Ijcp 12698

SYSTEMATIC REVIEW
Comparative efficacy and safety of antidiabetic drug

regimens added to stable and inadequate metformin
and thiazolidinedione therapy in type 2 diabetes
W. J. Saulsberry,* C. I. Coleman,* E. S. Mearns,* E. Zaccaro, Y. Doleh, D. M. Sobieraj*
SUMMARY
Review criteria
Aims: Determine the efficacy and safety of antidiabetic agents added-on to metformin and a thiazolidinedione (TZD) in patients with inadequately controlled type
2 diabetes (T2D). Methods: MEDLINE and CENTRAL were searched for randomised controlled trials (RCTs) evaluating the addition of an antidiabetic agent in
patients with T2D inadequately controlled on stable, optimised metformin and TZD
therapy ( 1500 mg metformin and 50% maximum TZD dose for 4 weeks).
Frequentist network meta-analysis was performed on identified studies. Results:
Eleven RCTs evaluating dipeptidyl peptidase-4 inhibitors (linagliptin, sitagliptin),
sulfonylureas (SUs) (glibenclamide, glimepiride), glucagon-like peptide-1 (GLP-1)
analogues (exenatide, liraglutide, dulaglutide, taspoglutide) and sodium-glucose
cotransporter2 (SGLT2) inhibitors (canagliflozin, empagliflozin) were identified. The
mean reduction in HbA1c from baseline was significant for all agents (range,
0.551.17%) vs. placebo. SUs were associated with weight gain (range, 3.31
7.29 kg), while weight loss was seen with all GLP-1 analogues (range, 1.53
2.20 kg) and SGLT2 inhibitors (range, 2.082.95 kg) vs. placebo. Relative risk of
hypoglycaemia was increased with dulaglutide, exenatide and glimepiride vs. placebo (RR range, 2.656.17); and trended higher with all other agents except linagliptin. GLP-1 analogues and canagliflozin reduced systolic blood pressure vs.
placebo (range, 2.395.05 mmHg). No agent with available data increased the
risk of urinary or genital tract infection vs. placebo. Conclusion: When added to
stable, optimised metformin and TZD, all evaluated antidiabetic agents reduced
HbA1c; albeit not to the same degree. Moreover, agents differed in their effects
on body weight, hypoglycaemia and systolic blood pressure.
Introduction
Effective glycaemic control plays an important role
in preventing chronic complications of type 2 diabetes (T2D) (1). The American Diabetes Association
(ADA) and European Association for the Study of
Diabetes (EASD) recommend metformin and lifestyle
modifications for initial pharmacological therapy of
T2D (2). However, T2D is a progressive disease and
most patients experience deterioration in glycaemic
control over time necessitating the use of combination therapy. Guidelines recommend a patient-centred approach when adding agents to metformin
considering both efficacy and side effect profiles (2).
Thiazolidinediones (TZDs) are one of several options
A systematic literature search was conducted in

MEDLINE and Cochrane CENTRAL through 8
January 2015.
Randomised controlled trials (RCTs) evaluating
adults with type 2 diabetes inadequately
controlled on metformin plus a thiazolidinedione
(TZD) were included if the trial compared a noninsulin drug or long-acting, once-daily basal
insulin to another of these therapies or placebo
and reported change in glycosylated haemoglobin
A1c.
Frequentist
network
meta-analyses
performed on 11 identified RCTs.
Correspondence to:
Diana M. Sobieraj, School of
Pharmacy, University of
Connecticut, 69 North
Eagleville Road, Storrs, CT
06269, USA
Tel.: + 860-972-2429
Fax: + 860-545-2277
Email:
diana.sobieraj@hhchealth.org
were
Message for the clinic
Department of Pharmacy
Practice, School of Pharmacy,
University of Connecticut,
Storrs, CT, USA
Ten therapies were evaluated in the following

classes: dipeptidyl peptidase-4 (DPP-4) inhibitors,
sulfonylureas (SUs), glucagon-like peptide-1 (GLP1) analogues and sodium-glucose cotransporter2
(SGLT2) inhibitors.
All individual agents reduced HbA1c when added
to stable metformin plus TZD therapy, but not to
the same degree.
SGLT2 inhibitors and GLP-1 agonists significantly
reduced weight compared with DPP-4 inhibitors,
SUs and placebo. No agent resulted in an
increased risk of urinary or genital tract infection
compared with placebo.
*These authors contributed

equally in the performance of
this meta-analysis
Disclosure
CIC has received grant funding
from Boehringer Ingelheim
Pharmaceuticals and Janssen
Pharmaceuticals. No other
authors have any conflicts
germane to this manuscript to
report.
when combination therapy is necessary and have the

advantages of high efficacy in glycosylated haemoglobin A1c (HbA1c) lowering, low risk of hypoglycaemia and durability of response in comparison to
other options. If glycaemic control targets are still
not met, triple drug therapy is needed.
There are numerous drugs available for potential
additive therapy after failure of metformin and a
TZD; however, randomised controlled trials (RCTs)
directly comparing their efficacy and safety are
sparse. Furthermore, patients failing optimised metformin and TZD therapy may differ from those with
uncontrolled T2D while receiving other metforminbased combination therapy [e.g. metformin and a
sulfonylurea (SU)] either in individual patient
2015 John Wiley & Sons Ltd

Int J Clin Pract, November 2015, 69, 11, 12211235. doi: 10.1111/ijcp.12698
1221
1222
Comparative efficacy and safety of antidiabetic drug regimens
characteristics or their disease progression; thereby

affecting their response to different third-line drugs.
Therefore, we performed a network meta-analysis
(NMA) of antidiabetic agents when used as third-line
therapy in patients with T2D not adequately controlled on stable and optimised metformin and TZD
combination therapy.
Material and methods

