diabetes research and clinical practice 173 (2021) 108689

Contents available at ScienceDirect

Diabetes Research
and Clinical Practice
journal homepage: www.elsevier.com/locat e/dia bre s

Differential gut microbiota composition between

type 2 diabetes mellitus patients and healthy
controls: A systematic review

Fatin Umirah, Chin Fen Neoh *, Kalavathy Ramasamy *, Siong Meng Lim
Collaborative Drug Discovery Research (CDDR) Group, Faculty of Pharmacy, Universiti Teknologi MARA (UiTM) Kampus Puncak
Alam, Cawangan Selangor, 42300 Bandar Puncak Alam, Selangor Darul Ehsan, Malaysia

A R T I C L E I N F O A B S T R A C T

Article history: Aims: This systematic review summarised the latest findings on differential composition
Received 15 April 2020 of gut microbiota in T2DM.
Received in revised form Methods: Literature search was performed using electronic databases. Relevant studies
24 December 2020 were identified, extracted and assessed for risk of bias. The primary outcome of this sys-
Accepted 25 January 2021 tematic review was the composition of gut microbiota in healthy controls and T2DM while
Available online 4 February 2021 the secondary outcomes included the correlation of gut microbiota with metabolic param-
eters.
Results: Thirteen case-control studies involving 575 T2DM and 840 healthy controls were
Keywords:
included. T2DM patients exhibited a marked increase in lactobacilli. Six studies found lac-
Dysbiosis
tobacilli to predominate the gut of T2DM patients; however, this could be confounded by
Metagenome
the types of antihyperglyacemic medications. Conversely, butyrate producers dominate
Microbiome
the gut of healthy controls. In T2DM patients, butyrate producers were surprisingly higher
Type 2 diabetes mellitus
in those taking metformin intake than those not taking the drug. Whilst lactobacilli were
found to be higher with increased plasma glucose, conflicting correlations were observed
between various genera and anthropometric measurements, dietary intake, lipid profiles
and inflammatory markers. There were moderate to strong significant positive correlations
between the class Clostridia and phylum Firmicutes with pro-inflammatory IFN-c as well as
between Negativicutes and IL-6.
Conclusions: Altogether, butyrate-producing bacteria are negatively correlated to glycaemic
parameters. Lactobacilli are higher in T2DM patients and Firmicutes is correlated with
inflammation.
Ó 2021 Elsevier B.V. All rights reserved.

1. Introduction tified which include age, diet, family history, sedentary life-
style and obesity [1]. Alteration of gut microbiota (i.e.
Type 2 diabetes mellitus (T2DM) is a complex, multifactorial dysbiosis) could play an important role in the development
disorder which can be attributed to either insulin insensitivity and progression of T2DM [15]. Opportunistic pathogens and
or insulin deficiency [13]. Various risk factors have been iden- endotoxin-producing bacteria were found to increase

* Corresponding authors.
E-mail addresses: [email protected] (C.F. Neoh), [email protected] (K. Ramasamy).
https://doi.org/10.1016/j.diabres.2021.108689
0168-8227/Ó 2021 Elsevier B.V. All rights reserved.
2 diabetes research and clinical practice 173 (2021) 108689

whereas SCFA-producing bacteria decreased in T2DM when surements. Search strategy and data extraction of all poten-
compared to healthy controls [1,12]. tial studies were independently performed by two authors
It is increasingly being recognised that frequent intake of (FU and CFN). Discrepancies on study inclusion were resolved
high-fat and/ or high-fructose diet causes dysbiosis of gut through discussion and consensus between all authors.
microbiota [16]. Dysbiosis disrupts intestinal tight junction
proteins and increases gut permeability [16]. This allows 2.3. Assessment of risk of bias
Gram negative bacteria-derived lipopolysaccharide (LPS) to
transfuse into blood stream through the leaky gut, and trigger Two authors (FU and CFN) had independently assessed the
the onset as well as progression of low-grade inflammation, risk of bias for the shortlisted studies using the criteria
leading to metabolic endotoxaemia and oxidative stress [16]. adapted from the Newcastle-Ottawa Scale (NOS) [18]. The
As LPS increases in the blood, it increases activation of assessments included selection bias (i.e. case and control def-
inflammatory pathways, which causes insulin resistance inition), comparability bias (any adjustment to confounding
[17], and thereby triggers the onset and progression of factors) and exposure bias (i.e. ascertainment of exposure
T2DM [16]. and method used). The studies were deemed low risk of bias
To date, 13 case-control studies that met the inclusion cri- when overall quality score of
3 or high risk of bias when
teria had evaluated differential gut microbiota compositions overall quality score of <3 [19]. Discrepancies on these assess-
between healthy controls and T2DM across various countries ments were resolved through discussion and consensus
[1–12,14]. Nevertheless, the results from these case control between all authors.
studies seem to vary. As such, the present systematic review
had systematically analysed evidence of differential composi- 3. Results
tion of faecal microbiota between T2DM and healthy controls
as well as its correlation with metabolic parameters. 3.1. Description of shortlisted studies

