BMJ Open: first published as 10.1136/bmjopen-2015-009417 on 24 February 2016. Downloaded from

PEER REVIEW HISTORY

BMJ Open publishes all reviews undertaken for accepted manuscripts. Reviewers are asked to complete a checklist review form (http://bmjopen.bmj.com/site/about/resources/checklist.pdf) and are provided with free text boxes to elaborate on their assessment. These free text comments are reproduced below.

This paper was submitted to a another journal from BMJ but declined for publication following peer review. The authors addressed the reviewers’ comments and submitted the revised paper to BMJ Open. The paper was subsequently accepted for publication at BMJ Open.

ARTICLE DETAILS

TITLE (PROVISIONAL) SGLT-2 receptor inhibitors for treating patients with type 2 diabetes mellitus- a systematic review and network meta-analysis AUTHORS Shyangdan, Deepson; Uthman, Olaken; Waugh, Norman

VERSION 1 - REVIEW

REVIEWER Xin Sun Chinese Cochrane Center REVIEW RETURNED 05-Sep-2014

GENERAL COMMENTS This study, using systematic review and network meta-analysis method, compared the effects of alternative sodium glucose transporter 2 (SGLT2) inhibitors for patients with type 2 diabetes mellitus (T2DM). The comments are as below.

1. The eligibility criteria and the search for relevant studies do not seem complete, and may pose high risk of selective inclusion of http://bmjopen.bmj.com/ trials. First, this review compared a selective group of SGLT-2 inhibitors (dapagliflozin, canagliflozin, empagliflozin and luseogliflozin) for patients with T2DM. Why did the authors not include other SGLT-2 inhibitors (e.g. ipragliflozin, ertugliflozin, tofogliflozin, remogliflozin) for the assessment? Second, the search strategy of this study does not seem comprehensive. The terms that the authors used have been limited. For example, as the intervention

of interest is SGLT-2 inhibitors, the authors may feel desired to use on September 26, 2021 by guest. Protected copyright. “sodium-glucose transport proteins” as a MeSH term, “sodium- glucose cotransporter 2 inhibitors” as key words for their search. Instead, they included the names of the few agents for the search and consequently identified 10 trials. It was not completely clear why the current review, addressing a same issue, included a much smaller number of trials. In addition, this review included trials that exclusively compared to placebo, while a few active-control trials were recently published. The incomplete search might have not effectively identified those active-control trials

2. In the assessment of the risk of bias, why did the authors not address the issue of blinding of participants and care providers?

3. In their analyses, the authors built a network of comparisons based on the placebo-controlled trials. As the e-figures 1 and 2 showed, all the included SGLT-2 inhibitors compared directly to placebo, but no comparison was made between any of active agents. Thus, the conclusion that the authors made – efficacy of BMJ Open: first published as 10.1136/bmjopen-2015-009417 on 24 February 2016. Downloaded from

alternative SGLT-2 inhibitors – was completely grounded on the indirect comparison. This poses a serious limitation, and implication of the findings was thus compromised.

4. In addition, in the construction of the network, the authors treated comparisons of alternative doses of SGLT-2 inhibitors (e.g. canagliflozin 100 mg and 300 mg) versus placebo – which came from a same trial (e.g. canagliflozin 100 mg vs. canagliflozin 300 mg vs. placebo) – as if the data were independent. I am not a statistician expert, but don’t think this approach is ideal.

5. The layout of the manuscript, particularly the results section, is confusing. The figure and table legends should be included at the end of the manuscript, instead of the mid of the text. Also, it would be better if the authors reported each of the baseline characteristics one at a column for each trial.

6. It would be desirable to include the study design “systematic review and network meta-analysis” in the title. The manuscript writing needs to be improved. In some of the places, the descriptions are confusing.

REVIEWER Young Min Cho Associate Professor REVIEW RETURNED 09-Sep-2014

GENERAL COMMENTS Although this article provides indirect evidence based on the results of network metaanalysis, it provides us with valuable insight comparing the efficacy of different SGLT2 inhibitors. In the absence of adequate head-to-head trial, the results of this study would be the best evidence as of yet. However, as the authors implied, head-to-

head trials should be conducted to claim which one is superior to http://bmjopen.bmj.com/ another. I have a few comments as shown below.

1. For the HbA1c target, it is unclear whether the authors used “less than 7%” or “less than or equal to 7%”.

2. eTable 1. Ferrannini et al 2010 and Stenlof et al 2013 show “+ metformin” values, even though these two studies were

monotherapy trials. on September 26, 2021 by guest. Protected copyright.

