Risk of Any Hypoglycemia with New Antihyperglycemic Agents in Patients with : A Systematic Review and Meta-Analysis

by

Sanaz Kamalinia

A thesis submitted in conformity with the requirements for the degree of Master of Science Institute of Medical Sciences University of Toronto

© Copyright by Sanaz Kamalinia (2019)

Risk of Any Hypoglycemia with New Antihyperglycemic Agents in Patients with Type 2 Diabetes: A Systematic Review and Meta- Analysis

Sanaz Kamalinia

Master of Science

Institute of Medical Sciences University of Toronto

2019 Abstract

Background : Evaluation of hypoglycemia risk relative to placebo with new

antihyperglycemic agents (AHA) including the dipeptidyl peptidase-4 inhibitors (DPP4i),

glucagon-like peptide-1 receptor agonists (GLP1RA) and sodium-glucose co-transporter-

2 inhibitors (SGLT2i) remains inconclusive.

Objective: This systematic review and meta-analysis aimed to assess risk of any and

severe hypoglycemia with new AHA relative to placebo by excluding studies with

background and .

Methods: Randomized, placebo-controlled studies, 12 weeks or greater in duration were

considered for inclusion. Studies allowing background use of any other AHA, apart from

, were excluded. This study is registered with PROSPERO

(CRD42018095458).

Results: 141 studies included in the meta-analysis demonstrate that relative to placebo,

risk of any and severe hypoglycemia did not significantly differ for any new AHA.

ii

Acknowledgments

First and foremost, I wish to express my sincere gratitude to my program advisor committee members. To my supervisor Dr Tobe, thank you for accepting me as your student and presenting me with this challenge. I thoroughly enjoyed it. Especially given your positive words of encouragement and insightful guidance for every step of this journey. To my program advisory committee members, Drs Robert Josse and Baiju Shah, I am so grateful for the collaboration. As icons in your field, thank you for instilling your trust in me.

A special note of appreciation to my independent reviewers Lindsay Leduc and Patrick J. Donio for taking on this massive project while in medical school. Their dedication and commitment to detail were key to the success of this analysis. It is reassuring to know the next generation of physicians will be of such high caliber.

Finally, to my family. My husband and best friend Fardin, thank you for never questioning and encouraging my mid-life decision to quit my job and return to school. To my precious children, son Kayan and daughter Chloe, I am so blessed for your unconditional love. I hope this will encourage you to never stop learning.

iii

Contributions

My supervisor, Dr Sheldon Tobe, originally conceived of the research question for the meta- analysis. My program advisor committee members including Dr Josse and Dr Shah provided instrumental support and guidance on the protocol, methodology and interpretation of the study results. They also provided valuable feedback on this thesis. Lindsay Leduc was an independent reviewer for study screening and selection phase. Pat Donio was an independent reviewer for the data extraction phase. I am truly grateful for everyone’s dedication and collaboration.

