Introduction

Type 2 diabetes mellitus (T2DM), an endocrine and metabolic disease, is influenced by host physiology and environmental factors [1]. More than 500 million people are living with diabetes globally, and this number is expected to increase to 783 million by 2045 [2]. T2DM is a common disease that accounts for approximately 90% of all cases of diabetes [3], and it may cause reduced life expectancy and life-threatening and costly complications [4]. There is no radical cure for T2DM [5, 6], and its treatment relies on the long-term use of anti-diabetic drugs [7, 8]. Therefore, it is crucial to explore new methods that may effectively delay or even reverse the progression of T2DM.

Recent studies have shown that the gut microbiota plays a key role in the maintenance of host homeostasis and pathogenesis of T2DM [9, 10]. Probiotics are microbial dietary supplements that alter the gut microbiota. Some randomised controlled trials (RCTs) have investigated the effects of probiotic interventions on glycaemic control in T2DM patients. However, evidence from clinical trials on the effects of probiotic supplementation on glycaemic control remains inconsistent. Asemi et al. [11] conducted a randomised double-blind placebo-controlled clinical trial involving 54 T2DM patients, which revealed that multi-species probiotic (mixture of Lactobacillus and Bifidobacterium) supplementation prevented an increase in the fasting blood glucose (FBG) level from baseline in these patients. Meanwhile, Razmpoosh et al. [12] randomly assigned 60 T2DM patients into two groups to take either a probiotic (mixture of Lactobacillus and Bifidobacterium) or a placebo intervention, and the results showed no significant differences in insulin or insulin resistance levels between the two groups. In 2016, Li et al. performed a systematic review and meta-analysis of 12 RCTs with 714 individuals and reported that probiotic supplementation could alleviate FBG, but no significant differences were observed in the haemoglobin A1c (HbA1c) level or homeostatic model assessment of insulin resistance (HOMA-IR) score between the probiotic and control groups of T2DM patients [13]. In 2020, Tao et al. systematically summarised 15 RCTs with 902 individuals, and the results of the meta-analysis indicated that probiotic supplementation reduced HbA1c, FBG and insulin resistance levels in T2DM patients [14]. However, some related RCTs (n = 11, including 630 patients) were not included in their study. Since then, more RCTs (n = 6) of the effects of probiotic supplementation on glycaemic control, including a total of 511 T2DM patients, have been reported [15, 16]. Controversy still exists regarding the effects of probiotics on glycaemic control in T2DM patients. Variations in participant (e.g. race) and intervention characteristics (e.g. dose, probiotic genus, and duration) in different studies may have given rise to the contradictory results. No study has detected differences in the effects of probiotic supplementation on glycaemic control according to the participant and intervention characteristics.

In this systematic review and meta-analysis, we aimed to evaluate the effects of a probiotic intervention on glycaemic control in T2DM patients and to evaluate the variations in these effects due to participant characteristics, e.g. race and baseline body mass index (BMI), and intervention characteristics, e.g. the probiotic dose, the duration of the intervention, the probiotic genus, and the type of vehicle used to deliver the probiotics.

Methods

This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement [17] (Additional file 1: Table S1). The protocol for this study has been registered at the International Prospective Register of Systematic Reviews (registration number: CRD42022370226).

Search strategy

Two reviewers (Guang Li and Yan-Jun Deng) independently searched PubMed, Web of Science, Embase, and Cochrane Library databases from their inception until October 2022 using various probiotic-related words and Medical Subject Heading terms in combination with ‘T2DM’ (Additional file 2: Table S2). No language or other restrictions were applied during the search, and all relevant studies were found to be published in English. A manual search was also performed to identify relevant studies from the references of the included studies.

Inclusion and exclusion criteria

Studies were included in the analysis if: (1) the participants were T2DM patients aged ≥ 18 years; (2) the study design was an RCT; (3) the intervention was the intake of probiotics from supplements and/or food; (4) the control group received a placebo intervention; and (5) the main outcomes included the glycaemic profile, e.g. FBG, insulin, and HbA1c levels and the HOMA-IR score. Studies were excluded from the analysis if: (1) the participants had other types of diabetes, e.g. gestational diabetes or type 1 diabetes or (2) the participants were concurrently receiving other interventions, e.g. synbiotics, herbs, prebiotics, or micro- nutrients.

