Effect of lactic acid fermented foods on glycemic control in diabetic adults: a systemic review and meta-analysis of randomized controlled trials

Abstract Lactic acid bacteria (LAB) fermented foods are reported to have potential in managing glycemic control. This systematic review aimed to evaluate the effectiveness of LAB-fermented foods on improving glycemic control in adults with prediabetics or type 2 diabetes mellitus (T2DM). Randomized controlled trials (RCTs) on LAB fermentation-related foods were searched on PubMed, Cochrane, Excerpta Medica database (EMBASE), Cumulative Index to Nursing and Allied Health Literature (CINAHL), and Web of Science. Sixteen RCTs were included, and the results concluded LAB-fermented food had significant effects in HbA1c (Z = 6.24, MD = −0.05, CI: −0.07 to −0.04, p ≤ 0.00001), fasting plasma glucose (Z = 2.50, MD = −0.16, CI: −0.29 to −0.04, p = 0.01) and fasting serum insulin (Z = 2.51, MD = −0.20, CI: −0.35 to −0.04, p = 0.01). There were significant effects on lipid profile, inflammatory markers, and body mass index in secondary analyses. Subgroup analysis suggests LAB-fermented consumption with a longer duration, younger age group and adults with T2DM, had a larger effect size. Clinicians could offer LAB-fermented food as dietary recommendations for prediabetic and diabetic adults. Larger trials are warranted to verify LAB-fermented food benefits on glycemic control. Systematic Review Registration: PROSPERO Registration No. CRD42022295220