iv

Table of Contents Table of Contents

Acknowledgments...... iii

Contributions...... iv

Table of Contents ...... v

List of Tables ...... xix

List of Figures ...... xx

List of Abbreviations ...... xxi

List of Appendices ...... xxii

INTRODUCTION ...... 1

TYPE 2 DIABETES ...... 3

2.1 EPIDEMIOLOGY ...... 3

2.1.1 Global ...... 3

2.1.2 Canada...... 3

2.1.3 United States of America ...... 4

2.1.4 Rest of the World ...... 5

2.2 PATHOPHYSIOLOGY ...... 5

2.3 RISK FACTORS ...... 6

2.3.1 Multifactorial ...... 6

2.3.2 Adiposity/Obesity ...... 6

2.3.3 Diet ...... 7

2.3.4 Particulate Matter ...... 8

2.3.5 Race...... 9

2.3.6 Sex/Gender ...... 9

2.4 COMPLICATIONS ...... 11 v

2.4.1 All-Cause Mortality ...... 12

2.4.2 Cardiovascular Disease ...... 12

2.4.3 Dementia ...... 13

2.4.4 Muscle Strength ...... 14

2.4.5 Falls ...... 15

2.4.6 Fractures ...... 16

2.4.7 Sex-Gender Differences in Complications ...... 18

2.4.8 Insurance / License / Employment ...... 19

2.5 COST ...... 19

2.5.1 Canada...... 19

2.5.2 USA...... 20

2.6 SUMMARY ...... 20

MANAGEMENT ...... 21

3.1 PREVENTION ...... 21

3.1.1 Lifestyle Modification ...... 21

3.2 DIAGNOSIS ...... 22

3.2.1 Prediabetes ...... 22

3.2.2 Diabetes...... 23

3.3 GLYCEMIC TARGETS ...... 23

3.4 INTENSIVE VS LESS INTENSIVE GLYCEMIC CONTROL ...... 24

3.4.1 UKPDS ...... 25

3.4.2 ADVANCE ...... 26

3.4.3 VADT ...... 26

3.4.4 ACCORD ...... 27

3.4.5 Meta-analysis of Intensive Glucose Control Trials ...... 31

vi

3.4.6 Intensive Glucose Control in Acute Coronary Syndrome ...... 31

3.5 HbA1c and MORTALITY ...... 32

3.6 ATTAINMENT OF GLYCEMIC TARGETS...... 32

3.6.1 Canada...... 32

3.6.2 US ...... 32

3.6.3 Europe ...... 32

3.6.4 Glycemic Target Attainment By Hypoglycemia Risk ...... 33

3.7 TEMPORAL CHANGES IN AHA USE ...... 34

3.7.1 Canada...... 34

3.7.2 US ...... 34

3.8 COMORBIDITY ...... 35

3.9 SUMMARY ...... 35

HYPOGLYCEMIA ...... 36

4.1 CLINICAL SYNDROME ...... 36

4.2 GUIDELINE DEFINITIONS ...... 37

4.2.1 ADA ...... 38

4.2.2 Diabetes Canada...... 39

4.2.3 European Medicines Agency (EMA) ...... 42

4.2.4 IDF ...... 42

4.3 SYMPTOMS ...... 42

4.3.1 Autonomic Neurogenic ...... 43

4.3.2 Neuroglycopenic ...... 43

4.3.3 Malaise ...... 43

4.4 PHYSIOLOGY ...... 44

4.4.1 Counterregulatory Responses in People Without Diabetes ...... 44

vii

4.4.2 Counterregulatory Responses in Patients with T1DM...... 46

4.4.3 Counterregulatory Responses in Patients with T2DM...... 47

4.4.4 Hypoglycemia-Associated Autonomic Failure ...... 49

4.5 PROTECTIVE EFFECT OF HAAF ...... 50

4.5.1 HAAF on All-Cause Mortality ...... 50

4.5.2 HAAF on Preconditioning of the Brain ...... 51

4.6 IMPACT OF AVOIDING HYPOGLYCEMIA ...... 51

4.7 HETEROGENEITY OF HYPOGLYCEMIA ...... 52

4.7.1 Definitions...... 52

4.7.2 Study Type ...... 52

4.7.3 Study Duration ...... 53

4.7.4 Study Eligibility ...... 53

4.7.5 Patient Attrition ...... 53

4.7.6 Ascertainment ...... 54

4.7.7 Patient Recall ...... 54

4.7.8 Patient Reluctance to Notify / Admit ...... 55

4.7.9 Exclusion of Study Participants During Screening or Active-Trial Period ...... 55

4.8 GUIDANCE FOR IMPROVING REPORTING OF HYPOGLYCEMIA IN THE CLINICAL LITERATURE ...... 55

4.9 EPIDEMIOLOGY ...... 56

4.9.1 Intensive vs Less Intensive Glucose Lowering RCTs ...... 56

4.9.2 Incidence in Patients with T1DM vs T2DM ...... 57

4.9.3 Incidence of Severe Hypoglycemia ...... 58

4.9.4 Incidence of Non-Severe Hypoglycemia ...... 59

4.9.5 Incidence of Nocturnal Hypoglycemia ...... 60

4.9.6 Incidence of Hypoglycemia Unawareness ...... 60 viii

4.9.7 Duration of Hypoglycemia ...... 62

4.9.8 Incidence with SMBG...... 63

4.9.9 Incidence with CGM ...... 63

4.9.10 CGM Incidence with Sulfonylureas...... 63

4.9.11 Real-World Incidence ...... 64

4.9.12 Real-World Incidence with Insulin ...... 64

4.9.13 Real-World Incidence with Sulfonylureas ...... 66

4.9.14 Incidence of Hypoglycemia in the Cardiovascular Outcomes Trials ...... 66

4.9.15 Incidence in Racial or Ethnic Groups ...... 67

4.9.16 Incidence by Frailty ...... 68

4.9.17 Regional Incidence...... 68

4.10 TEMPORAL TRENDS ...... 69

4.10.1 Canada...... 69

4.10.2 US ...... 69

4.10.3 Europe ...... 70

4.10.4 Asia ...... 70

4.11 RISK FACTORS ...... 71

4.11.1 Age ...... 71

4.11.2 Antecedent Hypoglycemia ...... 71

4.11.3 Cardiovascular Disease and Hypoglycemia...... 72

4.11.4 Coronary Artery Disease and Hypoglycemia ...... 73

4.11.5 Chronic Kidney Disease and Hypoglycemia ...... 73

4.11.6 Clinical Complexity / Frailty Status and Hypoglycemia ...... 73

4.11.7 Dementia and Hypoglycemia ...... 74

4.11.8 Depression and Hypoglycemia ...... 74

ix

4.11.9 Duration of Diabetes and Hypoglycemia ...... 75

4.11.10 Duration of Treatment and Hypoglycemia ...... 75

4.11.11 Food Insecurity and Hypoglycemia ...... 75

4.11.12 HbA1c Levels and Hypoglycemia ...... 75

4.11.13 Glycemic Variability ...... 78

4.11.14 Health Literacy ...... 79

4.11.15 Seasonal Predictors ...... 79

4.11.16 Sex ...... 80

4.12 CONSEQUENCES OF HYPOGLYCEMIA ...... 80

4.12.1 Attainment of HbA1c Target ...... 80

4.12.2 Increased Self-Monitoring ...... 80

4.12.3 Impact on Adherence to Therapy ...... 80

4.12.4 Dose Reduction ...... 81

4.12.5 Dosing Irregularities / Missed Doses ...... 81

4.12.6 Patient Fear ...... 81

4.12.7 Physician Fear ...... 83

4.12.8 Family Members Worry ...... 83

4.12.9 Depression...... 84

4.12.10 QoL ...... 84

4.12.11 Motor Vehicle and Other Accidents ...... 85

4.12.12 Social Implications ...... 85

4.12.13 Employment Implications ...... 86

4.12.14 Work Productivity and Burden of Non-Severe Hypoglycemia ...... 87

4.12.15 Burden of Non-Severe Nocturnal Hypoglycemia ...... 87

4.13 COMPLICATIONS OF HYPOGLYCEMIA ...... 88

x

4.13.1 Mortality Risk ...... 88

4.13.2 Biomarkers of Severe Hypoglycemia and Death (ACCORD) ...... 92

4.13.3 Prediction of All-Cause Mortality in Patients with Hypoglycemia ...... 93

4.13.4 MACE (CV Death, Nonfatal MI, Stroke) Risk Following Hypoglycemia...... 93

4.13.5 Progression of Atherosclerosis ...... 99

4.13.6 Blood Coagulation Abnormalities ...... 103

4.13.7 Inflammation ...... 103

4.13.8 Cognitive Function...... 103

4.13.9 Dementia ...... 104

4.13.10 Renal decline ...... 105

4.13.11 Falls and Fall-Related Fractures ...... 105

4.14 COST ...... 106

4.14.1 Cost of Severe Hypo, including Hospitalization...... 106

4.14.2 Cost of Non-Severe Hypoglycemia ...... 108

4.15 PERFORMANCE MEASURES ...... 108

4.16 SUMMARY ...... 109

Antihyperglycemic Agents ...... 111

5.1 METFORMIN ...... 112

5.1.1 Mechanism of Action ...... 112

5.1.2 UKPDS 34 ...... 112

5.1.3 Risk of All-Cause Mortality and MACE ...... 113

5.1.4 Efficacy ...... 114

5.1.5 Safety & Tolerability ...... 114

5.2 INSULIN ...... 114

5.2.1 Mechanism of Action ...... 114

xi

5.2.2 Risk of All-Cause Mortality and MACE ...... 115

5.2.3 Efficacy ...... 116

5.2.4 Safety & Tolerability ...... 116

5.2.5 Risk of Hypoglycemia by Insulin Type ...... 116

5.2.6 Risk of Hypoglycemia vs Active Controls ...... 117

5.2.7 Risk of Hypoglycemia Co-administered with DPP4i ...... 118

5.2.8 Risk of CV Event after Hypoglycemia ...... 119

5.2.9 Risk of All-Cause Mortality after Hypoglycemia in Insulin-treated Patients ...... 119

5.3 ...... 120

5.3.1 Mechanism of Action ...... 120

5.3.2 Risk of All-Cause Mortality...... 121

5.3.3 Risk of CV Death ...... 121

5.3.4 Characteristics of Different SU’s ...... 122

5.3.5 Efficacy ...... 123

5.3.6 Risk of Hypoglycemia - vs Glyburide vs and 123

5.3.7 Risk of Hypoglycemia vs Active Controls ...... 124

5.3.8 Sex-Gender Differences in Hypoglycemia Risk ...... 124

5.3.9 Risk of MACE after Hypoglycemia ...... 124

5.3.10 Healthcare Utilization SU vs other therapies ...... 125

5.4 INCRETIN ...... 125

5.4.1 The Incretin Effect in Healthy Subjects ...... 125

5.4.2 The Incretin Effect in Patients with T2DM ...... 126

5.4.3 Incretin-Based Therapies for Patients with T2DM ...... 127

5.5 GLUCAGON LIKE PEPTIDE-1 RECEPTOR AGONISTS ...... 127

5.5.1 Mechanism of Action ...... 127

xii

5.5.2 Risk of All-Cause Mortality...... 128

5.5.3 Risk of MACE ...... 128

5.5.4 Efficacy ...... 129

5.5.5 Safety & Tolerability ...... 129

5.5.6 Hypoglycemia Counterregulation with GLP1 Infusion ...... 129

5.5.7 Hypoglycemia Counterregulation with GLP1RA ...... 130

5.5.8 Combination with Insulin ...... 132

5.5.9 Risk of Hypoglycemia in CVD ...... 133

5.5.10 Risk of MACE after Hypoglycemia ...... 133

5.6 DIPEPTIDYL PEPTIDASE IV INHIBITORS ...... 134

5.6.1 Mechanism of Action of Glucose Lowering ...... 134

5.6.2 Risk of MACE ...... 134

5.6.3 Efficacy ...... 134

5.6.4 Safety and Tolerability ...... 135

5.6.5 Risk of Pancreatitis ...... 135

5.6.6 Hypoglycemia Counterregulation with DPP4i ...... 135

5.6.7 Risk of Hypoglycemia vs Active Controls ...... 139

5.6.8 Risk of Hypoglycemia Co-administered with Sulfonylurea vs Placebo ...... 140

5.6.9 Risk of Hypoglycemia Co-administered with Insulin ...... 140

5.6.10 Risk of Hypoglycemia Co-administered with Insulin vs Placebo ...... 140

5.6.11 Sex-Gender MACE Outcomes with DPP4i ...... 141

5.6.12 Risk of Hypoglycemia in RCTs of CKD ...... 141

5.6.13 Risk of Hypoglycemia in SRMA of CKD ...... 142

5.6.14 Risk of Hypoglycemia in CVD ...... 142

5.6.15 Risk of MACE after Hypo ...... 144

xiii

5.7 SODIUM GLUCOSE COTRANSPORT 2 INHIBITORS ...... 147

5.7.1 Mechanism of Action ...... 147

5.7.2 Risk of All-Cause Mortality...... 149

5.7.3 Efficacy ...... 150

5.7.4 Safety and Tolerability ...... 151

5.7.5 Hypoglycemia Counterregulation with SGLT2i ...... 154

5.7.6 Risk of Hypoglycemia vs Active Controls ...... 154

5.7.7 Risk of Hypoglycemia Co-administered with Sulfonylurea vs Placebo ...... 155

5.7.8 Risk of Hypoglycemia Co-administered with Insulin ...... 155

5.7.9 Risk of Hypoglycemia Co-administered with Incretin-based therapies ...... 156

5.7.10 Risk of MACE after Hypoglycemia ...... 156

5.8 SUMMARY ...... 157

RATIONALE ...... 160

6.1 DPP4i ...... 160

6.1.1 Data from RCTs of DPP4i ...... 160

6.2 GLP1RA ...... 161

6.2.1 Data from RCTs of GLP1RA ...... 161

6.2.2 Data from SRMA of GLP1RA ...... 162

6.2.3 Data from Network Meta-Analyses of GLP1RA...... 162

6.3 SGLT2i ...... 162

6.3.1 Data from RCTs of SGLT2i ...... 162

6.3.2 Data from SRMA of SGLT2i ...... 163

6.4 Combination therapy ...... 163

SCOPING REVIEW ...... 163

METHODS ...... 165

xiv

8.1 OBJECTIVES ...... 165

8.2 ELIGIBILITY CRITERIA ...... 165

8.2.1 Study Design ...... 165

8.2.2 Population ...... 165

8.2.3 Interventions ...... 166

8.2.4 Comparator ...... 166

8.2.5 Outcomes ...... 166

8.2.6 Exclusion Criteria ...... 167

8.3 SEARCH STRATEGY ...... 167

8.4 STUDY SELECTION ...... 168

8.4.1 Dealing with Duplicate and Companion Publications ...... 168

8.5 DATA EXTRACTION ...... 168

8.6 RISK OF BIAS ...... 169

8.6.1 Selection Bias: Random Sequence Generation ...... 170

8.6.2 Selection Bias: Allocation Concealment ...... 170

8.6.3 Performance Bias: Blinding of Participants and Personnel ...... 170

8.6.4 Detection Bias: Blinding of Outcome Assessment ...... 170

8.6.5 Attrition Bias: Incomplete Outcome Data ...... 170

8.6.6 Reporting Bias: Selective Reporting of Outcomes ...... 171

8.6.7 Other Bias: Confounding by Rescue ...... 171

8.7 DATA SYNTHESIS ...... 171

8.8 STATISTICAL ANALYSIS ...... 171

8.8.1 STATISTICAL METHOD ...... 171

8.8.2 EFFECT MEASURE ...... 173

8.8.3 ANALYSIS METHOD (FEM or REM) ...... 174

xv

8.8.4 BOTH ARM ZERO EVENTS (BA0E) ...... 175

8.8.5 Unit of Analysis ...... 177

8.8.6 Assessment of Heterogeneity ...... 178

8.8.7 Assessment of Reporting Bias ...... 178

8.8.8 Quality of Evidence ...... 178

RESULTS ...... 178

9.1 SEARCH RESULTS ...... 178

9.1.1 Prisma Flow ...... 178

9.1.2 Excluded Studies (that appeared at first to be eligible) ...... 182

9.2 CHARACTERISTICS OF INCLUDED STUDIES ...... 182

9.2.1 Study Duration ...... 182

9.2.2 Unit of Analysis ...... 182

9.2.3 Incomplete Outcome Data ...... 182

9.2.4 Funding ...... 182

9.3 CHARACTERISTIC OF THE STUDY POPULATION ...... 183

9.3.1 Number of Participants ...... 183

9.3.2 Mean Age ...... 183

9.3.3 Gender ...... 183

9.3.4 Race/Ethnicity ...... 183

9.3.5 Duration of Diabetes ...... 183

9.3.6 Baseline HbA1c ...... 184

9.3.7 Change in HbA1c ...... 184

9.4 HETEROGENEITY OF HYPOGLYCEMIA ...... 185

9.4.1 Definition of Hypoglycemia ...... 185

9.4.2 Definition of Severe Hypoglycemia ...... 185

xvi

9.4.3 Exclusion of Patients Experiencing or With a History of Hypoglycemia ...... 188

9.4.4 Rescue Medication ...... 189

9.4.5 Inclusion of Hypoglycemia Reports After Rescue Medication ...... 189

9.5 RISK OF BIAS ...... 192

9.5.1 Selection Bias: Random Sequence Generation ...... 192

9.5.2 Selection Bias: Allocation Concealment ...... 192

9.5.3 Performance Bias: Blinding of Participants and Personnel ...... 193

9.5.4 Detection Bias: Blinding of Outcome Assessment ...... 193

9.5.5 Attrition Bias: Incomplete Outcome Data ...... 193

9.5.6 Reporting Bias: Selective Reporting of Outcomes ...... 193

9.5.7 Other Bias: Confounding by Rescue Medication ...... 193

9.6 HYPOGLYCEMIA OUTCOMES ...... 194

9.6.1 Primary Outcome: Any Hypoglycemia ...... 194

9.6.2 Secondary Outcome: Severe Hypoglycemia ...... 195

9.6.3 Sensitivity Analysis: Inclusion of Studies with Zero Any Hypoglycemia ...... 195

9.6.4 Sensitivity Analysis: Inclusion of Studies with Zero Severe Hypoglycemia .....197

9.6.5 Post Hoc Analysis: Impact on Results if Defining Severe Hypoglycemia as ≤3.1 mmol/L ...... 199

9.7 STATISTICAL HETEROGENEITY ...... 199

9.7.1 Tau² ...... 199

9.7.2 Chi² ...... 199

9.7.3 I² ...... 200

9.7.4 Overall Effect (Z) ...... 200

9.8 PUBLICATION BIAS ...... 200

9.9 SUMMARY OF FINDINGS ...... 203

9.10 DISCUSSION ...... 204 xvii

9.10.1 Limitations ...... 205

9.11 CONCLUSION ...... 207

9.11.1 Future Research ...... 207

IMPLICATIONS ...... 208

10.1.1 TO HCP ...... 208

10.1.2 TO PATIENT...... 208

10.1.3 TO SYSTEM ...... 208

References ...... 210

Appendices ...... 264

Copyright Acknowledgements...... 421

xviii

List of Tables

Table 1: Risk of Bias Classification ...... 169

Table 2: Kappa Estimate ...... 179

Table 3: Studies with Multiple Comparison ...... 181

Table 4: Distribution of Single Hypoglycemia Count in Shared Placebo Arm ...... 181

Table 5: Summary of Characteristics of Included Studies...... 184

Table 6: Heterogeneity of Hypoglycemia - Definitions ...... 186

Table 7: Heterogeneity of Hypoglycemia - Rescue Therapy ...... 190

Table 8: Summary of Studies with Zero Any Hypoglycemia...... 196

Table 9: Summary of Studies with Zero Severe Hypoglycemia...... 197

xix

List of Figures

Figure 1: Summary Forest Plot of Any Hypoglycemia ...... 194

Figure 2: Summary Forest Plot of Severe Hypoglycemia ...... 195

Figure 3: Funnel Plot of DPP4i Monotherapy - Any Hypoglycemia ...... 201

Figure 4: Funnel Plot of DPP4i + Metformin - Any Hypoglycemia ...... 201

Figure 5: Funnel Plot of SGLT2i Monotherapy - Any Hypoglycemia ...... 202

Figure 6: Funnel Plot of SGLT2i + Metformin - Any Hypoglycemia ...... 202

xx

List of Abbreviations

ACM All-Cause Mortality AHA Antihyperglycemic Agents BA0E Both Arm Zero Events BG Blood Glucose CGM Continuous Blood Glucose Monitoring CAD Coronary Artery Disease CKD Chronic Kidney Disease CV Cardiovascular CVD Cardiovascular Disease CVOTs Cardiovascular Outcome Trials DPP4 Dipeptidyl Peptidase-4 DPP4i Dipeptidyl Peptidase-4 Inhibitors DSE Diabetes Education and Support ED Emergency Department ER Emergency Room GLP1 Glucagon like Peptide-1 GLP2 Glucagon like Peptide-2 GLP1RA Glucagon like Peptide-1 Receptor Agonists HbA1c Glycated hemoglobin HbA1c HHF Hospitalization for Heart Failure HR Hazard Ratio IDF International Diabetes Federation ILI Intensive Lifestyle Intervention MACE Major Adverse Cardiovascular Events MAGE Mean Amplitude of Glycemic Excursions NAFLD Non-Alcoholic Fatty Liver Disease NSH Non-Severe Hypoglycemia NSNH Non-Severe Nocturnal Hypoglycemia OR Odds Ratio pbo Placebo RA Requiring Assistance (medical or third party) RCT Randomized Controlled Trial RD Risk Difference RR Relative Risk or Risk Ratio SGLT Sodium Glucose Co-Transporter SGLT1 Sodium Glucose Co-Transporter-1 SGLT2 Sodium Glucose Co-Transporter-2 SGLT2i Sodium Glucose Co-Transporter-2 Inhibitors SoF Summary of Findings SRMA Systematic Review and Meta-Analysis SU Sulfonylureas T1DM Type 1 Diabetes T2DM Type 2 Diabetes USA United States of America

xxi

List of Appendices

Appendix I: Scoping Review Flow Diagram ...... 264

Appendix II: PRISMA 2015 Checklist ...... 265

Appendix III: Medline OVID Search Strategy ...... 268

Appendix IV: PRISMA Flow Diagram ...... 275

Appendix V: Excluded Studies ...... 276

Appendix VI: Characteristics of Included Studies ...... 282

Appendix VII: Forest Plots of Any Hypoglycemia ...... 392

Appendix VIII: Forest Plots of Severe Hypoglycemia ...... 398

Appendix IX: Forest Plots of Including Studies with Zero Any Hypoglycemia ...... 399

Appendix X: Forest Plots Including Studies with Zero Severe Hypoglycemia ...... 405

Appendix XI: Post hoc Analyses if Severe Hypoglycemia was defined as ≤3.1 mmol/L ...... 411

Appendix XII: Summary of Findings (SoF) Table ...... 416

xxii

INTRODUCTION

Diabetes is one of the greatest healthcare challenges of our era. Achieving glycemic control is the cornerstone of diabetes management, limited by fear of hypoglycemia. Patients who experience an episode of hypoglycemia report feeling lousy and are less likely to adhere to therapy. More severe episodes of hypoglycemia can lead to loss of consciousness and death.

For patients with type 2 diabetes (T2DM), three newer classes of antihyperglycemic agents (AHA) have increased in popularity and offer a low risk of hypoglycemia based on their glucose- dependent mechanism of action. These agents include the dipeptidyl peptidase-4 inhibitors (DPP4i), the glucagon like peptide-1 receptor agonists (GLP1RA) and the sodium glucose co- transporter2 inhibitors (SGLT2i). Relative to sulfonylureas (SU) and , a lower risk of hypoglycemia with newer classes of AHA has been demonstrated. However, it is unclear how these new AHA compare relative to placebo (pbo).

Most meta-analyses are efficacy-focused and assessment of hypoglycemia risk is often a secondary or tertiary endpoint. Moreover, use of background therapies such as SU or insulin increase estimates of hypoglycemia risk with new AHAs, making it difficult to estimate risk on their own. In meta-analyses of new antihyperglycemic agents compared to placebo, when an increased risk of hypoglycemia is observed, post hoc sensitivity analyses are often conducted to tease out studies which permit use of agents known to increase risk, such as SU and insulin. This is less than ideal given the importance of hypoglycemia for optimal diabetes management.

Thus, the objective of this systematic review and meta-analysis (SRMA) is to improve the evaluation of any and severe hypoglycemia risk with new AHA in patients with T2DM relative to placebo, by a priori excluding trials allowing use of any other therapy, apart from metformin. Risk of severe hypoglycemia is expected to be very low given the glucose mechanism of action with new AHA, yet still important to capture as a secondary endpoint.

This thesis will first provide a number of background chapters prior to presenting the SRMA. The background chapters review the latest evidence related to the patients, outcome and intervention of interest for this meta-analysis. In chapter 2 and 3, we review the disease of diabetes and its current clinical management. In chapter 4, our outcome of interest, 1

hypoglycemia, is examined, from physiology to epidemiology. The interventions of interest are debriefed in chapter 5, where we explore the mechanism of action of each class of AHA, a brief overview of their efficacy and overall safety as well as evidence of hypoglycemia risk from clinical trials. Finally, we delve into the rationale, scoping review, methods, results and potential implication of this SRMA from chapter 6 to 10, respectively.

2

TYPE 2 DIABETES

The systematic review and meta-analysis presented in the second half of this thesis involves new antihyperglycemic agents indicated in many countries globally for patients with type 2 diabetes. As such, we begin with a review of type 2 diabetes, examining the latest evidence relating to diabetes epidemiology, pathophysiology, complications and consequences of those suffering the disease.

2.1 EPIDEMIOLOGY

2.1.1 Global

The number of people living with diabetes has soared from 153 million in 1980, to 347 million in 2008 (Danaei et al. 2011). In 2017, 451 million people living were living with diabetes, representing 8.4% of the world’s population. By 2045, 9.9% or 629 million people will be afflicted. The Western Pacific region is home to the highest number of people globally living with diabetes at 37%. 79% of the global population living with diabetes reside in low to middle income countries (Cho et al. 2018).

Most countries do not have data sources for impaired glucose tolerance (IGT), gestational or undiagnosed cases. Nevertheless, it is estimated that half (49.7%) of the world’s population with diabetes remain undiagnosed. Not surprisingly, 81% of undiagnosed cases reside in low and middle-income countries. For instance, the highest number of undiagnosed cases is believed to reside in Africa (69.2 %), followed by South East Asia (57.5%) and then the Western Pacific region (54%). North America and the Caribbean represent the lowest rates of undiagnosed diabetes at 37.6%, followed closely by Europe at 37.8% (Cho et al. 2018).

2.1.2 Canada

The number of people diagnosed with diabetes in Canada has drastically increased in the last two decades. In 2000, 1.3 million (4.2%) Canadians were diagnosed with diabetes. In 2010, this number increased to 2.5 million (7.3%). By 2020, it is estimated that 3.7 million of the Canadian population (or 9.9%) will be diagnosed with diabetes (Pelletier et al. 2012).

3

In New Brunswick between 2001 and 2014, prevalence of T2DM increased by 120%. Coincidentally, during this period, the consumption of fruits and vegetables decreased while the prevalence of obesity, hypertension, prediabetes, alcohol consumption, immigration and urbanization increased. Contradictory to other reports, this study did not find that physical activity, smoking, socioeconomic status and education contributed to the increasing diabetes prevalence. Interestingly, those born in the youngest cohort (1961-1970) demonstrated a drastic increase compared to those born earlier. However, detection effect may have biased these results, as authors noted that the number of HbA1c tests also increased during this period. Of note was a dramatic decrease in mortality of patients with diabetes, from 42.0 per 1000-person- years in 2001 to 25.1 per 1000-person-years (Thibault et al. 2016).

In 2011, based on blood sample analyses, 20% of Canadians were living with the disease but undiagnosed (Pelletier et al. 2012). Estimates of diabetes differ depending on the diagnostic criteria used. In a study using national representative samples with biospecimen measures from Ontario, the prevalence of undiagnosed T2DM and diabetes was significantly higher using HbA1c compared with FPG levels (Rosella et al. 2016).

2.1.3 United States of America

A number of US sources have reported a stabilization of the incidence of diabetes in recent years (Geiss et al. 2014; Weng et al. 2016) https://www.cdc.gov/diabetes/pdfs/data/2014-report- estimates-of-diabetes-and-its-burden-in-the-united-states.pdf . Data from n=664,969 adults 20-79 years of age was analyzed for the period between 1980-2012. In the 1980’s, a stable age- adjusted prevalence and incidence of diagnosed diabetes was observed. Between 1990 and 2008 however, the prevalence increased from 3.5 to 7.9 per 100 persons, and the incidence from 3.2 to 8.8 per 1000 persons years. A levelling off was noted between 2008 and 2012 in most individuals, except non-Hispanic blacks and Hispanics. In addition, rate of increase was highest for adults with a high school education or less (Geiss et al. 2014). Also of concern, is the 26% of Americans aged 65 and older living with diagnosed or undiagnosed diabetes http://www.diabetes.org/diabetes-basics/statistics/.