Data extraction and quality assessment

Two researchers (Guang Li and Yan-Jun Deng) independently performed the literature search and data extraction, and disagreements were resolved by a third senior researcher (Su-Mei Xiao). Basic information (e.g. first author, year, and country of the study and the age, sex, and BMI of the participants), the study design, intervention information (probiotic genus and dose and duration of the intervention), and outcomes were extracted from the included studies. Two reviewers (Xiao-Bao Wang and Qiong Zhang) evaluated the quality of the included studies using the Cochrane risk-of-bias assessment tool. The risk of bias in the included studies was classified as low, unclear, or high.

Data synthesis and statistical analysis

The change in glycaemic control parameters was the primary outcome in this study. It was calculated as the final measurement value minus the baseline measurement value in each group. The mean and standard deviation (SD) of the change in glycaemic control parameters for the control group and the intervention group were extracted, respectively. If the study provided the standard error (SE) of mean change, the SE was converted to SD based on the sample size. For studies that did not directly report SD of mean change, the SDs of the baseline and final measurement values and the correlation coefficient (Corr) were used to calculated SDEffect,change (SDE,change) according to the following formula [18]:

$${\mathrm{SD}}_{\mathrm{E},\mathrm{change}}=\sqrt{{SD}_{E,baseline}^{2}+{SD}_{E,final}^{2}-(2 * Corr * {SD}_{E,baseline} * {SD}_{E, final})}$$

Corr is the correlation coefficient between the baseline and final measurement values. For the pretest–posttest design, presumably the correlation is at least 0.5. This was the Corr estimate value being used to impute the missing SDs of mean change in this study [18, 19]. If the study presented data in medians and quartiles, the mean and SD values were estimated [20, 21]. If the intervention included multiple time points, the longest intervention time was included in the analysis.

The standardised mean difference (SMD) with the 95% confidence interval (CI) was used to assess the effects of probiotic interventions on glycaemic control in T2DM patients. The boundary values of the SMD were set at 0.2, 0.5, and 0.8, corresponding to small, medium, and large effects, respectively [22]. Heterogeneity was assessed using Cochrane’s Q statistic (chi-square). The inverse variance (I2) was used to assess the size of the heterogeneity. A fixed-effects model was used for the meta-analysis when I2 ≤ 50%, and a random-effects model was used when I2 ≥ 50%. Subgroup analysis was used to explore the possible sources of heterogeneity. Subgroup analyses were performed for race (Asian vs. Caucasian), probiotic dose (≤ 1 × 1010 colony-forming units (CFU)/day vs. > 1 × 1010 CFU/day), the duration of the intervention (≤ 8 weeks vs. > 8 weeks), probiotic genus (Lactobacillus, Bifidobacterium, or Lactobacillus and Bifidobacterium), type of vehicle used to deliver the probiotics (food vs. non-food (powder/capsule/tablet), and baseline BMI (< 30 kg/m2 vs. ≥ 30 kg/m2). The leave-one-out approach was used in the sensitivity analysis. Funnel plots and Egger’s test were used to appraise the possible publication bias in this study.

Results

Study characteristics

The database search yielded 4,048 records, and one additional record (a conference paper [23]) was obtained from the manual search of the references of the included RCTs. A total of 1,125 records were then excluded due to duplication, leaving 2,924 articles for screening. After the screening based on the titles and abstracts, 2,821 articles were further excluded (e.g. reviews, protocols, animal studies, etc.). The full texts of the remaining 103 potentially relevant studies were assessed according to the inclusion and exclusion criteria. Finally, thirty RCTs were included in this systematic review and meta-analysis (Fig. 1).