2. Materials and methods Of the 3485 studies that were being identified, 258 were dupli-
cates and therefore removed (Fig. 1). A total of 3206 studies
2.1. Literature search strategy were further excluded based on the following reasons: 94
were in vivo or in vitro studies, 1662 had evaluated other
Literature search was performed by using electronic data- pathologies, 31 had described only the protocol, 402 were
bases (i.e. PubMed/Medline, EMBASE and Web of Science) review papers, 663 did not compare between T2DM and
with restriction to English language only. Both medical sub- healthy controls, 132 were randomised clinical trial (e.g. pro-
ject headings (MeSH term) and free text strings (Supplemen- biotics, prebiotics), 54 had evaluated T1DM and the remaining
tary material) were employed. The main search terms used 168 were non-English papers. Altogether, 13 articles were
were as follow: ‘diabetes mellitus’, ‘type 2 diabetes’, ‘meta- included in the present systematic review.
genome’, ‘microbiome’ and ‘next generation sequence’. The
terms were combined with Boolean operators to capture all 3.2. Study characteristics
studies related to T2DM and gut microbiota. The last search
was performed in February 2020 without restriction to publi- Table 1 summarises the characteristics of studies that were
cation date. This systematic review has been registered with shortlisted for this systematic review. All were case-control
International Prospective Register of Systematic Reviews studies, involving 575 T2DM patients and 840 healthy con-
(PROSPERO) (CRD42018076374). trols. The cases were T2DM patients while the controls were
made up of healthy individuals or those with normal glucose
2.2. Study selection tolerance. 13 studies had looked into the differences of gut
microbiota compositions between T2DM patients and healthy
The inclusion criteria were as follow: a) case-control studies controls [1–12,14] (Table 1). Seven were from untargeted
that had included T2DM (case) and healthy adults (control) approach [1–5,12,14]. Of the seven studies, two were con-
and b) studies with analyses of faecal gut microbiota. The ducted in China [2,4], three in Europe [1,3,5], one in Brazil
studies were excluded if they were a) presented only as [12] and Mexico [14] respectively. On the other hand, another
abstracts with no full report of findings, b) on-going clinical, six studies were from targeted approach [6–11]. Two studies
in vivo or in vitro studies, c) review papers, non-English litera- were conducted in Japan [6,7] two in Iran [9,11] and two in
tures or protocol papers, d) Gestational Diabetes Mellitus China [8,10]. In general, out of 13 included studies, seven
(GDM), Type 1 Diabetes Mellitus (T1DM) or metabolic diseases had no gender limit in participant selection [2,4,5,8,10,12,14],
other than T2DM and e) studies that did not compare gut four had similar number of male and female participants
microbiota in T2DM to that of healthy controls. Titles, [6,7,9,11], one included only women [1] and one included only
abstracts and keywords were screened to determine publica- men [3]. Participants in both groups were predominantly
tions of relevance. Whilst the primary outcome of this study female between 40 and 70 years old.
was the composition of gut microbiota in both healthy con- Ten studies defined T2DM patients based on their HbA1c
trols and T2DM, the secondary outcomes included the corre- values [1,2,5–7,9–12,14], two studies were based on their OGTT
lation of gut microbiota with glycaemic parameters, lipid values [3,4] while no information was obtained from one
profiles, inflammatory markers and anthropometric mea- study [8]. Of the 13 included studies, seven studies reported
 diabetes research and clinical practice 173 (2021) 108689 3

In vivo In vitro

Fig. 1 – Study selection PRISMA flowchart.

that T2DM patients were on antihyperglycaemic medications position of gut microbiota was reported as primary
[1,5–8,10,12] while no information was obtained from the outcome (Table 1).
remaining studies [2–4,9,11,14]. One of the studies further
stratified the T2DM patients according to the metformin 3.3. Differential gut microbiota compositions between
treatment status [5]. T2DM and healthy controls
It is important to note that gut microbiota composition
was analysed using different techniques. Whilst seven Fig. 2 showed the compositional changes of gut microbiota at
studies had adopted untargeted approaches; three had genus level in T2DM as opposed to that of healthy controls.
used the shotgun metagenomic sequencing [1,2,5], another One of the most notable genus that consistently showed a sig-
two used 16S rRNA sequencing (Illumina MiSeq) [12,14] and nificant domination (p < 0.05) in the gut of T2DM was lacto-
yet another two used pyrosequencing (i.e. Tag-encoded bacilli while genera Faecalibacterium and Roseburia increased
FLX-amplicon [3] and 454 GS FLX [4]), another six studies in healthy controls in two or more studies. On the other hand,
had used targeted approaches [i.e. quantitative real-time Bifidobacterium, was however, inconsistent.
polymerase chain reaction (qPCR) [8–11], reverse transcrip-
tion qPCR (RT-qPCR) [6] and PCR [7]]. The differential com-
 4
Table 1 – Study characteristics of healthy controls and T2DM (n = 13).
Author (year) Sample size Country Gender (M/F), Age (mean ± SD) Glycaemic control Antihyperglycaemic Method for metagenomics Outcome measures
parameter (T2DM) treatment (class)
Healthy control T2DM Healthy control T2DM Sequencing used Quantification Primary outcome Secondary outcomes

Untargeted approach (n = 7)
Zhang, et al. [4] 44 13 China M/F:12/32, Age: M/F:7/6), Age: OGTT, median: NR 454 GS FLX – Faecal microbiota compositions Association between gut
55.0 ± 9.0 52.0 ± 9.0 13.5 mmol/l microbiota diversity and clinical
indicators
Qin, et al. [2] 174 171 China M/F: 84/90, Age: M/F: 107/65, Age: HbA1c, mean: NR Illumina GAIIx and HiSeq – Faecal microbiota compositions –
42.0 ± 14.0 53.0 ± 13.0 48.0 mmol/mola 2000

diabetes research and clinical practice

Karlsson, et al. [1] 43 53 Sweden Age: 70 (mean ± SD not reported) HbA1c, mean ± SEM: 47.1 Yes (Met, SU) Illumina HiSeq 2000 – Faecal microbiota compositions Correlations between serum
Gender: All female ± 1.3 mmol/mol biomarkers and species or MGCs
Forslund, et al. [5] 554 (277[5] 106e (17[5] Denmark M/F: NR M/F: 28/78e HbA1c, mean ± SD: 8.5 ± 1.8e Yes (Met) Illuminag – Faecal microbiota Correlations between T2DMf with
+ 185[1] + 92[2]) + 33[1] + 56[2]) Age: 54.2 ± 13.1 47/46f 8.1 ± 2.1f compositions gut microbiota
93f (58[5] Age: 58.8 ± 12.8e
+ 20[1] + 15[2]) 61.1 ± 9.7f
Larsen, et al. [3] 18 18 Denmark Age: 56.0 ± 13.0 OGTT, mean ± SD: NR Tag-encoded FLX Amplicon Faecal microbiota –
Gender: All male 16.6 ± 5.4 mmol/l compositions
Leite, et al. [12] 22 20 Brazil M/F: 12/10, Age: M/F: 9/11, Age: HbA1c, mean ± SD: Yes (Met, DPP-4i, Illumina MiSeq – Faecal microbiota Correlations between systemic
55.7 ± 8.3 58.9 ± 8.4 Not stated insulin) compositions inflammation cytokines and
plasma LPS concentration with
bacteria
Lambeth, et al. [14] 15 14 Mexico M/F:5/10, Age: M/F:6/8, Age: HbA1c, mean ± SD: NR Illumina MiSeq Pico green Faecal microbiota compositions Relationship between gut
55.5 ± 13.7 62.0 ± 10.0 7.9 ± 1.7% dsDNA assay microbiota compositions and
HbA1c levels
Targeted approach (n = 6)
Adachi, et al. [7] 59 59 Japan M/F: 34/25, Age: 62.0 M/F: 34/25, Age: 64.0 HbA1c, median: Yesc PCR (16S rDNA) PCR Faecal microbiota compositions –
(59.0–69.0)b (57.5–69.0)b 54.0 mmol/mol (51.0–60.0)b
Sato, et al. [6] 50 50 Japan M/F:26/24, Age: M/F:26/24, Age: HbA1c, mean: Yesd RT-qPCR (16S rRNA) R T-qPCR Faecal microbiota compositions Association of faecal microbiota
60.2 ± 12.9 62.5 ± 10.8 72.0 mmol/mola with various clinical parameters
Navab-Moghadam, et al. [9] 18 18 Iran M/F: 7/11, Age: M/F: 7/11, Age: HbA1c, mean ± SD: NR RT-PCR (16S rRNA) qPCR Faecal microbiota compositions Correlations between studied
52.1 ± 7.56 54.3 ± 7.63 6.7 ± 1.0% bacterial species with BMI
Sedighi, et al. [11] 18 18 Iran M/F: 7/11, Age: M/F: 7/11, Age: HbA1c, mean ± SD: NR RT-PCR (16S rRNA) qPCR Faecal microbiota compositions Correlations between studied
52.1 ± 7.56 54.3 ± 7.63 6.7 ± 1.0% bacterial species with BMI
Lê, et al. [10] 30 50 South China M/F:17/13, Age: M/F: 23/27, Age: HbA1c, mean: 21/50 were on NR qPCR Faecal microbiota compositions Anthropometry and metabolic
41.0 ± 11.0 60.0 ± 8.0 <64.0 mmol/mola medication quantifications
(Met, a-GI, SU)
Wu, et al. [8] 12 16 China M/F: Not stated, M/F:10/6, Agea: 48– NR 4/16 were on PCR-DGGE qPCR Faecal microbiota compositions Identification of dominant