4

2.1.4 Rest of the World

The most dramatic rise in the number of people with diabetes exists in developing countries. For example, in South Asia, the rise is predicted to be more than 150% from 2000 to 2035. The majority of people with diabetes live in Asia, with almost half in India and China (Nanditha et al. 2016). The Middle East has the highest prevalence of diabetes at 11% and China will soon match this. In 1995, the prevalence of diabetes in China was 2.0%. By 2009, this number more than quadrupled to 9.7% (Tuomi et al. 2014).

2.2 PATHOPHYSIOLOGY

Originally, the pathophysiology of T2DM was described as a combination of insulin resistance in the muscle and liver coupled with beta-cell failure (DeFronzo 1988). Twenty-years later, more complex details about the etiology of T2DM have emerged. The roles of five additional organs have been added to the trifecta of pancreatic beta-cell, liver and muscle. This octet now includes adipose tissue for its role in increased lipolysis, gastrointestinal tract for incretin deficiency, alpha cell for hyperglucagonemia, kidney for increased glucose reabsorption and the brain for insulin resistance (Defronzo 2009).

(Reproduced with permission)

5

2.3 RISK FACTORS

2.3.1 Multifactorial

An umbrella review of 86 meta-analyses found a wide range of biomarkers, conditions, diet, lifestyle, environmental and psychosocial factors to be related to an increased risk of diabetes (Bellou et al. 2018).

(Permission not required – Open Access)

2.3.2 Adiposity/Obesity

When discussing risk factors for diabetes, adiposity or obesity is often the first that comes to mind. Obesity is associated with many chronic diseases of our time such as diabetes and cancer (Heymsfield and Wadden 2017). Although the risk factor for developing diabetes are complex, multifactorial and intersectional, the correlation to obesity is staggering.

6

http://www.cdc.gov/diabetes/data

(Permission not required – Open access)

In Canada, 47.5% of individuals with diabetes were obese compared to 19.1% without diabetes (Pelletier et al. 2012). The astonishing increase in the prevalence of diabetes in China may also be related to increases in BMI. The BMI at time of diagnosis in patients with T2DM in Europe and the US is usually 30 kg/m², whereas in China, the mean value is 25.9 kg/m² (Tuomi et al. 2014). The Nurses Health Study illustrated that even at the same BMI, Asians have more than double the risk of developing diabetes than Caucasians (Shai et al. 2006). One explanation for this paradox has been that Asians have more abdominal obesity than Caucasians. The IDF, which acknowledges this discrepancy, has reduced the waist circumference cut off for Asians for the criteria of metabolic syndrome (Alberti, Zimmet, and Shaw 2005).

2.3.3 Diet

In most developed countries, there has been a global shift in energy source to ultra-processed foods rich in sugar and saturated fats. This shift has been attributed to the obesity epidemic and its complications such as diabetes (Zobel et al. 2016). There is now strong evidence to support benefits of increasing intake of fruits, non-starchy vegetables, nuts, legumes, fish, vegetable oils, 7

yogurt and minimally processed whole grains. Similarly, reducing consumption of red meats, processed or sodium preserved meats, foods rich in refined grains, starch, added sugars salt and trans fats is recommended. However, more data is required to inform on the effects of phenolics, dairy fat, probiotics, fermentation, coffee, tea, cocoa, eggs, specific vegetable and tropical oils, vitamin D, individual fatty acids and diet microbiome interactions. Evidence to support benefits of local, organic, grass-fed, farmed/wild or non-genetically modified foods is also limited (Mozaffarian 2016).

A recent study has described how the North American Free Trade Agreement of January 1st , 1994 has drastically increased the supply and consumption of high fructose corn syrup in Canada. Since the tariff reductions, the daily consumption of sweeteners increased by 42 kilocalorie per capita, from 21.2 kcal to 62.9 kcal post NAFTA. The comparator, Australia and the UK did not show a similar increase, and results remained significant even with additional sensitivity analyses (Barlow et al. 2017).

Animal and prospective observational studies have demonstrated a strong association between fructose sugars and weight gain, diabetes and cardiovascular disease. Meta-analyses of randomized controlled trials, however, have been unable to tease out the putative effects of fructose in comparison to other carbohydrates, mainly due to an inability to control for energy. Experts recommend considering the entire diet as opposed to single energy sources (Khan and Sievenpiper 2016).

2.3.4 Particulate Matter

The Normative Aging study, a long-term prospective cohort of non-diabetic individuals was established in 1963 in the Greater Boston area. Every 3-5 years, participants undergo brief examinations by questionnaires and blood samples. Starting from 2000, estimated concentrations of particulate matter with aerodynamic diameter of ≤2.5 µm were obtained from a hybrid spatiotemporal prediction model in 656 participants. They found that among these non- diabetics, particulate matter ≤2.5 µm were significantly associated with higher fasting blood glucose (BG) (Peng et al. 2016).

8

2.3.5 Race

Around the world, Indigenous people have some of the highest rates of T2DM. In Canada, diabetes was virtually unknown in the Canadian Indigenous population prior to 1940’s (Young and Roche 1990). With the loss of traditional lifestyles following the government imposed reserve system and the impact of residential schools, rates of diabetes dramatically increased. While the age-standardized prevalence of the general population is 5%, First Nations individuals living on reserves is 17.2% and 10.3% for those living off-reserve. Among Métis it is 7.3% (Pelletier et al. 2012). Moreover, Indigenous individuals are diagnosed with diabetes an average of 14 years earlier compared to non-Indigenous people. The longer exposure to diabetes leads to more severe complications, such as retinopathy, neuropathy/amputation and end-stage renal disease (Crowshoe et al. 2018).

South Asians have also been identified as a population at increased risk for developing diabetes, whether they have migrated to a westernized country or not. In the CANHEART Study, a big data initiative from Ontario, Canada evaluating over 800,000 first generation immigrants, found that South Asians have the highest age-standardized risk for diabetes at 15.0, followed by Blacks at 12.3 and compared to all immigrants at 8.8 (Tu et al. 2015).

The Nurses Health Study was a large US prospective cohort from 1980-2000 of n=75,584 white women and n=801 Asians, n=613 Hispanics and n=1,421 blacks. They reported on the risk of developing T2DM based on ethnicity and BMI. After a 20 year follow up (1,294,799 person- years), the risk of developing diabetes was higher in Asians, Hispanics and Blacks compared to Whites, before and after adjusting for BMI. Weight gain was also found to be most detrimental in Asians on risk of developing diabetes (Shai et al. 2006). Unlike most individuals with diabetes, East Asians are reported to have a defect in insulin secretion, rather than insulin resistance, eluding again to the complex heterogeneous nature of diabetes (Nauck and Meier 2016).

2.3.6 Sex/Gender

According to the Canadian Institute of Health Research (CIHR), sex is defined as the biological attributes of males and females, such as the physical and physiological features. Gender on the other hand refers to socially constructed roles. For instance, gender reflects the behaviours, 9

expression and identities of people ( http://www.cihr-irsc.gc.ca/e/50836.html ). However, in medicine, sex and gender cannot be distinguished and some prefer the term “Sex-Gender” (Campesi et al. 2017). Sex-specific risk factors for diabetes is an evolving field and points to clinically important differences. Both physiological, pathological and psychosocial differences between the sexes contribute to clinical risk factors, symptomology, development and complications of diabetes.

Physiological differences include women typically having greater hypothalamus-pituitary axis (HPA Axis) activity, arcuate nucleus proopiomelanocortin (ARC POMC), increased adiponectin and leptin, increased central leptin sensitivity, decreased energy expenditure and less muscle mass. On the other hand, men typically have higher food intake, decreased brown adipose tissue (BAT) mass and activity, higher visceral fat mass, more fatty liver and decreased peripheral insulin sensitivity (Kautzky-Willer, Harreiter, and Pacini 2016).

Pathological differences include women having a higher incidence of impaired glucose tolerance (IGT), whereas men have a higher incidence of impaired fasting glucose (IFG). Although men are typically diagnosed with diabetes at an earlier age and lower body mass index, obesity is more common in women, especially after age 45. Additional risk factors for women include gestational diabetes, polycystic ovarian syndrome and high uric acid, whereas in men they include smoking, hypogonadism, high blood pressure and alcohol abuse, to name a few (Campesi et al. 2017).

Sex hormones may explain some of the differences as they significantly affect energy metabolism, body composition and vascular function. Studies have demonstrated that a higher testosterone level and low SHBG is associated with diabetes risk in women than men. Men are more likely to develop foot ulcers, neuropathy and vascular disease, whereas women are more likely to die after an amputation and renal replacement. Women also suffer from more fatal and non-fatal stroke and cardiovascular events than men (Kautzky-Willer, Harreiter, and Pacini 2016).

Psychosocial factors also tend to impact females more than males. For instance, growing evidence suggests low education, low socioeconomic status and lack of leisure time for physical activity are more common in women then men and are associated with increased prevalence of

10

diabetes and obesity. Evidence also suggests women have a higher prevalence of depression, anxiety, cognitive limitations and falls. Greater awareness and research into the sex and gender pathophysiological mechanisms and its complications can lead to improved personalized diabetes care (Kautzky-Willer, Harreiter, and Pacini 2016). Especially given the fact that women attain glycemic targets less often than men (Abbate et al. 2012).

2.4 COMPLICATIONS

High BG concentrations lead to wide range of disorders including cardiovascular, cerebrovascular, peripheral vascular, retinal, renal, nerve, cancer, cognitive decline, liver and skeletal disease, depression and sleep apnea (Gerstein and Werstuck 2013).

(Reproduced with permission)

Careful glycemic lowering, while maintaining a low risk of hypoglycemia, can reduce the risk of complications (Riddle et al. 2018). Below, we review some of the more common complications of diabetes, many of which will be revisited in the hypoglycemia chapter, as they also correlate with hypoglycemia risk and complications.

11

2.4.1 All-Cause Mortality

Patients with diabetes die earlier than those without (Wang and Liu 2016). In fact, chronic diseases such as diabetes are the leading cause of mortality in the US (Heymsfield and Wadden 2017). In 2013, the IDF estimated that 1 in 12 deaths were attributable to diabetes (Atlas 2016).

2.4.2 Cardiovascular Disease

It is well known that patients with diabetes are at an increased risk of developing cardiovascular disease (CVD) (Kannel and McGee 1979). Even in people without diabetes, elevated BG increases the risk of CVD. For instance, in a study of people without diabetes undergoing percutaneous coronary intervention, abnormal glucose metabolism was detected in 1 of 3 patients and independently associated with a four-fold higher risk of events (von Birgelen et al. 2018). Theories to explain the deleterious relationship of elevated BG levels on CVD include a direct toxic effect of glucose on the vasculature, insulin deficiency, hypertension, atherosclerosis and inflammation (Punthakee, Werstuck, and Gerstein 2007).

2.4.2.1 Silent Myocardial Infarction

Diabetes increases the risk of silent myocardial infarction (MI), possibly by damaging nerves that normally transmit symptoms. As a result, the diagnosis of heart disease is often delayed and with worse prognosis in patients with diabetes than in those without diabetes (Le Feuvre, Jacqueminet, and Barthelemy 2011).

CAD is believed to further increase the risk of silent MI’s, especially in patients with diabetes. In one study, patients with diabetes vs no diabetes with ischemia were evaluated by exercise thallium scintigraphy. Although clinical characteristics, tread mill test results and extent of infarction and ischemia were similar between patients with and without diabetes, angina was less likely to be reported in patients with diabetes compared to those without. The authors concluded that angina was a poor marker of MI in patients with diabetes and CAD (Nesto et al. 1988).

The DIAD study evaluated the prevalence of silent MI in n=1,123 patients aged 50-75 years with diabetes but no known CAD. Using adenosine technetium-99m sestamibi single photon emission-computed tomography, they found 1/5 (113 patients, 22%) with silent ischemia. Interestingly, typical cardiovascular (CV) risk factors were not associated with an abnormal 12

stress test. Rather, a significant predictor of ischemia in patients with diabetes was cardiac autonomic dysfunction (Wackers et al. 2004).

Silent MI coupled with silent hypoglycemia (ie undetected symptoms of low BG) can further increase risk of cardiovascular death. Therefore, prevention of hypoglycemia is particularly important in people with diabetes who are more likely than the general population to have CAD.

2.4.3 Dementia

In the general population, cognitive function begins to decline after about 40 years of age. T2DM is associated with a greater risk of cognitive decline and a greater rate of cognitive function (Cukierman, Gerstein, and Williamson 2005). Older patients with diabetes are at a 1.5 to 2.0-fold increased risk of dementia as compared with patients without diabetes. This is especially concerning given that severe hypoglycemia is also increased in older patients with diabetes (Biessels 2009).

In patients with diabetes, a decline in cognitive function is consistent across age groups, including declines in verbal memory, information processing speed, attention and executive functioning. Similar decrements are observed even in patients with prediabetes. The incidence of dementia is increased in patients with T2DM, compared to those without diabetes. Vascular damage underlies both decrements in cognitive function and dementia (Biessels et al. 2014).

13

(Reproduced with permission)

Thus far, meta-analysis of intensive vs less intensive glucose lowering studies have failed to demonstrate a correlation with an increase or decrease cognitive decline (Tuligenga 2015). However, it seems intuitively correct that avoiding hypoglycemia will help prevent additional risk of cerebral damage.

2.4.4 Muscle Strength

The Health Aging and Body Composition showed lower muscle grip strength, knee extensor strength and muscle mass in n=485 patients with T2DM compared to those without diabetes (Park et al. 2006). In another cross-sectional analysis of Bostonian men, those with diabetes had

14

lower muscle mass and strength, yet similar BMD compared to men without diabetes (Akeroyd et al. 2014).

2.4.5 Falls

Causes of falls are multiple and in patients with diabetes, this number can increase. For example, the most common causes of falls include impaired vision, muscle weakness, sedentary lifestyle, poorly fitted shoes and medication use. For patients with diabetes, risk of cerebrovascular disease, postural stability and fall risk are also of concern. In normal individuals, when loss of balance is perceived, hip and knees flex and arms extend to prevent a fall. In individuals with mild cerebrovascular disease, perception of the loss of balance is delayed. Wrist fracture are common since there is only sufficient time to extend the hand to break the impact of the fall (Birge 2008).

A cross-sectional study from Japan demonstrated that the risk of falls was significantly associated with a longer duration of diabetes, presence of neuropathy, peripheral artery disease and a history of any fractures. Conversely, risk of any fracture was significantly associated with a lower BMI, presence of neuropathy and risk of falls, which was an independent predictor. When risk of fracture by site was analyzed, only vertebral fractures remained significantly associated to falls (Yokomoto-Umakoshi et al. 2017).

Patients with diabetes compared to people without diabetes, have an increased risk of falls even without hypoglycemia (Lu et al. 2015). The Study of Osteoporotic Fractures prospectively followed n=9,249 women age 67 or greater, of which 6.8% (n=629) had diabetes. They found women with diabetes had increased risk of falling, irrespective of insulin use, although risk more the doubled in insulin-treated patients. Whether an episode of hypoglycemia was associated with the fall was however, not evaluated (Schwartz et al. 2002).

15

(Reproduced with permission)

2.4.5.1 Heterogeneity in Fall Risk by Race (DM or non-DM)

A recent Kaiser Permanente survey of n=6,277 older (65-90) women from Northern California, found risk of falls to be significantly less for Black and Asian women compared to White or Hispanic women. Interestingly, hip fracture is also higher in White women which may be part attributable to the greater risk of falls. After adjusting for multiple variables including age, poor health, mobility limitation, and comorbidities such as diabetes, arthritis and history of stroke, the odds of less falls in Blacks and Asians remained significantly lower (Geng et al. 2017).

2.4.5.2 Risk of Falls in Diabetes on Insulin

In a subgroup analysis of the Longitudinal Health Aging and Body Composition study, risk of falls in patients with diabetes and insulin was evaluated. Diabetes overall had significantly higher risk of injurious falls requiring hospitalization compared to no diabetes. In comparison to patients without diabetes, patients with diabetes not taking insulin trended for a higher nonsignificant risk of falls. On the other hand, patients with diabetes on insulin treatment were at a higher risk of falls even after multiple adjustments for confounders were made including age, sex race, study site, education and BMI (Yau et al. 2013). Avoiding hypoglycemia is particularly important in reducing the risk of falls.