Fig. 1
figure 1

PRISMA flowchart for search strategy and study selection process. RCT, randomised controlled trial; T2DM, type 2 diabetes mellitus; PRISMA, preferred reporting items for systematic reviews and meta-analyses; FBG, fasting blood glucose; HbA1c, haemoglobin A1c; HOMA-IR, homeostasis model of assessment of insulin resistance

Full size image

For the included 30 RCTs, all of them reported FBG, 17 RCTs reported HOMA-IR, 17 RCTs reported insulin, and 23 RCTs reported HbA1C (Fig. 1). Table 1 shows the basic information for the included 30 studies. Nine studies were conducted in Asian patients (three in China [15, 16, 24] and one each in India [25], Indonesia [26], Thailand [27], Japan [28], Malaysia [29], and Korea [30]), 19 studies were conducted in Caucasian patients (12 in Iran [11, 12, 31,32,33,34,35,36,37,38,39,40] and one each in Ukraine [41], Turkey [23], Sweden [42], Saudi Arabia40 [43], Egypt [44], Denmark [45], and Australia [46]) and two studies were conducted in other races (two in Brazil [47, 48]). In the 30 RCTs, there were a total of 1,827 subjects, with 922 in the probiotic group and 905 in the control group. The dose of probiotics used in the 30 studies ranged from 2 × 107 to 1 × 1012 CFU/day, the duration of the probiotic interventions ranged from 4 to 36 weeks, and the baseline BMI ranged from 23.1 to 35.9 kg/m2. The probiotics were consumed as food (n = 13) or non-food (powder/capsule/tablet; n = 15) forms, and the probiotic genera were mainly Lactobacillus (n = 11), Bifidobacterium (n = 2), and Lactobacillus and Bifidobacterium (n = 14; Table 1).

Table 1 Characteristics of the included studies (n = 30)
Full size table

Risk of bias assessment of the included RCTs

The Cochrane risk-of-bias assessment tool was used to assess the bias of the 30 included studies. Approximately half of the studies (53%) were randomised, but 14 studies did not clearly report the randomisation process. The methods of allocation concealment were described in 43% of the included RCTs, and the majority of the studies (87%) described the blinding method. Approximately 40% of the studies provided information about the blinding outcome assessment. Most of the included studies had a low risk of attrition bias (73%), a low risk of reporting bias (93%), and a low risk of other types of bias (70%). Overall, four of the studies were classified as high quality (all terms were assessed as low risk), 19 studies were classified as moderate quality (no term was assessed as a high risk and one or more terms were assessed as unclear risks), and seven studies were classified as low quality (one or more terms were assessed as a high risk). The general and individual risks of bias are shown in Additional file 3: Fig. S1.

Effects of probiotic supplementation on glycaemic control

Effects on FBG

Thirty studies including a total of 1,827 T2DM patients were used to evaluate the effects of probiotic supplementation on FBG level. The pooled effects of probiotic supplementation indicated a significant decrease in FBG level in the probiotic group (SMD = − 0.331, 95% CI  − 0.424 to − 0.238, Peffect < 0.001), and the heterogeneity was low (I2 = 29%, Pheterogeneity = 0.070; Fig. 2a). Leave-one-out sensitivity analysis confirmed that the pooled effects of probiotic supplementation on FBG level were stable and reliable (Additional file 4: Fig. S2a).

Fig. 2
figure 2figure 2

Forest plots of the effects of probiotics on a FBG, b Insulin, c HBA1c and d HOMR-IR. FBG, Fating blood glucose; HbA1c, Haemoglobin A1c; HOMA-IR, Homeostsis model of assessment of insulin resistance

Full size image

Subgroup analyses for FBG were performed according to race, probiotic intervention dose, probiotics genus, type of vehicle used to deliver the probiotics, and baseline BMI. As shown in Table 2, the significant subgroup differences (Psubgroup < 0.050) were observed for races (Asian vs. Caucasian), genus of probiotics (Lactobacillus vs. Bifidobacterium vs. Lactobacillus and Bifidobacterium), and baseline BMI (< 30 kg/m2 vs. ≥ 30 kg/m2). A stronger beneficial effect of the probiotic intervention was observed on FBG level in the Caucasian subgroup (SMD = − 0.448, 95% CI  − 0.575 to − 0.322, Peffect < 0.001, Psubgroup = 0.020), in the Bifidobacterium subgroup (SMD = − 0.626, 95% CI  − 1.221 to − 0.030, Peffect = 0.039, Psubgroup = 0.040), and in the high-baseline-BMI (≥ 30 kg/m2) subgroup (SMD = -0.490, 95% CI  − 0.644 to − 0.336, Peffect < 0.001, Psubgroup = 0.007). No differences were observed between the subgroups of probiotic dose, intervention duration, or type of vehicle used to deliver the probiotics (Table 2, Psubgroup > 0.050).