173 (2021) 108689

Agea: 50–65 62 medicationc (16S rRNA) bacteria within-groups

Abbreviations: T2DM, type 2 diabetes mellitus; NGT, normal glucose tolerance; NR, not reported; aSD values not stated; bthe medians (inter-quartile ranges); cdrug class of the antihyperglycaemic
medication is not provided by the authors; dMet, DPP-4i, SU, a-GI, thiazolidine, glinide, GLP-1 receptor agonist, insulin; e without Met; fwith Met; gtype of sequence not mentioned. DGGE, denaturing
gradient gel electrophoresis; qPCR, quantitative polymerase chain reaction; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; PCR, polymerase chain reaction; rRNA, ribosomal
ribonucleic acid; Met, metformin; DPP-4i, dipeptidyl peptidase 4 inhibitors; SU, sulphonylurea; a-GIs, a-glucosidase inhibitors. Note: Whilst anthropometric measurements such as BMI and waist
circumference were recorded in all studies, habitual diet was only assessed in five studies [6–9,11] (Data not shown).
 diabetes research and clinical practice 173 (2021) 108689 5

A)

Abiotrophia
Sporobacter Anaerostipes
Acidaminococcus
Subdoligranulum Clostridium Blautia
Anaerotruncus
Parabacteroides Eubacterium Clostridiales
Bacillus
Prevotella Streptococcus Coprococcus
Bulleidia
Bifidobacterium Alistipes Faecalibacterium
Dorea
Collinsella Bacteroides Lachnospira
Holdemania
Eggerthella Escherichia Megamonas
Lactobacillus
Bilophila Akkermansia Megasphaera
Oscillobacter
Enterobacteriaceae Roseburia
Peptostreptococcus
Desulfovibrio Veillonella
Ruminococcus
Haemophilus

B)

Lactobacillus
Clostridium
Staphylococcus
Prevotella Bacteroides
Enterobacteriaceae
Bifidobacterium Atopobium
Pseudomonas
Fusobacterium

C)

Roseburia
Lactobacillus
Subdoligranulum
Escherichia Intestinibacter Paraprevotella
Clostridiales
Haemophilus
6 diabetes research and clinical practice 173 (2021) 108689