Introduction
Diabetes mellitus is a chronic metabolic disorder involving abnormally high blood glucose levels (American Diabetes 2021).According to the International Diabetes Federation (IDF), diabetes is one of the largest health concerns and causes of death globally, accounting for 4.2 million deaths in 2019 (IDF 2019).Global expenditure on diabetes is expected to reach $825 billion by 2030, up from $760 billion in 2018 (IDF 2019).There was an estimated 463 million adults with diabetes in 2019 and is projected to increase to 700 million by 2045 (IDF 2019).Type 2 diabetes mellitus (T2DM) is characterized by insulin resistance, where the body does not respond to insulin effectively (Khan et al. 2019).Prediabetes is a condition of intermediate hyperglycemia with blood glucose levels below the T2DM diagnostic threshold (Hostalek 2019).Its global prevalence was estimated to be 352.1 million individuals and is anticipated to increase to 587 million by 2045 (Hostalek 2019).However, due to the complexities of prediabetes identification, 90% of prediabetic individuals are unaware of their condition, leaving many undiagnosed and untreated (Hostalek 2019).Hence, T2DM and prediabetes will present an exponential increase in socioeconomic and health burden globally, requiring effective measures to assess and combat this diabetes epidemic.
Glycemic markers are essential in routine practice and clinical trials to guide and assess the efficacy of diabetic treatments on glycemic control.Glycosylated hemoglobin (HbA1c) is considered the gold standard for diabetes diagnosis.It is more accurate and stable to predict long-term outcomes of diabetes-related complications (Chehregosha et al. 2019).Fasting plasma glucose (FPG) test uses enzymatic methods to produce highly precise screening results rapidly (Gurung and Jialal 2021).Fasting serum insulin (FSI) measures the insulin in the body.Insulin is vital for maintaining glucose homeostasis and monitoring insulin resistance (Gutch et al. 2015).The Homeostasis Model Assessment for Insulin Resistance (HOMA-IR) assesses pancreatic beta-cell function by the calculation of both fasting glucose and insulin concentration levels.Therefore, these glycemic markers are essential in guiding and assessing the efficacy of diabetic treatments on glycemic control (Kohnert et al. 2015).
Prediabetes may be reversible through lifestyle modifications with a healthy diet, which could reduce the risk of T2DM by 40-70% (Colberg et al. 2016).However, rapid urbanization and increase economic growth caused a shift in lifestyle habits favoring caloric-dense diets and sedentary living (Cheng et al. 2019).Dietary choices and level of diabetes control can affect the lipid profile of an individual.Lipid abnormalities are prevalent in diabetes as insulin resistance can disrupt lipid metabolism pathways (Athyros et al. 2018).Lipid profile is a direct measure of blood components such as total cholesterol (TC), low-density lipoproteins (LDL), high-density lipoproteins (HDL), and triglycerides (Schaefer et al., 2016).Research suggests a causal effect of obesity on diabetes prevalence (Gupta and Bansal 2020).A moderate increase in body mass index (BMI) is associated with an increased onset of diabetes development (Gray et al. 2015).Inflammatory markers such as tumor necrosis factor-alpha (TNF-α), interleukin-6 (IL-6), and high-sensitivity C-reactive protein (hs-CRP) are associated with diabetes.Hyperglycemia can stimulate inflammatory cytokines release, inducing hs-CRP secretion from the liver and indicating T2DM development (Lal 2021).Studies suggest an intricate relationship between diabetes and the gut microbiome, where diet plays a decisive role in the gut microbiome that affects glucose metabolism (Bezirtzoglou et al. 2021).The gut microbiome is a potent mediator of gut-derived serotonin synthesis, as approximately 90% of total body serotonin is synthesized from enterochromaffin cells in the linings of the gut wall (Martin et al. 2019).This peripheral source of serotonin is itself an important regular of glucose homeostasis (Martin et al. 2019).Hence, gut-derived peripheral serotonin provides a signaling nexus between diet, gut microbiome, and metabolism (Martin et al. 2017).
Lactic acid fermentation is an anaerobic process where sugars are broken down into lactic acid and other products (Şanlier, Gökcen, and Sezgin 2017).Lactic acid bacteria (LAB) have an enzymatic capacity for fermentation.LAB consists of a broad category of bacteria species, including Lactobacillus, Streptococcus, and Bifidobacterium, and are usually measured as colony forming unit (CFU) to estimate the number of viable microbial cells per sample (Pasolli et al. 2020).Many LAB species are present in fermented foods, including yogurt, kefir, sauerkraut, kimchi, cheonggukjang, sourdough, kombucha, miso, and natto (Chilton, Burton, and Reid 2015).LAB-fermented foods are well-known for their health benefits toward preventing and managing metabolic disorders, cognitive improvement, immune enhancement, and so on (Şanlier, Gökcen, and Sezgin 2017).Many fermented foods are considered functional foods with probiotic properties, therefore LAB-fermented foods could represent a safe, inexpensive, and reliable dietary intervention (Castellone et al. 2021).LAB-fermented foods can modulate the gut microbiome in its functionality and response to glycemic control (Castellone et al. 2021).During fermentation, LAB can decompose macromolecular components in food like degradation of indigestible polysaccharides, while synthesizing health-promoting metabolites (Wang et al. 2021).Major products of LAB fermentation such as short-chain fatty acids (SCFAs), bacteriocins, vitamins, and exopolysaccharides (EPS) have antidiabetic, anxiolytic and anti-inflammatory effects (Sahab et al. 2020).Hence, studies suggest LAB-fermented foods possess anti-obesity, anti-diabetic, and anti-carcinogenic properties (Evivie et al. 2017).
To date, there are numerous primary publications on fermented food in relation to glycemic control in both healthy and T2DM subjects, however, limited reviews or meta-analyses aggregate these individual studies (Melini et al. 2019).Notably, there are similar reviews that studied the therapeutic effect of LAB-fermented food on glycemic control (Barengolts et al. 2019;Salari et al. 2021).However, these reviews were limited to heterogeneous outcome analysis due to the poor quality of studies used (Salari et al. 2021), and did not report on prediabetes or outcomes like inflammatory markers and lipid profile (Barengolts et al. 2019;Salari et al. 2021).It will be scientifically valuable to fill the gaps and provide better insights into potential therapeutic alternatives, justifying LAB-fermented food inclusion on health dietary guidelines for glycemic control and beyond.Hence, this systematic review aims to examine all available good-quality randomized controlled trials (RCTs) and assess the effectiveness of LAB-fermented foods on glycemic control in both T2DM and prediabetic adults.

Methods
This systematic review followed the guidelines of the preferred reporting items for systematic review and meta-analyses (PRISMA) (Supplemental Table S1) (Page et al. 2021).A review protocol was developed a priori.The protocol was registered on PROSPERO (CRD42022295220).

Eligibility criteria
The selection for eligible studies was established according to population, intervention, comparison, outcomes, study design (PICOS) (Table 1).HbA1c, FPS, FSI, and HOMA-IR were classified as primary outcomes, while lipid profile, inflammatory markers, and BMI as secondary outcomes.The selection criteria included: adults with T2DM or prediabetes; examining LAB-fermented food effects on any primary or secondary outcomes; placebo as comparison and RCT studies; studies were published in English; and studies were peer-reviewed.The selection criteria excluded: subjects with type 1 diabetes or gestational diabetes; studies with no usable data for meta-analysis; review or discussion papers, non-experimental; or qualitative studies and animal studies.