2.4.6 Fractures

In patients with T1DM, but not necessarily T2DM, bone mineral density (BMD) is decreased. In both patients with T1DM and T2DM, bone turnover is decreased and bone material properties are altered. This is especially so in patients with diabetes and microvascular disease. The

16

pathophysiological mechanisms underlying bone fragility in diabetes are complex. They include hyperglycemia, oxidative stress and advanced glycated end-products that alter collagen properties, marrow adiposity and osteocyte functionality, to name a few (Napoli et al. 2017).

Mechanisms Underlying Bone Loss and Fractures in T2DM:

(Reproduced with permission)

In 2007, a meta-analysis was conducted to assess the association of diabetes and fractures. Sixteen RCTs of n=836,941 patients with diabetes and 139,531 cases of fractures were evaluated. Risk of hip fractures in men with T2DM (RR=2.8) was found to be slightly higher than in women with T2DM (RR=2.1). Patients with T1DM (RR=6.3) had a higher risk of fractures than patients with T2DM (RR=1.7). Fractures at other sites had only a weak association in patients with T2DM (Janghorbani et al. 2007).

The Freemantle Diabetes Study from Australia recently accounted for competing risk of death to evaluate the influence of hip fractures on mortality in n=1,291 patients with T2DM. To date, no other study has considered the competing risk of death on hip fracture risk in patients with diabetes. They found that risk of hip fracture was non-significant after adjusting for the competing risk of death (Hamilton et al. 2017).

(Reproduced with permission)

17

2.4.7 Sex-Gender Differences in Complications

In patients without diabetes, risk of CVD is higher in men. However, diabetes reverts women’s advantage making them more susceptible to a CV event than men with diabetes (Campesi et al. 2017).

(Reproduced with permission)

In a gender specific meta-analysis, diabetes and CVD were compared as risk equivalents. Despite limited studies that reported gender specific HRs for mortality (only 0.17% of 5,425 studies), a sex-gender specific difference was noted. In men, CVD conferred a higher risk of mortality than diabetes. The reverse was true for women; although not significant, diabetes conferred a greater mortality risk than CVD (Lee et al. 2012).

In a Ukrainian nationwide diabetes registry, risk of nonfatal stroke in patients with T2DM was assessed. Overall, men demonstrated a higher risk of stroke than women when on oral antihyperglycemic agents or lifestyle modification but equal risk to women if on insulin (Khalangot et al. 2009).

18

2.4.8 Insurance / License / Employment

Individuals with diabetes are considered to have a disability according to the Americans with Diabetes Disability Act. This is because diabetes constitutes an impairment of the endocrine system. Individuals with diabetes have a harder time getting life and other kinds of insurance, and often pay higher premiums. The Diabetes Disability Act aims to protect individuals with diabetes from employment discrimination (American Diabetes, Anderson, et al. 2014). However, an individual’s history such as hospitalization or other health care encounters due to hypoglycemia can impact employment opportunities.

2.5 COST

The economic burden of diabetes is staggering and continues to rise. Depending on the cost model, the cost of diabetes can vary greatly, especially related to the contribution of indirect costs such as the sequelae from hypoglycemia.

2.5.1 Canada

In 2009, Diabetes Canada estimated that the direct and indirect cost of diabetes in 2010 to be at $12.2 billion and $16.9 billion by 2020 (Association 2014).

2.5.1.1 Cost of Test Strips

In August 2013, consulting experts from the CDA and Ontario public drug plan introduced limits of BG test strips. Annual reimbursement limits depend on type of treatment; 3000 strips per year for insulin-treated patients, 400 for those treated with oral antihyperglycemic agents that may cause hypoglycemia (such as SUs) and 200 for all others with diabetes. In the case of drug to drug interactions, failure to attain glycemic targets over a 3-month period, or certain safety sensitive occupations, additional test strips can be reimbursed. In a recent cross-sectional time- series analysis between August 2010 and July 2015, of all patients 65 years and older receiving test strips, the associated costs were evaluated for 4 treatment groups (insulin, non-insulin but hypoglycemia inducing treatment, non-hypoglycemia inducing agents and no treatment). During this period, 657,338,117 test strips were dispensed to those 65 years of age and older in Ontario, costing $482.3 million (CDN). One-year post policy implementation, test strip decreased by 22.2%, resulting in a net cost reduction of $24 million. During this period, the study also found

19

that less people were treated with hypoglycemia causing therapies, however, more were also treated with insulin. Since the implementation of test strip policy, other provinces have also followed suit, including British Columbia’s provincial drug plan and Health Canada’s non- insured health benefits program (Gomes et al. 2016). If patients with T2DM could be initially managed with minimal risk of hypoglycemia, then for this initial period, safety monitoring with BG test strips might not be necessary.

2.5.2 USA

The American Diabetes Association estimates of the cost of diabetes in 2017 at $327 billion, of which $237 billion relate to direct medical costs and $90 billion in indirect costs (Association 2018).

2.6 SUMMARY

Diabetes continues to rise globally at an alarming rate around the world. Between 1980 and 2017, the percentage of people with the disease nearly tripled. By 2045, it is estimated that 1 in 10 will be living with diabetes, a number Canada will reach by 2020. Most individuals living with diagnosed and undiagnosed diabetes reside in low-to-middle income countries.

The pathophysiology of the diabetes is much more complex than initially believed with many more organs involved than the trifecta of insulin resistance in the liver and muscles and the progressive beta-cell failure of the pancreas. It is now believed that multiple organs contribute to hyperglycemia and the pathology of the disease, including the gut, brain and kidneys. Similarly, a wide host of risk factors increases the risk of diabetes onset; from classic metabolic syndrome risk factors to adiposity, dietary factors and psychosocial factors such as socioeconomic status and low educational status. Global heat maps depicting the rise of diabetes prevalence strikingly mirror those demonstrating the global rise of obesity.

The diagnosis of diabetes is based on elevated BG concentrations which reek havoc on the vasculature. Patients with diabetes compared to those without, are at a much greater risk of dying and experiencing a wide range of complications. Patients with diabetes suffer from complications relating to the heart, kidney, brain, nerves, eyes, skeleton as well as infections, depression and cancer. A diagnosis of diabetes also affects employment opportunities and higher

20

premiums for health insurance coverage. Attainment of glucose control early after diabetes diagnosis can attenuate risk of complications.

Diabetes continues to be a major strain on health care resources around the world. Prevention of diabetes is key and in patients with diabetes, reducing the risk of complications and burden on the health care system continues to be of critical importance.

MANAGEMENT

In this chapter, we review how diabetes is diagnosed, managed and the evidence to support intensive glycemic lowering.

3.1 PREVENTION

3.1.1 Lifestyle Modification

Lifestyle modification is crucial for patients with diabetes as it can lead to weight reduction, less need for , reduced risk of sleep apnea and improved well being (Gerstein 2013). The Look Ahead Trial tested whether lifestyle changes can result in improved CV outcomes in overweight or obese patients with diabetes. Overweight or obese patients with diabetes (n=5,145) were randomized to an intensive lifestyle intervention (ILI) targeting 7% weight loss or a conventional arm who received diabetes education and support (control arm, aka DSE). The intervention arm successfully lost 8.6% vs. 0.7% at one year and 6.0% vs. 3.5% at study end for intensive lifestyle intervention vs the control arm, respectively. However, the trial was stopped early at 9 years due to futility for its primary CV endpoints (Wing et al. 2016).

Given the long duration between obesity onset and the development of diabetes, it should not be a surprise that the Look AHEAD had to be stopped due to futility. Perhaps a larger study with a longer duration of follow up than 9 years could have demonstrated a signal. Further, the Look AHEAD participants were at much lower risk of experiencing a CV event than the current cardiovascular outcome trials (CVOTs) mandated for new diabetes therapies by the FDA. Study entry criteria in Look AHEAD required a pass on an exercise stress test and only 14% had established CVD at baseline. In addition, Look AHEAD participants were younger with better glycemic control and a shorter duration of diabetes than seen in most CVOT trials. Despite the

21

neutral findings of LOOK AHEAD, the importance of lifestyle modification should not be overlooked (Belalcazar and Ballantyne 2017).

Rates of severe hypoglycemia over the entire study period did not differ between the groups. Rate of severe hypoglycemia were 0.49/100 person-years and 0.51/100 person-years for the intensive and control (DSE), respectively. However, severe hypoglycemia was greater in the first year for the ILI arm, despite a 3% reduction of insulin use. Regressional analyses demonstrated that severe hypoglycemia was higher with baseline insulin, SU or glitinide use. Of the insulin users, hypoglycemia rates were similar in the control arm compared to the ILI whose weight loss was in the lower 50%, with rates of 0.52 and 0.58 events /100 patient-years in the control and intensive arms, respectively. However, patients who lost greater weight in the ILI arm experienced less hypoglycemia and lower percent insulin use compared to the control arm, possibly protecting them from hypoglycemic events (Greenway 2016).

3.2 DIAGNOSIS

3.2.1 Prediabetes

Prediabetes includes both impaired fasting glucose (IFG) and impaired glucose tolerance (IGT). IFG is primarily believed to be the result of hepatic insulin resistance, seen early in the course of beta-cell dysfunction (Abdul-Ghani, Tripathy, and DeFronzo 2006). The diagnosis of IFG is a fasting plasma glucose of ≥5.6 mmol/L and <7.0 mmol/L.

Peripheral insulin resistance and advanced beta-cell dysfunction is thought to contribute to IGT. After a 75g oral glucose tolerance test, IGT is diagnosed as part of a 2-hour plasma glucose concentration with BG values ≥7.8 mmol/L and <11.1 mmol/L.

Interestingly, isolated IFG and IGT may define differences in racial group, gender and age. Recent evidence suggests that isolated IGT may be more common in women and IFG more common in men. As a result, estimates of the prevalence of prediabetes and diabetes may differ based on which tests were conducted (Jaacks et al. 2016). The Global Burden of Disease (GBD) study found that for each decade between 1980 and 2008, the mean fasting plasma glucose increased by 0.07 mmol/L in men and 0.09 mmol in women in the general population (Danaei et al. 2011).

22

3.2.2 Diabetes

Diabetes is diagnosed with either a fasting plasma glucose of ≥7.0 mmol/L, a 2-hour plasma glucose value after a 75g oral glucose tolerance test of ≥11.1 mmol/L or HbA1c ≥6.5%. These levels predict the development of retinopathy.

(Punthakee, Goldenberg, and Katz 2018)

(Reproduced with permission)

3.3 GLYCEMIC TARGETS

Diabetes Canada places emphasis on individualizing treatment targets based on patient frailty, functional dependence and life expectancy. They recommend most adults with T2DM should strive for an HbA1c ≤7%. Tighter targets (≤6.5%), to reduce the risk of chronic kidney disease (CKD) and retinopathy are recommended, but only if the risk of hypoglycemia is believed to be low. More relaxed targets (HbA1c 7.1 to 8.5) are recommended for those at risk of hypoglycemia, or for frail individuals with limited life expectancy.

23

(Imran et al. 2018)

(Reproduced with permission)

3.4 INTENSIVE VS LESS INTENSIVE GLYCEMIC CONTROL

It is important to revisit why intensification of treatment to achieve glycemic targets is important in the first place, given the fear of experiencing hypoglycemia. This section will review the landmark randomized controlled trials investigating intensive vs less intensive glycemic control in patients with T2DM. The below table provides a nice summary of the key characteristics of the studies that will be reviewed.

24

(Pistrosch and Hanefeld 2015)

(Reproduced with permission)

In all intensive vs less intensive treated studies, hypoglycemia risk increased (Ray et al. 2009). Yet, it will remain unknown, if risk of hypoglycemia would still be elevated if the same intensive vs less intensive studies were conducted using today’s new AHA with less risk of hypoglycemia, at least relative to the agents used at the time (sulfonylurea and insulin).

3.4.1 UKPDS

The United Kingdom Prospective Diabetes Study was the first trial to demonstrate a significant benefit in organ damage with glycemic lowering. The trial randomized n=3,867 newly diagnosed T2DM patients to intensive treatment with a SU or insulin or the conventional arm of diet only. At 10-year follow up, the HbA1c in the intensive arm was 7.0% compared to 7.9% in the conventional arm. This difference translated into to 12% significant reduction in diabetes- related endpoints, which was mainly due to a 25% reduction in microvascular endpoints such as the new onset of retinopathy or nephropathy (Group 1998b).

Ten years after the trial, despite having discontinued the randomized interventions and similar HbA1c levels, patients in the intensive glucose lowering arm demonstrated sustained reduction 25

of risk of complications, a phenomenon that was termed the legacy effect. During this posttrial follow up, the reduction in any diabetes-related endpoints and microvascular disease continued to be significant relative to patients in the conventional arm. More importantly, MI and ACM which failed to demonstrate a benefit in the first 10 years of the trial, reached statistical significance (Holman et al. 2008).

3.4.2 ADVANCE

The Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation study evaluated over n=11,000 patients with T2DM and micro- or macrovascular disease to intensive lowering with gliclazide or conventional therapy. After a median of 5 years, the intensive arm achieved an HbA1c of 6.5% compared to 7.3% in the standard arm. This translated into a 10% reduction in the combined endpoint of micro- and macrovascular disease for intensively controlled patients (Patel et al. 2008).

Similar to the metabolic memory observed with the post-trial follow up of UKPDS patients, the ADVANCE-ON trial followed participants for a median 5.4 years post-trial. Despite a loss of HbA1c differences, patients originally randomized to the intensive arm continued to demonstrate a lower risk of end-stage kidney disease. Risk of severe hypoglycemia for the in-trial period was significantly greater for patients in the intensive arm at 2.7% and 1.5% for the standard arm, with a risk ratio 1.85 (95% CI 1.42 -2.42). For the overall study period, rates of severe hypoglycemia remained significantly higher for patients in the intensive arm at 8.0% compared to 6.7% for those in the standard arm and translating to a risk ratio of 1.19 (95% CI 1.04-1.36) (Wong et al. 2016).

3.4.3 VADT

The Veteran Affairs Diabetes Trial (VADT) randomized n=1,791 military veterans to an intensive vs standard glucose-lowering arm. By 5.6 years of follow-up, the HbA1c % attained was 6.9% in the intensive arm and 8.4% in the standard arm. Rates of any hypoglycemia was significantly higher in the intensive arm at 24.1% compared to 17.6% in the control arm. Only progression of albuminuria was significantly reduced, while the rates of major CV events, death or microvascular complications did not differ (Duckworth et al. 2009).

26

3.4.3.1 VADT Outcomes in patients with lower calcified CAD

RACED (Risk Factors Atherosclerosis and Clinical Events in Diabetes) evaluated coronary artery calcium in a subset (n=301) of the VADT (n=1,791) patients for 7.5 years for the development of CV events. Although intensive glycemic control did not reduce MACE outcomes in this cohort, baseline CAC scores suggested differing responses. Multivariable adjusted HR for intensive treatment in those whose CAC was >100 was not significant at HR 0.74 p=0.21, yet was in CAC of ≤100, HR 0.08 p=0.03 (Reaven et al. 2009).

3.4.4 ACCORD

Of all the intensive glucose lowering trials, Action to Control Cardiovascular Risk in Diabetes Study (ACCORD) may have raised more questions than it answered, particularly for the impact on hypoglycemia. The primary outcome in ACCORD of CV death, nonfatal MI and nonfatal death was numerically, but not significantly less in the intensively treated patients (HR 0.90 CI 0.78-1.04). However, due to an excess of deaths in the intensively treated arm (HR 1.22 CI 1.01- 1.46 p=0.04), the trial was discontinued early (Action to Control Cardiovascular Risk in Diabetes Study et al. 2008). Risk of ischaemic heart disease, as suggested by the primary composite HR trended lower in the intensive than standard therapy, except for a non-significant increase in fatal MI.

27

(Gerstein et al. 2014)

(Reproduced with permission)

Since hypoglycemia (both recognized and unrecognized) was more common in the intensively treated arm, it was initially considered as the main culprit. The incidence per 100 person-years of first severe hypoglycemia requiring medical assistance was 3.14 for patients in the intensive arm and 1.03 for patients in the standard arm, a finding that was statistically significant (P<0.0001). Post hoc analyses also demonstrated that hypoglycemia risk was greater in women, African American, lesser education, and insulin use at baseline (Seaquist et al. 2012).