3.4. Correlations between gut microbiota compositions (r = 0.46, p < 0.01) and 2PPG level [3]. Prevotella spp., on the
and clinical parameters other hand, were presented with a weak, negative correlation
with HbA1c (r = 0.27, p = 0.038) [7]. No significant correlation
Table 2 shows the correlations between gut microbiota com- was noted between Prevotella spp. and 2PPG level [3].
position and clinical parameters across all studies. Correla-
tion values (r) of 0.10  r  0.39, 0.40  r  0.69, 0. 3.4.2. Anthropometric measurements
70  r  0.89 and 0.90  r  1.00 indicate weak, moderate, Five studies had reported the effect of anthropometry mea-
strong and very strong correlation, respectively [20]. surements on gut microbiota profiles [1,3,6,9,11] (Table 2).
Weak, negative correlations were only found between C. coc-
3.4.1. Glycaemic parameters coides count (r = 0.31, p = 0.03) [6] and Atopobium cluster
There were six studies that had related gut microbiota com- (r = 0.32, p = 0.02) [6] with BMI. Elsewhere, a weak, negative
position to glycaemic parameters in T2DM [1,3,4,7,12,14] correlation was observed between B. intestinalis and waist cir-
(Table 2). Two studies had evaluated the effects of B:F ratio cumference (r = 0.29, p = 0.04) [1]. No significant correlations
[3,4] on plasma glucose levels [fasting blood glucose (FBG) were found in the remaining three studies [3,9,11].
and 2-hour postprandial glucose (2PPG)]. There was a moder-
ate but significant positive correlation between B:F ratio and 3.4.3. Dietary intake
2PPG (r = 0.47, p = 0.04) [3]. However, no significant correlation Three studies had evaluated of the differential gut microbiota
was found between B:F ratio and FBG (r = 0.01, p > 0.05) [4]. composition between T2DM and healthy controls by docu-
Lactobacilli that belong to the phylum Firmicutes (data menting dietary intake [6–8]. C. coccoides count appeared to
was not available/ reported) and L. gasseri were positively cor- have a weak, negative correlation with energy (r = -0.33,
related with glycaeted haemoglobin (HbA1c) (r = 0.40, p = 0.02) and saturated fatty acid intake (r = 0.31, p = 0.03)
p = 0.0034) and FBG (r = 0.40, p = 0.0047) [1]. There was a weak, [6]. Similar trend was observed with order Lactobacillales and
but positive correlation between lactobacilli and a 2PPG level Bifidobacterium spp., both of which showed a negative correla-
(r = 0.33, p = 0.05) [3]. tion with protein (r = 0.28, p = 0.035) [7] and carbohydrate
In general, the main butyrate-producing bacteria (i.e. intake (r = 0.42, p = 0.001) [7], respectively. On the other
Clostridium spp., family Ruminococcaceae and Roseburia spp.) hand, protein (r = 0.36, p = 0.005) and fat (r = 0.30, p = 0.021)
were negatively correlated with most of the glycaemic param- intakes showed weak positive correlations with Clostridium
eters. Bacteria from the family Ruminococcaceae were reported cluster XI as well as carbohydrates intake (r = 0.27,
to decrease significantly with increased HbA1c level (r = 0.69, p = 0.042) with Clostridium cluster VI while fat intake showed
p = 0.02) [12]. It is noteworthy that there were weak, negative otherwise (r = 0.26, p = 0.046) [7].
correlations between Clostridium spp. and insulin (r = 0.33,
p = 0.02) [1] as well as C-peptide test (r = 0.36, p = 0.02) [1]. 3.4.4. Lipid profiles
Whilst only C. clostridioforme showed a weak, positive correla- Three studies had investigated the changes in gut microbiota
tion with C-peptide level (r = 0.28, p = 0.02), Clostridium cluster composition with lipid profiles [1,6,7]. Order Lactobacillales was
XVIII showed a weak negative correlation with HOMA-IR negatively correlated with total cholesterol (TC) (r = 0.32,
(r = 0.30, p = 0.022) [7]. Clostridium spp. faecal count was p = 0.016) [7]. There were weak, positive correlations between
reported to be moderately, positively correlated with adipo- Clostridium spp. (r = 0.35, p = 0.0098) [1], Clostridium cluster XI
nectin (r = 0.47, p = 0.00000875) [1]. On the other hand, no sig- (r = 0.28, p = 0.034) [7] with high density lipoprotein (HDL),
nificant correlations were found between Clostridium spp. and between C. clostridioforme clusters and triglycerides (TG) level
HbA1c (r = 0.31, p = 0.06) [1], FBG (r = 0.26, p = 0.10) [1] or (r = 0.29, p = 0.00980) [1] and between Clostridium cluster XIVa
2PPG level (r = 0.42, p = 0.06) [3]. Almost similar correlation and TC (r = 0.32, p = 0.014) [7], respectively. Karlsson, et al. [1],
was noted between Roseburia spp. and 2PPG (r = 0.52, however, was the only study that indicated a weak, negative
p = 0.06) level [3]. correlation between Clostridium spp. and TG level (r =0.33,
Bacteroides intestinalis and Bacteroides spp. showed a weak, p = 0.00605) [1]. Atopobium cluster was positively correlated
negative correlation with insulin level (r = 0.28, p = 0.04) [1] with HDL (r = 0.35, p = 0.01) [6].
and FBG (r = 0.27, p = 0.043) [7], respectively. There were,
however, positive correlations between Betaproteobacteria 3.4.5. Inflammatory markers
spp. relative abundance (r = 0.46, p = 0.04), Bacteroides- There were three studies that had described the relationship
Prevotella to C. coccoides-E. rectale group ratio (r = 0.38, between inflammatory markers with gut microbiota composi-
p = 0.03) and Bacteroides-Prevotella to C. coccoides subgroup ratio tion. Class Clostridia was positively correlated with interferon

3
Fig. 2 – Differential. composition of gut microbiota at genus level between T2DM when compared to healthy controls (n = 13)
in A) untargeted approach and B) targeted approach. *p < 0.05 in
two studies. Blue colour arrow represented gut microbiota
were higher in T2DM while increased gut microbiota compositions in healthy controls were indicated by red colour arrow.
Gut microbiota compositions that were found to be in inconsistent pattern across all studies were shown in the middle circle.
C) Differential genera abundances between non-metformin T2DM and metformin-T2DMa; between non-metformin T2DM
and controlb (significance at FDR < 0.1) [5]; bp, butyrate producers. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)
Table 2 – Correlation between gut microbiota composition and relevant parameters (n = 13).
B:F Lactobacilli Butyrate-producing bacteria Others

Untargeted Targeted Untargeted Targeted Untargeted Targeted Untargeted Targeted

Glycaemic HbA1c "* Lactobacilli (r = NA) [1] ; Clostridium spp. (r = -0.31) [1] M microbiome diversity (r = 0.07) [14] ;* Prevotella spp. (r = -0.27) [7]

diabetes research and clinical practice

Parameters "* L. gasseri (r = 0.40) [1] ;* Ruminococcaceae (r = -0.69) [12]
FBG " (r = 0.01) [4] "* Lactobacilli (r = NA) [1] ; Clostridium spp. (r = -0.26) [1] M Gut microbiota abundance (r = NA) [12] ;* Bacteroides spp. (r = -0.27) [7]
"* L. gasseri (r = 0.40) [1]
2PPG "* (r = 0.47) [3] " Lactobacilli (r = 0.33) [3] ; Clostridia (r = -0.42) [3] " Prevotella (r = 0.32) [3]
; Roseburia (r = -0.52) [3] "* Betaproteobacteria (r = 0.46) [3]
"* Bacteroides-Prevotella : C. coccoides-E. rectale (r = 0.38) [3]
"* Bacteroides-Prevotella : C. coccoides subgroup (r = 0.46) [3]
Insulin / C- ;* Clostridium spp. [r = -0.33 / r = -0.36) [1] ;* B. intestinalis (r = -0.28) [1] ;* Clostridium cluster XVIII (r = -0.30) [7]
peptide / "* C. clostridioforme (r = 0.28) [1]
HOMA-IR
Adinopectin "* Clostridium spp. (r = 0.47) [1]
Diet Energy / ;* C. coccoides (r = -0.33 / r = -0.31) [6] "* Clostridium cluster XI (r = 0.30) [7]
saturated ;* Clostridium cluster VI (r = -0.26) [7]
fatty acid
intake
Carbohydrates
"* Clostridium ;* Bifidobacterium spp. (r = -
cluster VI 0.42) [7]
(r = 0.27) [7]
Protein ;* Lactobacillales (r = -0.28) [7]
"* Clostridium cluster XI (0.36) [7]
Inflammatory hs-CRP ;* C. coccoides (r = -0.39) [6] ;* Atopobium cluster (r = -0.39) [6]
Markers IFN-c "* Clostridia (r = 0.69) [12] "* Firmicutes (r = 0.75) [12]
"* Prevotella (r = 0.66) [12]
;* Bacteroides (r = -0.70) [12]
IL-17A "* Enterobacteriaceae (r = 0.58) [12]
IL-6 "* Negativicutes (r = 0.59) [12]