3. The rate of renal glucose filtration follows the function of plasma glucose levels and eGFRs. As shown in eTable 2 and eTable 3, baseline HbA1c levels are not much different except Bolinder et al 2012. However, few studies reported baseline eGFR values, which might affect the HbA1c lowering efficacy. So, the lack of data on baseline eGFR is another limitation of this study.

REVIEWER Sonal Singh Johns Hopkins REVIEW RETURNED 09-Sep-2014

GENERAL COMMENTS The authors have conducted an NMA of SGLT-2 inhibitors. Their methods are relatively ok but have limitations.

1. The question is scoped too narrowly for readers of BMJ. BMJ Open: first published as 10.1136/bmjopen-2015-009417 on 24 February 2016. Downloaded from

Comparison among SGLT-2 is less relevant, then their actual role if any in second line therapy. The SR should include a full range of second line oral options for T2 DM. 2. The outcomes are limited in scope. Infact there is no evaluation of AEs in the trial. The outcome of glycated hemoglobin followed for 26 weeks in most studies is insufficient for the evaluation of a new diabetic drug. 3. They rate these AEs as minor but for canaglifozin --more than one in 10 women had UTIs or candidial infection making it a significant effect on quality of life. 4. There is no comparison of direct MA to the NMA and compare the results. 5. If there are no head to head studies, then sponsors should be asked to conducted them.

VERSION 1 – AUTHOR RESPONSE

Referee 1. Responses

The eligibility criteria and the search for relevant Trials of some of the newer flozins have studies do not seem complete, and may pose appeared and are included. high risk of selective inclusion of trials. First, this review compared a selective group of SGLT-2 inhibitors (dapagliflozin, canagliflozin, We have clearly stated our inclusion criteria empagliflozin and luseogliflozin) for patients with under the section ‘Study selection’. We have T2DM. Why did the authors not include other included randomised controlled trials assessing SGLT-2 inhibitors (e.g. ipragliflozin, ertugliflozin, the efficacy of flozins in monotherapy in patients tofogliflozin, remogliflozin) for the assessment? with type 2 diabetes inadequately controlled with

diet and exercise, and in dual therapy in patients http://bmjopen.bmj.com/ with inadequate control on metformin monotherapy. Other inclusion criteria included minimum duration of 24 weeks, and those reporting at least one of the following outcomes - % of patients achieving HbA1c level of ≤ or mean change in HbA1c, body weight and SBP from baseline to 24 weeks. on September 26, 2021 by guest. Protected copyright.

Second, the search strategy of this study does We have revised our search strategy and found not seem comprehensive. The terms that the other flozins. The relevant studies have been authors used have been limited. For example, included. Using the names of the drugs is the as the intervention of interest is SGLT-2 best way of finding the trials. If we wanted to inhibitors, the authors may feel desired to use search for reviews, we agree that using the “sodium-glucose transport proteins” as a MeSH MeSH term would be worthwhile. term, “sodium-glucose cotransporter 2 inhibitors” as key words for their search. In addition, we have been involved in three Instead, they included the names of the few NICE appraisals, STAs of dapagliflozin and agents for the search and consequently empagliflozin in combination therapy, and those identified 10 trials. and canagliflozin in monotherapy, and we have the manufacturers’ submission as an additional check on completeness.

In addition, this review included trials that The referee misses the point. To do an NMA, BMJ Open: first published as 10.1136/bmjopen-2015-009417 on 24 February 2016. Downloaded from exclusively compared to placebo, while a few we need common comparators, and the active- active-control trials were recently published. The control trials (of which there are two, with incomplete search might have not effectively linagliptin and sitagliptin) would not provide that identified those active-control trials

2. In the assessment of the risk of bias, We have taken this into consideration since one why did the authors not address the issue of of the items of the Cochrane risk of bias is blinding of participants and care providers? related to blinding and binding takes into account into whether participants or care providers or both have been blinded in the study

3. In their analyses, the authors built a Yes – we needed a common comparator which network of comparisons based on the placebo- had to be placebo. There are no head to head controlled trials. As the e-figures 1 and 2 trials of one flozin against another so an indirect showed, all the included SGLT-2 inhibitors comparison is inevitable. compared directly to placebo, but no comparison was made between any of active agents. Thus, the conclusion that the authors made – efficacy of alternative SGLT-2 inhibitors – was completely grounded on the indirect comparison. This poses a serious limitation, and implication of the findings was thus compromised.

In addition, in the construction of the network, This approach is correct and it is a common http://bmjopen.bmj.com/ the authors treated comparisons of alternative approach to compare treatment arms in a doses of SGLT-2 inhibitors (e.g. canagliflozin network 100 mg and 300 mg) versus placebo – which came from a same trial (e.g. canagliflozin 100 mg vs. canagliflozin 300 mg vs. placebo) – as if the data were independent. I am not a statistician expert, but don’t think this approach on September 26, 2021 by guest. Protected copyright. is ideal.

