The Effect of Probiotics and Synbiotics on Risk Factors Associated with Cardiometabolic Diseases in Healthy People—A Systematic Review and Meta-Analysis with Meta-Regression of Randomized Controlled Trials

We aimed to systematically review the effectiveness of probiotic/synbiotic formulations to counteract cardiometabolic risk (CMR) in healthy people not receiving adjunctive medication. The systematic search (PubMed/MEDLINE/Embase) until 1 August 2019 was performed for randomized controlled trials in >20 adult patients. Random-effect meta-analysis subgroup and meta-regression analysis of co-primary (haemoglobin A1c (HbA1C), glucose, insulin, body weight, waist circumference (WC), body mass index (BMI), cholesterol, low-density lipoproteins (LDL), high-density lipoproteins (HDL), triglycerides, and blood pressure) and secondary outcomes (uric acid, plasminogen activator inhibitor-1–PAI-1, fibrinogen, and any variable related to inflammation/endothelial dysfunction). We included 61 trials (5422 persons). The mean time of probiotic administration was 67.01 ± 38.72 days. Most of probiotic strains were of Lactobacillus and Bifidobacterium genera. The other strains were Streptococci, Enterococci, and Pediococci. The daily probiotic dose varied between 106 and 1010 colony-forming units (CFU)/gram. Probiotics/synbiotics counteracted CMR factors (endpoint data on BMI: standardized mean difference (SMD) = −0.156, p = 0.006 and difference in means (DM) = −0.45, p = 0.00 and on WC: SMD = −0.147, p = 0.05 and DM = −1.21, p = 0.02; change scores on WC: SMD = −0.166, p = 0.04 and DM = −1.35, p = 0.03) in healthy persons. Overweight/obese healthy people might additionally benefit from reducing total cholesterol concentration (change scores on WC in overweight/obese: SMD: −0.178, p = 0.049). Poor quality of probiotic-related trials make systematic reviews and meta-analyses difficult to conduct and draw definite conclusions. “Gold standard” methodology in probiotic studies awaits further development.


Introduction
Cardiovascular diseases (CVD) are the most prevalent noncommunicable disorders, with cardiometabolic risk factors (CMRF) including obesity [1], abnormal lipid profile and hypertension [2], insulin resistance, and aberrant glycaemia [3], playing a role in the pathogenesis. Increased consumption of unhealthy, high-calorie foods combined with a sedentary lifestyle further contribute to their poor outcomes [4,5]. In healthy persons, modestly skewed metabolic parameters may stand for the early onset CMRF [2].
Metabolic malfunctions of diverse nature, with epigenetic, hormonal, and infectious factors, are involved in the pathogenesis [6,7]. Intestinal microbiota actively participating in metabolism is an important factor regulating body metabolism [8]. Microorganisms, primarily bacteria, inhabiting our digestive tract actively participate in the digestion of nutrients and, through its metabolites, can regulate not only energy recovery from food but also lipogenesis or fat formation [9]. The mechanisms by which the gut microbiota can contribute to the pathogenesis of metabolic disorders include the short chain fatty acids (SCFAs) biosynthesis to triglycerides and glucose as well as the phenomenon of endotoxemia leading to increased blood levels of liposaccharide (LPS), which aggravates the process of systemic inflammation [10]. Both LPS and LPS-related inflammation have been linked to metabolic diseases, e.g., diabetes and insulin resistance (IR) [11].
The microbiota communicates with the host via toll-like receptors, nuclear factor-kB, and mitogen-activated protein kinase [12], which were shown to improve serum and glucose lipid concentration, to reduce insulin resistance [13,14], and to induce hypocholesterolemic effects [13]. Also, the products of the metabolic activity of the microbiota-predominant SCFAs were shown to regulate various metabolic processes [15]. These molecules after binding to G-protein-coupled receptors make the secretion of peptide YY, which lowers gut motility and augments nutrient absorption [16]. Also, butyrate serves as a source of energy for intestinal cells and improves tissue sensitivity to insulin, counteracting the development of type 2 diabetes. Together with propionic acid, it can stimulate the production of satiety hormones. Of note, butyrate can also stimulate the formation of fat cells and the storage of fat droplets in these cells, presumably through increased glucose uptake or participation in lipid formation. On the other hand, it may also inhibit lipolysis, which, together with stimulating glucose uptake and triglyceride synthesis, makes it a potential therapeutic agent in the fight against hyperglycemia and hyperlipidemia [17].
Considering these facts, metabolic impairment is at least a consequence of gut microbiota alteration. The use of probiotics and synbiotics to counteract metabolic disturbances has been reported. Probiotics are "live microorganisms that, when administered in adequate amounts, confer a health benefit on the host", which has been confirmed in properly controlled studies [18]. Synbiotics are combinations of probiotics and prebiotics. Prebiotics are substrates that are selectively utilized by host microorganisms conferring a health benefit, which must be scientifically documented [19].
A few meta-analyses evaluating the efficacy of probiotics and synbiotics in persons diagnosed with diabetes or hypertension have been published [20][21][22]. However, early-onset CMRF have never been meta-analysed and reported in the literature. Therefore, we conducted the first systematic review and meta-analysis in healthy individuals. We hypothesized that probiotics/synbiotics would be superior to placebo yet would result in greater improvement of some metabolic indices-possibly via microbiota and/or inflammatory as well as gut barrier related pathways as assessed by biochemical parameter alterations-with very few adverse effects. We included studies in which clinically healthy people including those with excess body weight, those who are overweight, and those who are obese.