Search strategy
A comprehensive search strategy was conducted using PubMed, Cochrane, Excerpta Medica database (EMBASE), cumulative index to nursing and allied health literature (CINAHL), and Web of Science (Supplemental Table S2).This review topic is related to the focus of these five databases and their combination could provide extensive coverage of biomedical and healthcare-related RCTs, without compromising the validity of the research (Bramer et al. 2017).These searches were conducted from the inception of data-based through December 2021 and were limited to the English language.The keywords and Medical Subject Headings (MeSH) terms were used: lactic acid fermented foods ("fermented" OR "lactic acid bacteria," OR "lactobacillales") AND diabetes mellitus ("diabetes mellitus, type 2" OR "diabetes mellitus, type II" OR "prediabetes" OR "impaired fasting glucose") AND glycemic control ("HbA1c" OR "fasting plasma glucose," OR "fasting serum insulin").Two reviewers independently examined the search strategy (WZT and JYS).Specific keywords and MeSH terms were developed and truncated according to the syntax guidelines of each database.Unpublished clinical trials from trial registries and grey literature sources from ProQuest Dissertation and Theses, and Google Scholar were explored to optimize potential trials.Hand searching was done from the reference lists of relevant systematic reviews or included studies to identify potential additional studies.

Selection process
The bibliographical software EndNote X20 was used to manage all the relevant studies.Titles and abstracts of studies in the search were screened and duplications were removed.The full texts of relevant studies were retrieved and reviewed independently by the reviewers, against the eligibility criteria.Disagreements were resolved through discussion and with a third reviewer (VXW).Excluded studies were recorded under the "Characteristics of excluded studies" table (Table S3 in Supplemental information).

Data extraction
A standardized search outcome table was used to perform data extraction based on the Cochrane Handbook for Systematic Reviews of Interventions guidelines (Higgins et al. 2021).Data extractions were presented based on a pre-developed summary table (Supplemental Table S4), heavily screened and reviewed independently by the reviewers (WZT and JYS).Discrepancies were resolved through discussion with the third reviewer.Data were extracted based on the author, publication year, study design, study duration, population characteristics, sample size and information of intervention and control.Studies with more than one publication were screened thoroughly to prevent duplication of data count.Authors of studies were contacted through email for any missing information or queries, but to no avail.Since different International System of Units (SI) units were reported among the included studies, the outcome units were standardized using conversion factors: HbA1c to percentage; FPS to mmol/L; FSI to μU/ml; TC, LDL, HDL, and triglycerides to mg/dl; CFU to CFU/mL.

Risk of bias
The risk of bias (RoB) was assessed based on the Cochrane Risk of Bias tool, to evaluate the risk of potential bias and to improve the appraisal of evidence (Higgins et al. 2021).Two reviewers independently appraised the internal validity of each full-text included studies based on six domains, namely, random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome assessment, and selective reporting.These domains were graded as low-risk, high-risk, or unclear accordingly.Consequently, each study was assigned an overall bias according to the worst RoB among the six domains.These domains assessments were based on the reliability and interpretation of the result (Supplemental Table S5).Discrepancies were resolved through discussion with the third reviewer.Publication bias among the included studies was also evaluated after assessing the degree of symmetry of a funnel plot.The grading of recommendations, assessment, development, and evaluation (GRADE) was used to determine the quality of outcomes in the included studies and their overall strength of evidence (Guyatt et al. 2008).Two reviewers independently assessed and rated the included studies from very-low-risk to high-risk based on GRADE domains, namely, RoB, imprecision, inconsistency, indirectness, and publication bias, using GRADEpro GDT software.

Data analysis and synthesis
Statistical analysis was performed based on the Cochrane Handbook for Systematic Reviews of Interventions guidelines (Higgins et al. 2021).RevMan software (Review Manager Version 5.4.1) was used for data analysis and synthesis.Cohen's kappa was used to test the inter-rater agreement between the reviewers (McHugh 2012).Continuous data reported from the outcomes were evaluated and derived as mean difference (MD) accompanying 95% confidence intervals (CIs), using inverse-variance method.Due to the variations in the interventional approach, heterogeneity was observed in the included studies, thus a random-effect model was employed to better reflect real-world synthesis where conditions within studies vary.A p-value of <0.05 was considered statistically significant, and Cohen's d was used to assess the overall effect size of LAB-fermented foods, where small (d = 0.2), medium (d = 0.5) and large (d = 0.8) (Lakens 2013).Sensitivity analysis determines the causes of heterogeneity, where it highlighted the robustness of the results from included studies that introduced heterogeneity, and thus were removed (Higgins et al. 2021).Assessment of heterogeneity was performed through the forest plots and utilization of a Cochran Q (Chi-square test) with a 5% level of statistical significance.I 2 statistic describes the proportion of variation in point estimates due to heterogeneity rather than sampling error.The heterogeneity among the included studies was evaluated based on the Chi-square (χ 2 ) test and I 2 statistics.I 2 statistics were calculated with values of 0-40%, 30-60%, 50-90%, and 75-100% representing low, moderate, substantial and high levels of heterogeneity, respectively (Higgins et al. 2021).Subgroup analysis was conducted as there were sufficient data available to determine the effects of LAB-fermented foods on subtypes of diabetes, duration, and age categories.