Table 2 Subgroup analysis for the effects of probiotics on FBG
Full size table

Effects on insulin

Eight hundred and eighty-six patients in 17 RCTs were included in the meta-analysis of the effects of probiotic intake on insulin level. Probiotic supplementation in T2DM patients led to a significant reduction in insulin level (SMD = − 0.185, 95% CI  − 0.313 to − 0.056, Peffect = 0.004) without heterogeneity (Fig. 2b, I2 = 0%, Pheterogeneity = 0.500). Sensitivity analysis also supported the robustness of the results for insulin level (Additional file 4: Fig. S2b).

As shown in Table 3, the magnitude of the reduction was significantly greater in the subgroup of patients taking food-type probiotics (SMD = − 0.386, 95% CI  − 0.592 to − 0.180, Peffect < 0.001, Psubgroup = 0.014) than in the subgroup taking non-food (powder/capsule/tablet) types. In addition, no differences were observed between the subgroups of races, probiotic dose, intervention duration, probiotic genus, or baseline BMI (Table 3, Psubgroup > 0.050).

Table 3 Subgroup analysis for the effects of probiotics on insulin
Full size table

Effects on HbA1c

The effects of probiotic interventions on HbA1c level were evaluated in 23 RCTs including 1,466 T2DM patients. A significant decrease was observed in the HbA1c level in the probiotic group (Fig. 2c, SMD = − 0.421, 95% CI  − 0.583 to − 0.258, Peffect < 0.001) with moderate heterogeneity (I2 = 56%, Pheterogeneity < 0.001). Sensitivity analysis showed that the results for HbA1 level were stable and reliable (Additional file 4: Fig. S2c).

The subgroup analysis was performed for HbA1c according to races (Asian vs. Caucasian), genera of probiotics (Lactobacillus vs. Bifidobacterium vs. Lactobacillus and Bifidobacterium), types of vehicle used to deliver the probiotics (food vs. non-food (powder/capsule/tablet)), and baseline BMI (< 30 kg/m2 vs. ≥ 30 kg/m2). As shown in Table 4, a significantly greater reduction was observed in the HbA1c level in the subgroups of Caucasians (SMD = − 0.465, 95% CI  − 0.672 to − 0.257, Peffect < 0.001, Psubgroup = 0.032), Bifidobacterium probiotics (SMD = − 0.913, 95% CI  − 1.387 to − 0.438, Peffect < 0.001, Psubgroup = 0.001), food-type probiotics (SMD = − 0.524, 95% CI  − 0.800 to − 0.249, Peffect < 0.001, Psubgroup = 0.047), and baseline BMI ≥ 30 kg/m2 (SMD = − 0.485, 95% CI  − 0.783 to − 0.188, Peffect = 0.001, Psubgroup = 0.018). No differences were observed between the subgroups of probiotic dose or intervention duration (Psubgroup > 0.050).

Table 4 Subgroup analysis for the effects of probiotics on HbA1c
Full size table

Effects on the HOMA-IR score

The results of the meta-analysis of 17 RCTs (n = 1,116) suggested significant effects of probiotic interventions on reducing the HOMA-IR scores in T2DM patients (SMD = − 0.224, 95% CI  − 0.342 to − 0.105, Peffect < 0.001). The heterogeneity (I2 = 41%, Pheterogeneity = 0.040) of these RCTs was moderate (Fig. 2b). Sensitivity analysis showed that the pooled effects of probiotic supplementation on HOMA-IR scores did not significantly change, suggesting that the meta-analysis results were stable and reliable (Additional file 4: Fig. S2b).

No statistically significant differences were observed in the HOMA-IR score between subgroups (Table 5, Psubgroup > 0.050). However, an effective reduction in the HOMA-IR score was observed in the subgroups of Caucasians (SMD = − 0.308, 95% CI  − 0.471 to − 0.146, Peffect < 0.001, Psubgroup = 0.173), high baseline BMI (≥ 30 kg/m2; SMD = − 0.320, 95% CI  − 0.615 to − 0.026, Peffect = 0.033, Psubgroup = 0.144), and Bifidobacterium probiotics (SMD = − 0.248, 95% CI  − 0.387 to − 0.109, Peffect = 0.004, Psubgroup = 0.345).