173 (2021) 108689

Anthropometric Waist circumference
Measurements ; Clostridium spp. (r = -0.30) ;* B. intestinalis (r = -0.29) [1]
[1]
BMI ; (r = -0.32) [3] " Lactobacilli (r = 0.30) ; Clostridium spp. (r = 0.22) ;* C. coccoides (r = -0.31) [6] " Bacteroides-Prevotella : C. coccoides-E. ;* Atopobium cluster (r = -0.32) [6]
[11] [1] ; F. prausnitzii (r = -0.35) [9] rectale (r = 0.06) [3] " Fusobacterium (r = 0.83) [11]
; Bifidobacterium (r = -0.06) [11]
; Prevotella (r = -0.08) [11]
; B. longum (r = -0.21) [9]
; B. fragilis (r = -0.01) [9]

Lipid Profiles HDL "* Clostridium spp. (r = 0.35) [1] "*Atopobium cluster (r = 0.35) [6]
"*Clostridium cluster XI (r = 0.28) [7]
TG ;* Clostridium spp. (r = -0.33) [1]
"* C. clostridioforme (r = 0.29) [1]
TC "* Clostridium cluster XIVa (r = 0.32) [7] ;* Order Lactobacillales (r = -0.31) [7]

Abbreviations: *p < 0.05 when compared to control; ", positive correlation; ;, negative correlation; M, no correlation; B:F, Bacteroidetes to Firmicutes ratio; HbA1c, glycaeted haemoglobin; FBG, fasting
blood glucose; 2PPG, 2-hour postprandial glucose; HOMA-IR, homeostatic model assessment of insulin resistance; BMI, body mass index; HDL, high density lipoprotein; TG, triglyceride; TC, total
cholesterol; hs-CRP, high sensitive C-reactive protein; IFN-c, interferon gamma; IL-17A, interleukin 17A; IL-6, interleukin 6; r, correlation coefficient value; NA, not available.

7
 8diabetes research and clinical practice
Table 3 – Newcastle-Ottawa quality assessment scale for included case-control studies (n = 13).
Quality assessment criteria Acceptable (*) Leite, et al. [12] Lambeth, Zhang, Karlsson, Qin, Forslund, Larsen, Adachi, et al. [7] Navab-Moghadam, Sedighi, et al. [11] Sato, et al. [6] Lê, et al. [10] Wu, et al. [8]
et al. [14] et al. [4] et al. [1] et al. [2] et al. [5] et al. [3] et al. [9]

Is the case definition adequate? Yes, with independent validation * * * * * * * *

Representativeness of cases? Consecutive or obviously *
representative series of cases
Selection of controls? Community controls? * * * * * * * * *
Definition of controls? No history of T2DM * * * * * *
Study controls for one factor * * * * * * * * *
Study controls for * * * * * * *
additional factors
Ascertainment of exposure? Secure record, Structured *
interview where blind to case/
control status
Same method of ascertainment Yes * * * * * * * * * * * *
of cases and controls?

173 (2021) 108689

Non-response rate? Same for both groups * * * *
Overall Quality Score (Maximum 9) 4 3 4 6 2 5 5 4 6 6 3 6 3

Abbreviations: *represents fulfilled individual criterion within the subsection.