Results
The search process is presented in the PRISMA flow diagram (Supplemental Figure S1).A total of 2343 results were generated, and 1080 duplicate articles were removed using EndNote.The remaining 1263 articles were screened independently by two reviewers.A total of 1230 articles were then excluded by their titles and abstracts.Full-text versions of the remaining 33 articles were retrieved and reviewed based on the eligibility criteria (kappa = 0.59; moderate agreement).However, 17 articles were excluded with the reasons provided.Hence, 16 full-text articles were included for this systematic review and meta-analysis (kappa = 0.88; almost perfect agreement).
GRADE assessment was performed on both primary and secondary outcomes with more than 10 included studies.The overall quality of the evidence was presented under the "GRADE summary of findings" table, with the justifications in the footnotes (Table 3).The quality of evidence for the primary outcomes were low for HbA1c and very low for fasting plasma glucose and FSI.Whereas the quality of evidence for the secondary outcomes were very low for total cholesterol and low for both LDL and HDL.
Supplemental Figure S3 illustrates publication bias results for HbA1c using a funnel plot, where each dot represents one study.The vertical line represents the pooled effect size.The funnel plot appears asymmetrical.The cluster of studies is plotted near the pooled effect size are considered high precision, while the remaining studies are considered of a lower precision as they are spread symmetrically (Simmonds 2015).We concluded the possibility of publication bias involved subgroup analysis between prediabetic and T2DM studies, where the differences in effect between the larger prediabetic studies and the smaller T2DM studies could introduce heterogeneity, leading to asymmetr y (Simmonds 2015).

Primary outcomes
This review examined the effect of LAB-fermented food on glycemic outcomes in the following sections, using the random-effects model.
e selection bias (35%) and reporting bias (31%) were unclear.only one study had high risk of Performance bias (8%) Most information was from studies still at low or unclear risk of bias.overall potential limitations remained unlikely to lower confidence in the estimate of effect.f Majority of the trials (85%) used less than sample size of 50 per group, with wide confidence intervals.g the funnel plot seems asymmetrical, suggesting some evidence of publication bias for egger's regression asymmetry test.h selection bias (27%), and reporting bias (27%) were unclear.only one study had high risk of Performance bias (9%).Most information was from studies still at low or unclear risk of bias.overall potential limitations remained unlikely to lower confidence in the estimate of effect.i selection bias (38%) and reporting bias (38%) were unclear.
only one study had high risk of Performance bias (13%).Most information was from studies still at low or unclear risk of bias.overall potential limitations remained unlikely to lower confidence in the estimate of effect.j Majority of the trials (75%) used less than sample size of 50 per group, with wide confidence intervals.k selection bias (25%) and reporting bias (20%) were unclear.only one study had high risk of Performance bias (10%).Most information was from studies still at low or unclear risk of bias.overall potential limitations remained unlikely to lower confidence in the estimate of effect.l Majority of the trials (90%) used less than sample size of 50 per group, with wide confidence intervals.

Subgroup analysis
Subgroup analyses were conducted on HbA1c outcomes because HbA1c had an overall significant effect size with a sufficient number of studies eligible for meta-analysis.Detailed results are provided in Table 4 and Supplemental Figure S5-S7.In summary, subgroup differences were significant in the group between prediabetes and T2DM, whereas intervention duration (less-than-8-weeks and more-than-12-weeks) and age groups (less than 51-years-old and above 51-years-old) were not.The results showed prediabetic, longer intervention duration and younger age groups, had a greater effect of intervention than other respective subgroups.The duration of intervention was categorized into less-than-8-weeks and more-than-12-weeks was because the included studies had varied intervention durations of 2, 4, 8, and 12 weeks, therefore when there was a more even split of participants in the chosen category, a subgroup analysis was then possible.In addition, the decision behind categorizing the age group as 51-years-old was because five included studies had participants with a mean age less than 51-years-old while the remaining five studies had a mean age above 51-years-old, therefore eligible studies were split evenly to conduct a subgroup analysis.