Table 5 Subgroup analysis for the effects of probiotics on HOMA-IR
Full size table

Publication bias analysis

Potential publication bias was assessed using funnel plots and Egger’s test. A visual inspection of the funnel plots revealed no publication bias for FBG, insulin, or HbA1c levels or the HOMA-IR score (Additional file 5: Fig. S3). Egger’s test results showed no publication bias for FBG (P = 0.349), insulin (P = 0.260) or HbA1c (P = 0.108) levels or the HOMA-IR score (P = 0.391).

Discussion

This systematic review and meta-analysis summarised data from 30 RCTs, including a total of 1,827 individuals, to evaluate the effects of probiotic supplementation on glycaemic control in T2DM patients. The results revealed that probiotic supplementation significantly decreased FBG, insulin, and HbA1c levels and HOMA-IR scores in T2DM patients. Further subgroup analyses showed that the effect was larger in the subgroups of Caucasians, high baseline BMI (≥ 30.0 kg/m2), Bifidobacterium probiotics, and food-type probiotics.

This study supported the notion that probiotics improve glycaemic control in T2DM patients. This is inconsistent with the results reported by the systematic review and meta-analysis of 12 RCTs in 2016 [13]. They found no significant differences in the HbA1c level and HOMA-IR score between the probiotic and control groups of T2DM patients. For their study, the meta-analysis of HbA1c and HOMA-IR were conducted with limited number of RCTs (n = 6), and five of them had the participants’ baseline BMI less than 30 kg/m2. In this study, the subgroup analysis found that the effect was larger in individuals with higher baseline BMI (≥ 30.0 kg/m2). These may partially explained the differences between the two studies. The gut microbiota is largely involved in the metabolic, nutritional, physiological, and immune functions of the host [49,50,51]. A previous study showed that T2DM patients are characterised by a decrease in the abundance of certain butyrate-producing bacteria and the enrichment of other microbial functions conferring sulphate reduction and oxidative stress resistance [52]. Changes in the gut microbial composition may be a mechanism whereby probiotic supplementation improves glycaemic control. Probiotic supplementation may modulate and increase the abundance of intestinal flora that are beneficial to glycaemic control [53, 54]. Moreover, the gut microbiota may regulate glucagon-like peptide 1, which promotes the secretion of insulin from islet β cells, and reduces the secretion of glucagon from islet α cells, resulting in a reduction in gastric emptying time, gastrointestinal peristalsis, and loss of appetite [55, 56]. Previous studies have found that probiotics may stimulate the production of short-chain fatty acids, especially butyrate, which increase insulin sensitivity and thus improve glycaemic control [57,58,59].

The subgroup analyses suggested that Bifidobacterium have greater effects than other probiotic genera. Probiotics that colonise the gut may change the host’s gut microbiota. According to a 5-year follow-up study, Bifidobacterium longum, a member of the core microbiota of the human gut, can stably colonise the gut [60]. Another study reported that oral supplementation with B. longum persists in the gut for 6 months in 30% of subjects [61]. Moreover, Xiao et al. (2020) found that Bifidobacterium appears to have a better ability to colonise the gut than Lactobacillus [62]. This may explain the finding that Bifidobacterium had a larger effect than other probiotic genera on glycaemic-control-related parameters (e.g. FBG and HbA1c levels) in T2DM patients, to some extent, in this study.

Food-type probiotics (e.g. yogurt and fermented milk) may have greater effects than other types of probiotics on glycaemic control in T2DM patients. Gastric acidity is thought to be one of the main obstacles to gut colonisation [63, 64]. Food-type probiotics (e.g. yogurt and fermented milk) may buffer the stomach acid, allowing the probiotics to better colonise the gut [65]. An in vitro study assessed the tolerance of probiotics in the human gastrointestinal tract by evaluating the effects of food addition on the viability of probiotics in simulated pH 2.0 gastric juices, revealing that adding soymilk or a liquid breakfast greatly enhanced the survival of the probiotics [66].

