the interventions were assessed in this review. These were sodium butyrate capsules and starch powder (intervention a), and starch capsules and starch powder (control). One publication reported no significant differences in FBG, PBG at 2 hours, FI, HbA1c, and HOMA-IR values postintervention compared to the control group (52). The other reported QUICKI results, which were not significantly different from those in the control group following the 45-day intervention (P = 0.137) (53). Compared to the control group, GLP-1 secretion significantly increased following the butyrate intervention (by 22.57 pg/ml; P = 0.008) when adjusted for the baseline value, BMI, and blood pressure (52).

Propionate
The characteristics of the eligible propionate studies are summarized in Table 5. RCT, randomized controlled trial; XO, crossover. Abbreviations: HV, healthy volunteers; NI, no information; PBG, postprandial blood glucose; PI, postprandial insulin; RCT, randomized controlled trial; XO, crossover. 2 The second enema was given 3 hours after the first infusion and given with 75 g of oral glucose load to represent the postprandial state.
Forest plots of the pooled effects of propionate acute interventions on glycemic outcomes are shown in     Homogeneity via a Galbraith plot, publication bias via a funnel plot, and a sensitivity analysis via a leave-1-out plot were assessed and are reported in Supplemental Figures 9 and 10.
Forest plots are shown in Figures 13-16. Chronic interventions with propionate had no significant effect on the PBG ( Figure 13 Homogeneity via a Galbraith plot, publication bias via a funnel plot, and a sensitivity analysis via a leave-1-out plot were assessed and are reported in Supplemental Figures 11-14.

Mixed SCFAs
The characteristics of the eligible studies using mixed SCFAs are summarized in Table 6.

Acute administration
Five studies (36,(65)(66)(67)(68) were found to investigate the effects of acute, mixed SCFA administration on glycemic control, all of which were given via the ileum or rectum. In 6 healthy individuals, Wolever and colleagues (65) rectally infused       compared to acetate alone, independently of changes in glucagon, free fatty acids, and total cholesterol.
Another study in healthy, lean males who received a 18hour ileal perfusion of SCFAs (acetate at 60 mmol/L, propionate at 25 mmol/L, and butyrate at 15 mmol/L) (67). This was followed by a saline solution at 12 hours. The study showed that mixed SCFA administration did not have an effect on insulin sensitivity, basal hepatic glucose production, or concentrations of triacylglycerol, total cholesterol, and insulin between the 3 conditions (67).
Canfora and team (68) used a rectal infusion of 200 mmol/L of a high-acetate, -propionate, or -butyrate solution or a placebo solution in healthy participants, followed by a 75-g glucose load. There was no differential effect on blood glucose or insulin levels between the 3 SCFA mixtures.

Certainty of evidence
The certainty of evidence ranged from low to very low for all glycemic outcomes in all SCFAs investigated. GRADE    143). A random-effects model was used to calculate Hedge's g (squares), 95% CIs (horizontal lines), and summary effects (diamond). The study weight (expressed as a percentage) indicates the relative contribution of an individual study to the overall pooled effect size. Between-study heterogeneity was calculated using the I 2 statistic. A P value ≤ 0.05 was considered statistically significant. Abbreviations: iAUC, incremental AUC; H, healthy volunteers; T2D, type 2 diabetes.  assessments of each outcome can be found in Tables 7-11. The certainty of evidence was downgraded due to high risks of bias, imprecision due to small sample sizes and wide CIs, and differences in study methodologies, dose sizes, and study populations. Furthermore, for many of the studies the glycemic response was not the primary outcome, reducing the likelihood that researchers would be able to detect an effect.

Confounders of glycemic control
The studies were assessed for controlling for known confounders of glycemic control (body weight and fat change; standard evening meal; whether fiber, strenuous exercise, and alcohol were avoided; and whether an overnight fast was completed prior to the study visit), which are summarized in Supplemental Table 3.
In acute studies, 1 acetate (28) and 3 vinegar (31, 33, 41) studies instructed participants to consume a standard evening meal. Participants in 4 vinegar studies (31,33,34,41) and 2 propionate studies (54,57), avoided strenuous physical activity. One acetate study (28), 3 vinegar studies (33,34,41), and 3 propionate studies (54, 57, 59) discouraged participants from consuming alcohol. Three acetate studies (28,35,36), 1 vinegar study (54), 2 propionate studies (36,54), and 2 mixed vinegar    123). A random-effects model was used to calculate standardized mean differences (squares), 95% CIs (horizontal lines), and summary effects (diamond). The study weight (expressed as a percentage) indicates the relative contribution of an individual study to the overall pooled effect size. Between-study heterogeneity was calculated using the I 2 statistic. A P value ≤ 0.05 was considered statistically significant. When interpreting effect sizes, values <0. 40 = 117). A random-effects model was used to calculate standardized mean differences (squares), 95% CIs (horizontal lines), and summary effects (diamond). The study weight (expressed as a percentage) indicates the relative contribution of an individual study to the overall pooled effect size. Between-study heterogeneity was calculated using the I 2 statistic. A P value ≤ 0.05 was considered statistically significant. When interpreting effect sizes, values <0.40 were categorized as having a small effect size, values 0.40 to 0.70 as having a moderate effect size, and values >0.70 as having a large effect size. Abbreviation: iAUC, incremental AUC. studies (36, 67) prescribed a low-fiber diet before the study visit. All acute studies, excluding 1 which provided no information (29), required participants to fast for more than 6 hours before the study visit.