3.4.4.1 Absolute levels of HbA1c and Hypoglycemia Risk (ACCORD)

The crude incidence of hypoglycemia requiring medical assistance was consistently higher in the intensively treated patients than standard, whether baseline HbA1c, updated average HbA1c (which included the baseline HbA1c measurement), or most recent HbA1c were considered. Overall, a higher annual incidence of hypoglycemia was noted for updated average HbA1c and most recent HbA1c. For instance, for every 1% unit higher baseline HbA1c, patients in the standard treatment arm experienced a significant 30% higher risk of hypoglycemia requiring medical assistance. For updated average HbA1c, the risk of severe hypoglycemia requiring medical assistance increased by 76% for those in the standard arm. The relationship of baseline HbA1c and hypoglycemia requiring medical assistance was not found patients in the intensively treated arm (HR 1.01, 95% CI 0.94 to 1.10). However, similar to the standard arm, every 1% unit increase in updated HbA1c resulted in a significant increase (15%) in hypoglycemia requiring assistance (HR 1.15, 95% CI 1.02 to 1.29). For every 1% decline in HbA1c, a reduction of hypoglycemia risk was observed in both arms (Miller et al. 2010).

3.4.4.2 Change in HbA1c and Hypoglycemia Risk (ACCORD)

It is important to note that in both the standard and intensive arms, a 1% fall in HbA1c was associated with a 35% and 15% reduced risk of hypoglycemia requiring medical assistance, respectively. Moreover, this association was not dependent on baseline HbA1c levels (p=0.2569 for interaction) (Miller et al. 2010). Thus, patients who attempted but were unsuccessful in

28

lowering BG levels were at greatest risk of experiencing severe hypoglycemia. This may be counterintuitive to some, as HbA1c lowering is often believed to increase hypoglycemia risk.

3.4.4.3 Hypoglycemia and Mortality Risk (ACCORD)

In patients who never experienced a hypoglycemic event, mortality rates were higher in the intensively treated (1.2%/year) arm than in the standard arm (1.0%/year). However, in patients with at least one previous severe event, defined as requiring any, medical, or non-medical assistance, mortality was lower in the intensively treated arm (2.8%/year vs 3.7%/year). Adjusted HR of mortality was 1.41 (CI 1.03-1.93) in the intensive arm vs 2.30 (CI 1.46-3.65) in the standard arm in patients who had experienced at least one severe event vs no previous severe events. Subgroup analyses demonstrated an even more robust relationship between hypoglycemia and mortality when the severity of hypoglycemia was greater (ex. hypoglycemia requiring medical assistance). The risk of death in those experiencing hypoglycemia requiring medical assistance in the intensive treated arm was 2.8%/year compared to 4.9%/year in the standard arm, a statistically significant difference (adjusted HR 0.55 (CI 0.31-0.99)) (Bonds et al. 2010).

(Reproduced with permission)

Post hoc analyses of HbA1c and its relation to mortality were conducted and provided surprising results. A 1% increase in HbA1c was associated with a 20% increase risk of death. Interestingly, mortality and HbA1c were directly proportional in the intensively treated patients,

29

yet a U-shape was observed in the standard arm (Riddle et al. 2010). Similar associations by treatment arm were observed when analyzed by age (Miller et al. 2014).

(Reproduced with permission)

The paradox of increased mortality in the intensively treated patients, yet a lower HbA1c and reduced risk of dying after a severe hypoglycemia episode remains to be fully understood. However, the minority of patients in the intensive treated arm with HbA1c >7% accounted for the excess risk. To further support this notion, the excess risk of death in the intensive arm was associated with little reduction in HbA1c from baseline to the first 4 or 12 months of treatment. Risk of hypoglycemia requiring medical assistance according to a 1% decline in HbA1c from baseline to 4-month visit was reduced by 28% and 14% in the intensive vs standard arms. Further, risk of hypoglycemia requiring medical assistance increased with a 1% increased in the

30

average HbA1c (standard arm HR 1.76 CI 1.50 – 2.06 and intensive arm HR 1.15 CI 1.02 – 1.21). Note positive value reflects a drop in HbA1c (Riddle 2010). To this day, the cause of increased mortality in the intensive treatment group of ACCORD is still not fully understood.

(Reproduced with permission)

3.4.5 Meta-analysis of Intensive Glucose Control Trials

A recent meta-analysis using individual patient data from n=27,049 participants from ACCORD, ADVANCE, UKPDS, and VADT found that intensive glucose control reduced the risk of kidney and eye disease but not nerve events (Zoungas et al. 2017).

3.4.6 Intensive Glucose Control in Acute Coronary Syndrome

A meta-analysis of RCTs of intensive vs less intensive glucose control and reporting on mortality in patients with T2DM and a recent MI was conducted. Included in the analysis was data from n=2,113 patients from three RCTs (Digami, Digami 2 and the Hi-5 trial). While the primary endpoint of ACM was neutral (RR 0.94), the risk of hypoglycemia was drastically increased in the intensive treatment groups (RR 13.40 CI 3.69-48.61). Further, mortality correlated with longer duration of treatment in these patients with intensive glucose control (Chatterjee et al. 2013).

31

3.5 HbA1c and MORTALITY

A retrospective analysis of the UK Clinical Practice Research Datalink demonstrated a U-Shaped association between achieved HbA1c and mortality. More importantly, a differing pattern was observed when class of AHA was considered according to its tendency to cause hypoglycemia. AHA known to increase risk of hypoglycemia was associated with a higher risk of mortality at the lower range of HbA1c, whereas AHA with a low hypoglycemia risk was associated with an increased risk of mortality at higher HbA1c levels (Currie et al. 2018).

3.6 ATTAINMENT OF GLYCEMIC TARGETS

3.6.1 Canada

Three Canadian observational studies conducted between 2003 to 2013 show approximately half of patients with diabetes failed to achieve HbA1c targets despite the importance of attaining glycemic control (Harris et al. 2005; Braga et al. 2012; Leiter et al. 2013).

3.6.2 US

NHANES data reported an improvement in HbA1c in all patients with diabetes from 7.6% in 1999 to 7.2% in 2010 (Ali et al. 2013).

3.6.3 Europe

The DPV Registry from Germany and Austria assessed trends of diabetes management in patients with T1DM (n=35,144) and T2DM (n=149,720) from 2002 to 2014. HbA1c increased over time until 2011 in both T1DM and T2DM and then started to fall. Over the same period, severe hypoglycemia also increased and then levelled off and fell by 2014. An explanation of either trend is not available (Bohn et al. 2016).

32

(Reproduced with permission)

3.6.4 Glycemic Target Attainment By Hypoglycemia Risk

The Diabetes Unmet Need with Basal Insulin Evaluation (DUNE) prospectively evaluated patients with T2DM (newly or recently <12 months) initiating basal insulin from Europe and the US, using electronic medical records. Over a two year period, they assessed pre-defined factors associated with glycemic control by considering short term hypoglycemia incidence in the first three months after insulin initiation, which was reflective of hypoglycemia risk throughout the remainder of the trial. Of the n=3,319 eligible participants, hypoglycemia across the countries was 3.8%, albeit with varying definitions. Physician set HbA1c targets were concordant with

33

low vs high hypoglycemia risk. However, target attainment of <7% was not achieved in either group (Mauricio et al. 2017).

3.7 TEMPORAL CHANGES IN AHA USE

Based on new drug availability, study results and clinical practice guidelines, the frequency of use of AHA is changing over time in different jurisdictions.

3.7.1 Canada

In London Ontario, from 2002 to 2013, the percentage of people prescribed glyburide and prescriptions declined, whereas the metformin, gliclazide, DPP4i as well as combination treatment increased. Insulin treatment remained stable during this period (Clemens et al. 2015).

3.7.2 US

An electronic health record system from the Cleveland Clinic for patients with new onset of T2DM between 2008 and 2013 demonstrated that metformin was the most prescribed medication, SUs remained unchanged and the use of decreased substantially (Pantalone et al. 2016). The age and sex-standardized proportion of patients with T2DM who filled prescriptions for glucose-lowering agents between 2006 and 2013 showed an increased use of metformin, DPP4i, insulin and a decline in SUs and thiazolidinediones (Lipska et al. 2017).

34

(Reproduced with permission)

3.8 COMORBIDITY

A US electronic medical record database estimated that 97.5% of all patients with T2DM had at least one other comorbid condition, and 88.5% had at least two (Iglay et al. 2016). Studies suggest the comorbidity in patients with T2DM is not declining. Newly diagnosed patients with T2DM between 2008 and 2013 recorded in the electronic health record system of the Cleveland Clinic also showed that hypertension, CVD and neuropathy were common comorbidities. Between 2008 and 2013, the prevalence of comorbidities increased; 71.5% and 71.0% of patients with hypertension, 1.3% and 3.2% with retinopathy, 4.4% and 7.2% with nephropathy, 9.6% and 16.2% with neuropathy, 3.9% and 5.2% with cerebrovascular disease, 14.7% and 15.8% with cardiovascular disease and 2.0% and 3.8% with peripheral vascular disease, respectively (Pantalone et al. 2016).

3.9 SUMMARY

In this chapter, we reviewed the clinical management of diabetes, from diagnosis to the importance of achieving glycemic targets. A diabetes diagnosis can be conferred by three methods: 1. A fasting plasma glucose of ≥7.0 mmol/L; 2. a 2-hour plasma glucose value of ≥11.1 mmol/L after a 75g oral glucose tolerance test or an HbA1c ≥6.5%.

The importance of intensive glycemic lowering on reducing adverse diabetes outcomes stems primarily from the UKPDS trial. Subsequent trials of intensive glycemic lowering aimed to evaluate various HbA1c targets. However, unlike UKPDS which was conducted in newly diagnosed patients, ADVANCE, VADT and ACCORD enrolled patients with a longer duration of diabetes and with more comorbid conditions. These latter studies demonstrated the importance of glycemic lowering on reducing mainly microvascular complications and in longer- term follow-up, reductions in macrovascular endpoints.

Antihyperglycemic agents used at the time and thus in these trials included SU and insulin, both of which are notoriously known to increase risk of hypoglycemia. Patients in the intensive glucose control of these trials experienced greater rates of hypoglycemia than those enrolled in the standard treatment arm.

35

The excess risk of deaths in the intensive glucose lowering arm of the ACCORD trial remains unresolved. Hypoglycemia was initially believed to be the culprit, however, post hoc analyses were unable to demonstrate a correlation. In fact, a higher baseline or updated HbA1c correlated with a higher risk of hypoglycemia in either treatment arm. Conversely, a decline in HbA1c correlated with a lower risk of hypoglycemia. This suggested that hypoglycemia risk may be increased in patients whose BG levels are not well managed or who attempt but are unsuccessful in reducing their elevated BG levels (for whatever reason) at least with the traditional antihyperglycemic agents (SUs, , insulin and thiazolidinediones) used at the time.

Since these trials, a new era of antihyperglycemic agents now exist. In addition to metformin, which is recommended first line in guideline bodies in US, Canada and Europe, three new classes of antihyperglycemic agents have dramatically changed the treatment paradigm for patients with type 2 diabetes. They include the incretin-based therapies of dipeptidyl peptidase-4 inhibitors, glucagon like peptide-1 receptors and sodium glucose cotransporter-2 inhibitors. These new antihyperglycemic agents boast effective BG lowering without the increased risk of hypoglycemia, seen with SUs and insulin. However, before we explore the mechanism of action and evidence of hypoglycemia risk for these agents, a deeper dive into the latest understanding of hypoglycemia is necessary.

HYPOGLYCEMIA

Given the outcome of interest, it is only befitting that the latest evidence on hypoglycemia is examined. We begin with a review of how hypoglycemia is defined across the major diabetes guideline bodies. Next, the symptoms, physiology and counterregulation of hypoglycemia are explored. Prior to the epidemiological appraisal, the various nuances contributing to the heterogeneity of hypoglycemia in the clinical literature are reviewed. Finally, we discuss the risk factors for and complications of hypoglycemia, as well as its impact on health care expenditure.

4.1 CLINICAL SYNDROME

The clinical syndrome of hypoglycemia is most commonly documented by Whipple’s Triad; symptoms of hypoglycemia, a low plasma glucose concentration and relief of symptoms upon carbohydrate ingestion (Cryer et al. 2009).

36

4.2 GUIDELINE DEFINITIONS

National diabetes associations vary in their definitions of hypoglycemia, although some progress has recently been made in standardizing the definitions. Some argue that defining hypoglycemia by a specific glucose concentration when thresholds for symptoms and counterregulatory responses vary within individuals is inappropriate. Factors such as poor glycemic control may shift thresholds to higher glycemic concentrations where as more intensive therapy may shift thresholds to lower glucose concentrations. As a result, the American Diabetes Association defines hypoglycemia non-numerically as, “all episodes of an abnormally low plasma glucose concentration that exposes the individual to potential harm” (Seaquist et al. 2013). This definition is difficult to quantify for patients and to classify episodes (Shaefer, Hinnen, and Sadler 2016).

The International Hypoglycemia study group believes it is important to identify a glucose level that should be avoided based on significant clinical and economical consequences. They suggest a glucose concentration of <3.0 mmol/L (<54 mg/dL) be reported in diabetes clinical trials, where as those ≤3.9 mmol/L need not be. Glucose values of <3.0 mmol/L do not occur under normal physiological conditions in individuals without diabetes and in patients with diabetes, it has been shown to be associated with CV events and increased mortality. The group stressed that standardizing the classification of hypoglycemia is necessary and would allow for more appropriate meta-analyses of treatment comparisons, which has been a major obstacle to date. They further recommend that a glucose value of ≤3.9 mmol/L alert patients and health care providers to consider dose adjustment. The American Diabetes Association has followed their recommendations (International Hypoglycaemia Study 2017).

Table 1. International Hypoglycemia Study Group: Classification of Hypoglycemia

Level Glycemic criteria Description

Glucose alert value (level 1) ≤70 mg/dL (≤3.9 Sufficiently low for treatment mmol/L) with fast-acting carbohydrate and dose adjustment of glucose-lowering therapy

37

Clinically significant <54 mg/dL (<3.0 Sufficiently low to indicate hypoglycemia (level 2) mmol/L) serious, clinically important hypoglycemia

Severe hypoglycemia (level 3) No specific threshold Hypoglycemia associated with severe cognitive impairment requiring external assistance for recovery

(Open Access - Permission to reproduce not required)

Defining severe hypoglycemia as requiring external or third-party assistance without a glycemic value, as proposed by the ADA, has gained wider acceptance. The International Hypoglycemia Study Group has maintained this definition (International Hypoglycaemia Study 2017).

4.2.1 ADA

The 2017, ADA Standards of Medical Care in Diabetes followed the recommendations made by the International Hypoglycemia Study Group, which also uses the ADA’s definition of severe hypoglycemia of requiring external assistance (Association 2017).

It should be noted that in 2004, The American Diabetes Association assembled a working group to advise the FDA on how hypoglycemia should be defined, reported and what makes a clinically meaningful reduction of hypoglycemia for evaluation in clinical trials of new treatments (Association 2005). The report was updated in 2013 to also include the Endocrine Society. Here, they suggest that hypoglycemia can be classified as:

1) Severe hypoglycemia. Severe hypoglycemia is an event requiring assistance of another person to actively administer carbohydrates, glucagon, or take other corrective actions. Plasma glucose concentrations may not be available during an event, but neurological recovery following the return of plasma glucose to normal is considered sufficient evidence that the event was induced by a low plasma glucose concentration.

38

2) Documented symptomatic hypoglycemia. Documented symptomatic hypoglycemia is an event during which typical symptoms of hypoglycemia are accompanied by a measured plasma glucose concentration ≤70 mg/dL (≤3.9 mmol/L).

3) Asymptomatic hypoglycemia. Asymptomatic hypoglycemia is an event not accompanied by typical symptoms of hypoglycemia but with a measured plasma glucose concentration ≤70 mg/dL (≤3.9 mmol/L).