Compared to the baseline BMI < 30 kg/m2 subgroup, the stronger beneficial effects of a probiotic intervention were also observed on FBG and HbA1c levels in the baseline BMI ≥ 30.0 kg/m2 subgroup. This may be due to gut dysbiosis in obese individuals. In 2021, Liu et al. summarised the characteristics of the gut microbiota in obesity. Obese individuals were observed to have an increased Firmicutes/Bacteroidetes ratio at the phylum level and decreased abundances of the genera Lactobacillus and Bifidobacterium [67]. Probiotic supplementation may alleviate gut dysbiosis [68]. These findings indicate that obese individuals may be more sensitive to probiotic interventions. In addition, this may partly explain the observed racial differences, i.e. the effect was larger in Caucasians than in Asians. In this study, the average baseline BMI (30.3 kg/m2) was higher in Caucasians than in Asians (26.2 kg/m2).

In addition, no significant difference was observed between the longer-term intervention (> 8 weeks) and the shorter-term intervention (≤ 8 weeks) groups. In 2020, an RCT was conducted in 150 new-borns (38–40 weeks gestational age). In that study, the intervention group received probiotic supplementation containing 2 × 106 CFU/day of B. breve PB04 and L. rhamnosus KL53A. The stool samples from days 5, 6, and 30 were collected for an analysis of the gut microbiome. The results showed that L. rhamnosus and B. breve colonised rapidly, generally on days 5 and 6 [69]. This ability of the probiotics to rapidly colonise the gut may have resulted in the very small difference between the short and long intervention durations.

Furthermore, no significant differences were found between the higher-dose (> 1 × 1010 CFU/day) and lower-dose (≤ 1 × 1010 CFU/day) probiotic intervention groups. Several studies have reported similar results. Ibarra et al. (2018) performed a randomised double-blind, placebo-controlled trial to determine the effects of 4 weeks of supplementation with 1 × 109 or 1 × 1010 CFU of B. animalis subsp. lactis HN019 on adults diagnosed with functional constipation. The results showed no significant difference between the two groups with different doses of probiotics [70]. However, Whorwell et al. (2006) conducted a multi-centre clinical trial of 362 patients with irritable bowel syndrome (IBS) and found that 1 × 108 CFU of B. infantis 35,624 significantly alleviated the symptoms of IBS and that its effect was superior to that of the administration of 1 × 106 CFU/day and 1 × 1010 CFU/day of B. infantis 35624 [71]. In all of the included RCTs, the probiotic intervention doses were higher than 1 × 106 CFU/day, and only one RCT had a probiotic intervention dose lower than 1 × 108 CFU/day. Thus, these two doses were not used as the limits for subgroup analysis in this systematic review and meta-analysis. Further studies are warranted to determine the optimal dose of probiotics for glycaemic control in T2DM patients.

This study systematically and comprehensively evaluated the effects of probiotic supplementation on glycaemic control in T2DM patients. To the best of our knowledge, this is the first systematic review and meta-analysis study to investigate the differences in the effects of probiotic interventions on glycaemic control in T2DM patients according to participant characteristics (e.g. race, baseline BMI), and intervention characteristics, (e.g. probiotic doses, probiotic genus, treatment duration, and types of vehicles used to deliver the probiotics). However, this study also has some limitations. First, as 12 of the included studies (40%) were conducted in Iran, some racial and ethnic groups may be underrepresented. This may have resulted in a limited racial representation. Second, the number of RCTs in some subgroup analyses was low. For example, in the subgroup analysis of HbA1c level, the number of RCTs in the Bifidobacterium subgroup was only two. Third, the duration of most of the RCTs included in the analysis was from 4 to 24 weeks, and only one RCT was longer than 24 weeks (a 36-week intervention). Therefore, the long-term effects could not be explored in this study.

Conclusions

The findings of this study indicate that probiotic supplementation had favourable effects on glycaemic control in T2DM patients. Bifidobacterium and food-type probiotics had greater glucose-lowering effects than other probiotic genera and types of vehicle used to deliver the probiotics. Patients with a higher BMI may gain more glycaemic control benefits from a probiotic intervention. The administration of probiotics may be a promising adjuvant therapy for glycaemic control in T2DM patients.