 diabetes research and clinical practice 173 (2021) 108689 9

gamma (IFN-c) (r = 0.69, p = 0.02) [12] while C. coccoides count another group of butyrate producing bacteria [22] was
was negatively correlated with high sensitivity C-reactive pro- observed. Study found that individuals that lack butyrate-
tein (hs-CRP) (r = 0.39, p = 0.0066) [6]. Leite et al. [12] reported producers in the body are prone to develop T2DM [23]. Buty-
moderate to strong positive correlations between the relative rate, one of the preferred sources of energy for colonic epithe-
abundance of phylum Firmicutes (r = 0.75, p = 0.01) and Prevo- lial cells, maintains the health of the colons [24]. Its reduction
tella spp. (r = 0.66, p = 0.02) with IFN-c, Enterobacteriaceae spp. may intensify intestinal permeability and endotoxaemia.
with interleukin 17A (IL-17A) (r = 0.58, p = 0.04) and Negativi- Depletion of these butyrate producers are associated with
cutes spp. with interleukin 6 (IL-6) (r = 0.59, p = 0.04), respec- inflammation pathogenesis, as observed in T2DM [25]. Like-
tively. However, genus Bacteroides showed a strong, negative wise, a moderate positive correlation was observed between
correlation with IFN-c (r = 0.70, p = 0.01) [12] while Atopobium Clostridium spp. (a butyrate producing bacteria) and adiponec-
cluster had a weak, negative correlation with hs-CRP tin [1]. It has been suggested that butyrate enhanced adipo-
(r = 0.39, p = 0.0058) [6]. nectin by increasing the glucose uptake by skeletal and
cardiac muscle as well as inhibition of hepatic gluconeogene-
3.5. Antihyperglycaemic agents sis which improved insulin efficacy [26].
Other than that, T2DM was presented with significantly
Three studies had described the effects of antihyperglycaemic high level of lactobacilli and the order Lactobacillales [1,3,5–
agents (e.g. metformin and a-glucosidase inhibitors) on the 7,10,11]. More specifically, lactobacilli showed moderate posi-
gut microbiota compositions [1,5,6]. Clostridium spp. and tive correlation with HbA1c and FBG [1] while weak positive
Eubacterium spp. faecal counts were significantly decreased correlation was observed with 2PPG [3]. Lactobacilli is
in metformin-T2DM patients [1]. On the other hand, Enter- famously known for its probiotics properties [27,28], lacto-
obacteriaceae spp. count [1,6] and Staphylococuus spp [6] were bacilli represents a heterogeneous group with well-
significantly increased in patients who were taking met- documented immune-modulating properties. On the other
formin. When comparing the gut microbiota changes hand, some lactobacilli may also potentially contribute to
between metformin-naive and metformin-treated T2DM chronic inflammation in diabetic patients [29]. The controver-
patients, an increase in Escherichia spp. abundance (r = 0.25, sial effects of lactobacilli could be due to species- or strain-
q = 0.07) while a decrease in Intestinibacter spp. (r =0.40, specific, and therefore, the role of lactobacilli in T2DM
q = 0.01) were noted in metformin-treated T2DM patients remains inconclusive [30].
[5]. T2DM who were taking a-glucosidase inhibitors were pre- B:F ratio, Bacteroides-Prevotella:C. coccoides subgroup ratio
sented with significantly higher number of genera Bifidobac- and class Betaproteobacteria were moderately correlated with
terium, Lactobacillus and Enterococcus [6]. In the study by 2PPG [3]. Bacteroides and Prevotella belong to the phylum Bac-
Forslund, et al. [5], the authors reported a depletion of known teroidetes while Betaproteobacteria belongs to the phylum
butyrate producers i.e. Roseburia spp., Subdoligranulum spp. Proteobacteria. These observations suggested that Gram neg-
and butyrate-producing Clostridiales spp. in metformin- ative phyla Bacteroidetes and Proteobacteria might initiate
untreated T2DM when compared to healthy controls. Of note, T2DM through endotoxin-induced inflammatory responses
the increase in lactobacilli in T2DM patients was not seen in due to the presence of high concentrations of endotoxins
the study by Forslund, et al. [5] when the analysis was further and LPS [31].
controlled for metformin treatment status. The inverse relationship between Bifidobacterium spp. with
carbohydrates was observed in one study in this systematic
3.6. Risk of bias and publication bias review [7]. Bifidobacterium spp. are, however, believed to play
an important role in carbohydrate fermentation in the lower
Table 3 summarises the outcomes of the NOS assessment for gut [32]. A study had reported about increased abundance of
all case-control studies included in this systematic review. Bifidobacterium spp. after ingestion of carbohydrates, suggest-
Except for one study with high risk of bias (<3 scores) [2], all ing that carbohydrate has bifidogenic effect that can be used
studies were presented with low risk of bias (
3 scores) [1,3– as prebiotics [33]. There are two possible reasons to these con-
12,14]. trasting results. Firstly, the ability of Bifidobacteria to metabo-
lise particular carbohydrates (e.g. cellulose, lactose) depends
on their species and strains as well as the amount and nature
4. Discussion of the carbohydrates present in the diet [34]. Secondly, it could
be due to the homeostatic imbalance between gut microbiota
It is evident from this systematic review (Fig. 2) that among and systems in T2DM that had led to such outcome [7].
the consistently reported findings at genus level, Faecalibac- The ferocious cycle between gut microbiota dysbiosis and
terium [1,2,4,12] and Roseburia [1–5,12] which are butyrate pro- activated low-level inflammation is considered as a downturn
ducers with anti-inflammatory properties [21] were factor in the development of T2DM. The present findings
significantly higher in healthy controls in two or more studies which had indicated a moderate to strong significant positive
when compared to T2DM. Lactobacilli, that is often used as a correlation between the class Clostridia and phylum Firmi-
probiotic on the other hand was found to be present in the gut cutes with pro-inflammatory cytokines IFN-c as well as
of T2DM in three untargeted approach studies [1,3,5]. Further- between Negativicutes (class of bacteria in the phylum Firmi-
more, butyrate producing bacteria seem to be negatively cor- cutes) and IL-6 [12]. The reason for this could be due to the
related with glycaemic parameters (Table 2). A negative but fact that overweight individuals were correlated with high rel-
strong correlation between Ruminococcaceae and HbA1c [7,12]; ative abundance of Firmicutes [35] and the majority T2DM
10 diabetes research and clinical practice 173 (2021) 108689

patients from Leite, et al. [12] were overweight. Thus, it can be used may produce inconsistent findings that may not exert
suggested that increase Firmicutes stimulated the secretion the real microbial community towards a particular condition
of pro-inflammatory markers by macrophages, leading to (e.g. T2DM).
chronic inflammation [36]. Increased Enterobacteriaceae, and Next, since the 13 included case control studies were
Prevotella were also associated with metabolic endotoxaemia selected from all around the world, their environment varies.
and inflammation, mainly mediated by pro-inflammatory Geographical, ethnicity and lifestyle differences (e.g. diet,
cytokines [12]. In contrast, the relative abundance of genus physical activities) among the participants could have influ-
Bacteroides showed a significant negative correlation with enced the overall gut microbiota compositions. Other discrep-
IFN-c [12]. The reason could be possibly due to the different ancy of gut microbiota between studies can be explained by
Bacteroides spp. and strains that may play different roles in intake of medications which are likely to influence the gut
controlling the thickness of intestinal mucus and glycocalyx, microbiota compositions that lead to inconsistent findings.
thus influencing the intestinal permeability [37] as well as The present systematic review had identified only three out
intake of hypoglycaemic medication as mentioned by Hung of 13 studies [1,5,6] that had looked into the relationship
and Hung [38]. between gut microbiota and antihyperglycaemic medications,
With regards to anthropometric measurements and lipid of which the lack of information makes it difficult to evaluate
profiles, no consistent patterns were observed between said as to whether these findings were, by and large, confounded
parameters and gut microbiota. Although the results were by medications.
significant, however, the correlations were weak. More stud- Even so, the abovementioned limitations may act individ-
ies need to be done to confirm the findings. ually or together to bring their effects on the gut microbiota
Research has identified that drugs can have an impact on that will influence the condition of T2DM. Whilst there were
the function and composition of gut microbiota [39]. In partic- inconsistent findings among these case control studies, this
ular, there is evidence stated that intake of oral hypogly- systematic review has collected a number of useful informa-
caemia agents influence the composition of gut bacteria tion that could pave the way towards a well-designed studies
[40]. Metformin administration could contribute to an that underline the importance of accounting for these limita-
increase in SCFA-producing bacteria [40]. Karlsson, et al. [1], tions to untwine the relationship between gut microbiota and
however, reported otherwise. Forslund, et al. [5] reported sig- T2DM since case control study designs are constricted by con-
nificant increases in butyrate producers (e.g. Roseburia, Sub- founding factors between diseased and control groups.
doligranulum and Clostridiales) in healthy controls when In any case, the present systematic review suggests that
compared to metformin-untreated T2DM patients. However, gut microbiota may act as an intermediary towards a number
the increase in Lactobacillus spp. was reversed or eliminated of risk factors for T2DM. Hence, it is perhaps essential to take
when controlled for metformin treatment status. Taken all into consideration the gut microbiota composition of T2DM
together, metformin enriched T2DM faeces in Enterobacteri- patients individually upon recommendation of treatment.
aceae spp. [1,6], Staphylococuus spp. [6] and Escherichia spp.
[5], while decreased Intestinibacter spp. count [5]. The mecha- 5. Authors’ contributions
nisms of action of metformin is not clearly understood, it
may probably have both direct and indirect effects on gut CFN and FU independently searched, extracted and assessed
microbiota; it may either be a growth or diminution factor all the potential studies. FU wrote the manuscript. CFN, SML
for some specific bacterial species [40]. and KR assisted with the manuscript revision. All authors
a-glucosidase inhibitors affect the nutrient sources of bac- read and approved the final manuscript.
teria by partitioning complex carbohydrates and thus influ-
ence gut microbiota composition [41]. This systematic Declaration of interest
review found administration of a-glucosidase inhibitors was
associated with significantly high number of Bifidobacterium, None.
Lactobacillus and Enterococcus spp. [6]. These bacteria may be
involved in the regulation of glucose metabolisms given that
Acknowledgements
Lactobacillus and Bifidobacterium are known for their probiotic
properties. Enterococcus spp., on the other hand, showed a sig-
This work was funded by Universiti Teknologi MARA (UiTM)
nificant negative correlation with LPS, suggesting that they
[Lestari grant: 600-IRMI 5/3/LESTARI (016/2019)].
may decrease inflammation [42]. Nevertheless, gut microbiota
profile of T2DM may be masked by oral hypoglycaemia agents
as the ameliorative effects of these drugs could contribute to
R E F E R E N C E S
SCFAs production as well as microbiome shifts [5].
This systematic review had identified several limitations
from the shortlisted studies. Firstly, not all recruited T2DM
[1] Karlsson FH, Tremaroli V, Nookaew I, et al. Gut metagenome
share similar BMI category. Therefore, it cannot be excluded
in European women with normal, impaired and diabetic
that gut dysbiosis occurred as a result of obesity. Secondly, glucose control. Nature 2013;498(7452):99–103. https://doi.
there was a wide variation in terms of sample size amongst org/10.1038/nature12198.
the included studies (10–200 subjects) that will influence the [2] Qin J, Li Y, Cai Z, et al. A metagenome-wide association study
effect size. In addition, diverse approaches on the sample pro- of gut microbiota in type 2 diabetes. Nature 2012;490:55.
cessing such as various microbial DNA extraction methods https://doi.org/10.1038/nature11450.
 diabetes research and clinical practice 173 (2021) 108689 11