Search strategy
We performed a systematic literature search for all
relevant articles from the earliest possible date
through 8 January 2015 in MEDLINE and Cochrane
CENTRAL. The search strategy combined the medical subject heading and keywords for metformin
with terms for T2D (type 2 diabetes mellitus, T2DM,
noninsulin dependent diabetes, NIDDM) and for glycosylated hemoglobin HbA1c (glycosylated hemoglobin, hemoglobin A1c, HbA1c, A1c). The results of
this search were then limited to RCTs. Our MEDLINE search strategy is included in Appendix 1. We
also performed a manual search of references from
reports of clinical trials and review articles to identify
additional relevant studies. Results of identified studies were supplemented when possible with data from
the grey literature including www.clinicaltrials.gov,
regulatory agency reports and contacting investigators for clarification or additional data. Two investigators reviewed all potentially relevant citations
independently (ESM, CIC).
Study inclusion
To be included in this meta-analysis, studies had to
be (i) published in English; (ii) parallel-designed
RCTs in adult patients with T2D; (iii) compare at
least one antidiabetic drug (including non-insulin
drugs and long-acting, once-daily basal insulin) to
another antidiabetic drug or placebo/control; (iv)
included only patients who showed inadequate
response to stable, optimised metformin and TZD
combination therapy at randomisation; (v) treated
patients for at least 12 weeks but no more than
52 weeks after randomisation; and (vi) reported
change in HbA1c as an end-point. For our metaanalysis, the criterion of stable therapy was considered to be met if a study included patients who
received at least 1500 mg/day (or maximum tolerated dose) of metformin (or 1000 mg/day as long as
the mean dose in the study was > 1500 mg/day) and
at least 50% of the maximal dose of TZD. Moreover,
patients had to be on combination therapy for at
least the preceding 4 weeks before randomisation or
the study had to explicitly state in the methods that
patients failed stable therapy (3).
Data extraction
Two investigators (YD, EZ), through use of a standardised tool, independently extracted all data with
disagreements resolved by discussion or a third
investigator (CIC). The following information was
sought from each trial: (i) author identification; (ii)
year of publication; (iii) study design and methodological quality information needed to complete the
Cochrane Collaborations tool for assessing risk of
bias (4, 5); (iv) sample size; (v) inclusion/exclusion
criteria; (vi) duration of follow-up; (vii) metformin
and experimental antidiabetic agents, doses and
schedules used; and (viii) baseline characteristics
(age, gender, anthropometrics, baseline HbA1c, duration of T2D). End-point data collected included
mean change from baseline in HbA1c, body weight
(BW) and systolic blood pressure (SBP) and the
number/proportion of patients experiencing confirmed hypoglycaemia (confirmed by finger stick),
urinary tract infections (UTIs) and genital tract
infections (GTIs).
Bias assessment
To assess the quality of each study, the Cochrane
Collaboration risk of bias tool was used (4). This
tool evaluates the characteristics of seven domains
within each trial to determine if the risk of bias in
each domain is low, high or unclear. Domains
include random sequence generation, allocation concealment, blinding of participants and personnel,
blinding of outcomes assessment, incomplete outcome data, selective reporting and any other bias
(Appendix 2). Two reviewers independently assessed
each included trial and resolved any discrepancy
through discussion.
Statistical analysis
Traditional meta-analyses analysing changes in
HbA1c, BW and SBP as continuous variables were
performed. Separate pair-wise analyses were performed for each antidiabetic drug therapy, combining data from approved doses of the same therapies
using the method recommended by the Cochrane
Collaboration (5). In all cases, weighted mean differences (WMDs) and associated 95% confidence intervals (CIs) were calculated using a random-effects
model. Net changes in each of these study end-points
were calculated as the difference between treatment
groups in the changes (baseline minus follow-up) in
these mean values. In instances where variances for
net changes were not reported directly, they were calculated from CIs, p-values or individual variances.
The proportion of patients experiencing confirmed
hypoglycaemia, UTIs and GTIs on each drug therapy
Int J Clin Pract, November 2015, 69, 11, 12211235
were meta-analysed as dichotomous end-points using

a random-effects model, with weighted averages
reported as relative risks (RRs) and associated 95%
CIs. The likelihood of statistical heterogeneity in traditional meta-analyses was assessed using the I2
statistic (I2 > 50% considered representative of
important statistical heterogeneity) (6). When more
than three studies were available in an analysis, publication bias was assessed using the Eggers weighted
regression statistic p-value (p < 0.05 suggesting a
higher likelihood of publication bias) (7). Traditional
pair-wise meta-analyses were performed using StatsDirect version 2.7.8 (StatsDirect Ltd, Cheshire, UK).
In addition to traditional meta-analyses, NMA was
performed (8). NMA is a generalisation of traditional
pair-wise meta-analysis that compares all pairs of
treatments within a set of treatments for the same
disease state (in this case T2D). Along with analysing
the direct within-trial comparisons between two
treatments, the NMA framework enables incorporation of indirect comparisons constructed from two
trials that have one treatment in common (8). This
type of analysis safeguards the within-trial randomised
treatment comparison of each trial while combining
all available comparisons between treatments. We used
the package netmeta (version 0.5-0) in R (version
3.0.2, The R Foundation for Statistical Computing) to
perform NMA (9). The package uses a novel graph
theory methodology that exploits the analogy between
treatment networks and electrical networks to construct a NMA model accounting for the correlated
treatment effects in multi-arm trials. This approach
has been found to be equivalent to other Frequentist
approach to NMA (10). We implemented a randomeffects model assuming common heterogeneity across
all comparisons. To assess for the presence of whole
network heterogeneity as well as inconsistency in our
NMA, we utilised generalised Cochran Q statistics for
multivariate meta-analysis as described by Krahn and
colleagues (11).
Results
[1222]. All 11 RCTs reported change in HbA1c and

change in BW, five trials reported SBP, nine trials
reported confirmed hypoglycaemia, three trials
reported incidence of UTI and one trial reported
incidence of GTI (Table 2). Figure S2 illustrates the
network of included RCTs allowing for direct
comparison of specific agents and placebo. Ten
treatments, as well as placebo were analysed including
dipeptidyl peptidase-4 (DPP-4) inhibitors (linagliptin,
sitagliptin),
SUs
(glibenclamide,
glimepiride),
glucagon-like peptide-1 (GLP-1) analogues (exenatide,
liraglutide, dulaglutide, taspoglutide) and sodiumglucose cotransporter2 (SGLT2) inhibitors (canagliflozin, empagliflozin). No RCTs evaluating long-acting,
once-daily basal insulin were identified.
Of the 11 RCTs included, none proved to have a
low risk of bias for all seven domains (Figure S3).
Four studies demonstrated uncertainty regarding
random sequence generation (1215) and two studies were unclear in methods of allocation concealment (15,16). One study was open-label (13) and
one did not specify the manner of blinding participants and personnel (17). A majority of the studies
were judged to be unclear regarding blinding of outcomes assessment (12,1420) as only three studies
confirmed low risk of bias for this domain
(13,21,22). Four studies were judged to have a high
risk of bias concerning incomplete outcome data
(13,15,20,21) and two studies were determined to
selectively report data (16,19) while the remaining
five studies were determined to have low risk of bias
for these two domains (12,14,17,18,22). Regarding
the final domain, half of the studies were found to
have other identifiable bias defined by the investigators as not receiving stable metformin and a TZD
for at least 12 weeks prior to randomisation and/or
the study not being published in a peer-reviewed
journal (12,17,2022).
For all outcomes, the results obtained upon traditional meta-analyses are reported in Tables S1S6.
Given how few pair-wise analyses were evaluated, the
results reported hereafter focus on those obtained
from the NMA.
Search results
A total of 1024 nonduplicate citations were identified
from the primary literature search (including four
additional RCTs identified through other sources).
After we screened title and abstracts, 313 full-text
articles were screened for eligibility. Upon assessment
of the full-text articles, 302 articles were excluded for
reasons depicted in Figure S1. A total of 11 RCTs
(n = 5353 participants; mean [range] trial age: 53
58 years; 4468% men; mean [range] trial duration,
33 [2454] weeks and mean [range] trial baseline
HbA1c: 7.18.8%) met all inclusion criteria (Table 1)
Network meta-analysis
All antidiabetic agents were associated with statistically significant reductions in HbA1c (range, 0.55
1.17%) compared with placebo (Table 3). When
comparing active agents, glibenclamide was associated with statistically significant reductions in HbA1c
compared with all other active agents except dulaglutide, glimepiride and liraglutide (Figure 1). Glimepiride was associated with statistically significant
reductions in HbA1c compared with empagliflozin
and exenatide. Dulaglutide and liraglutide were each
1223
24
26
Bajaj 2014,
N = 272
Forst 2014,
N = 342
52
54
DeRosa 2013,
N = 453
Dobs 2013,
N = 278
24
26
Wysham, 2014,
N = 978
Kovacs 2014,
N = 499
Follow-up,
weeks
Author year,
N*
HbA1c: 7.510%; BMI:

45 kg/m2, Stable therapy
for 12 weeks
Metformin dose: 1500 mg/
day or MTD; Pioglitazone
dose: 45 mg/day
HbA1c: 710.5%; Stable
therapy for 12 weeks
day (1500 mg if unable to
tolerate higher doses);
Pioglitazone dose: 3045 mg/
day
HbA1c: 710%; BMI: 45 kg/
m2; Stable therapy for
12 weeks; Metformin dose:
> 1500 mg/day or MTD;
day or MTD
HbA1c: > 8%; BMI: 2530 kg/
m2; stable therapy for
12 months; Metformin dose:
1850 mg/day; Pioglitazone
dose: 15 mg/day; at
randomisation, metformin and
pioglitazone were up-titrated
to 2200 mg/day and 30 mg/
day respectively
HbA1c: 7.511%; stable
therapy for 6 weeks
day; Rosiglitazone dose:
4 mg/day or pioglitazone
HbA1c: 711%; BMI: 23

45 kg/m2; Stable therapy for
15003000 mg/day or MTD;
day or MTD
Inclusion criteria
Sitagliptin 100 mg
daily
Placebo
Glibenclamide 5 mg
tid
Sitagliptin 100 mg
daily
Empagliflozin
10 mg daily
Empagliflozin
25 mg daily
Placebo
Canagliflozin
100 mg daily
Canagliflozin
300 mg daily
Placebo
Dulaglutide 1.5 mg
weekly
Dulaglutide
0.75 mg weekly
Exenatide 10 lg
BID
Placebo
Linagliptin 5 mg
daily
Placebo
Interventions
evaluated
8.8 (1.0)
8.7 (1.0)
54.4 (8.8), 56.0

54.8 (9.5), 60.0
7.1 (0.7)
NR, 49.6
8.2 (0.9)
54.6 (10.5), 44.2
7.3 (0.8)
8.1 (0.8)
54.2 (8.9), 50.6
NR, 50.7
8.1 (0.9)
54.7 (9.9), 50.3
8.0 (1.0)
58.3 (9.6), 66.1
8.5 (0.8)
55.2 (8.4), 55.1
7.9 (0.9)
8.1 (1.3)
8.4 (0.8)
55.0 (10.0), 59.0

53.1 (9.7), 45.4
57.0 (10.2), 55.3
8.1 (1.3)
55.0 (10.0), 57.0
8.0 (0.9)
8.1 (1.2)
56.0 (9.0), 60.0
56.7 (10.4), 68.1
8.1 (1.3)
Baseline HbA1c,
% (SD)
56.0 (10.0), 58.0
Population
characteristics, age,
years (SD); males, %
NR
NR
NR
NR
NR
NR
NR
128.2 (12.3)
126.7 (12.0)
126.4 (12.3)
NR
125.0 (14.0)
NR
127.0 (15.0)
127.0 (15.0)
127.0 (15.0)
Baseline SBP,
mmHG (SD)
86.8 (20.0)
82.5 (19.7)
78.9 (7.9)
78.4 (7.5)
78.1 (20.1)
78.9 (19.9)
78.0 (19.1)
93.8 (22.4)
94.4 (25.9)
94.2 (22.2)
74.4 (19.8)
94.0 (19.0)
74.4 (20.4)
97.0 (19.0)
96.0 (21.0)
96.0 (20.0)
Baseline
weight,
kg (SD)
30.8 (5.6)
30.1 (6.2)
27.6 (2.4)
27.3 (2.1)
29.3 (5.4)
29.1 (5.5)
29.2 (5.6)
32.5 (6.4)
32.8 (7.7)
32.3 (6.2)
28.1 (5.5)
33.0 (6.0)
28.2 (5.2)
34.0 (5.0)
33.0 (6.0)
33.0 (5.0)
Baseline BMI,
kg/m2 (SD)
Table 1 Baseline characteristics of randomised controlled trials evaluating antidiabetic drugs in addition to metformin and thiazolidinedione in type 2 diabetes mellitus
9.4 (6.8)
9.3 (5.9)
NR
NR
> 5 years: 41.3%
> 5 years: 44.7%
> 5 years: 46.1%
10.1 (6.6)
11.0 (7.6)
10.5 (6.6)
> 5 years: 6.7%
9.0 (6.0)
> 5 years: 12.3%
9.0 (6.0)
9.0 (5.0)
9.0 (6.0)
Duration of
DM, years (SD)
1224


26
52
24
26
26
Fonseca 2013,
N = 313
Rosenstock
2013,
N = 1189
Henry 2012,
N = 326
Zinman 2009,
N = 533
Roberts 2005,
N = 170
30 mg/day; FPG: 6.7

15.6 mmol/l
HbA1c: 7.511%; BMI: 20
1500 mg/day; Pioglitazone
dose: 30 mg/day
HbA1c: 710%; BMI: 25
45 kg/m2; stable therapy for
1500 mg/day; TZD dose:
4 mg/day rosiglitazone or
30 mg/day pioglitazone;
weight: stable (5% variation)
for 3 months
HbA1c: 710%; BMI: 25
45 kg/m2; stable therapy for
1500 mg/day or MTD;
day; Weight: stable (5%
variation) for 12 weeks
HbA1c: 711%; BMI: 45 kg/
m2; stable therapy for
2000 mg/day; Rosiglitazone
dose: 8 mg/day; FBG 7.5
12.8 mmol/l
HbA1c: 7.59.5%; BMI: 26
10002500 mg/day; TZD
dose: 48 mg/day
rosiglitazone or half maximum
45 mg/day pioglitazone;
FPG 7.213.1 mmol/l
Inclusion criteria
Glimepiride 8 mg
daily
Placebo
Liraglutide 1.2 mg
daily
Liraglutide 1.8 mg
daily
Placebo
Taspoglutide 10 mg
weekly
Taspoglutide 20 mg
weekly
Placebo
Taspoglutide 10 mg
weekly
Taspoglutide 20 mg
weekly
Exenatide 10 lg
bid
Sitagliptin 100 mg
daily
Placebo
Interventions
evaluated
8.2 (0.8)
56.4 (10.0), 62.3
8.4 (1.2)
55.0 (10.0), 62.0
8.2 (0.7)
8.6 (1.2)
55.0 (11.0), 51.0
56.5 (9.8), 61.0
8.5 (1.2)
8.1 (0.9)
54.3 (9.6), 50.0
55.0 (10.0), 57.0
8.1 (0.9)
55.5 (10.1), 53.0
8.1 (0.9)
55.0 (9.9), 49.0
8.2 (1.0)
8.1 (0.9)
56.0 (10.0), 52.0
52.5 (10.3), 59.0
8.1 (0.9)
8.7 (1.0)
56.4 (9.4), 62.8
56.0 (9.6), 58.0
8.8 (1.0)
Baseline HbA1c,
% (SD)
55.7 (8.7), 61.8
Population
characteristics, age,
years (SD); males, %
NR
NR
128.0 (14.5)
126.0 (14.2)
129.0 (14.8)
NR
NR
NR
131.2 (14.3)
132.0 (14.3)
132.1 (14.3)
NR
NR
Baseline SBP,
mmHG (SD)
96.3 (19.2)
100.9 (19.0)
NR
NR
NR
88.5 (20.6)
93.5 (21.8)
94.0 (22.3)
94.5 (18.6)
93.2 (18.9)
95.5 (20.0)
83.8 (19.1)
82.1 (19.1)
Baseline
weight,
kg (SD)
32.8 (5.1)
34.0 (5.2)
33.9 (5.2)
33.5 (5.1)
33.2 (5.4)
32.0 (5.3)
33.0 (5.0)
32.8 (5.3)
33.8 (5.2)
33.1 (5.3)
33.5 (5.2)
30.0 (5.2)
29.9 (5.2)
Baseline BMI,
kg/m2 (SD)
8.7 (6.8)
7.9 (4.9)
9.0 (6.0)
9.0 (6.0)
9.0 (6.0)
7.5 (5.8)
8.3 (5.3)
7.3 (4.6)
6.5 (5.4)
7.0 (5.7)
6.3 (5.2)
10.2 (6.1)
9.4 (5.8)
Duration of
DM, years (SD)
BID, twice daily; BMI, body mass index; DM, diabetes mellitus; FPG, fasting plasma glucose; MTD, maximally tolerated dose; n, number of patients; NR, not reported; SBP, systolic blood pressure; SD, standard
deviation; TZD, thiazolidinedione. *Number of study participants randomised. Baseline demographic data include the entire study population, i.e. patients on metformin + pioglitazone background therapy as well
as patients only on pioglitazone background therapy.
Follow-up,
weeks
Author year,
N*
Table 1 Continued