4) Probable symptomatic hypoglycemia. Probable symptomatic hypoglycemia is an event during which symptoms typical of hypoglycemia are not accompanied by a plasma glucose determination but that was presumably caused by a plasma glucose concentration ≤70 mg/dL (≤3.9 mmol/L).

5) Pseudo-hypoglycemia. Pseudo-hypoglycemia is an event during which the person with diabetes reports any of the typical symptoms of hypoglycemia with a measured plasma glucose concentration >70 mg/dL (>3.9 mmol/L) but approaching that level.

(Seaquist et al. 2013).

(Reproduced with permission)

4.2.2 Diabetes Canada

Diabetes Canada defines hypoglycemia by:

1) Development of autonomic or neuroglycopenic symptoms

2) Plasma glucose level <4.0 mmol/L for patients on insulin or insulin secretagogues and,

3) Symptoms responding to the administration of carbohydrate

The severity of hypoglycemia is classified according to clinical manifestations:

39

(Yale, Paty, and Senior 2018)

(Reproduced with Permission)

Recently, hypoglycemia definitions from the 109 studies included in CADTH’s (Canadian Agency for Drugs and Technologies in Health) second and third line agents were evaluated. Sixty percent of studies did not provide a definition, while 20% reported results for hypoglycemia without defining it. When definitions were provided, BG cut-off values and inclusion of symptoms varied amongst the studies. Further, definitions of hypoglycemia across studies within a class of medications differed. Severe hypoglycemia was reported in 52% of the included studies although only half of these provided a definition. Heterogeneity was also found in the definition of severe hypoglycemia. BG thresholds were specified in only half of the studies reporting severe hypoglycemia and values ranged from 2.0 to 3.3 mmol/L. In contrast, some studies relied on the presence of neuroglycopenic symptoms with or without documented BG measurements. Of concern was the scarcity of nocturnal hypoglycemia included in studies. Of those that did include nocturnal hypoglycemia, only 5% provided a definition each with its own unique characterization. The authors concluded that the heterogeneity of hypoglycemia definitions makes comparison amongst studies and classes of medications extremely challenging and meta-analyses of limited use. Nevertheless, meta-analyses comparing hypoglycemia should consider statistical techniques such as sensitivity analyses to evaluate differing definitions (Balijepalli et al. 2017).

40

(Reproduced with permission)

In response to the critique of meta-analyses on hypoglycemia risk by Balijepali et al., it is timely to briefly describe our intended approach. As outlined in greater detail in the Methodology Chapter, our meta-analysis aims to overcome the issue of hypoglycemia heterogeneity across studies by estimating the primary endpoint of any hypoglycemia, irrespective of the definitions or thresholds used. Our secondary endpoint of severe hypoglycemia aims to capture all major episodes; using a threshold of <3.0 mmol/L, or any episode requiring third party assistance, as this captures the criteria for most studies and guideline bodies, including the recent recommendations made by the International Hypoglycemia Study group. Further, since our outcome is an adverse event, we will take the more conservative approach of including studies with events in at least one treatment arm and conduct additional sensitivity analyses using risk difference to allow for inclusion of studies which attempted but did not observe any events in either arm. Additional sensitivity analyses will also be conducted to evaluate the impact of

41

defining severe hypoglycemia using a more conservative threshold of ≤3.1 mmol/L. With these considerations in mind, we believe that an estimation of hypoglycemia risk can and should be tackled given the diabetes epidemic, popularity of new AHA and importance of this outcome on diabetes management.

4.2.3 European Medicines Agency (EMA)

The most recent EMA guidelines from 2012 recommends a standardized approach to the definition of hypoglycemia and uses the ADA, IDF and Diabetes Canada classification of ≤3.9 mmol/L (≤70 mg/dL). In fact, IDF and Diabetes Canada’s glycemic threshold is technically <4.0 mmol/L. Given the original EMA definition of hypoglycemia was set at 3.1 mmol/L (56 mg/dL) (EMA 2012 guidelines), this recent change was met with much resistance due to employment and driving licence implications (Kenny 2014).

4.2.4 IDF

Similar to the ADA, the International Diabetes Federation defines hypoglycemia as; “all episodes of an abnormally low plasma glucose concentration (with or without symptoms) that expose an individual to harm”. Biochemically it is defined as a BG concentration of <4.0 mmol/L (70 mg/dL) (Group 2014).

4.3 SYMPTOMS

As BG values decline, the autonomic nervous system triggers neurogenic symptoms of hunger, sweating, palpitations, tremors and anxiety. These symptoms trigger patient awareness of an episode. Symptoms of hypoglycemia vary in individuals, by type of diabetes and by age. Symptoms of hypoglycemia are generally described within three categories: autonomic, neuroglycopenic and general malaise (Frier 2014).

Diabetes Canada Clinical Practice Guidelines 2018 lists symptoms of hypoglycemia as neurogenic or neuroglycopenic.

42

(Yale, Paty, and Senior 2018)

(Reproduced with permission)

4.3.1 Autonomic Neurogenic

Autonomic symptoms of hypoglycemia include tremor, anxiety, palpitations, sweating, tingling, hunger and trembling.

4.3.2 Neuroglycopenic

Neuroglycopenic symptoms include headache, blurred vision, confusion, dizziness, irritability, incoordination and speech disturbance. Since the brain is unable to store or produce glucose, a continuous supply to the brain from the circulation is vital. Only during periods of prolonged fasting, when glucose concentrations are depleted, the energy sources of the brain switches to other sources of energy. However, ketone bodies and amino acids are available in limited quantities, and free fatty acids are not readily transported through the blood brain barrier (Alsahli and Gerich 2013).

4.3.3 Malaise

Hypoglycemic symptoms of malaise include headache and nausea (Iqbal and Heller 2016).

43

4.4 PHYSIOLOGY

(Cryer 2013)

(Reproduced with permission)

4.4.1 Counterregulatory Responses in People Without Diabetes

The counterregulatory response to a low BG involves a behavioural component of carbohydrate ingestion. The physiological defense entails a decrease in insulin secretion and an increase in glucagon secretion. If this fails, epinephrine secretion increases as glucose levels reach below physiological range. Increases in epinephrine secretion from the adrenal medulla leads to an increase in emergency glucose production by the liver from stored glycogen and kidneys and a decrease in glucose clearance by the kidney, muscle and adipocytes.

44

It is the neurological symptoms of hypoglycemia, namely the sympathetic neural activation, that alerts the individual to recognize their low BG symptoms and ingest carbohydrates. In the setting where a decrease in insulin secretion and an increase in glucagon secretion is absent, such as in patients with T1DM or advanced T2DM, the increase in epinephrine secretion is critical (Cryer 2013).

Counterregulatory response thresholds –Individuals without diabetes (Iqbal and Heller 2016)

Glucose mmol/L

4.5 Inhibition of endogenous insulin

3.8 Production of glucagon and adrenaline

3.2 Autonomic symptoms

3.0 Cognitive dysfunction

2.8 Neuroglycopenic symptoms

Glycemic thresholds for counterregulatory hormones in healthy subjects are summarized below.

45

First, endogenous insulin secretion is inhibited at a BG value of approximately 4.6 mmol/L. At 3.8 mmol/L, glucagon is secreted, followed closely by adrenaline and growth hormone release. Below 3.2 mmol/L cortisol levels increase and below 2.6 mmol/L, cognitive dysfunction occurs. If glucose levels continue to decline (around 1.5 mmol/L), coma and seizures are likely to occur.

(Yun and Ko 2015)

(Permission not required – Open Access)

4.4.1.1 Sex-Gender Differences in Hypoglycemia Counterregulation

In a hypoglycemic insulin clamp study in healthy age-matched non-obese subjects, greater epinephrine, growth hormone, cortisol, glucagon and norepinephrine responses were found in males than in females. This may partly be explained by the counterregulatory responses occurring at a lower threshold in women than in men (Diamond et al. 1993).

4.4.2 Counterregulatory Responses in Patients with T1DM

In patients with T1DM, when BG levels fall, insulin and glucagon responses are absent and epinephrine secretion is often attenuated. An impairment in the glucagon response is apparent as early as 2 years after diagnosis and established by 5 years after diagnosis. It is believed that the autoimmune system progressively destroys the ‘cross-talk’ between alpha and beta-cells of the pancreas (Gerich et al. 1973).

The sympathoadrenal response is also progressively impaired and response to repeated hypoglycemia shifts to a lower glucose threshold. The adrenal response and secretory capacity is however undisturbed by stimuli other than a low BG (Meneilly, Cheung, and Tuokko 1994).

In comparison to patients without diabetes, whose epinephrine responses are normal, the risk of severe hypoglycemia is increased by at least 25-fold in patients with T1DM (Bolli et al. 1984). The attenuated sympathoadrenal, mainly sympathetic neural response, is responsible for the clinical syndrome of impaired hypoglycemia awareness in patients with T1DM (Cryer 2013). 20-25% of patients with T1DM are believed to have impaired awareness of hypoglycemia, which increases to approximately 50% after 25 years of disease (Iqbal and Heller 2016). Patients with

46

T1DM and impaired awareness of hypoglycemia have a six-fold increased risk experiencing a severe event (Geddes et al. 2008).

4.4.3 Counterregulatory Responses in Patients with T2DM

Counterregulatory responses to hypoglycemia includes inhibition of endogenous insulin secretion and stimulation of glucagon, catecholamines (norepinephrine and epinephrine), cortisol and growth hormone; all stimulating hepatic glucose production and decreased peripheral glucose utilization.

(Martin-Timon and Del Canizo-Gomez 2015)

(Permission not required – Open Access)

47

Early in the disease, the counterregulatory responses to hypoglycemia are intact in patients with T2DM.

(Zammitt and Frier 2005)

(Reproduced with permission)

As the duration of diabetes increases and endogenous insulin levels fall, the glucagon and the sympatho-adrenal responses are attenuated (Segel, Paramore, and Cryer 2002). However, irrespective of disease duration in patients with T2DM, poor glycemic control can impair the counterregulatory response and occur even at normal glucose levels (Spyer et al. 2000).

It is estimated that eight to ten percent of patients with T2DM have an impaired awareness of hypoglycemia. However, it remains unclear if oral agents including SUs contribute to an inability to perceive hypoglycemic symptoms (Iqbal and Heller 2016). Insulin-treated patients with T2DM have a 17-fold increased risk of incurring a severe hypoglycemic event compared to T2DM with intact awareness of their symptoms (Schopman, Geddes, and Frier 2010).

48

4.4.4 Hypoglycemia-Associated Autonomic Failure

(Cryer 2005)

(Reproduced with permission)

Hypoglycemic episodes of sufficient depth and duration impair counterregulatory responses leading to a vicious cycle of more frequent episodes and at a declining glycemic threshold to elicit defensive counterregulation. Cryer has denoted this sequence of events as Hypoglycemia- Associated Autonomic Failure (HAAF). This is not to be confused with the common neuropathic complication of diabetes (Cryer 1992). Interestingly, sleep and exercise can also cause a similar phenomenon (Cryer 2005). Even in patients without diabetes, an antecedent hypoglycemic episode can diminish the counterregulatory response for up to a week later (George et al. 1995).

The absent insulin and glucagon responses in patients with T1DM and advanced T2DM are well understood. However, the precise mechanism underlying HAAF, namely the attenuated sympathoadrenal response, is still not understood. Many theories have been postulated, which include alterations in afferent or efferent components of the sympathoadrenal response,

49

decreased adrenomedullary capacity to secrete epinephrine, cortisol, brain fuel transport, brain metabolism and the cerebral network hypothesis (Cryer 2013).

Recurrent hypoglycemia can lead to cellular adaptation and HAAF as depicted by Martin-Timon (Martin-Timon and Del Canizo-Gomez 2015).

(Permission not required – Open Access)

4.5 PROTECTIVE EFFECT OF HAAF

In contrast to the maladaptive view of HAAF, some have suggested that repeated hypoglycemia induces tolerance of a low BG by preconditioning. From an evolutionary perspective, this would improve survival during periods of insufficient caloric intake (McCrimmon and Sherwin 2010). In rat brains (at least), recurrent hypoglycemia led to less seizures during and less necrosis of neurons and cognitive impairment following a hypoglycemic episode (Puente et al. 2010).

4.5.1 HAAF on All-Cause Mortality

In a recent animal study, severe hypoglycemia induced lethal cardiac arrhythmias. Interestingly, however, mortality rats were highest in the streptozotocin diabetic rats (36%), followed by

50

controls (21%) and lowest in rats that were exposed to recurrent moderate hypoglycemia (4%) whom also had a lower epinephrine response (Reno et al. 2013).

4.5.2 HAAF on Preconditioning of the Brain

In rats, severe hypoglycemia induced brain damage and cognitive dysfunction, including impairments in spatial learning and memory. Interestingly rats exposed to recurrent moderate hypoglycemia had 62-74% less brain cell death and were protected from these cognitive impairments. This suggests that hypoglycemia, if moderate, may provide a beneficial adaptive response against the brain damage of severe hypoglycemia (Puente et al. 2010).

HAAF may very well be both adaptive and maladaptive. It may protect patients from the detrimental consequences of subsequent hypoglycemia, but also increase their risk of morbidity and mortality.

4.6 IMPACT OF AVOIDING HYPOGLYCEMIA

Generally, a glucose value of <3.0 mmol/L causes defective glucose counterregulation and impaired awareness and its avoidance can help reverse some of these aspects (International Hypoglycaemia Study 2017).

Promising results in patients with T1DM have been reported by 3 separate labs demonstrating that 2-3 weeks of scrupulous hypoglycemia avoidance can reverse the impaired awareness by increasing epinephrine levels (Fanelli et al. 1993; Dagogo-Jack, Rattarasarn, and Cryer 1994; Cranston et al. 1994).

Although the adaptive or maladaptive responses to hypoglycemia is debated, it is clear that avoidance of hypoglycemia in patients with diabetes is of utmost importance. Antihyperglycemic agents that do not increase risk of hypoglycemia in patients with T2DM relative to placebo can thus help ensure the pre-conditioning responses to a progressively lower threshold of hypoglycemia do not occur.

51

4.7 HETEROGENEITY OF HYPOGLYCEMIA

Reporting of hypoglycemia in the clinical literature is heterogeneous. Definition and classification of hypoglycemia, if provided, vary. Various nuances in study methodology also make comparisons of hypoglycemia risk challenging. Here we review some of these variables to consider for our meta-analysis.

4.7.1 Definitions

A standardized definition of hypoglycemia does not exist. Diabetes Organizations, pharmaceutical companies and clinical trials have differed in their definition of hypoglycemia. It has proven difficult to achieve consensus on a single definition that encompasses the different descriptors of hypoglycemia severity, presence or absence of the perception of symptoms, need for BG documentation, method of measurement or BG thresholds. In part, this is because glucose thresholds for hypoglycemia symptom perception and counterregulation vary amongst individuals with diabetes. As a result, estimates of hypoglycemia incidence and its meta- analyses have been difficult to ascertain and discussions related to even severe types have been limited (Iqbal and Heller 2016).

For severe hypoglycemia, a precise blood glucose threshold for clinical trial reporting has been debated. The internationally accepted terminology for pharmaceutical regulatory purposes (The Medical Dictionary for Regulatory Activities (MedDRA) includes a glucose level of <2.8mmol/L, a value recognized in Canada (Yale, Paty, and Senior 2018) and Japan (Namba et al. 2018). Yet, the American Diabetes Association (Association 2019) and the European Medicines Agency (EMA) (https://www.ema.europa.eu/.../draft-guideline-clinical-investigation- medicinal-product ) consider <3.0 mmol/L.

4.7.2 Study Type

Evaluation of hypoglycemia incidence may differ by study type. For instance, relative to observational studies, some may believe that randomized trials provide a higher estimate of hypoglycemia risk as they conduct more frequent and formal screening of adverse events throughout the active phase. Alternatively, however, the inclusion criteria may forbid patients with a history of hypoglycemia to enroll in the study or exclude patients experiencing events.

52

Interestingly, a recent systematic review found higher rates of hypoglycemia were observed in real-world data in comparison to randomized controlled trials in insulin-treated patients with T1DM and T2DM (Elliott et al. 2016).