Adverse events
The adverse events (AEs) reported for each intervention arm for each study category were assessed (Supplemental Table 4). AE data were not disclosed in all publications.  (63) was reported in the meta-analysis. A random-effects model was used to calculate standardized mean differences (squares), 95% CIs (horizontal lines), and summary effects (diamond). The study weight (expressed as a percentage) indicates the relative contribution of an individual study to the overall pooled effect size. Between-study heterogeneity was calculated using the I 2 statistic. A P value ≤ 0.05 was considered statistically significant. When interpreting effect sizes, values <0. 40 = 67). A random-effects model was used to calculate standardized mean differences (squares), 95% CIs (horizontal lines), and summary effects (diamond). The study weight (expressed as a percentage) indicates the relative contribution of an individual study to the overall pooled effect size. Between-study heterogeneity was calculated using the I 2 statistic. A P value ≤ 0.05 was considered statistically significant. When interpreting effect sizes, values <0.40 were categorized as having a small effect size, values 0.40 to 0.70 as having a moderate effect size, and values >0.70 as having a large effect size. Abbreviations: GIT, gastrointestinal tract; iAUC, incremental AUC.
In acute interventions, no AEs were reported for acetate nor vinegar (35,41,44,45). Two studies reported AEs for propionate interventions: 1 case of nausea was reported (57) and 1 study documented no AEs (54). Two studies reported AEs for mixed SCFA administration, and 1 had up to 6 incidences of belching, in both the intervention and control groups (65), while the other reported no AEs (68).
In chronic interventions with propionate, there were 3 incidences of flatulence when consuming inulin-propionate ester bread (61). Another study with propionate reported 6 cases of nausea, 2 cases of constipation, 1 case of flatulence, and 4 cases of vomiting in the intervention group, although nausea, constipation, and flatulence were also reported in the control group (64). One study with propionate assessed AEs but did not have any to report (63). Three vinegar studies reported AEs (47,48,50), but only 1 reported 1 case of nausea, stomachache, and headache (48).

Compliance
Study adherence to the intervention for each study category was assessed (Supplemental Table 5). Withdrawals were defined as participants that dropped out after randomization. One acute study reported a 40% withdrawal rate, because researchers failed to clip the catheter to the colonic mucosa, which meant the SCFA could not be administered (35). Studies acutely supplementing propionate had 100% adherence, whereas chronic interventions had an average withdrawal rate of 12%, and 1 study reported a participant withdrawal due to nausea (64). Studies using chronic butyrate had a 1.7% withdrawal rate due to losses to follow-up (52,53).

Discussion
A total of 43 publications with 46 studies and 913 participants were incorporated into our analysis, which showed

FIGURE 15
Forest plots for randomized controlled trials of chronic propionate on fasting blood glucose. Chronic interventions with propionate had a main effect of −0.14 (95% CI, −0.47 to 0.19; P = 0.41) on the postintervention fasting blood glucose (n = 67). A random-effects model was used to calculate standardized mean differences (squares), 95% CIs (horizontal lines), and summary effects (diamond). The study weight (expressed as a percentage) indicates the relative contribution of an individual study to the overall pooled effect size. Between-study heterogeneity was calculated using the I 2 statistic. A P value ≤ 0.05 was considered statistically significant. When interpreting effect sizes, values <0. 40 67). A random-effects model was used to calculate standardized mean differences (squares), 95% CIs (horizontal lines), and summary effects (diamond). The study weight (expressed as a percentage) indicates the relative contribution of an individual study to the overall pooled effect size. Between-study heterogeneity was calculated using the I 2 statistic. A P value ≤ 0.05 was considered statistically significant. When interpreting effect sizes, values <0.40 were categorized as having a small effect size, values 0.40 to 0.70 as having a moderate effect size, and values >0.70 as having a large effect size. that acute vinegar administration had a favorable effect on blood glucose in subjects with T2D or IGT and healthy participants. Acute and/or chronic administrations of acetate, vinegar, propionate, butyrate, and mixed SCFAs had no effect on glycemic measures, including FBG, FI, PBG, and PI. A summary of the results of the meta-analyses can be found in Table 12.

