Supplementary Online Content
Zheng SL, Roddick Aj, Aghar-Jaffar R. Association between use of sodium- glucose cotransporter-2 inhibitors, glucagon-like peptide-1 agonists, and dipeptidyl peptidase-4 inhibitors with all-cause mortality in patients with type 2 diabetes: a systematic review and metaanalysis. JAMA. doi:10.1001/jama.2018.3024
eMethods 1. Search strategy eMethods 2. Drug doses eMethods 3. Event definitions eMethods 4. Detailed Statistical Methods eMethods 5. Changes in protocol eTable 1. DIC for model selection eTable 2. Baseline Characteristics (All studies) eTable 3. Baseline characteristics for cardiovascular outcome trials eTable 4. Risk of bias of individual trials eTable 5. All-cause mortality network meta-analysis by individual drug type eTable 6. Clinical Endpoints in Cardiovascular Outcome Trials eTable 7. Sensitivity analysis (Bayesian fixed-effect) eTable 8. Frequentist network meta-analysis eFigure 1. Risk of bias summary eFigure 2. Funnel plot eFigure 3. Network plots eFigure 4. Forest plots and ranking plots for additional secondary outcomes eFigure 5. Network plot for individual drugs eFigure 6. Forest plot for all-cause mortality for individual drugs eFigure 7. Forest plots and ranking plots for safety outcomes eFigure 8. Forest plots of drug-class specific adverse effects of interest eFigure 9. Breakdown of direct and indirect evidence eMethods 1. Search strategy
This supplementary material has been provided by the authors to give readers additional information about their work.
© 2018 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 10/02/2021 eMethods 1: Search strategy
SGLT2 inhibitor
Embase (via OVID): 1. (Diabetes mellitus or type 2 diabetes or type ii diabetes).af 2. Sodium glucose cotransporter 2 inhibitor.af or (sodium glucose transporter 2 inhibitor* or sodium glucose transporter ii inhibitor* or SGLT2 inhibitor*).af 3. (Canagliflozin or Dapagliflozin or Empagliflozin or Ertugliflozin or Tofogliflozin or Ipragliflozin or Remogliflozin).af 4. (randomi*ed controlled trial or randomi*ed trial).af 5. 2 or 3 6. 1 and 4 and 5
MEDLINE: o Diabetes mellitus or type 2 diabetes or type ii diabetes o AND o Sodium glucose cotransporter 2 inhibitor OR (sodium glucose transporter 2 inhibitor* or sodium glucose transporter ii inhibitor* or SGLT2 inhibitor*) o OR o Canagliflozin or Dapagliflozin or Empagliflozin or Ertugliflozin or Tofogliflozin or Ipragliflozin or Remogliflozin o AND o randomized controlled trial[Publication Type]) OR controlled clinical trial[Publication Type]) OR randomized[Title/Abstract]) OR placebo[Title/Abstract]) OR drug therapy[MeSH Subheading]) OR randomly[Title/Abstract]) OR trial[Title/Abstract]) OR groups[Title/Abstract])) NOT ((animals[mh]) NOT humans[mh]))
CENTRAL: o Diabetes mellitus or type 2 diabetes or type ii diabetes o (sodium glucose transporter 2 inhibitor* or sodium glucose transporter ii inhibitor* or SGLT2 inhibitor*) or (Canagliflozin or Dapagliflozin or Empagliflozin or Ertugliflozin or Tofogliflozin or Ipragliflozin or Remogliflozin)
© 2018 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 10/02/2021 GLP-1 analogue
Embase (via OVID): 1. (Diabetes mellitus or type 2 diabetes or type ii diabetes).af 2. (glucagon like peptide 1 receptor agonist).af or (glucagon-like peptide 1 receptor inhibitor* or glucagon-like peptide 1 receptor agonist* or glucagon- like peptide 1 inhibitor* or glucagon-like peptide 1 agonist* or GLP-1 receptor inhibitor* or GLP-1 receptor agonist* or GLP-1 inhibitor* or GLP-1 agonist*).af 3. Albiglutide or Dulaglutide or Semaglutide or Liraglutide or Lixisenatide or Taspoglutide or Exenatide 4. (randomi*ed controlled trial or randomi*ed trial).af 5. 2 or 3 6. 1 and 4 and 5