Search Strategy and Inclusion Criteria
Two independent authors (K.S.Z. and K.B.) searched PubMed/MEDLINE/Embase from database inception until 1 August 2019 for randomized controlled trials (RCTs) comparing probiotics and synbiotics with placebo/no-intervention/physical activity/diet to counteract cardiometabolic malfunctions in healthy people with normal weight or moderate/high-risk obesity (i.e., not exceeding 40 kg/m 2 ).
A manual review of reference lists from the most recent reviews followed the electronic search. Inclusion criteria were (i) full-text randomized controlled trial, (ii) populations containing >20 adult (>18 years old participants, excluding pregnant women), (iii) treatment with pro-/synbiotics for at least 4 weeks, (iv) randomization to probiotic/synbiotic vs. controls (placebo, no intervention, physical activity, and dietary elements, e.g., yoghurts and milk), and (v) available meta-analyzable change score/endpoint data on any of the following outcomes: HbA1C OR glucose OR OGTT OR insulin OR weight OR waist circumference OR BMI OR cholesterol OR LDL OR HDL OR triglycerides OR blood pressure OR SBP OR DBP OR uric acid OR Plasminogen activator inhibitor-1 OR fibrinogen OR any outcome related to inflammation/endothelial dysfunction. The exclusion criteria were as follows: (i) intervention with microbial agent and adjunctive medication aiming or known to prevent or counteract metabolic dysregulation, e.g., metformin, and (ii) disease, excluding morbid and super obese persons. Data from more than 2-arm studies were abstracted separately for particular comparators; however, placebos were preferentially selected, and regarding dietary comparators, products contained no lactic acid bacteria (e.g., milk vs. yoghurt).

Data Abstraction
We used the standard data extraction sheet according to our previous studies [23][24][25]. Due to a high number of studies included into metaanalysis, the abstraction stage was done by 4 independent authors. The study list was divided into two parts, and each was abstracted by 2 authors (the 1st part by K.S.Z. and K.B. and the 2nd part by D.M. and J.Ś.-D.). We abstracted data on the study design, the persons enrolled, and the probiotic intervention characteristics in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). For evaluation of the risk of bias (ROB) [26], we incorporated The Cochrane Collaboration's tool and reported the number of low-risk assessments [26]. This was done by one investigator (D.M.). If some data were missing or difficult to abstract (e.g., from figures) for the review, authors were contacted via email twice, one week apart. All inconsistencies were resolved by senior author (W.M. and I.Ł.) consensus. Data from figures was extracted by means of WebPlotDigitizer software (https://automeris.io/WebPlotDigitizer/).