[3] Larsen N, Vogensen FK, van den Berg FWJ, et al. Gut [19] Rotenstein LS, Ramos MA, Torre M, et al. Prevalence of
microbiota in human adults with type 2 diabetes differs from depression, depressive symptoms, and suicidal ideation
non-diabetic adults. PLoS ONE 2010;5(2). https://doi.org/ among medical students: A systematic review and meta-
10.1371/journal.pone.0009085 e9085. analysis. JAMA 2016;316(21):2214–36. https://doi.org/
[4] Zhang X, Shen D, Fang Z, et al. Human gut microbiota 10.1001/jama.2016.17324.
changes reveal the progression of glucose intolerance. PLoS [20] Schober P, Boer C, Schwarte LA. Correlation coefficients:
ONE 2013;8(8). https://doi.org/10.1371/journal.pone.0071108 appropriate use and interpretation. Anesth Analg 2018;126
e71108. (5):1763–8. https://doi.org/10.1213/ane.0000000000002864.
[5] Forslund K, Hildebrand F, Nielsen T, et al. Disentangling type [21] La Rosa SL, Leth ML, Michalak L, et al. The human gut
2 diabetes and metformin treatment signatures in the human Firmicute Roseburia intestinalis is a primary degrader of
gut microbiota. Nature 2015;528(7581):262–6. https://doi.org/ dietary b-mannans. Nat Commun 2019;10(1):905. https://doi.
10.1038/nature15766. org/10.1038/s41467-019-08812-y.
[6] Sato J, Kanazawa A, Ikeda F, et al. Gut dysbiosis and detection [22] Gomez-Arango LF, Barrett HL, McIntyre HD, Callaway LK,
of ‘‘live gut bacteria” in blood of Japanese patients with type 2 Morrison M, Dekker Nitert M (2016) Connections between the
diabetes. Diabetes Care 2014;37(8):2343–50. https://doi.org/ gut microbiome and metabolic hormones in early pregnancy
10.2337/dc13-2817. in overweight and obese women. 65(8): 2214-2223. https://doi.
[7] Adachi K, Sugiyama T, Yamaguchi Y, et al. Gut microbiota org/10.2337/db16-0278.
disorders cause type 2 diabetes mellitus and homeostatic [23] Xu J, Lian F, Zhao L, et al. Structural modulation of gut
disturbances in gut-related metabolism in Japanese subjects. microbiota during alleviation of type 2 diabetes with a
J Clin Biochem Nutr 2019;64(3):231–8. https://doi.org/10.3164/ Chinese herbal formula. ISME J 2015;9(3):552–62. https://doi.
jcbn.18-101. org/10.1038/ismej.2014.177.
[8] Wu X, Ma C, Han L, et al. Molecular characterisation of the [24] Barcenilla A, Pryde SE, Martin JC, et al. Phylogenetic
faecal microbiota in patients with type II diabetes. Curr relationships of butyrate-producing bacteria from the human
Microbiol 2010;61(1):69–78. https://doi.org/10.1007/s00284- gut. Appl Environ Microbiol 2000;66(4):1654–61. https://doi.
010-9582-9. org/10.1128/aem.66.4.1654-1661.2000.
[9] Navab-Moghadam F, Sedighi M, Khamseh ME, et al. The [25] Gevers D, Kugathasan S, Denson LA, et al. The treatment-
association of type II diabetes with gut microbiota naı̈ve microbiome in new-onset Crohn’s disease. Cell Host
composition. Microb Pathog 2017;110:630–6. https://doi.org/ Microbe 2014;15(3):382–92. https://doi.org/10.1016/
10.1016/j.micpath.2017.07.034. j.chom.2014.02.005.
[10] Lê K-A, Li Y, Xu X, et al. (2013) Alterations in fecal [26] Lihn AS, Pedersen SB, Richelsen B. Adiponectin: action,
Lactobacillus and Bifidobacterium species in type 2 diabetic regulation and association to insulin sensitivity. Obes Rev
patients in Southern China population. Front Physiol 3: 496. 2005;6(1):13–21. https://doi.org/10.1111/j.1467-
https://dx.doi.org/10.3389%2Ffphys.2012.00496 789X.2005.00159.x.
[11] Sedighi M, Razavi S, Navab-Moghadam F, et al. Comparison of [27] Barrons R, Tassone D (2008) Use of Lactobacillus
gut microbiota in adult patients with type 2 diabetes and probiotics for bacterial genitourinary infections in women: a
healthy individuals. Microb Pathog 2017;111:362–9. https:// review. 30(3): 453-468. https://doi.org/10.1016/
doi.org/10.1016/j.micpath.2017.08.038. j.clinthera.2008.03.013.
[12] Leite AZ, Rodrigues NdC, Gonzaga MI, et al. (2017) Detection [28] Ciorba MA. A gastroenterologist’s guide to probiotics. Clin
of increased plasma interleukin-6 levels and prevalence of Gastroenterol Hepatol 2012;10(9):960–8. https://doi.org/
Prevotella copri and Bacteroides vulgatus in the feces of type 10.1016/j.cgh.2012.03.024.
2 diabetes patients. Front Immunol 8: 1107. https://dx.doi. [29] Abo Ali FH, Ashour ZA, Shahin RY, Zaki WK, Ragab SB, Attia
org/10.3389%2Ffimmu.2017.01107 MY. Role of intestinal microflora (Lactobacillus Acidophilus)
[13] Carrera Boada CA, Martinez-Moreno JM. Pathophysiology of in phagocytic function of leukocytes in type 2 diabetic
diabetes mellitus type 2: beyond the duo ‘‘insulin resistance- patients. Egyptian Journal of Medical Human Genetics
secretion deficit”. Nutr Hosp 2013;28(2):78–87. https://doi.org/ 2013;14(1):95–101. https://doi.org/10.1016/j.
10.3305/nh.2013.28.sup2.6717. ejmhg.2012.10.003.
[14] Lambeth SM, Carson T, Lowe J, et al. Composition, diversity [30] Gurung M, Li Z, You H, et al. Role of gut microbiota in type 2
and abundance of gut microbiome in prediabetes and type 2 diabetes pathophysiology. EBioMedicine 2020;51. https://doi.
diabetes. J Diabetes Obes 2015;2(3):1–7. 10.15436%2F2376- org/10.1016/j.ebiom.2019.11.051 102590.
0949.15.031. [31] Aw W, Fukuda S. Understanding the role of the gut ecosystem
[15] Gérard C, Vidal H. Impact of Gut Microbiota on Host Glycemic in diabetes mellitus. J Diabetes Investig 2018;9(1):5–12.
Control. Front Endocrinol 2019;10(29):1–13. https://doi.org/ https://doi.org/10.1111/jdi.12673.
10.3389/fendo.2019.00029. [32] Pokusaeva K, Fitzgerald GF, van Sinderen D (2011)
[16] Cani PD, Bibiloni R, Knauf C, et al. Changes in gut microbiota Carbohydrate metabolism in Bifidobacteria. Genes Nutr 6(3):
control metabolic endotoxemia-induced inflammation in 285-306. https://dx.doi.org/10.1007%2Fs12263-010-0206-6.
high-fat diet-induced obesity and diabetes in mice. Diabetes [33] Depeint F, Tzortzis G, Vulevic J, I’Anson K, Gibson GR.
2008;57(6):1470–81. https://doi.org/10.2337/db07-1403. Prebiotic evaluation of a novel galactooligosaccharide
[17] Pedro MN, Magro DO, da Silva EUPP, et al. Plasma levels of mixture produced by the enzymatic activity of
lipopolysaccharide correlate with insulin resistance in HIV Bifidobacterium bifidum NCIMB 41171, in healthy humans: a
patients. Diabetol Metab Syndr 2018;10(1):5. https://doi.org/ randomized, double-blind, crossover, placebo-controlled
10.1186/s13098-018-0308-7. intervention study. Am J Clin Nutr 2008;87(3):785–91. https://
[18] Wells GA, Shea B, O’Connell D, et al. (2014) The Newcastle- doi.org/10.1093/ajcn/87.3.785.
Ottawa Scale (NOS) for assessing the quality of [34] Watson D, O’Connell Motherway M, Schoterman MH, van
nonrandomised studies in meta-analyses. Available from Neerven RJ, Nauta A, van Sinderen D. Selective carbohydrate
http://www.ohri.ca/programs/clinical_epidemiology/oxford. utilization by Lactobacilli and Bifidobacteria. J Appl Microbiol
asp. Accessed 3 April 2017. 2013;114(4):1132–46. https://doi.org/10.1111/jam.12105.
12 diabetes research and clinical practice 173 (2021) 108689