1225
1234
Appendix 1 Medline Search Strategy
1. Metformin.mp or Metformin/
2. Type 2 diabetes mellitus.mp or Diabetes Mellitus, Type 2/
3. T2DM.mp
4. Noninsulin dependent diabetes.mp
5. NIDDM.mp
6. Type 2 DM.mp
7. Glycosylated hemoglobin.mp or Hemoglobin A, Glycosylated/
8. Glycated hemoglobin.mp
9. Hemoglobin a1c.mp
10. Hba1c.mp
11. A1c.mp
12. OR/2-11
13. 1 AND 12*
*Limited to randomised controlled trials/
Appendix 2 Grading Parameters for the Cochrane Risk of Bias Assessment
Risk of bias domain
Assessment criteria
Random sequence generation
Low risk
High risk
Unclear risk
Allocation concealment
Low risk
High risk
Blinding of participants and personnel
Unclear risk
Low risk
High risk
Unclear risk
Blinding of outcome assessment
Incomplete outcome data
Low risk
High risk
Unclear risk
Low risk
Participants had an equal opportunity to be placed in each arm of the study. Investigators described a random
component of sequence generation (e.g. computer random number generation, shuffling cards or envelopes)
Participants were pre-determined to be placed in a particular arm of the study. Investigators described a
non-random component in sequence generation (e.g. odd or even date of birth, date of admission, judgment
of the clinician or participant)
There was no mention of the sequence generation process. This included studies that only stated randomization
was performed without any further information about the sequence generation process
Participants and investigators enrolling participants could not foresee assignment due to the use of proper
allocation concealment methods (e.g. central allocation including telephone, web-based and pharmacycontrolled randomization, sequentially numbered, opaque, sealed envelopes or IVRS)
Participants or investigators enrolling participants could foresee assignments into the arms of the study because
of improper allocation concealment methods (e.g. use of an open random allocation schedule, alternation or
rotation or assignment envelopes without appropriate safeguards)
There was no mention or insufficient information to permit judgment of low or high risk
The study indicated that the participants and personnel were blinded throughout the study. The study had to
mention double-blinding and also indicate who was blinded and the method of blinding (e.g. matching
placebo, double-dummy design)
No or incomplete complete blinding of both participants and personnel
There was insufficient information to determine whether both participants and personnel were sufficiently
blinded
The study specifically indicated the outcome assessor was blinded
The study indicated the outcome assessor knew which arm of the study the participant was allocated
There was insufficient evidence to determine whether the outcome assessor was blinded
The study was considered low risk if there was no missing outcome data or if the following criteria were met:
The study filled the gaps in data (missing data) with imputation (e.g. last observation carried forward).
The study utilised an intention to treat (included all randomised patients) or modified intention to treat
analysis (included all randomised patients who received one dose and/or had one follow-up visit)
methodology.
The study did not have more than 20% of participants dropout of the trial after randomization.
The study did not have more than a 10% differential in dropout rate between study arms
High risk
Unclear risk
The above-mentioned criteria were not met

Insufficient reporting of attrition and/or exclusions to allow for judgment of low or high risk assessment

1235
Appendix 2 Continued
Risk of bias domain
Assessment criteria
Selective reporting
Low risk
High risk
Other bias
Unclear risk
Low risk
High risk
Unclear risk
The study reported all end-points discussed in the methods section and the study included all outcomes
expected of a type 2 diabetes study and specific antidiabetic medication classes:
All studies reported HbA1c changes and confirmed hypoglycaemia;

SGLT2 inhibitor studies reported the incidences of UTIs and GTIs;
Sulfonylurea, meglitinide and basal insulin studies reported weight change
The study did NOT report all end-points discussed in the methods section and/or did NOT include all outcomes
to be expected of the disease state and medication classes discussed above
There is insufficient information to assess low or high risk
Participants were stable on maximal or near-maximal metformin ( 1500 mg/day) and at least half-maximal
dose of TZD for at least 12 weeks prior to the start of the study. Study was published in a peer-reviewed
journal
Participants were either not on stable, maximal or near-maximal metformin and at least half-maximal dose
of TZD for less than 12 weeks prior to randomization, and/or the study was not published in a
peer-reviewed journal
There was insufficient information to assess low or high risk
GTI, genital tract infection; HbA1c, glycated haemoglobin; IVRS, interactive voice response system; SGLT2, sodium-glucose cotransporter 2; SU, sulfonylurea; TZD,
thiazolidinediones; UTI, urinary tract infection.