Data Synthesis and Statistical Analysis
We conducted a random-effects [27] meta-analysis of outcomes for which ≥3 studies contributed data, using Comprehensive Meta-Analysis V3 (http://www.meta-analysis.com). We explored study heterogeneity using the chi-square test of homogeneity, with p < 0.05 indicating significant heterogeneity. All analyses were two-tailed with alpha = 0.05.
Group differences in continuous outcomes were analysed as the pooled standardized mean difference (SMD) in either endpoint scores (preferred) or change scores from endpoint to baseline (if endpoint scores were not available) using observed cases (OC). For continuous metabolic outcomes, standardized mean difference (SMD) and, where applicable, differences in means (DM) were calculated. The additional analyses included studies with participants with proper BMI value (20-25 kg/m 2 ) and trials including overweight and obese persons (BMI > 25 kg/m 2 , not exceeding 45 kg/m 2 ) To understand the relationship between effect sizes and various study-level predictors, we fit random-effect meta-regression (multiple) models without interaction term using DerSimonian-Laird estimator estimation of the amount of heterogeneity. The test statistics of the individual coefficient (and confidence intervals) for predictors were based on standard normal distribution (z), and the overall test was based on the chi-square distribution (Q statistics following the chi-square distribution with degrees of freedom representing the number of predictors). Meta-regression variables included (i) number of low ROB assessments, (ii) study duration, (iii) mono-vs. multi-strain probiotic intervention, (iv) sample size (analysed persons), and (v) age of participants (mean). Finally, we evaluated funnel plots and conducted Egger's regression test [28] to detect whether publication bias could have influenced the results we obtained.

Search Results
The initial search yielded 2813 hits. Almost 97% (n = 2727) of screened studies were excluded, being duplicates and/or after evaluation on the title/abstract level. Two (n = 2) additional articles were identified via hand search. After exclusion of duplicates between the initial search and hand search results, 88 (n = 88) full-text articles were reviewed. Of those, a total of 27 (n = 27) papers were excluded due to not fitting the inclusion criteria. The primary reasons for exclusion were wrong study aim (n = 10); non-healthy participants (n = 7); no probiotic treatment (n = 5); too few participants (n = 3); too short a study duration (n = 2); unavailability of full texts (n = 2); and another language other than English, German, and Polish (n = 1), yielding 61 (n = 61) studies that were included in the meta-analysis ( Figure 1).
All studies included healthy subjects (including overweight and obese but excluding morbidly obese persons), with a total of 6820 subjected to randomization and 5422 subjected to analysis. The overall mean age was 44.26 ± 12.87 (range: 21.43-71.9) years. The majority of studied persons were females (n = 2934, 57.22%). Baseline metabolic parameters of included persons are presented in Table S1, and the smoking status and diet along with physical activity are in Table S2. When analysing discontinuation events being consequences of adverse events, we found that the probiotic intervention was linked to very few adverse effects, the majority of which were of gastrointestinal origin. Apart from the most common bowel discomforts, i.e., nausea, diarrhea, constipation, and flatulence, there were also cardiac-related events, dental infections, chest tightness, sleep dysregulation, as well as hives. The details on are presented in Supplementary Table S3). DB-double blind, SB-single blind, TB-triple blind, N-no, Y-yes, NA-not applicable, CFU-colony-forming units, ROB-risk of bias, SD-standard deviation, URTI-upper respiratory tract infection, CIDs-common infectious diseases, LDL-low-density lipoprotein, TRAP-total reactive antioxidant potential, EPA-eicosapentaenoic acid, DHA-decosahexaenoic acid, ND-not determined, PBO-placebo, PRO-probiotic, FOS-fructooligosaccharides.

Risk of Bias Assessment
As evaluated by means of a ROB assessment tool, the mean number of low risks of bias assessment was 3 (median 2.5). The highest score, i.e., 7 low ROB assessments was detected in only one study [35] and 6 low ROB assessments were detected in two studies only [73,82]. Additionally, while analysing the papers, we detected a number of unclear risks of bias. The exact ROB evaluation in particular domains is in Table S4.

Effects on Metabolic Indices
Out of all the metabolic indices that we evaluated, BMI and waist circumference decrease were significantly lower with the probiotic compared to controls.  (Figures 2-7). In one case, Egger's test did indicate publication bias (DM for BMI: t value = 2.37, p = 0.02). For complete results, see Figures S1-S6.