Secondary outcomes
This review examined the effect of LAB-fermented food on lipid profile, inflammatory markers, and BMI.Detailed results are provided in Table 5 and Supplemental Figure S8-S15.In summary, the results yielded significant differences in TC, LDL, HDL, triglycerides, TNF-α, IL-6 hs-CRP, and BMI after sensitivity analysis.

Discussion
The results of this systematic review and meta-analysis provided evidence on the beneficial effects of LAB-fermented food consumption on glycemic control among persons with diabetes.It is worth mentioning the benchmark against clinical changes seen from pharmacological treatments such as oral diabetic medications where HbA1c would drop significantly by 0.3-1.1% (Fang et al. 2022).While it is fair to say the magnitude of change induced by LAB-fermented food would not be as great as with drug treatments, the results from this review can be interpreted with great interest from a non-pharmacological point of view.Overall, HbA1c, FPG, FSI, and HOMA-IR showed statistically significant improvements in glycemic control.Our findings are consistent with a meta-analysis (Salari et al. 2021) that observed significant differences in FPG and FSI but inconsistent with other meta-analyses (Barengolts et al. 2019;Salari et al. 2021) that observed no significant difference in HbA1c.A possible discrepancy could be explained by the differences in characteristics of study samples.A possible reason for these findings is that LAB-fermented foods are lower in glycemic index and rich in fiber, which facilitates easier digestion and absorption, decreasing insulin requirements, and improving glycemic control (Melini et al. 2019).Additionally, due to the enzymatic activity of fermentation, most fermented foods are already pre-digested into amino acids and nutrient-rich simple sugars, making them easier for absorption by the body (Melini et al. 2019).Metabolites such as SCFAs produced during fermentation in LAB-fermented food also block carbohydrates from entering the blood and becoming blood sugar.SCFAs strongly influence glucose metabolism by promoting glucagon-like peptide-1 (GLP-1) expression in enteroendocrine-L cells (Supplemental Figure S16), which suppresses glucagon secretion, and promotes beta-cell proliferation, intensifies glucose-induced insulin release from beta-cell while protecting beta-cells from apoptosis (Cunningham, Stephens, and Harris 2021).SCFAs are associated with activation of intestinal gluconeogenesis via a cyclic adenosine monophosphate (cAMP)-dependent mechanism (Supplemental Figure S16), triggering pancreatic secretion to increase glucagon and insulin levels, reducing blood glucose levels (Boulange et al. 2016).LAB-fermented foods have probiotic properties that confer health benefits on the host (Melini et al. 2019).Genome-wide analysis has concluded that closely-related LAB strains exist in both the food and gut microbiome, providing substantial evidence that LAB foods can be considered a viable source of LAB for the gut microbiome (Pasolli et al. 2020).A healthy gut microbiome encourages better absorption of carbohydrates into the intestine and enhances insulin sensitivity, which prevents blood sugar spikes (Li et al. 2016).
Subgroup analysis revealed statistically significant effects of LAB-fermented food on HbA1c between prediabetes and T2DM groups.Studies found Proteobacteria and Escherichia Shigella are higher in the T2DM group than in the  s5 in supplemental information. 2duration of intervention were categorized into less-than-8-weeks and more-than-12-weeks.Forest plots presented in Figure s6 in supplemental information. 3age groups were categorized into less-than-51-years-old and more-than-51-years-old. we used 51-years-old because five included studies had participants with a mean age less than 51-years-old while the remaining five studies had a mean age above 51-years-old, ensuring eligible studies were split evenly.Forest plots presented in Figure s7 in supplemental information.prediabetes group but were not statistically significant.These gram-negative bacteria contain lipopolysaccharides (LPS), which can cause metabolic endotoxemia, increasing inflammation and inducing insulin resistance (Zhang et al. 2021).
As such, inflammatory cytokines levels could predict T2DM development through diminishing insulin sensitivity (Wang et al. 2016).However, this discrepancy is still unclear in this subgroup analysis, thus recommended for further study.
Nevertheless, this analysis is consistent with the possibility of the gut microbiome being a modifiable determinant of both T2DM and prediabetes closely-related to dietary interventions, hence LAB-fermented food can be recommended for both groups (Wu et al. 2020).Subgroup analyses were not significant for intervention duration and age group.Nevertheless, the longer duration had larger effect sizes than the shorter duration in reducing HbA1c.A longer intervention duration would have allowed for the gut microbiome to adequately adapt to the increase in fiber-rich LAB-fermented food.Although studies suggest short-term intake may alter microbiome diversity quickly, the alteration is transient and does not persist for long (Leeming et al. 2019).A longer duration of habitual LAB-fermented food consumption could be more effective in promoting increased microbiome diversity (Leeming et al. 2019).This may explain the pronounced effect of LAB-fermented food on glycemic control observed in a longer duration.Meanwhile, the effect size for the younger age group was much larger than the older age group on reducing HbA1c.The gut microbiome composition has been found to differ across age groups and has been shown to decrease gut biodiversity with age (Xu, Zhu, and Qiu 2019).Aging is often associated with gut dysbiosis, presumably due to physiological changes of the colon, which could increase intestinal permeability to bacterial products, inducing systemic inflammation which affects insulin sensitivity (Thevaranjan et al. 2017).This may explain that a younger gut microbiome can have more effective intestinal gluconeogenesis.Hence, LAB-fermented food can be recommended to younger groups to preserve and to older groups to restore microbiome diversity.