(2.9313.00)
(0.061.89)
(0.813.44)
(0.713.95)
(0.816.15)
(1.699.20)
CI, confidence interval; GTI, genital tract infection; HbA1c, glycosylated haemoglobin A1c; NA, not applicable; RR, relative risk; UTI, urinary tract infection; WMD, weighted mean difference.
2.33)
( 7.77 to
( 4.80 to 0.03)
0.72)
( 5.35 to
(0.273.02)
(0.333.51)
0.10 (0.011.77)
NA
NA
NA
NA
NA
NA
NA
NA
NA
0.68
1.26
NA
0.90
NA
NA
NA
NA
1.08
NA
1.29)
1.01)
Canagliflozin
Dulaglutide
Empagliflozin
Exenatide
Glibenclamide
Glimepiride
Linagliptin
Liraglutide
Sitagliptin
Taspoglutide
0.70
0.97
0.55
0.58
1.17
0.98
0.57
1.00
0.77
0.87
(
(
(
(
(
(
(
(
(
(
0.88 to 0.52)
1.13 to 0.81)
0.73 to 0.37)
0.74 to 0.41)
1.33 to 1.00)
1.2 to 0.76)
0.92 to 0.22)
1.24 to 0.76)
0.92, 0.61)
1.04, 0.70)
2.95
1.72
2.08
2.20
7.29
3.31
0.17
2.10
0.29
1.53
( 4.01 to 1.89)
( 2.54 to 0.89)
( 2.69 to 1.47)
( 2.98 to 1.43)
(6.727.86)
(1.844.78)
( 1.11 to 0.77)
( 2.83 to 1.37)
( 0.22 to 0.81)
( 2.25, 0.80)
1.77
2.65
NA
3.94
NA
6.17
0.32
1.67
1.68
2.23
(0.388.32)
(1.126.23)
3.76
3.33
NA
3.04
NA
NA
NA
5.05
NA
2.39
( 6.23 to
( 5.65 to
(0.241.91)
(0.443.63)
Incidence of
GTI RR (95% CI)
Agent vs.
Placebo
Change in HbA1c, %
WMD (95% CI)
Change in body
weight, kg WMD
(95% CI)
Confirmed hypoglycaemia
RR (95% CI)
Change in systolic blood

pressure, mmHg WMD
(95% CI)
Incidence of UTI
RR (95% CI)
pressure and incidence of UTI and GTI
Table 3 Results of network meta-analysis comparing antidiabetic agents with placebo on change in HbA1c, change in body weight, confirmed hypoglycaemia, change in systolic blood
1228
Canagliflozin, dulaglutide, exenatide and sitagliptin

were not associated with an increased risk of UTI
compared with placebo or in comparison to each
other (Figure 5). There were no data to evaluate the
other agents for this outcome. Only one trial
reported GTI data. This study compared canagliflozin to placebo and only a small number of GTIs
were observed [nine events in the canagliflozin arms
and no events the placebo arm (RR 10.0, 95% CI
0.56100.0)].
Heterogeneity, publication bias and

inconsistency
Heterogeneity and publication bias within the traditional meta-analyses could not be assessed because
there were no pair-wise comparisons in which more
than two trials were identified. For each end-point
analysed in the NMA, we observed minimal heterogeneity in the whole network, as well as no significant statistical inconsistency.
Discussion
When advancing to dual therapy for glycaemic control in T2D, TZDs have been shown to be a viable
option evidenced by their proven efficacy and are
considered to have a high degree of efficacy relative
to other pharmacologic classes according to most
recent guidelines (2). Like all antihyperglycaemia
agents, TZDs carry adverse effects which must be
considered in the light of additional glucose lowering
potential. Most commonly, TZDs are associated with
fluid retention, increased weight and bone fracture
risk. Although some evidence suggests negative
cardiovascular effects of rosiglitazone and bladder
cancer risk with pioglitazone, by and large the evidence is inconclusive. The guidelines emphasise the
importance of personalisation of antihyperglycaemic
therapy based on efficacy, potential adverse effects,
patient age and comorbidities. Therefore, while TZDs
are unlikely to be used in patients with heart failure
or bone fracture history, they may be preferred in a
patient who does not want to inject medications or
who would rather avoid side effects known to alternative therapies such as gastrointestinal disturbance.
As T2D progresses, a third agent is often needed
to adequately maintain glycaemic goals. Although
insulins are associated with the greatest potential for
glycaemic control, they are unattractive options to
patients because of fear of injection, inconvenience
and their side effect profile (most importantly,
weight gain and hypoglycaemia). Current guidance
only briefly discusses triple therapies because of the
paucity of trials addressing their comparative efficacy
and safety and again stresses the importance of drug
Figure 1 Effect of antidiabetic agents on change in A1c. Therapies are reported in alphabetical order. The values in the
matrix represent the weighted mean difference and 95% confidence interval for the row-defining treatment vs. the columndefining treatment. Abbreviations: CANA=canagliflozin; DULA=dulaglutide; EMPA=empagliflozin; EXEN=exenetide;
GLIB=glibenclamide; GLIM=glimepiride; LINA=linagliptin; LIRA=liraglutide; PLC=placebo; SITA=sitagliptin;
TASPO=taspoglutide
selection based on patient and disease-specific characteristics (2). Our NMA utilised both direct and
indirect evidence to assess efficacy and safety of a
third agent added to metformin and TZD dual therapy; including two therapies from the newest pharmacologic class, SGLT2 inhibitors.
Our NMA illustrates that addition of any one of
the evaluated active therapies to metformin and a
TZD is more effective in reducing HbA1c compared
with placebo. The extent to which HbA1c was lowered did vary with each agent, ranging from a reduction of 0.551.17%. Comparing active therapies to
each other demonstrated that there are clinically
meaningful differences between drugs when applying
the FDA established threshold of a change in HbA1c
of 0.3% (23). Differences within drug classes were
also observed, for example, both dulaglutide and
liraglutide lowered HbA1c to a greater extent than
exenatide, by approximately 0.4%. Differences in
HbA1c change between the two SUs were also
observed in our analysis.
Though efficacy of antidiabetic agents is measured by HbA1c reduction, this benefit is often
weighed against the risk of hypoglycaemia. Hypoglycaemia can be debilitating and dangerous, especially
in elderly people, frail or those with multiple

comorbidities. Drugs that stimulate insulin secretion, such as SUs, are strong culprits for causing
hypoglycaemia because of their actions independent
of blood glucose concentration (2). In contrast,
drugs with a mechanism of action dependent upon
the presence of elevated blood glucose, like GLP-1
analogues, DPP-4 inhibitors and SGLT2 inhibitors,
are less likely to cause hypoglycaemia because they
activate a more controlled release of insulin or do
not stimulate insulin release at all (24). Our results
were consistent with these expectations as glimepiride demonstrated the greatest risk in confirmed
hypoglycaemia vs. placebo. In addition, glimepiride
increased the risk of hypoglycaemia compared with
the DPP-4 inhibitors studied and liraglutide. We
found the SGLT2 inhibitor canagliflozin and the
DPP-4 inhibitors linagliptin and sitagliptin did not
significantly increase hypoglycaemia risk compared
with other active therapies or placebo. The results
of the GLP-1 agonists were mixed with dulaglutide
and exenatide increasing hypoglycaemia risk compared with placebo; while liraglutide and taspoglutide did not. Despite some therapies being less
likely to individually cause hypoglycaemia, the
1229
1230
Figure 2 Effect of antidiabetic agents on change in weight. Therapies are reported in alphabetical order. The values in the
matrix represent the weighted mean difference and 95% confidence interval for the row-defining treatment vs. the columndefining treatment. Abbreviations: CANA=canagliflozin; DULA=dulaglutide; EMPA=empagliflozin; EXEN=exenetide;
GLIB=glibenclamide; GLIM=glimepiride; LINA=linagliptin; LIRA=liraglutide; PLC=placebo; SITA=sitagliptin;
TASPO=taspoglutide
Figure 3 Effect of antidiabetic agents on change in systolic blood pressure. Therapies are reported in alphabetical order.
The values in the matrix represent the weighted mean difference and 95% confidence interval for the row-defining
treatment vs. the column-defining treatment. Abbreviations: CANA=canagliflozin; DULA=dulaglutide; EXEN=exenetide;
LIRA=liraglutide; PLC=placebo; TASPO=taspoglutide
ADA/EASD guidelines particularly emphasise the