Effects on Metabolic Indices Regarding Obesity Status
When conducting analysis by BMI status, i.e., in persons with BMI within normal (BMI: 20-25 kg/m 2 ) and abnormal (BMI: >25 kg/m 2 ) range, we were able to demonstrate that probiotic intake significantly affected total cholesterol (endpoint analyses) in persons with normal BMI value (SMD: −0.974; 95% CI: −1.661 to −0.286, p = 0.006). However, Egger's test did indicate publication bias (SMD for total cholesterol (endpoint data): t value = 5.38, p = 0.000006). On the other hand, the analyses on the same parameter but regarding change scores depicted that probiotics significantly lowered the parameter in persons with abnormal BMI only (SMD: −0.206, 95% CI: −0.395 to −0.018, p = 0.032). In this case, no publication bias was detected (SMD for total cholesterol (change sores): t value = 1.64, p = 0.137). At last, we evaluated that WC (change score) was significant also in persons with abnormal BMI (SMD: −0.178, 95% CI: −0.354 to −0.001, p = 0.049). In this case, no publication bias was detected (SMD for total cholesterol (change sores): t value = 1.29, p = 0.265).

Discussion
In past years, many studies revealed that probiotics and synbiotics, through interactions with hosts, could affect nutrient metabolism and energy balance. Our current meta-analysis of 61 clinical trials and 5422 persons exclusively investigated the impact of probiotic and synbiotic interventions to reduce cardiovascular risk factors in otherwise healthy adults. The only factor we decided not to exclude was overweight and obesity, as their prevalence is worldwide and as they impact human's health [91]. Morbidly obese persons (BMI ≥ 45 kg/m 2 ) were excluded from the present analysis. We also decided to exclude studies with adjunct medications with reported efficacy against metabolic dysregulation (e.g., metformin [92]). Similarly, we excluded patients with diagnosed diseases, as meta-analyses in such patients have already been published [93][94][95] The results of the present meta-analysis indicated that probiotics may reduce the BMI by 0.5 unit (provide stats) and decrease waist circumference by more than 1.5 cm (stats). The effect sizes were/were not affected by meta-regression statistics. The up-to-date published data indicate that probiotics may reduce body weight, BMI, and other anthropometric indices, e.g., fat mass and waist circumference, via several mechanisms. While restoring the microecological ecosystem, probiotics diminish the inflammation responsible for insulin sensibility in the hypothalamus [96]. This in turn, together with increased concentration of glucagon-like peptide-1 (GLP-1) as well as peptide YY (PYY), improve satiety and suppress appetite by delaying gastric emptying. It should be emphasized that gut-derived GLP-1 is able to attenuate gut motility and to facilitate the aggregation of the constitutive flora to ferment more polysaccharides [97]. Furthermore, healthy microbiomes within the gut upregulate the expression of fasting-induced adipocyte factor (FIAF) and thus limits the degradation of lipoproteins and the deposition of free fatty acids in adipose tissue. Together with reduced food intake, the abovementioned healthy microbiome can promote reduction of body weight [96,98]. The systematic review by Crovesy et al. [96] indicated that strains of Lactobacillus gasseri and Lactobacillus amylovorus may promote decrease of body weight in the overweight population. The meta-analysis by John et al. [97] confirmed that probiotic therapy was associated with a significant reduction of BMI and, thus, body weight and fat mass. The study group consisted of overweight and obese persons. Notwithstanding, another systematic review and meta-analysis in a similar group of subjects showed that administration of probiotics was related to reduction of body weight in comparison to the placebo; however, the effect sizes were small (weighted mean difference (95% confidence interval); −0.60 (−1.19, −0.01) kg, I 2 = 49%), BMI (−0.27 (−0.45, −0.08) kg m −2 , I 2 = 57%) and fat percentage (−0.60 (1.20, −0.01) %, I 2 = 19%). Similarly to our findings, the effect of probiotics on fat mass was not significant (−0.42 (−1.08, 0.23) kg, I 2 = 84%) [99]. Also, a study by Depommier et al. [100] demonstrated that supplementation with Akkermansia Muciniphila in overweight and obese human volunteers improved insulin sensitivity and total plasma cholesterol with a small reduction of body mass compared to controls. In contrast, in healthy, but overweight subjects, the administration of Lactobacillus amylovorus and Lactobacillus fermentum strains reduced this body fat mass [101].
The current meta-analysis did not confirm the efficacy of probiotics administration in reduction of other cardiovascular risk in healthy people. Of note, carbohydrate and lipid metabolism was not significantly affected by this type of intervention. In contrary to diabetic patients, we did not find any effect of probiotic therapy on carbohydrate metabolism. A study by Raygan et al. [102] which was conducted in patients with type 2 diabetes mellitus (T2DM) and coronary heart disease found that the intervention, during which the strains of Bifidobacterium bifidum, Lactobacillus casei, and Lactobacillus acidophilus were ingested for 12 weeks, significantly decreased the plasma glucose and insulin resistance. In a meta-analysis by Samah et al., [103] moderately hypoglicaemic properties (lower levels of fasting blood glucose) of microbial agents were confirmed. As in previously quoted studies, the meta-analysis cohort coincided with T2DM patients. Probiotics were demonstrated to affect glucose metabolism via several mechanisms, including antioxidant activity, and thus diminished gut-barrier integrity disruption, enhanced NK cells activity in the liver cells, and diminished insulin resistance by modulating the expression of proinflammatory cytokines and NF-kB-binding activity.