MEDLINE: o Diabetes mellitus or type 2 diabetes or type ii diabetes o AND o (glucagon like peptide 1 receptor agonist).af or (glucagon-like peptide 1 receptor inhibitor* or glucagon-like peptide 1 receptor agonist* or glucagon- like peptide 1 inhibitor* or glucagon-like peptide 1 agonist* or GLP-1 receptor inhibitor* or GLP-1 receptor agonist* or GLP-1 inhibitor* or GLP-1 agonist*) o OR o Albiglutide or Dulaglutide or Semaglutide or Liraglutide or Lixisenatide or Taspoglutide or Exenatide o AND o randomized controlled trial[Publication Type]) OR controlled clinical trial[Publication Type]) OR randomized[Title/Abstract]) OR placebo[Title/Abstract]) OR drug therapy[MeSH Subheading]) OR randomly[Title/Abstract]) OR trial[Title/Abstract]) OR groups[Title/Abstract])) NOT ((animals[mh]) NOT humans[mh]))
CENTRAL: o Diabetes mellitus or type 2 diabetes or type ii diabetes o (glucagon-like peptide 1 receptor inhibitor* or glucagon-like peptide 1 receptor agonist* or glucagon-like peptide 1 inhibitor* or glucagon-like peptide 1 agonist* or GLP-1 receptor inhibitor* or GLP-1 receptor agonist* or GLP-1 inhibitor* or GLP-1 agonist*) OR o Albiglutide or Dulaglutide or Semaglutide or Liraglutide or Lixisenatide or Taspoglutide or Exenatide
© 2018 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 10/02/2021 DPP-4 inhibitor
Embase (via OVID): 1. (Diabetes mellitus or type 2 diabetes or type ii diabetes).af 2. (dipeptidyl peptidase IV inhibitor).af 3. (dipeptidyl peptidase IV inhibitor* or dipeptidyl-peptidase 4 inhibitor* or DPP4 inhibitor* or DPP 4 inhibitor or DPP IV inhibitor).af 4. (Alogliptin or Anagliptin or Gemigliptin or Linagliptin or Omarigliptin or Saxagliptin or Sitagliptin or Teneligliptin or Vildagliptin).af 5. (randomi*ed controlled trial or randomi*ed trial).af 6. 2 or 3 7. 1 and 4 and 5
MEDLINE:
o Diabetes mellitus or type 2 diabetes or type ii diabetes AND o (dipeptidyl peptidase IV inhibitor) or (dipeptidyl peptidase IV inhibitor* or dipeptidyl-peptidase 4 inhibitor* or DPP4 inhibitor* or DPP 4 inhibitor or DPP IV inhibitor) OR o Alogliptin or Anagliptin or Gemigliptin or Linagliptin or Omarigliptin or Saxagliptin or Sitagliptin or Teneligliptin or Vildagliptin AND o randomized controlled trial[Publication Type] OR controlled clinical trial[Publication Type] OR randomized[Title/Abstract] OR placebo[Title/Abstract] OR drug therapy[MeSH Subheading] OR randomly[Title/Abstract] OR trial[Title/Abstract] OR groups[Title/Abstract] NOT o ((animals[mh]) NOT humans[mh]))
CENTRAL: o Diabetes mellitus or type 2 diabetes or type ii diabetes o (dipeptidyl peptidase IV inhibitor* or dipeptidyl-peptidase 4 inhibitor* or DPP4 inhibitor* or DPP 4 inhibitor or DPP IV inhibitor) OR o Alogliptin or Anagliptin or Gemigliptin or Linagliptin or Omarigliptin or Saxagliptin or Sitagliptin or Teneligliptin or Vildagliptin
© 2018 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 10/02/2021 eMethods 2: Drug doses Data only extracted and analysed for study groups receiving either marketed drug doses or doses evaluated in phase 3 clinical trials SGLT-2 inhibitors o Empagliflozin total daily dose 10 to 25mg PO o Canagliflozin total daily dose 100 to 300mg PO Dapagliflozin total daily dose 5 to 10mg PO o Ipragliflozin total daily dose 50 to 100mg PO o Luseogliflozin total daily dose 2.5 to 5mg PO o Remogliflozin excluded as no phase 3/marketing dose o Tofogliflozin excluded as no phase 3/marketing dose Ertugliflozin total daily dose 5 to 10mg PO GLP-1 analogues o Dulaglutide 0.75-1.5mg SC once weekly o Semaglutide 0.5-1mg SC once weekly o Liraglutide 1.2 to 1.8mg SC daily o Lixisenatide 10 to 20mcg SC daily o Taspoglutide 10 to 20mg SC once weekly o Exenatide 5 to 10mcg SC twice daily o Albiglutide 30 to 50mg SC once weekly DPP-4 inhibitors o Alogliptin total daily dose 12.5 to 25mg PO o Saxagliptin total daily dose 2.5 to 5mg PO o Sitagliptin total daily dose 25 to 100mg PO o Linagliptin total daily dose 5mg PO o Vildagliptin total daily dose 50 to 100mg PO o Omarigliptin excluded as no phase 3/marketing dose o Tenelegliptin excluded as no phase 3/marketing dose o Gemigliptin excluded as no phase 3/marketing dose o Evogliptin excluded as no phase 3/marketing dose
© 2018 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 10/02/2021 eMethods 3: Event definitions All-cause mortality: Death from any cause occurring during the duration of the randomised and controlled period.