Furthermore, secondary outcomes indicated statistically significant effects of LAB-fermented food on TC, LDL, HDL, triglycerides, TNF-α, IL-6, hs-CRP, and BMI levels.Our findings are consistent with a meta-analysis (Ziaei et al. 2021), showing a significant reduction in TC and LDL levels after probiotic fermented milk consumption.In contrast to our findings, HDL and triglycerides levels were not significant (Ziaei et al. 2021).Ziaei et al. mentioned the discrepancies could be due to the different baseline and intervention characteristics of the included studies and suggested fermented products using both Lactobacilli and Bifidobacteria could be more effective in improving triglycerides and HDL levels.Literature revealed Lactobacilli have cholesterol-lowering effects by deconjugation of bile salts and ferment refined sugars, whereas Bifidobacteria are major producers of SCFAs (Kim et al. 2017;Vlasova et al. 2016).The mixture of Lactobacilli and Bifidobacteria has shown more success in improving the overall lipid profile and ameliorating obesity-related metabolic disorders (Mazloom, Siddiqi, and Covasa 2019).
Nonetheless, our review supports the beneficial effects of LAB-fermented food on lipid profile.Studies have shown EPS produced from LAB, can have hypocholesterolemic effects.Studies observed bile acids were scavenged by EPS, which translates to more utilization of blood cholesterol to produce new bile acids, reducing overall blood cholesterol  s8 in supplemental information.b Four studies were removed after sensitivity analysis (ahn et al. 2018;ejtahed et al. 2011;Jung et al. 2014;oh et al. 2014).c using post-intervention ldl levels.Forest plots presented in Figure s9 in supplemental information.d six studies were removed after sensitivity analysis (ahn et al. 2018;ejtahed et al. 2011;Hove et al. 2015;naito et al. 2018;oh et al. 2014;ostadrahimi et al. 2015).e using post-intervention Hdl levels.Forest plots presented in Figure s10 in supplemental information.f two studies were removed after sensitivity analysis (ejtahed et al. 2011;Hove et al. 2015).g using post-intervention triglyceride levels.Forest plots presented in Figure s11 in supplemental information.h Four studies were removed after sensitivity analysis (ahn et al. 2018;ejtahed et al. 2011;Jung et al. 2014;oh et al. 2014).i using post-intervention tnF-α levels.Forest plots presented in Figure s12 in supplemental information.j one study was removed after sensitivity analysis (toshimitsu et al. 2020).k using post-intervention il-6 levels.Forest plots presented in Figure s13 in supplemental information.l one study was removed after sensitivity analysis (toshimitsu et al. 2020).m using post-intervention hs-CrP levels.Forest plots presented in Figure s14 in supplemental information.n sensitivity analysis was not performed due to low heterogeneity.o using post-intervention BMi values.Forest plots presented in Figure s15 in supplemental information.p one study was removed after sensitivity analysis (naito et al. 2018).(Cabello-Olmo et al. 2019).LAB-fermented foods are found to have bile salt hydrolase (BSH) activity present, which catalyzes the hydrolysis of conjugated bile acids into deconjugated free bile salts and amino acid residues (Supplemental Figure S16).Free bile acids are readily absorbed as there are specialized transporters that pump bile acids from the intestinal lumen into the blood, which then undergoes enterohepatic circulation (Dawson and Karpen 2015).As the conjugated form of bile acids cannot be absorbed, BSH activity in the gut is needed to release free bile acids for absorption, thus the gut microbiome generates a secondary bile acid pool, increasing enterohepatic recirculation (Foley et al. 2019).Circulating bile acids interact with bile acid receptors in major organs to impart signals that regulate metabolic and homeostatic processes such as lipid metabolism (Foley et al. 2019).Unconjugated bile acids are also less effective mediators of cholesterol absorption relative to conjugated bile acids, thus the elevation of BSH activity from LAB-fermented food is likely to directly reduce cholesterol uptake from the lumen (Joyce et al. 2019).SCFAs in LAB-fermented food also inhibit cholesterol synthesis through the inhibition of hydroxymethylglutarate coenzyme A (HMG-CoA) reductase (Supplemental Figure S16), which catalyzes the rate-controlling step in cholesterol production in the liver (Ziaei et al. 2021).
Our findings on TNF-α are consistent with a review (SaeidiFard, Djafarian, and Shab-Bidar 2020), showing a significant reduction after fermented food consumption.In contrast to our findings, IL-6 and hs-CRP findings were not significant (SaeidiFard, Djafarian, and Shab-Bidar 2020).A possible discrepancy could be the high heterogeneity present between the results from SaeidiFard et al.In line with our findings, studies found a significant reduction in IL-6 after consuming fermented soy food (Yang et al. 2018).Studies also found a significant reduction in hs-CRP after fermented milk consumption (Meilina, Anjani, and Djamiatun 2020).Nonetheless, our review supports the possible beneficial effects of LAB-fermented food on inflammatory markers.Prediabetes and T2DM are associated with elevated systematic levels of pro-inflammatory cytokines (Bezirtzoglou et al. 2021).Metabolic inflammation causes LPS-release from bacteria, which can increase intestinal permeability, inducing further systematic inflammatory response and oxidative stress (Scheithauer et al. 2020).Inflammatory markers, like IL-6 and TNF-α, can modulate inflammation, increase beta-cell apoptosis, and reduce insulin secretion, thus affecting glucose metabolism and insulin sensitivity (Gurung et al. 2020).TNF-α interferes with insulin receptor substrate-1 (IRS1) within the beta-cells, which affects insulin-signaling pathways, leading to insulin resistance (Scheithauer et al. 2020).IL-6 increases the suppressor of cytokine signaling 3 (SOCS3) activity, which inhibits other mediators of insulin receptor signaling, decreasing hepatic insulin sensitivity (Scheithauer et al. 2020).Inflammatory cytokines can stimulate hs-CRP production in the liver, inducing hepatic insulin resistance through the RhoA-activation signaling pathway, which involved the phosphorylation of IRS1 (Jeong et al. 2019).