Indeed, eubiosis within the gut may serve as a protective point for the preDM and DM onsets, diminishing low-grade inflammation which characterizes all metabolic diseases [104,105]. As concerns inflammation status, we did not find the relationship between common inflammatory markers (CRP and leukocytes count) as well as other indices associated with insulin resistance, including endothelial markers and uric acid. In T2DM patients, probiotics were found to lower the concentrations of hs-CRP, IL-6, and TNF-α [106]. Similar results, regarding hs-CRP, were demonstrated lately in a meta-analysis by Zheng et al. [107] and by Tabrizi et al. [108]. At last, the increase of the bioavailability of gliclazide regulating the intestinal absorption of glucose may also play a role [93].
In our study, we found that probiotics can decrease the total cholesterol level in persons with increased BMI, but other lipid parameters were not affected by probiotics and synbiotics administration. In Wang et al.'s meta-analysis including 32 randomized controlled trials (1971 participants with various metabolic entities), it was proved that probiotics significantly reduced serum total cholesterol (MD = −13.27, 95% CI (−16.74-9.80), p < 0.05) in comparison to controls [109]. Similar results were obtained in the meta-analyses by Chao et al. [110] and Shimizu et al. [111] (30 RCTs and 33 RCTs, respectively; hypocholesterolemic effects of probiotics-mean net change of total cholesterol: 7.8 mg/dL and 6.57 mg/dL, respectively, both in persons with mild lipid malfunctions). There are many hypotheses regarding mechanisms in which probiotics may lower the cholesterol level, such as binding of cholesterol to the probiotic cellular surface and incorporation of cholesterol molecules into the probiotic cellular membrane. However, the deconjugation of bile via bile salt hydrolase (BSH) activity seems to be the most profound mechanism in which probiotics reduce cholesterol level [112]. Bile salt hydrolase is the enzyme that catalyses the hydrolysis of glycine-and/or taurine-conjugated bile salts into amino acids residues and free bile acids. The most BSH-active probiotics belong to the genera of Lactobacillus, Lactococcus, and Bifidobacterium. These probiotics increase the production of bile salts from cholesterol in their colonized area and, as a consequence, contribute to reduced risk of coronary heart diseases [112].
The administration of probiotics improved blood pressure in humans, which was confirmed in Khalesi et al.'s meta-analysis including 9 randomized, controlled trials [113]. The consumption of probiotics significantly decreased systolic blood pressure by 3.56 mmHg and diastolic blood pressure by 2.38 mmHg in comparison to control groups (the duration of intervention is ≥8 weeks or daily dose > 10 11 CFU). In contrast to our study, the authors included studies evaluating people with metabolic syndrome, hypertension, and hypercholesterolemia. As the menopause period is a strong contributor of CVD [114], we looked for metabolic effects on probiotic intake in this particular subgroup of participants. We were able to demonstrate that probiotic intake decreased the vascular stiffness in obese postmenopausal women [80]. Also, as reported by Lambert et al. [62], probiotics significantly diminished vasomotor symptoms of menopause. In a study by Szulińska et al. [81] was found that probiotics administration favorably affected the risk factors in a dose-dependent manner, showing beneficial effects on the cardiometabolic parameters and gut permeability of obese postmenopausal women. However, Brahe et al. [36] did not record that metabolic index was affected by microbial agent administration. Only these three studies reported on metabolic effects in the perimenopausal period; thus, we did not conduct a subgroup analysis. More studies are needed to clarify if and how probiotics can affect CVD risk in women at the menopause period.
Last but not least, we abstracted data related to the influence of probiotic administration on gut microbiota and immunological markers. The most frequently studied variables were (i) the effects of probiotic administration on the composition of the microbiota and (ii) colonization with probiotics. Among microbial metabolites, mostly faecal SCFAs were analyzed. The authors analyzed also markers of gut-barrier integrity-mostly LPS and different cytokines as well as inflammatory markers. CRP measured in few studies can be considered as an inflammatory marker as well as a gut integrity marker. Based on the results obtained, no definite association can be found between the use of probiotics, microbiota changes, modulation of the immune system, and either presence or lack of clinical effects ( Table 2 and Table S6). Of note, the results cannot be subjected to meta-analysis due to very diverse methods used to analyze the microbiota. Therefore, the results are difficult to compare.
For this reason, in order to fully assess the causal relationship between the microbiota and the function of the immune system and gut-integrity markers with relation to cardiovascular risk prevention, a multifactorial analysis should be performed, which was not performed in the works described in this systematic review. In only one study, the correlation between microbiota changes and cardiovascular risk factors was demonstrated [48]; however, in this study, no preventive outcome of probiotics administration was observed. In addition, the results of metabolomic studies did not contribute to elucidation of the mechanism of action of probiotics studied. Therefore, it cannot be determined whether the effect of probiotics in cardiovascular risk prevention is related to their effect on microbiota or the immune system or gut-barrier function. The relationship observed in some studies is rather based on association and not causation. We conclude that mechanistic studies should be an important point in analysis of probiotics/synbiotics efficacy.