importance of hypoglycaemia when multiple agents
working
additively/synergistically
are
utilised
together to lower blood glucose (2). Together with

our results, this highlights the importance of close
blood glucose monitoring when adding a third
Figure 4 Effect of antidiabetic agents on relative risk of confirmed hypoglycaemia. Therapies are reported in alphabetical
order. The values in the matrix represent the relative risk and 95% confidence interval for the row-defining treatment vs.
the column-defining treatment. Abbreviations: CANA=canagliflozin; DULA=dulaglutide; EXEN=exenetide;
GLIM=glimepiride; LINA=linagliptin; LIRA=liraglutide; PLC=placebo; SITA=sitagliptin; TASPO=taspoglutide
Figure 5 Effect of antidiabetic agents on relative risk of urinary tract infections. Therapies are reported in alphabetical
order. The values in the matrix represent the relative risk and 95% confidence interval for the row-defining treatment vs.
the column-defining treatment. Abbreviations: CANA=canagliflozin; DULA=dulaglutide; EXEN=exenetide; PLC=placebo;
SITA=sitagliptin
agent to metformin and a TZD, even if the agent is

acting dependently on glucose.
Elevated BW and blood pressure increase cardiovascular risk in patients with T2D. The ADA recom 2015 John Wiley & Sons Ltd
mends weight loss of 510% of BW in patients with

T2D to reduce the risk of cardiovascular complications (2). Unfortunately, weight gain is a common
side effect of some antidiabetic agents while others
1231
1232
induce weight loss or are considered weight neutral

and may be preferred for this reason. In previous
meta-analyses involving metformin dual therapy,
results showed a significant increase in BW with SUs
compared with placebo and all other studied agents
(25). Our results showed an additional increase in
BW when SUs are added as a triple therapy agent.
SUs directly stimulate insulin release and store any
excess glucose as fat (26). Markedly, the new SGLT2
inhibitors and the GLP-1 agonists demonstrated significant BW reduction compared with DPP-4 inhibitors, SUs and placebo, making them favourable in
patients who could benefit from weight loss. SGLT2
inhibitor induced weight loss is likely a result of their
diuretic mechanism causing urinary glucose excretion, reduction in water weight and unabsorbed glucose calories (27). As for the GLP-1 agonists, weight
loss is likely induced by increased satiety and gastric
emptying time.
Canagliflozin, dulaglutide, exenatide and liraglutide
demonstrated significant SBP lowering compared with
placebo, although no one drug was found to significantly impact SBP more than another. These results
are limited by the fewer number of drugs within the
network compared with other outcomes evaluated in
this NMA. The mechanism behind the effects of GLP1 agonists on blood pressure are not fully understood,
however, Kim et al. suggest GLP-1 receptor activation
on cardiomyocytes induces atrial natriuretic peptide
release and corresponding smooth muscle relaxation
and natriuresis (28). The same mechanism for weight
loss (elimination of volume through urine) may also
be responsible for the blood pressure lowering effects
observed with SGLT2 inhibitors (29).
Concerns have been raised regarding SGLT2 inhibitors and their potential to increase rates of UTIs
and GTIs in patients due to increased glycosuria creating an ideal environment for bacterial and mycotic
growth (30). There was no significant increase in
UTI risk associated with the SGLT2 inhibitors, but a
possible increased rate of GTIs. As a result of the
small number of trials reporting UTI and GTI and
their relative infrequency, our NMA was likely
underpowered to assess these outcomes. Our NMA
has a number of advantages worth noting. This is
the first published NMA evaluating metformin plus
TZD in triple antihyperglycaemic therapy. Moreover,
our analysis included data on the newest class of
References
1 Holman RR, Paul SK, Bethel MA, Matthews DR,
Neil HAW. 10-year follow-up of intensive glucose
control in type 2 diabetes. N Engl J Med 2008; 359:
15771589.
agents, the SGLT2 inhibitors. Finally, to increase the

internal validity of our results, we required trials to
confirm hypoglycaemia objectively to be included in
our analysis.
A limitation of our NMA was the possibility of
heterogeneity and inconsistency in the network.
However, we used strict inclusion criteria to limit
occurrence, and model diagnostics did not suggest
either was of significant concern. Regardless, their
presence cannot be fully ruled out. Second, as in any
NMA, there is a risk of type 1 (as a result of multiple
hypothesis testing) and type 2 (due to underpowered
analyses) error. Next, publication bias is always a
concern in meta-analyses. We attempted to decrease
the likelihood of such bias by searching multiple grey
literature sources in additional to traditionally used
bibliographical databases. In addition, it is important
to note that statistically significant differences
between treatment options do not always equate to
clinical relevant differences. In interpreting the clinical significance of our NMAs results, clinicians
might choose to interpret changes greater than 0.3%,
45 kg and 5 mmHg as thresholds for important
clinical differences between treatments for HbA1c,
BW and SBP respectively (23). Finally, to increase
the amount of data available in our analysis, we
included some agents that have not yet been
approved in the USA.
In conclusion, the addition of a third agent to
stable, optimised metformin plus a TZD in patients
with inadequately controlled T2D can be beneficial
in treating hyperglycaemia; however, agents reduced
HbA1c to differing extents and have varying effects
on BW, hypoglycaemia and systolic blood pressure.
Therefore, patient-specific characteristics must be
considered in selecting an appropriate agent, and
patients should be closely monitored for both efficacy and safety.
Author contributions
CIC was responsible for the concept and design.
Data collection, analysis, interpretation and statistics
were performed by WJS, ESM, DMS, EZ and YD.
WJS, ESM, EZ, YD, CIC and DMS drafted the article
and DMS, ESM and CIC performed critical revisions.
All authors have approved the final article for submission.
2 Inzucchi SE, Bergenstal RM, Buse JB, Diamant M,

Ferrannini E, Nauck M, et al.; American Diabetes
Association (ADA); European Association for the
Study of Diabetes (EASD). Management of hyperglycemia in type 2 diabetes, 2015: a patient centered approach: update to a position statement of
the American Diabetes Association (ADA) and the
European Association for the Study of Diabetes

(EASD). Diabetes Care 2015; 38:140149.
3 Phung OJ, Scholle JM, Talwar M, Coleman CI. Effect of noninsulin antidiabetic drugs added to metformin therapy on glycemic control, weight gain,
and hypoglycemia in type 2 diabetes. JAMA 2010;
303: 14101418.