[35] Krajmalnik-Brown R, Ilhan ZE, Kang DW, DiBaise JK. Effects of [39] Walsh J, Griffin BT, Clarke G, Hyland NP. Drug-gut microbiota
gut microbes on nutrient absorption and energy regulation. interactions: implications for neuropharmacology. Br J
Nutrit Clin Pract 2012;27(2):201–14. https://doi.org/10.1177/ Pharmacol 2018;175(24):4415–29. 10.1111%2Fbph.14366.
0884533611436116. [40] Rodriguez J, Hiel S, Delzenne NM. Metformin: old friend, new
[36] Kim SJ, Kim S-E, Kim AR, Kang S, Park M-Y, Sung M-K. Dietary ways of action-implication of the gut microbiome?. Curr Opin
fat intake and age modulate the composition of the gut Clin Nutr Metab Care 2018;21(4):294–301. https://doi.org/
microbiota and colonic inflammation in C57BL/6J mice. BMC 10.1097/mco.0000000000000468.
Microbiol 2019;19(1):193. https://doi.org/10.1186/s12866-019- [41] Montandon SA, Jornayvaz FR. Effects of antidiabetic drugs on
1557-9. gut microbiota composition. Genes 2017;8(10):250. https://
[37] Devaraj S, Hemarajata P, Versalovic J. The human gut doi.org/10.3390/genes8100250.
microbiome and body metabolism: implications for obesity [42] Su B, Liu H, Li J, et al. Acarbose treatment affects the serum
and diabetes. Clin Chem 2013;59(4):617–28. https://doi.org/ levels of inflammatory cytokines and the gut content of
10.1373/clinchem.2012.187617. Bifidobacteria in Chinese patients with type 2 diabetes
[38] Hung W-W, Hung W-C (2020) How gut microbiota relate to the mellitus. J Diabetes 2015;7(5):729–39. https://doi.org/10.1111/
oral antidiabetic treatment of type 2 diabetes. 3: 100007. 1753-0407.12232.
https://doi.org/10.1016/j.medmic.2020.100007.