Within observational studies, estimates of hypoglycemia risk may also differ between prospective vs retrospective studies, as recall of events may be affected in retrospective studies. However, a literature survey of severe hypoglycemia frequency in insulin-treated patients with T2DM reported that retrospective studies reported a greater incidence and proportion of patients with hypoglycemia compared to prospective studies (Akram, Pedersen-Bjergaard, Borch- Johnsen, et al. 2006). Unfortunately, few studies exist that assess hypoglycemia prospectively (Frier 2014).

4.7.3 Study Duration

The DiaRegis study was a prospective multicentre observational cohort from Germany with n=3,810 patients with T2DM. The study demonstrated that 11.4% of patients experienced hypoglycemia in the first year, whereas, 17.8% experienced any hypoglycemia in the second year (Tschope et al. 2016). Contrary to this, the second-year hypoglycemia risk occurred at lower rate than the first year in both the and glipizide treatment arms. Although, rising HbA1c levels and/or less frequent study visits may have been contributing factors (Nauck, Del Prato, et al. 2014).

4.7.4 Study Eligibility

Rates of hypoglycemia reported in the literature may also differ depending on whether studies excluded patients with a previous history of hypoglycemia. The DiaRegis study analyzed incidence of hypoglycemia based on prior history. They found that while 14.9% of patients with T2DM without a prior history reported experiencing any hypoglycemia, the number was much higher in patients with a prior history at 40.3% (Tschope et al. 2016).

4.7.5 Patient Attrition

Patient attrition is especially important to consider in studies with a primary safety objective. The number of participants in a study for whom no outcome data is available can vary across studies, and more importantly, between treatment arms of a study. As such, evaluation of 53

participant disposition and reasons for discontinuation can help inform of risk of attrition bias (Higgins et al. 2011).

Intent-to-treat methods are predominantly used as the primary method of analysis within the clinical literature. By measuring outcome data in all participants, intention-to-treat is considered a more conservative estimation, given the efficacy objective of most AHA interventional studies. However, for safety endpoints, intention-to-treat may underestimate the risk of an adverse or unintended effect as patients may discontinue or be excluded from the trial. Per-protocol analysis, on the other hand, refers to inclusion of participants who were in full compliance with the study protocol. This approach may also underestimate risk of hypoglycemia if patients withdraw or are withdrawn after an event. Thus, for our meta-analysis, it will be important to capture the statistical methods used within each study to ensure fair comparisons.

4.7.6 Ascertainment

Quantification on the occurrence of hypoglycemia can vary drastically depending on whether diagnostic codes are used compared to clinical notes. In a recent large US EHR database, ascertainment of overall hypoglycemia with clinical notes doubled that of diagnostic codes and a 20-fold difference was found for non-serious events (Nunes et al. 2016).

4.7.7 Patient Recall

Patient recall of severe hypoglycemia is believed to be accurate for up to one year in patients with T1DM (Pramming et al. 1991) and T2DM (Akram, Pedersen-Bjergaard, Carstensen, et al. 2006). Episodes of severe hypoglycemia, however, may occur unrecognized. Further, amnesia is common in patients with severe hypoglycemia, also resulting in potential under-reporting. An accurate evaluation of episodes would require careful prospective monitoring. Incidence of milder forms of hypoglycemia are even more difficult to accurately depict since accurate recall is estimated to be valid for only one week (Frier 2014). For instance, when symptoms were perceived, patient recall of milder episodes of hypoglycemia were accurate for one-week past event in patients with T1DM (Pramming et al. 1991). In contrast, recall of a severe hypoglycemic episode was accurate for up to a year in patients with T1DM and T2DM (Akram, Pedersen-Bjergaard, Borch-Johnsen, et al. 2006).

54

4.7.8 Patient Reluctance to Notify / Admit

Patients with diabetes may underestimate the frequency of hypoglycemia episodes knowingly for fear of judgement and its social, employment or licence implications. Alternatively, patients may underestimate the frequency of episodes unknowingly, due to symptom unawareness (Broz et al. 2015; Ratzki-Leewing et al. 2018). Interestingly, when family members of patients with T1DM are questioned, they typically report double the number of episodes relative to the patient’s report (Lawton et al. 2014).

4.7.9 Exclusion of Study Participants During Screening or Active-Trial Period

How investigators handle participants with hypoglycemia history or experiencing hypoglycemia during a trial must be considered. Most studies are efficacy focused and in randomized controlled studies, investigators can typically discontinue participants for a number of concerns, including hypoglycemia. This can occur during the screening phase, or during the active ‘in- trial’ phase. Participants who are enrolled in a trial and experience a hypoglycemia event are most often not included in the efficacy or safety data analysis. Thus, rates of hypoglycemia as evidenced from randomized controlled trials, as with other safety endpoints may not reflect the real-world incidence. Alternatively, real-world data, in which patients are often asked to recall events from the past, may also reflect an underestimate.

For every reported episode of severe hypoglycemia, there is likely one or more nocturnal severe hypoglycemia and for other hypoglycemia reports, the reported number is likely only a fraction of what is actually occurring. Therefore, preventing the occurrence of hypoglycemia altogether is of critical importance.

4.8 GUIDANCE FOR IMPROVING REPORTING OF HYPOGLYCEMIA IN THE CLINICAL LITERATURE

The ADA and the European Association for the Study of Diabetes recently proposed the following glucose levels for reporting of hypoglycemia in clinical trials.

LEVEL ONE: A glucose alert value of 3.9 mmol/L (70 mg/dL) or less. This need not be reported routinely in clinical studies, although this would depend on the purpose of the study.

55

LEVEL TWO: A glucose level of <3.0 mmol/L (<54 mg/dL) is sufficiently low to indicate serious clinically important hypoglycemia. This is the recommended value to be included in clinical trials.

LEVEL THREE: Severe hypoglycemia, as defined by the ADA, denotes severe cognitive impairment requiring external assistance for recovery (Group 2017).

4.9 EPIDEMIOLOGY

4.9.1 Intensive vs Less Intensive Glucose Lowering RCTs

In RCTs with both T1DM and T2DM, more intensive glucose lowering was associated with a significant higher risk of experiencing hypoglycemia. These studies defined hypoglycemia as a plasma glucose level of <2.8 mmol/L (Rana, Byrne, and Greaves 2014).

(Reproduced with permission)

56

4.9.2 Incidence in Patients with T1DM vs T2DM

Incidence of hypoglycemia is higher in insulin-treated patients with T1DM vs those with T2DM, although a number of variables must be considered. For instance, when duration of disease is accounted for, the incidence of hypoglycemia in patients with T2DM is similar to patients with T1DM (Frier 2014). Similar incidences have also been reported when duration of insulin therapy is accounted for (Hepburn et al. 1993). However, Since there are more patients with T2DM than T1DM, more episodes of severe hypoglycemia occur in patients with T2DM (approximately 20X).

The Global Hat Study reported that hypoglycemia rates occur at a higher frequency than previously believed. Over a 6-month retrospective and 4-week prospective period, self assessment questionnaires and diaries from n=27,585 people with diabetes from 24 countries were collected. All patients were treated with insulin for greater than 12 months. The study demonstrated that the majority of patients experienced hypoglycemia and close to 50% of patients with T2DM experienced any form of hypoglycemia in a four-week period (Khunti et al. 2016).

T1DM T2DM

(n=8,022) (n=19,563)

Population Characteristics

Male/female, % 48/52 53/47

Mean age, years 42 61

Duration of diabetes, years 18 14

Duration of Insulin use, years 17 6

HbA1c, %) 7.9 8.0

57

6-month Retrospective

History of hypoglycemia, % 97 78

4-week Prospective

Reporting of Hypoglycemia, % 83.0 46.5

Estimated Annual Rates

Any events/patient-year 73.3 19.3

Nocturnal events/patient-year 11.3 3.7

Severe events/patient-year 4.9 2.5

(Permission not required – Open access)

In insulin-treated patients, the Canadian cohort of the Hypoglycemia Assessment Tool (HAT) recently reported on the prevalence of hypoglycemia during a 6-month retrospective and 4-week prospective period. They found the prevalence of hypoglycemia to be similar retrospectively (T1DM 92.3%, T2DM 63.5%) and prospectively (T1DM 95.2%, T2DM 64.2%) (Aronson et al. 2018).

The last decade has witnessed greater number of systematic studies on hypoglycemia, mainly the severe form. Despite this trend, an underestimation is likely given that many cases are not diagnosed. Estimates of severe hypoglycemia both from observational and RCT data have been well documented in patients with T1DM and T2DM on insulin, but less so for patients with T2DM on oral agents, and even less is known about the less severe forms (Halimi 2015). Patient recall of severe events is robust for up to a year, however, events that are unrecognized and amnesia are common (Zammitt and Frier 2005).

4.9.3 Incidence of Severe Hypoglycemia

Hospitalizations for older US adults was assessed using adverse-event data between 2007 to 2009. Second to hematologic agents, endocrine agents with hypoglycemia as the adverse-event 58

manifestation was the second therapeutic category for hospitalizations. Specifically, insulin and oral hypoglycemic agents accounted for 13.9% and 10.7% of all adverse medication related hospitalizations (Budnitz et al. 2011).

(Reproduced with permission from NEJM, Copyright Massachusetts Medical Society.)

4.9.4 Incidence of Non-Severe Hypoglycemia

Mild and moderate hypoglycemia, often referred to as non-severe hypoglycemia (NSH) is rarely reported in clinical trials. A recent meta-analysis found insufficient data to provide any estimates for NSH (Edridge et al. 2015). Nevertheless, other studies estimate that patients with T1DM experience an incidence of mild hypoglycemia of 1-2 episodes per week compared to 0.3- 0.7 episodes in patients with T2DM treated with insulin (Frier 2014).

A questionnaire of European patients with T1DM or insulin-treated T2DM was conducted to better understand the incidence of NSH. Of n=3,827 respondents, the mean number of events per respondent week was 1.8 per T1DM and 0.4-0.7 for T2DM patients. Sixty-three percent of T1DM and 49-64% of T2DM reported either impaired hypoglycemic awareness or unawareness. Furthermore, 65% of T1DM and 50-59% of T2DM respondents rarely or never informed their general practitioner or specialist about hypoglycemia. Sixteen percent of T1DM respondents and 26% of T2DM respondents reported not being asked about hypoglycemia during routine appointments (Ostenson et al. 2014).

59

4.9.5 Incidence of Nocturnal Hypoglycemia

Incidence of nocturnal hypoglycemia has been difficult to accurately describe in the literature. At night, symptoms of hypoglycemia are absent, and monitoring of BG is rarely performed. In both patients with and without diabetes, sleep has been demonstrated to impair the counterregulatory hormones to hypoglycemia. A rise in epinephrine, the main counterregulatory hormone of hypoglycemia is impaired during sleep (Jones et al. 1998).

In patients with T1DM, 50% of all severe episodes of hypoglycemia occur at night (Frier 2014). Similarly, in patients with T2DM, an Australian study of n=25 patients treated with SU revealed that 54% of hypoglycemia <2.8 mmol/L and 51% of borderline events (2.9 to 3.6) occurred between 10pm and 6am. Again, none of these episodes were recognized by the patients (Hay, Wilmshurst, and Fulcher 2003).

(Permission not required – open access)

A multicentered randomized CGM study (HypoDE) of n=126 patients from Germany with T1DM on MDI insulin therapy and a history of hypoglycemia was conducted. Bedtime glucose levels were found to be elevated, perhaps to avoid nocturnal hypoglycemia which worked in the first half of the night (10pm -2am), but not the second (2am-6am) (Heinemann et al. 2018).

4.9.6 Incidence of Hypoglycemia Unawareness

An inability to perceive symptoms of hypoglycemia is common, leading to underreporting in both T1DM and T2DM. In a study of n=13 middle-aged (39-64) and n=13 older (≥65 years)

60

patients with T2DM, hormonal subjective and cognitive responses were evaluated during a 30- min steady state hypoglycemia at 2.8 mmol/L. Similar BG and insulin concentrations were observed between groups at baseline during steady state hypoglycemia. In comparison to euglycemia, hypoglycemia caused an increase in both middle aged and older patients in epinephrine, norepinephrine ACH, cortisol, growth hormone and a reduction in glucagon. Counterregulatory hormones were generally and non-significantly lower in the elderly. At the beginning and end of the hypoglycemic clamp, patients were questioned on their autonomic (anxiety, palpitations, hunger, sweating, irritability, tremor) and neuroglycopenic (dizziness, tingling, blurred vision, difficulty to concentrate and faintness) symptoms. Despite similar patient reported symptoms between the two groups at the beginning of the hypoglycemic interval, symptoms drastically differed by the end. Where middle-aged patients reported an increase in symptoms, older patients failed to detect a difference from baseline in neuroglycopenic or autonomic symptoms during the hypoglycemic clamp. Seven of thirteen middle aged and only 1/13 older patients correctly estimated their BG levels to be <3.3 mmol/L. Furthermore, reaction time was significantly longer in older patients and remained prolonged in both groups even after 30 minutes of euglycemia had been achieved (on average 57 minutes for middle aged and 82 minutes for older patients) (Bremer et al. 2009).

(Reproduced with permission)

61

Continuous glucose monitoring measures the glucose levels in the interstitial space. CGM studies have demonstrated that frequency of asymptomatic events in patients treated with insulin is common (Schopman et al. 2014). For instance, in n=25 insulin-treated patients with longstanding T2DM with a history of CVD or CV risk factors, only 9% (3/34) of hypoglycemia episodes (defined as BG <3.5 mmol/L) were symptomatic. Such a population, given their CV status and long disease duration (17 years), is also likely accompanied by a blunted counterregulatory response (Chow et al. 2014).

In a German study of n=95 patients with T2DM treated with insulin or SU and established CVD wearing simultaneous holter and continuous glucose monitors, most hypoglycemic episodes were not perceived by patients. In fact, only 7 daytime (38.9%) and 4 nighttime (11.1%) episodes were symptomatic (Pistrosch and Hanefeld 2015). An Australian study of n=25 elderly patients with T2DM (>65 years of age, 21 men and 4 women) using continuous glucose monitoring, treated with SU (with or without metformin) found that 80% experienced hypoglycemic events (<2.8 mmol/L). None, of the 103 episodes however were recorded in their diaries as instructed (Hay, Wilmshurst, and Fulcher 2003). Finally, a pilot study from the Netherlands of n=23 frail elderly patients with T2DM treated with SU showed that 22% (5 patients) had 15 episodes of hypoglycemia of <3.0 mmol/L as confirmed by CGM. Eight patients had 25 events of levels <3.5 mmol/L. Again, none of the patients reported experiencing any symptoms related to hypoglycemia (van Dijk et al. 2017).

4.9.7 Duration of Hypoglycemia

In a study of T2DM patients >65 years of age (n=21 men) suggested that the average duration of hypoglycemia was 78 minutes. Patients were treated with SU (with or without background metformin) and wore continuous monitors for 12 days (Hay, Wilmshurst, and Fulcher 2003).

62

(Permission not required – open access)

4.9.8 Incidence with SMBG

Nocturnal glucose, but not daytime levels were found to be lower with CGM compared to SMBG (Zick et al. 2007).

4.9.9 Incidence with CGM

Capillary and interstitial BG measurements can vary, especially during the night, effecting the estimation of nocturnal hypoglycemia (McGowan, Thomas, and Moran 2002).

4.9.10 CGM Incidence with Sulfonylureas

A major limitation of continuous glucose monitoring studies is that many patients cannot tolerate them. One study found 52% of patients >65 years of age could not tolerate the device for a full 12 days and 40% stated an inability to cope with the monitor (Hay, Wilmshurst, and Fulcher 2003). Nevertheless, incidence of SU-induced hypoglycemia was evaluated in n=23 frail elderly patients with T2DM living in the Netherlands. After a median 97 hours of monitoring, five patients experienced 15 events with a BG of <3.0 mmol/L and eight patients experienced 25 events with a BG of 3.5 mmol/L. Concerningly, all episodes were asymptomatic. Although there were still a significant number of hypoglycemic events with gliclazide, numerically, it was associated with less events and less duration of events relative to glimepiride (van Dijk et al. 2017).