4 Higgins JP, Altman DG, Gtzsche PC, J

uni P,
Moher D, Oxman AD, et al.; Cochrane Bias Methods Group; Cochrane Statistical Methods Group.
The Cochrane Collaborations tool for assessing
risk of bias in randomised trials. BMJ 2011; 18
(343): d5928.
5 Higgins JPT, Green S (editors). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011]. The Cochrane
Collaboration, 2011. www.cochrane-handbook.org.
(accessed June 27, 2014).
6 Higgins JP, Thompson SG, Deeks JJ, Altman DG.
Measuring inconsistency in meta-analyses. BMJ
2003; 327: 55760.
7 Sterne JAC, Gavaghan D, Egger M. Publication and
related bias in meta-analysis: power of statistical
tests and prevalence in literature. J Clin Epidemiol
2000; 53: 11191129.
8 Hoaglin DC, Hawkins N, Jansen JP, Scott DA, Itzler R, Cappelleri JC, et al. Conducting indirecttreatment-comparison and network-meta-analysis
studies: report of the ISPOR Task Force on Indirect
Treatment Comparisons Good Research Practices:
part 2. Value Health 2011; 14: 429437.
9 Package netmeta. June 24, 2014. http://cran.rproject.org/web/packages/netmeta/netmeta.pdf
(accessed June 27, 2014).
10 R
ucker G. Network meta-analysis, electrical networks and graph theory. Research Synthesis Methods
2012; 3: 312324.
11 Krahn U, Binder H, K
onig J. A graphical tool for
locating inconsistency in network meta-analyses.
BMC Med Res Methodol 2103:13;35.
12 Fonseca V, Staels B, Morgan JD 2nd, Shentu Y,
Golm GT, Johnson-Levonas AO, et al. Efficacy and
safety of sitagliptin added to ongoing metformin
and pioglitazone combination therapy in a randomized, placebo-controlled, 26-week trial in
patients with type 2 diabetes. J Diabetes Complications 2013; 27: 17783.
13 Rosenstock J, Balas B, Charbonnel B, Bolli GB, Boldrin M, Ratner R, et al.; T-Emerge 2 Study Group.
The fate of taspoglutide, a weekly GLP-1 receptor
agonist, versus twice-daily exenatide for type 2 diabetes: the T-Emerge 2 trial. Diabetes Care 2013; 36:
498504.
14 Henry RR, Mudaliar S, Kanitra L, Woloschak M,
Balena R; T-Emerge 3 Study Group. Efficacy and
safety of taspoglutide in patients with type 2 diabetes inadequately controlled with metformin plus
pioglitazone over 24 weeks: T-Emerge 3 trial. J Clin
Endocrinol Metab 2012; 97: 23709.
15 Roberts VL, Stewart J, Issa M, Lake B, Melis R. Triple therapy with glimepiride in patients with type 2
diabetes mellitus inadequately controlled by metformin and a thiazolidinedione: results of a 30-week,
randomized, double-blind, placebo-controlled, parallel-group study. Clin Ther 2005; 27: 153547.
16 Derosa G, Cicero AF, Franzetti IG, Querci F, Carbone A, Piccinni MN, et al. A comparison between
sitagliptin or glibenclamide in addition to metformin + pioglitazone on glycaemic control and b-

17
18
19
20
21
22
23
24
25
26
27
cell function: the triple oral therapy. Diabet Med

2013; 30: 84654.
Wysham C, Blevins T, Arakaki R, Colon G, Garcia P,
Atisso C, et al. Efficacy and safety of dulaglutide
added onto pioglitazone and metformin versus exenatide in type 2 diabetes in a randomized controlled
trial (AWARD-1). Diabetes Care 2014; 37: 215967.
Bajaj M, Gilman R, Patel S, et al. Linagliptin
improved glycaemic control without weight gain or
hypoglycaemia in patients with type 2 diabetes
inadequately controlled by a combination of metformin and pioglitazone: a 24-week randomized,
double-blind study. Diabet Med 2014; 31: 150514.
Kovacs CS, Seshiah V, Swallow R, Jones R, Rattunde H, Woerle HJ, et al.; on behalf of the
EMPA-REG PIOTM Trial Investigators. Empagliflozin improves glycaemic and weight control
as add-on therapy to pioglitazone or pioglitazone
plus metformin in patients with type 2 diabetes: a
24-week, randomized, placebo-controlled trial. Diabetes Obes Metab 2014;16:14758.
Zinman B, Gerich J, Buse JB, Lewin A, Schwartz S,
Raskin P, et al.; LEAD-4 Study Investigators. Efficacy and safety of the human glucagon-like peptide-1 analog liraglutide in combination with
metformin and thiazolidinedione in patients with
type 2 diabetes (LEAD-4 Met+TZD). Diabetes Care
2009; 32: 122430.
Forst T, Guthrie R, Goldenberg R, Yee J, Vijapurkar U, Meininger G, et al. Efficacy and safety
of canagliflozin over 52 weeks in patients with type
2 diabetes on background metformin and pioglitazone. Diabetes Obes Metab 2014; 16: 46777.
Dobs AS, Goldstein BJ, Aschner P, Horton ES,
Umpierrez GE, Duran L, et al. Efficacy and safety
of sitagliptin added to ongoing metformin and
rosiglitazone combination therapy in a randomized
placebo-controlled 54-week trial in patients with
type 2 diabetes. J Diabetes 2013; 5: 6879.
Treadwell JR, Uhl S, Tipton K, Shamliyan T, Viswanathan M, Berkman ND, et al. Assessing equivalence and noninferiority. J Clin Epidemiol 2012; 65:
11449.
Lee YS, Jun HS. Anti-diabetic actions of glucagon-like
peptide-1 on pancreatic beta-cells. Metabolism 2014;
63:9, Demuth HU, McIntosh CH, Pederson RA. Type
2 diabetestherapy with dipeptidyl peptidase IV inhibitors. Biochim Biophys Acta 2005; 1751: 33.
Phung OJ, Scholle JM, Talwar M, Coleman CI. Effect of noninsulin antidiabetic drugs added to metformin therapy on glycemic control, weight gain,
and hypoglycemia in type 2 diabetes. JAMA 2010;
303: 14101418.
Proks P, Reimann F, Green N, Gribble F, Ashcroft
F. Sulfonylurea stimulation of insulin secretion. Diabetes 2002; 51 (Suppl. 3): S36876.
Sha S, Devineni D, Ghosh A, Polidori D, Chien S,
Wexler D, et al. Canagliflozin, a novel inhibitor of
sodium glucose co-transporter 2, dose dependently
reduces calculated renal threshold for glucose
excretion and increases urinary glucose excretion in
healthy subjects. Diabetes Obes Metab 2011; 13:
66972.
1233
28 Kim M, Platt MJ, Shibasaki T, Quaggin SE, Backx