5. The layout of the manuscript, The layout of the manuscript has been changed particularly the results section, is confusing. The figure and table legends should be included at the end of the manuscript, instead of the mid of the text. Also, it would be better if the authors reported each of the baseline characteristics one at a column for each trial.

6. It would be desirable to include the OK study design “systematic review and network meta-analysis” in the title. BMJ Open: first published as 10.1136/bmjopen-2015-009417 on 24 February 2016. Downloaded from

The manuscript writing needs to be improved. In We have revised the manuscript some of the places, the descriptions are confusing.

Referee 2

Although this article provides indirect evidence We agree but, in the absence of head to based on the results of network metaanalysis, it comparison of drugs, indirect comparison provides us with valuable insight comparing the method provides some evidence. efficacy of different SGLT2 inhibitors. In the absence of adequate head-to-head trial, the results of this study would be the best evidence as of yet. However, as the authors implied, head-to-head trials should be conducted to claim which one is superior to another. I have a few comments as shown below.

For the HbA1c target, it is unclear whether the The target HbA1c level is <7%. authors used “less than 7%” or “less than or equal to 7%”. eTable 1. Ferrannini et al 2010 and Stenlof et al Thank you pointing this out. We have rectified 2013 show “+ metformin” values, even though this. these two studies were monotherapy trials.

Referee 3.

The authors have conducted an NMA of SGLT-2 The term “have limitations” is very vague. No inhibitors. Their methods are relatively ok but response required. have limitations. http://bmjopen.bmj.com/

The question is scoped too narrowly for readers The referee misses the point of the study, which of BMJ. Comparison among SGLT-2 is less was to compare the potencies of the individual relevant, then their actual role if any in second flozins. No change required. line therapy. The SR should include a full range of second line oral options for T2 DM. on September 26, 2021 by guest. Protected copyright. The outcomes are limited in scope. In fact there HbA1c is the standard outcome in trials of new is no evaluation of AEs in the trial. The outcome drugs for type 2 diabetes, though nowadays of glycated hemoglobin followed for 26 weeks in longer-term studies of cardiovascular outcomes most studies is insufficient for the evaluation of are usually requested by the licensing bodies. a new diabetic drug. But results of those will not be available for years, and choice of drug needs to be considered now.

3. They rate these AEs as minor but for The authors of the study rate these events as canagliflozin --more than one in 10 women had minor therefore, the impact on qualify of life UTIs or candidial infection making it a significant might not be significant given the intensity of effect on quality of life those events.

There is no comparison of direct MA to the NMA There are no direct trials with which to create a and compare the results “direct MA” so this is not a sensible comment BMJ Open: first published as 10.1136/bmjopen-2015-009417 on 24 February 2016. Downloaded from

If there are no head to head studies, then Agreed, but this is not something we or even the sponsors should be asked to conducted them. BMJ can command.

VERSION 2 – REVIEW

REVIEWER Zhongxue Chen Indiana University Bloomington, USA REVIEW RETURNED 16-Aug-2015

GENERAL COMMENTS I read the reviewer's comments of a earlier version of this manuscript and the authors' responses. Overall I think the authors have addressed all the comments well and the manuscript is read to be accepted. However, the presentation of this manuscript can be improved to make it more readable. For example, in the Figure Legends section, Figure 1 was not listed there. For the Online Only Figures, it starts with eFigure 4. The Appendix should be given after the main text, instead of the beginning. Also it is better to give a sentence or two to describe the Appendix.

REVIEWER Avinesh Pillai Department of Statistics University of Auckland New Zealand REVIEW RETURNED 05-Nov-2015

http://bmjopen.bmj.com/

GENERAL COMMENTS There are many flozins being compared both for monotherapy and for dual therapy with metformin. The authors have done well in explaing how the analyses were done in the 'Data synthesis and model implementation' section.

The number of limitations of the study are highlighted well in the discussion, and puts the conclusions of the study in proper context; that there are few clinically significant differences amonst the drugs. on September 26, 2021 by guest. Protected copyright. As the authors state, the strength of their study is that it allows for comparisons between SGLT-2 inhibitors that have not been compared head-to-head in RCTs. However, the issues of trials being included where the risk of bias could not be determined due to lack of information, and the fact that in the trials included patients who were randomised to canagliflozin 300 mg instead of 100 mg first, highlight the issues with the analysis and comparisons being made. Also, we have no indication of what the safety data for SGLT-2 inhibitors froim these trials looks like. This is the best evidence we have of the comparison of different doses of the different SGLT2 inhibitors in lieu of direct comparisons of the drugs...some evidence is better that none!

Minor comment: Page 38, first sentence: ...which was 'ran' should be 'run'.