Limitations
Several limitations of this meta-analysis need to be underlined. These include (i) a relatively small number of high-quality double-blinded studies comparing probiotic intervention to controls with a wide range within the number of participants preceded by no sample size calculations; (ii) heterogeneous study inclusion criteria (various age, profession of participants, and dietary and physical activity add-on interventions), and (iii) various type of strains and duration of probiotic intervention. In studies incorporated into the present meta-analysis, the association between the probiotic effect in relation to supplement dose and treatment duration was not analyzed. At last, most of the studies were financed by the industry and include products combined with different ingredients. These all are confounding factors for probiotic efficacy, which may have resulted in some publication bias as evaluated by Eagerr's test and funnel plots [115]. Consequently, in order to draw some evidence-based conclusions and to give some guidelines regarding probiotic intake in healthy adults, strict inclusion criteria and homogenous intervention protocols are needed. Lastly, during meta-analysis, we did not use intent-to-treat data but adopted per-protocol evaluation as the majority of studies reported on that. We could have introduced potential bias during the review process and could have missed studies not clearly aimed at reducing cardiovascular risk but possibly reporting such outcomes.

Conclusions
Probiotics may counteract some CMRF (e.g., BMI and waist circumference) in clinically healthy participants. Overweight/obese persons might benefit from the reduction of total cholesterol serum concentration. Poor quality of probiotic-related trials make systematic reviews and meta-analyses difficult to conduct and draw exact conclusions. "Gold standard" methodology in probiotic studies awaits further development.
Supplementary Materials: The following are available online at http://www.mdpi.com/2077-0383/9/6/1788/s1, Table S1. Baseline metabolic parameters in studies persons, Table S2. Smoking status and physical activity, Table S3. Trial discontinuations and adverse effects, Table S4. Risk of bias, Table S5. Effect sizes regarding metabolic outcomes, Table S6. Summary of the preventive outcome and changes in microbial composition and metabolites as well as anti-inflammatory effects and gut-barrier markers associated with probiotics administration; Figure S1. Funnel plot for endpoint BMI (SMD) in the present meta-analysis, Figure S2. Funnel plot for endpoint BMI (DM) in the present meta-analysis, Figure S3. Funnel plot for endpoint WC (SMD) in the present meta-analysis, Figure S4. Funnel plot for endpoint WC (DM) in the present meta-analysis, Figure S5. Funnel plot for WC change scores (SMD) in the present meta-analysis, Figure S6. Funnel plot for WC change scores (DM) in the present meta-analysis.