PH, Seino S, et al. GLP-1 receptor activation and
Epac2 link atrial natriuretic peptide secretion to control of blood pressure. Nat Med 2013; 19: 567575.
29 Baker WL, Smyth LR, Riche DM, Bourret EM,
Chamberlin KW, White WB. Effects of sodium-glucose co-transporter 2 inhibitors on blood pressure:
a systematic review and meta-analysis. J Am Soc
Hypertens 2014; 8, 26275.e9.
30 Scholes D, Hooton TM, Roberts PL, Gupta K, Stapleton AE, Stamm WE. Risk factors associated with
acute pyelonephritis in healthy women. Ann Intern
Med 2005; 142: 207.
Supporting Information
Additional Supporting Information may
be found in the online version of this
article:
Table S1. Results of traditional metaanalysis comparing antidiabetic therapies
effects on HbA1c.
effects on body weight.
effects on systolic blood pressure.
effects on experiencing confirmed hypoglycaemia.
effects on experiencing urinary tract
infections.
effects on experiencing genital tract infections.
Figure S1. Results of the literature
search.
Figure S2. Network diagram of randomised controlled trials evaluating
antidiabetic drugs in addition to metformin and a thiazolidinedione in type 2
diabetes mellitus.
Figure S3. Risk of bias assessment of randomised controlled trials.
Paper received March 2015, accepted June 2015
1234
Appendix 1 Medline Search Strategy
1. Metformin.mp or Metformin/
2. Type 2 diabetes mellitus.mp or Diabetes Mellitus, Type 2/
3. T2DM.mp
4. Noninsulin dependent diabetes.mp
5. NIDDM.mp
6. Type 2 DM.mp
7. Glycosylated hemoglobin.mp or Hemoglobin A, Glycosylated/
8. Glycated hemoglobin.mp
9. Hemoglobin a1c.mp
10. Hba1c.mp
11. A1c.mp
12. OR/2-11
13. 1 AND 12*
*Limited to randomised controlled trials/
Appendix 2 Grading Parameters for the Cochrane Risk of Bias Assessment
Risk of bias domain
Assessment criteria
Random sequence generation
Low risk
High risk
Unclear risk
Allocation concealment
Low risk
High risk
Blinding of participants and personnel
Unclear risk
Low risk
High risk
Unclear risk
Blinding of outcome assessment
Incomplete outcome data
Low risk
High risk
Unclear risk
Low risk
Participants had an equal opportunity to be placed in each arm of the study. Investigators described a random
component of sequence generation (e.g. computer random number generation, shuffling cards or envelopes)
Participants were pre-determined to be placed in a particular arm of the study. Investigators described a
non-random component in sequence generation (e.g. odd or even date of birth, date of admission, judgment
of the clinician or participant)
There was no mention of the sequence generation process. This included studies that only stated randomization
was performed without any further information about the sequence generation process
Participants and investigators enrolling participants could not foresee assignment due to the use of proper
allocation concealment methods (e.g. central allocation including telephone, web-based and pharmacycontrolled randomization, sequentially numbered, opaque, sealed envelopes or IVRS)
Participants or investigators enrolling participants could foresee assignments into the arms of the study because
of improper allocation concealment methods (e.g. use of an open random allocation schedule, alternation or
rotation or assignment envelopes without appropriate safeguards)
There was no mention or insufficient information to permit judgment of low or high risk
The study indicated that the participants and personnel were blinded throughout the study. The study had to
mention double-blinding and also indicate who was blinded and the method of blinding (e.g. matching
placebo, double-dummy design)
No or incomplete complete blinding of both participants and personnel
There was insufficient information to determine whether both participants and personnel were sufficiently
blinded
The study specifically indicated the outcome assessor was blinded
The study indicated the outcome assessor knew which arm of the study the participant was allocated
There was insufficient evidence to determine whether the outcome assessor was blinded
The study was considered low risk if there was no missing outcome data or if the following criteria were met:
The study filled the gaps in data (missing data) with imputation (e.g. last observation carried forward).
The study utilised an intention to treat (included all randomised patients) or modified intention to treat
analysis (included all randomised patients who received one dose and/or had one follow-up visit)
methodology.
The study did not have more than 20% of participants dropout of the trial after randomization.
The study did not have more than a 10% differential in dropout rate between study arms
High risk
Unclear risk
The above-mentioned criteria were not met

Insufficient reporting of attrition and/or exclusions to allow for judgment of low or high risk assessment

1235
Appendix 2 Continued
Risk of bias domain
Assessment criteria
Selective reporting
Low risk
High risk
Other bias
Unclear risk
Low risk
High risk
Unclear risk
The study reported all end-points discussed in the methods section and the study included all outcomes
expected of a type 2 diabetes study and specific antidiabetic medication classes:
All studies reported HbA1c changes and confirmed hypoglycaemia;

SGLT2 inhibitor studies reported the incidences of UTIs and GTIs;
Sulfonylurea, meglitinide and basal insulin studies reported weight change
The study did NOT report all end-points discussed in the methods section and/or did NOT include all outcomes
to be expected of the disease state and medication classes discussed above
There is insufficient information to assess low or high risk
Participants were stable on maximal or near-maximal metformin ( 1500 mg/day) and at least half-maximal
dose of TZD for at least 12 weeks prior to the start of the study. Study was published in a peer-reviewed
journal
Participants were either not on stable, maximal or near-maximal metformin and at least half-maximal dose
of TZD for less than 12 weeks prior to randomization, and/or the study was not published in a
peer-reviewed journal
There was insufficient information to assess low or high risk
GTI, genital tract infection; HbA1c, glycated haemoglobin; IVRS, interactive voice response system; SGLT2, sodium-glucose cotransporter 2; SU, sulfonylurea; TZD,
thiazolidinediones; UTI, urinary tract infection.


Ijcp 12698

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Ijcp 12698

Hochgeladen von

Copyright:

Verfügbare Formate

SYSTEMATIC REVIEW

Comparative efficacy and safety of antidiabetic drug

A systematic literature search was conducted in

Message for the clinic

Ten therapies were evaluated in the following

*These authors contributed

when combination therapy is necessary and have the

2015 John Wiley & Sons Ltd

Comparative efficacy and safety of antidiabetic drug regimens

characteristics or their disease progression; thereby

Material and methods

Comparative efficacy and safety of antidiabetic drug regimens

were meta-analysed as dichotomous end-points using

[1222]. All 11 RCTs reported change in HbA1c and

HbA1c: 7.510%; BMI:

HbA1c: 711%; BMI: 23

54.4 (8.8), 56.0

54.6 (10.5), 44.2

54.2 (8.9), 50.6

54.7 (9.9), 50.3

58.3 (9.6), 66.1

55.2 (8.4), 55.1

55.0 (10.0), 59.0

57.0 (10.2), 55.3

55.0 (10.0), 57.0

56.0 (9.0), 60.0

56.7 (10.4), 68.1

56.0 (10.0), 58.0

> 5 years: 41.3%

> 5 years: 44.7%

> 5 years: 46.1%

> 5 years: 6.7%

2015 John Wiley & Sons Ltd

2015 John Wiley & Sons Ltd

30 mg/day; FPG: 6.7

56.4 (10.0), 62.3

55.0 (10.0), 62.0

55.0 (11.0), 51.0

56.5 (9.8), 61.0

54.3 (9.6), 50.0

55.0 (10.0), 57.0

55.5 (10.1), 53.0

55.0 (9.9), 49.0

56.0 (10.0), 52.0

52.5 (10.3), 59.0

56.4 (9.4), 62.8

56.0 (9.6), 58.0

55.7 (8.7), 61.8

Comparative efficacy and safety of antidiabetic drug regimens

Comparative efficacy and safety of antidiabetic drug regimens

Appendix 1 Medline Search Strategy

Appendix 2 Grading Parameters for the Cochrane Risk of Bias Assessment

Risk of bias domain

Random sequence generation

Blinding of participants and personnel

Incomplete outcome data

The above-mentioned criteria were not met

2015 John Wiley & Sons Ltd

Comparative efficacy and safety of antidiabetic drug regimens

Risk of bias domain

All studies reported HbA1c changes and confirmed hypoglycaemia;

2015 John Wiley & Sons Ltd

Change in systolic blood