Acute SCFA administration
The effects of acute SCFA administration upon glycemic responses has been explored using acetate, vinegar, propionate, and mixed SCFAs. Doses varied widely, from 12 to 200 mmol/L, and durations ranged from 60 to 1080 minutes. Our findings, which suggest that acute vinegar influences PBG, correspond   6,7,8 Vinegar may reduce or have little to no effect on fasting blood glucose, but the evidence is very Vinegar may increase or have little to no effect on fasting insulin, but the evidence is very uncertain  The evidence suggests that mixed SCFA results in little to no difference in acute postprandial insulin. One study saw a small, significant decrease in insulin secretion after mixed SCFA administration 1 In the GRADE Working Group grades of evidence, low certainty indicates our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Abbreviations: GRADE, Grading of Recommendations, Assessment, Development, and Evaluation; HV, healthy volunteers; IGT, impaired glucose tolerance; RCT, randomized controlled trial; T2D, type 2 diabetes. 2 The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). 3 It is difficult to generalize the results due to the small sample sizes. Rectal and gastric infusions mean that the interventions are not very applicable, replicable, or tolerable, reducing transferability. The studies tend to have small sample sizes, increasing imprecision. although digestion could be modulated by phenolic compounds rather than by the presence of acetic acid in vinegar (74). Increased fecal bulk was also reported after propionate administration (55), suggesting that undigested starch could be reaching the colon, thereby reducing the glycemic load. Furthermore, acetate and propionate administration promoted gluconeogenesis in rodent studies (75,76), but gastric administration of propionate was not shown to have any effect (36). One study (66) reported   (68). Reductions in circulating free fatty acids have been observed after propionate and acetate administration, which could suggest acetate found in the peripheral circulation may influence fat oxidation (28,35,36,59,60,65,68). However, when acetate was administered via constant gastric infusion in rats in another study, researchers reported increases in lipogenesis and insulin resistance (78).
In the few acute studies that recorded AEs, nausea and belching were reported. The lack of AE reporting makes it difficult to determine the tolerability and safety of SCFA administration. Reported withdrawals were due to methodological issues, noncompliance, unpleasant tastes of interventions, and participant availability. Although there were AEs, the results suggest SCFAs could be tolerated by participants. Unfortunately, this cannot be confirmed, as most studies did not report compliance or withdrawals.
Acute vinegar supplementation was shown to significantly improve PBG in all participants. However, the GRADE certainty of evidence for all acute outcomes, except for mixed SCFAs, was very low. Caution should be taken when interpreting these results, as there is still little to no certainty that acute SCFA administration has any effect on PBG and PI. The low certainty of evidence and lack of significant results in this systematic review stems from variability in the route of administration, dosage, participant health, and sample size, demonstrated by the high heterogeneity of some outcomes [PBG in healthy volunteers (HV) and PI in unhealthy volunteers after vinegar supplementation] and wide I 2 95% CIs. Furthermore, publication bias was detected in several outcomes [excluding PBG for vinegar (HV), propionate, and mixed SCFAs and PI for propionate]. Sensitivity analyses also indicated that 3 studies (29,31,38) were driving the outcome for acute PBG of vinegar (HV), and 1 study (40) had a significant influence on the outcome of acute PI in HV. These factors influenced the precision and directness of the outcomes, which downgraded the quality of evidence.