Lactobacillus plantarum in LAB-fermented foods can help suppress inflammatory cytokines stimulation such as LPS (Gurung et al. 2020).SCFA from LAB fermentation can inhibit histone deacetylase action and represses the nuclear transcription factor-kappaB (NF-κB) (Supplemental Figure S16), disrupting pro-inflammatory molecules production and release (Boulange et al. 2016).Given the probiotic properties of LAB-fermented food, it may decrease the expression of cyclooxygenase-2 (COX-2) enzyme that catalyzes prostaglandins from arachidonic acid (Supplemental Figure S16), which stimulates cell proliferation and pro-inflammatory process (SaeidiFard, Djafarian, and Shab-Bidar 2020).Through supplementation of LAB-fermented food, studies observed an increase in microbiome diversity, which decreases inflammatory cytokines activation (Shahbazi et al. 2021).
Finally, our findings coincide with a meta-analysis (Mohammadi et al. 2021), showing a significant reduction in BMI from probiotic fermented milk consumption.The mechanism involved could be associated with BMI, obesity, and gut microbiome (Boulange et al. 2016).Visceral fat accumulation increases inflammatory cytokines, which reduces hepatic insulin sensitivity and causes beta-cell dysfunction (Szotkowská et al. 2021).SCFA acetate from LAB-fermented food may mediate appetite suppression in the central nervous system (CNS) (Supplemental Figure S16), suggesting a mechanism that induces increased activation of the hypothalamic arcuate nucleus (ARC), leading to a rise in lactate and γ-aminobutyric acid (GABA) production, which favors a reduction in appetite and body weight (Frost et al. 2014).This is the first meta-analysis to evaluate the effectiveness of LAB-fermented food on glycemic control, lipid profile, and inflammatory markers.The results suggest glycemic benefits of LAB-fermented food.Subgroup analyses involving types of diabetes, intervention duration, and age groups, are unique to this meta-analysis.Other meta-analyses have similar interventions like the use of probiotic capsules, however, those interventions did not undergo a prior fermentation process to produce metabolites and enzymes responsible for glycemic control.Probiotic capsules supplements can be expensive and limited in availability, therefore LAB-fermented food, which is readily available and inexpensive, can become a growing dietary choice.
There are some limitations of this review.The search strategy was limited to the English language only, which can be subjected to language bias.The present meta-analysis has included a few studies with small sample sizes and relatively short intervention duration.Some included studies have comparators like conventional yogurt and milk, which might potentially alter glycemic outcomes.In addition, the source and amount of LAB-fermented foods have differing CFU consumed, with an average of 10 6 CFU/mL, among participants which may dictate the efficacies of the intervention.As such, this meta-analysis is unable to further analyze the CFU consumption threshold for a health effect.Differences in lifestyle choices like physical activity and diet, and comorbidities were not accounted for in most included studies, which could contribute to glycemic control.No included studies examined the composition of the gut microbiome at baseline and post-intervention to fully analyze the influence of preexisting flora on later outcomes.Finally, this meta-analysis was unable to assess the differences in BMI categories as only two RCTs involved obese or overweight subjects (Naito et al. 2018;Taniguchi-Fukatsu et al. 2012).
This meta-analysis has several implications for clinicians to incorporate LAB-fermented food as part of their dietary advice for diabetic patients.Since it has shown beneficial effects for both prediabetes and T2DM individuals.LAB-fermented food can be emphasized more in the dietary recommendations of the general public as a preventive health measure for the onset of diabetes.LAB-fermented food can be recommended to the younger population and for longer periods to prevent diabetes progression.Fermented foods are commonly consumed in many cultures around the world and are readily accepted as part of the regular diet, hence this review can contribute to policy changes and increases the inclusion of LAB-fermented foods in food guides by health authorities (Chilton, Burton, and Reid 2015).Dietitians can also recommend the dietary composition of LAB-fermented foods on personalized dietary guidelines, such as the quantity of serving, for prediabetics and persons with T2DM, which would enable more accurate predictions of an individual glycemic response to LAB-fermented food (Zeevi et al. 2015).
This review underlined the need to conduct more primary studies with extensive recruitment sample sizes to verify the effectiveness of LAB-fermented food on glycemic outcomes.Future research should focus more on standardizing intervention components, like duration and total CFU consumed, for attaining effective glycemic control among prediabetics and persons with T2DM to aid in the conversion of research evidence into a comprehensive dietary recommendation.Future research may explore whether the baseline composition of the gut microbiome affects the primary and secondary outcomes since it is associated with the immune system in diabetes pathophysiology.
In conclusion, this systematic review and meta-analysis presented beneficial effects of LAB-fermented food in glycemic control, lipid profile, and inflammatory markers in diabetic individuals.Subgroup analyses also concluded improvements in glycemic control.Clinicians could incorporate cost-effective LAB-fermented food into dietary recommendations to promote self-management and preventive habits in combating diabetes.High-quality randomized trials are warranted to recruit larger sample sizes with more consistent study characteristics and evaluate LAB-fermented food effects at different durations and age groups to achieve effective glycemic control.