Definitions of hypoglycaemia were: - Any: any episode not meeting criteria for major hypoglycaemia, study-defined minor hypoglycaemia - Major: any episode resulting in hospitalization, medical assistance or trial withdrawal, study-defined major or serious hypoglycaemia Note: Trial definitions of hypoglycaemia were variable and did not necessarily require recording of capillary blood glucose levels.
Adverse event category (any, serious, leading to trial withdrawal) were based on individual trial definitions.
From NCT registry: Additional events were identified from the NCT database trial entry for studies which had NCT reporting. The following are a list of definitions for events.
Myocardial infarction: • Acute myocardial infarction • Myocardial infarction • Acute coronary syndrome • Coronary artery occlusion Stroke: • Ischaemic stroke • Cerebrovascular accident • Transient ischaemic attack • Lacunar infarction • Brainstem infarction • Brainstem stroke • Cerebral infarction Heart failure: • Cardiac failure congestive • Cardiac failure • Cardiac failure acute • Left ventricular failure • Acute left ventricular failure • Cardiogenic shock • Congestive cardiomyopathy
© 2018 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 10/02/2021 Note: For heart failure events it was unclear from reporting if patients required hospitalization. Therefore, the outcome is for heart failure events and not heart failure hospitalisation. Analysis of heart failure hospitalisation events in cardiovascular outcome trials is presented in eTable 6 of the Supplement.
© 2018 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 10/02/2021 eMethods 4: Detailed statistical methods
Several studies included in this analysis had open-label extensions for additional safety analysis. For these studies, data was extracted only from the randomized controlled phase to minimise bias.
Studies with multiple treatment group were incorporated into the network as multi- arm studies where appropriate, allowing for multiple pairwise comparisons from a single study to be incorporated into the network without duplication of data. In cases where this was not appropriate (for example where study groups contained different baseline anti-diabetic medication), these treatment groups were incorporated as independent pairwise comparisons, such that no data was duplicated.
Bayesian network meta-analysis
A Bayesian network meta-analysis approach was undertaken1 using the gemtc package in R (version 3.4.1)2 and JAGS (version 4.3.0)3.
Deviance information criterion was adopted to determine the model-used (fixed vs. random-effects) for individual outcomes4 using the approach in Kew et al5. For each outcome, the model with the lowest DIC value was used. Using this approach, fixed- effects model was used for primary and secondary outcomes, and random-effects model was used for safety outcomes. The DIC values are listed in the eTable 1 of the Supplement.
For studies reporting hazard ratio, the log hazard ratio and corresponding standard error were extracted for inclusion in the analysis. For studies that did not report hazard ratio, the number of events and length of follow-up (in person-years) for each group were extracted to permit comparison of studies with different durations of follow-up. It should be noted that this assumes a constant event rate over time in these trials.
For each outcome, a network was generated incorporating all studies that reported at least one event in at least one study group. Studies in which no events occurred were excluded from the network, as these studies do not contribute to the treatment effect estimate1.
© 2018 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 10/02/2021 Consistency models were generated from each network using a Poisson likelihood and log link, incorporating both data from event counts and from hazard ratios. Non- informative vague priors were used for all parameters. The Markov Chain Monte Carlo (MCMC) method was used with 5000 adaptation iterations followed by 100,000 iterations of 4 chains. Convergence was assessed using the Brooks-Gelman-Rubin, using a cut-off of 1.056.
Unless otherwise specified, all data are presented as hazard ratio (HR) with 95% credible intervals (CrI). 95% CrI that do not cross 1 were taken to represent statistical significance.
I2 statistic was used to determine heterogeneity, with a cut-off of ≤0.25 used to determine low heterogeneity7.
Inconsistency was evaluated by node-splitting analysis8, and quantitative assessment of inconsistency was achieved by comparing the difference in direct and indirect estimates using the following formula9:
Δ = � − �′