Chronic SCFA administration
The effects of chronic SCFA supplementation in the forms of vinegar, butyrate, and propionate have been investigated. Doses were from 12 to 200 mmol/L, and study durations ranged between 2 and 70 days. All interventions were orally administered, and some studies used alimentary vehicles, such as bread, smoothies, cheese, and dietary fibers. In this review and meta-analysis, chronic supplementation of SCFAs had no significant effect on PBG, PI, FBG, or FI.
Two studies administered more than 15 ml of vinegar per day and reported a significant reduction in FBG (46,50). Reducing the rate of gastric emptying, via vinegar administration, could have an effect on fasting glucose and insulin concentrations in the long term (65,71,72). However, this is yet to be extensively studied. Some studies attributed changes in fasting glycemia to reduced oxidative stress, which can be associated with both the acetic acid and phenolic compounds present in vinegar (79,80), but only 1 study reported an increase in 2,20-diphenyl-1picrylhydrazyl, a free radical (46). Oxidative stress is associated with reductions in insulin sensitivity and glycemic deterioration (81).
Chronic sodium butyrate supplementation showed no effect on glycemic measures compared to a starch placebo. Furthermore, a study using 13 C-labelled butyrate administered in the colon demonstrated that only 2% of butyrate is found systemically (82). At this concentration, it is unlikely that butyrate would exert an effect on metabolically active tissues, such as adipocytes and skeletal muscle, to alter glucose tolerance.
Chambers and colleagues (62) found that fasting insulin, Matsuda index (a measure of insulin sensitivity), and HOMA-IR values improved when propionate was supplemented compared to results in the cellulose group, but were not significantly different to the insulin control group. This perhaps reinforces previous findings suggesting that incubation with propionate inhibits apoptosis and stimulates insulin release in pancreatic β-cells in vitro (69). As inulin can be used for bacterial fermentation, it is not possible to distinguish whether the effect seen was due to increased propionate or overall SCFA production.
An increase in 2-hour postprandial GLP-1 levels was detected following the butyrate intervention (52). Butyrate and other SCFAs act as signaling molecules in the colon, binding to FFAR2 and FFAR3 expressed in the enteroendocrine L-cells and triggering the release of gut hormones such as GLP-1 (83). Chronic propionate administration to adipocytes expressing FFAR2 has been shown to inhibit lipolysis and reduce circulating nonesterified fatty acid concentrations (84). Maintaining low circulating concentrations of nonesterified fatty acids may prevent β-cell dysfunction and peripheral insulin resistance over time (85).
The AEs reported included nausea, flatulence, constipation, vomiting, and belching. Reported withdrawals were due to noncompliance, losses to follow-up, personal reasons, AEs both unrelated and related (vinegar, propionate) to the intervention, consent withdrawal, and the unpleasant taste of the intervention (vinegar). In some cases, withdrawal reasons were not reported; thus, whether they are related to the intervention is not known. These results could suggest that these interventions are not feasible or tolerable for chronic supplementation in a free-living population.
The GRADE certainty of evidence for all chronic outcomes was very low. Caution should be taken when interpreting these results, as there is still little to no certainty that chronic SCFA administration has any effect on PBG, PI, FBG, or FI. High heterogeneity was detected in chronic FBG and PBG responses to vinegar supplementation, and wide 95% CIs were seen for all heterogeneity results, excluding chronic vinegar supplementation and postprandial glucose responses, indicative of lower certainty of the heterogeneity value. Furthermore, publication bias was detected in all chronic outcomes. These factors influenced the precision, consistency, and directness of the outcomes, which downgraded the quality of evidence.

Future research on SCFAs and glycemic control
Further high-quality, double-blind, randomized controlled trials using standardized study methodology (sites of administration, types of interventions) and powered to detect changes in glycemic responses are required. Future studies should employ gold-standard methodology (e.g., euglycemic hyperinsulinemic clamps) to elucidate the impacts of SCFAs on insulin sensitivity. Studies should also account for and report on glycemic confounders, such as physical activity and body weight changes. Researchers should ensure test foods are blinded and palatable to avoid triggering delayed gastric emptying and confounding results. Further research should attempt to elucidate the effects of SCFA interconversion in the colon by gut microbiota on the host glucose metabolism. Furthermore, tracer and dose-response studies could help to determine optimal concentrations of SCFAs and the roles of SCFAs in different metabolic states. Lastly, improving the understanding of FFAR desensitization during chronic administration of SCFAs may provide insight into how chronic SCFA supplementation could play a therapeutic role in the future.

Strengths and limitations
One strength of this systematic review is that studies were separated according to categories to increase homogeneity. By separating individuals based on health status, we were able to demonstrate that vinegar supplementation may influence glycemia in metabolically compromised individuals; to our knowledge, this was previously unreported. One limitation of this review is that the paired nature of data from crossover studies was not accounted for, so these studies may be underweighted. Our restrictive inclusion criteria identified a small and heterogeneous pool of studies for each subject category. Some records did not report uniform glycemic measures, such as AUCs, for acute interventions and could not be included in the meta-analysis, which highlights the need to standardize glycemic outcome reporting to aid future meta-analyses. In this review, we analyzed the PBG and PI data via the iAUC, potentially masking any significant differences at specific time points (e.g., first-phase insulin response), which could be informative of metabolic disease progression for subjects with or at risk of T2D.

Conclusions
The present systematic review and meta-analysis found that acetate, butyrate, and propionate have no effect on glycemic control. Acetic acid, in the form of vinegar, acutely reduced blood glucose levels in adults with T2D and IGT and in healthy adults. This evidence comes from a very limited and heterogeneous number of studies for all categories, with moderate to high risks of bias and low to very low certainties of evidence. Future highquality research should be focused on investigating the effects of both acute and chronic interventions of SCFAs, with glycemic control measures as the primary outcome in subjects of all health statuses. Such studies should be controlled for confounders that can affect glycemia to ensure that high-quality evidence is produced.

Data Availability
Data described in the manuscript, code book, and analytic code will be made available upon request pending application and approval. The protocol for this systematic review and meta-analysis was registered in the international prospective register of systematic reviews (PROSPERO, registration number: CRD42021231115). Supplemental data is available via the "Supplementary data " link in the online posting of the article at https://academic.oup.com/ajcn/.