Figure 1 .
Figure1.Forest plots on the effect of laB-fermented food consumption on Hba1c levels before and after sensitivity analysis.
Salari et al.   had included studies that varied among individuals with diabetes, metabolic syndrome, overweight and healthy individuals, where the intervention also included aerobic exercise and metformin.Barengolts et al. had most included studies using only conventional yogurt as control, which might have provided glycemic improvements, obviating the intervention effects.Nevertheless, our review still supports the possible glycemic improvements of LAB-fermented foods.Animal studies also presented strong evidence of LAB-fermented food improving glucose metabolism and preventing T2DM development(Cabello-Olmo et al. 2019;Hyun et al. 2021;Park, Chang, and Lee 2021;Zulkawi et al. 2018).

Figure 2 .
Figure 2. Forest plots on the effect of laB-fermented food consumption on fasting plasma glucose levels before and after sensitivity analysis.

Figure 3 .
Figure 3. Forest plots on the effect of laB-fermented food consumption on fasting serum insulin levels before and after sensitivity analysis.

Table 2 .
Characteristics of included studies.

Table 3 .
Grade summary of findings.Most information was from studies still at low or unclear risk of bias.overall potential limitations remained unlikely to lower confidence in the estimate of effect.b there are variation in the interventions through types, amount, duration and frequency of laB Fermented Food used in the trials.c note: a selection bias (32%) and reporting bias (27%) were unclear.only one study had high risk of Performance bias (10%).

Table 4 .
subgroup analyses of laB-fermented food on Hba1c.
1 types of diabetes were categorized into prediabetes and t2dM.Forest plots presented in Figure

Table 5 .
results of secondary outcomes.