63

4.9.11 Real-World Incidence

The Supreme DM Study group, a network of US diabetes researchers from 11 organizations found the annual rate of severe hypoglycemia to range from 1.4 to 1.6 events per 100-person- years. This cohort consisted of 917,440 adults with T1DM and T2DM. It was assumed that 95% of the population consisted of patients with T2DM, however, given the use of electronic medical records, they were unable to delineate precise rates between patients with T1DM vs T2DM. The risk of severe hypoglycemia was found to be higher with advanced age, kidney disease, congestive heart failure, depression, higher HbA1c, and in patients taking insulin, insulin secretagogues or beta blockers (Pathak, Schroeder, Seaquist, Zeng, Lafata, Thomas, Desai, Waitzfelder, Nichols, Lawrence, Karter, Steiner, Segal, and O'Connor 2016).

4.9.12 Real-World Incidence with Insulin

The frequency of severe hypoglycemia is greater in patients with T1DM than in patients with T2DM. The difference, however, is much less when the duration of insulin treatment is taken into consideration (Frier 2014). This may reflect a decline in endogenous insulin production leading to a diminished release of glucagon and more variable insulin levels (Iqbal and Heller 2016).

A number of studies from around the world have reported on the incidence of insulin-induced hypoglycemia. In Tayside Scotland, a random sample of n=267 insulin-treated patients with T1DM or T2DM were selected from a diabetes registry to record symptoms of hypoglycemia in a diary. Patients with T1DM self-reported 42.9 events per patient per year for overall and 1.2 for severe hypoglycemia. In those with T2DM, overall hypoglycemia occurred in 16.4 events per patient per year and severe 0.4 events per patient per year (Donnelly et al. 2005). In Canada, the InHypo-DM cross sectional cohort study surveyed n=458 people with T2DM on either insulin monotherapy (35%), SU monotherapy (47%) or a combination of both (18%). They found the incidence of hypoglycemia to be highest in patients on combination therapy than either therapy alone, especially for severe episodes (3.7 events/person-year) (Ratzki-Leewing et al. 2018). In a UK study, GP records of n=5,974 T2DM patients over 75 years of age on insulin or SU were reviewed. Hypoglycemia events of significance were found in 4.9% of men and 5.1% of

64

women. Similar to ACCORD, prevalence of hypoglycemia increased with higher HbA1C (Heald et al. 2018).

Recently, a large systematic review and meta-analysis of population-based studies was conducted consisting of n=532,542 patients with T2DM taking oral agents and insulin. They estimated that severe hypoglycemia occurs in 21% of patients taking insulin, whereas mild-to- moderate hypoglycemia occurred in 52% of patients. In patients taking SU, severe hypoglycemia was estimated at 5% and mild-to moderate was estimated at 33%. Although, there was insufficient data to estimate mild-to-moderate hypoglycemia with use of non-insulin and non-sulfonylurea treatment, severe hypoglycemia occurred in a similar prevalence as SU users at 5%.

Hypoglycemia Mild-to- Severe Moderate

Overall

Prevalence 45% 6%

Incidence per person-year 19 0.8

Insulin (with or without oral glucose lowering therapies)

Prevalence 52% 21%

Incidence per person-year 23 1

Sulfonylurea (non-insulin but with or without other oral-glucose lowering therapies)

65

Prevalence 33% 5%

Incidence per person-year 2 0.01

Non-Sulfonylureas (non-insulin and non-sulfonylurea oral-glucose lowering therapies)

Prevalence Insufficient data 5%

Incidence per person-year Insufficient data Insufficient data

(Edridge et al. 2015).

(Permission not required – Open Access)

4.9.13 Real-World Incidence with Sulfonylureas

Data from Germany and Austria found the rate of severe hypoglycemia in SU-treated patients with T2DM to be 3.9/100 patient-years (Schloot et al. 2016). Although a recent study suggested that the prevalence of severe hypoglycemia with SUs is similar to non-SUs, at 5% (Edridge et al. 2015), data from Europe suggests otherwise. In the German DiaRegis registry, hypoglycemia risk of n=3,808 patients with T2DM were evaluated while on monotherapy vs dual oral therapy. Patients with hypoglycemia were more likely to be on SUs and less likely on metformin, TZD or DPP4i. SU use was also an independent predictor of hypoglycemia (OR 2.58 CI 2.03-3.29) (Tschope et al. 2011).

4.9.14 Incidence of Hypoglycemia in the Cardiovascular Outcomes Trials

Newer therapies, namely the DPP4i, GLP1RA and SGLT2i, in theory should not cause hypoglycemia, yet mixed results have been observed in the large CVOTs. Although all CVOTs were conducted on top of standard of care and thus background SU or insulin use was permitted, risk of major (severe) hypoglycemia was significantly increased in the SAVOR-TIMI 53 trial with relative to placebo (2.1 vs 1.7%, p=0.047). The primary endpoint of hospitalization for hypoglycemia was comparable between arms (Scirica et al. 2013). Post hoc subgroup analyses concluded that the increased hypoglycemic events were attributable to

66

background SU treatment. Yet, the placebo arm at baseline and throughout the study had greater use of sulfonylurea. Other CVOTs have shown no difference in hypoglycemic rates between treatment arms, whereas in the LEADER trial, a significant lower rate of hypoglycemia was observed with vs placebo (Marso, Daniels, Brown-Frandsen, Kristensen, Mann, Nauck, Nissen, Pocock, Poulter, Ravn, Steinberg, Stockner, Zinman, Bergenstal, Buse, et al. 2016).

In the ORIGIN trial, of n=12,537 individuals with diabetes, n=3,518 experienced at least one episode of hypoglycemia during a follow up period of 6.2 years. The majority of hypoglycemia events (74.3%) occurred in the glargine group, while 25.7% occurred in the standard group and n=472 participants experienced at least one episode of severe hypoglycemia (≤2.0 mmol/L), in the same proportion. This reflects an annual incidence of 0.9% and 0.3% for the glargine and standard group respectively (Investigators 2014).

(Reproduced with permission)

4.9.15 Incidence in Racial or Ethnic Groups

It is not well understood how the prevalence of hypoglycemia varies in different racial and ethnic groups. Most of the data on the epidemiology of hypoglycemia comes from high income countries in the West, and its applicability to the rest of the world is not known (Frier 2014). The SUPREME study, a large 7-year US surveillance study from 2005-2011 demonstrated a substantially greater risk of hypoglycemia in African Americans. Annual rates of severe hypoglycemia ranged from 1.8% to 2.1%. Over the 7-year period, concerning trends showed an increased rate only amongst African Americans (Karter et al. 2017).

67

(Reproduced with permission)

Emergency department (ED) visit rates for hypoglycemia was found to be on the decline in a Veteran’s Administration (VA) database analysis of older patients on insulin from 2004-2013. However, despite the decline, blacks continue to have higher rates of ED visits compared to whites, a pattern which persists in all comorbid subgroups (Tseng, Soroka, and Pogach 2018).

4.9.16 Incidence by Frailty

A cross sectional analysis of older participants with T1DM and T2DM (n=1,826) from the Atherosclerosis Risk in Community (ARIC) study, was conducted to determine the relationship between frailty scores and incidence of severe hypoglycemia. Frailty was defined as 3 or more of the following; weight loss, low physical activity, slow walking speed, exhaustion or weak grip strength and pre-frailty was defined as 1-2 of the components. Mean age was 75 years, 45% were male and 29% were black. Prevalence of hypoglycemia occurred in 6.0%, 3.2% and 1.2% of the frail, prefrail and robust participants (Lee et al. 2017).

4.9.17 Regional Incidence

In patients with T1DM, northern Europe, Canada and Latin America reported the highest rates of hypoglycemia in the Global Hat study. Russia and Eastern Europe reported the highest rates in patients with T2DM (Khunti et al. 2016).

68

4.10 TEMPORAL TRENDS

4.10.1 Canada

Using linked medical databases from Ontario between 2002 and 2013, Clemens et al. reported that the absolute number of hypoglycemia increased until mid-2006, while the overall percentage declined over the period from 2002 to 2013 (Clemens et al. 2015). This trend was corroborated using an administrative and census database from Ontario between 1994 and 1999. They found rates of hospitalization for hyper and hypoglycemia emergencies decreased by 32.5 and 76.9% respectively. In fact, overall diabetes-related emergency visits declined by 23.9% (Booth et al. 2005).

4.10.2 US

A retrospective analysis from n=33,953,331 Medicare beneficiaries aged 65 years or older from 1999-2011 found that rates of admissions for hyperglycemia declined by 38.6% whereas rates of hypoglycemia increased by 11.7%. More specifically, rates for females and whites declined, where as they remained high (but stable) for blacks and increased for all age 75 and over. However, after adjusting for the changing diabetes prevalence, admissions for hyperglycemia and hypoglycemia were found to be on the decline by 55.2% and 9.5% respectively (Lipska et al. 2014).

The same authors found that rates of severe hypoglycemia remained the same from 2006 to 2013 (1.3 per 100 person-years), despite a deterioration of glycemic control over the same period. Although use of agents that carry a low risk of hypoglycemia and a reduction in SU use was observed during this period, a significant increase in the use of insulin (mostly analogs) may explain why the overall rates of severe hypoglycemia did not improve. When analyzed by age, the oldest patients demonstrated a reduction in rates of severe hypoglycemia (from 2.9 to 2.3 per 100 person-years) (Lipska et al. 2017). However, even prior to the approval of new AHA, between 1993 and 2005, the visit rate for people with diabetes did not change. Interestingly, rates of hypoglycemia were higher in patients <45 and ≥75, in females (vs males) and in blacks (vs whites) (Ginde, Espinola, and Camargo 2008).

69

Evidence also exists to suggest an increase in the recent temporal trends of hypoglycemia in patients with diabetes. Between 2007-2013, in inpatient and outpatient hypoglycemia were evaluated 12 months after initiation of SU (n=230,556) or DPP4i (n=133,768) (monotherapy or combination with other AHA). An increase in outpatient hypoglycemia rates per 100-person- years was observed with both SU’s (4.4 to 9.3) and DPP4i’s (3.1 to 5.4). Inpatient hypoglycemia ranged from 1.0 to 1.3 in SU users and 0.2-0.3 in DPP4i’s (Tang et al. 2017).

4.10.3 Europe

The use of insulin analogues significantly increased in Germany and Austria between 2002 and 2014 (Bohn et al. 2016). In Germany, despite an increased use of new medications that presumably do not cause hypoglycemia, incidence of severe hypoglycemia increased. Between 2006 and 2011, use of SU and human insulin decreased, whereas use of incretin-based therapies and insulin analogues was on the rise. Although, HbA1c and causes of severe hypoglycemia was not detectable from the study design, the increased incidence may have been attributable to increase use of insulin analogues (Muller et al. 2017). Over the last 10 years in England, hospital admissions for hypoglycemia increased in absolute terms. However, after accounting for the diabetes prevalence, a reduction of admission rates was found. Further, an increase in same day discharge and a decrease in in-hospital mortality decreased (Zaccardi et al. 2016).

4.10.4 Asia

Emergency room (ER) visits from patients with T2DM in Korea between 2004 and 2009 showed an increase in the number of individuals with severe hypoglycemia. The study however did not consider the increasing prevalence of T2DM in Korea, which has steadily increased in the last 30 years (Kim et al. 2011). Taiwan has also reported increasing trends in ER visits for hypoglycemia in patients with T2DM, but again, the prevalence of diabetes was not accounted for (Chen, Yang, et al. 2015). In a Japanese study, patients with T2DM with severe hypoglycemia were divided into 3 groups based on their visit period of 2008-2009, 2010-2011 and 2012-2013. Overall, use of insulin, , glinides and DPP4i significantly increased, while use of SUs decreased. Patients with severe hypoglycemia on insulin, biguanides, glinides and DPP4i significantly increased, while those on SU experiencing a severe event significantly

70

decreased. Authors reported errors of drug use as a trigger that significantly increased risk of hypoglycemia (Ito et al. 2016).

Overall, studies on the temporal trends of hypoglycemia are highly heterogeneous and must account for the changing diabetes prevalence, use of insulin, SU and new AHA, age and ethnicity, to name a few key variables.

4.11 RISK FACTORS

4.11.1 Age

It is well recognized that elderly individuals are at an increased risk of hypoglycemia. In the SUPREME DM Study, an observational cohort of n=917,440 adults, advanced age was found be a strong risk factor for hypoglycemia (Pathak, Schroeder, Seaquist, Zeng, Lafata, Thomas, Desai, Waitzfelder, Nichols, Lawrence, Karter, Steiner, Segal, O'Connor, et al. 2016). In a retrospective analysis of n=887,182 patients with T2DM and hypoglycemia-related hospitalizations in the USA, age, SU, insulin and renal disease were found to be most associated with predicting hospitalizations associated with hypoglycemia (Fu et al. 2014).

4.11.2 Antecedent Hypoglycemia

An episode of hypoglycemia, irrespective of its severity, can beget a subsequent episode. In an observational study of elderly T2DM patients from Alberta, patients hospitalized for hypoglycemia were independently associated with subsequent hospitalizations in a dose- dependent manner, recurrent hypoglycemia and mortality (Majumdar et al. 2013). In the ORIGIN trial, NSH was independently associated with severe hypoglycemia (Investigators 2014). In addition, the PREDICTIVE study, a global prospective observational study with T1DM (n=7,420) and T2DM (n=12,981) starting was associated with frequency of NSH was significantly associated risk of severe hypoglycemia (Sreenan et al. 2014).

In a post hoc analysis of 15 insulin trials, baseline hypoglycemia, both nocturnal and total, significantly correlated with post baseline events. Adjusting for history of hypoglycemia reduced the standard of error and improved statistical power in T1DM or T2DM patients previously treated with insulin, but not in T2DM who are insulin-naïve. The authors also found

71

that trials with longer lead-in periods (ie from screening to randomization) strengthened the correlation between baseline to postbaseline hypoglycemia (Luo et al. 2013).

Recent US data from 100 million diverse Medicare beneficiaries analyzed readmission rates after severe hypoglycemia and hyperglycemia. Age was found to be the strongest independent predictor for hypoglycemia. Significant predictors of hyperglycemia were younger age and comorbidities. After an admission for hypoglycemia, 0.4% were admitted for hypoglycemia and 4.3% for hyperglycemia. Interestingly, after hyperglycemia admission the opposite pattern occurred; only 0.2% were readmitted for a recurrent hyperglycemia episode compared to 1.3% for new hypoglycemia (McCoy et al. 2018).

Thus, sufficient evidence exists to support that antecedent hypoglycemia or hyperglycemia begets a subsequent episode. Avoidance of the first event is of great importance, especially in populations with diabetes at greatest risk, such as the elderly and those with co-morbid conditions. AHA with no increased risk of hypoglycemia should be considered first-line.

4.11.3 Cardiovascular Disease and Hypoglycemia

TECOS was the first clinical trial to demonstrate a bidirectional relationship between CV events and hypoglycemia (see 4.26.21) (Standl et al. 2018). Observational studies have also demonstrated an association between CV events and hypoglycemia. For instance, a Japanese prospective cohort from 2001 to 2012 of n=624 patients, demonstrated CVD history to be a strong predictor of severe hypoglycemia. Other variables demonstrating a strong ability to predict severe hypoglycemia included age (per 10 years), diabetes duration (≥10 years), insulin use, retinopathy, or autonomic neuropathy. CVD history remained a significant predictor even after adjusting for sex, age, diabetes duration, hypertension, mean HbA1c, diabetic nephropathy, diabetic retinopathy, insulin, ACEi, ARB and CV autonomic neuropathy stage (Yun and Ko 2015).

Patients with cardiovascular disease need to be carefully evaluated for risk of hypoglycemia and antihyperglycemic agents that have demonstrated no increased risk should be considered. Given the potential bi-directional nature of hypoglycemia and CV events, avoidance of hypoglycemia can also reduce risk of a subsequent CV event.

72

4.11.4 Coronary Artery Disease and Hypoglycemia

Post hoc analysis from the VADT trial showed that patients who experienced severe hypoglycemia were more likely to have a higher baseline score of CAC, but not history of CVD. In addition, they were more likely to be treated intensively, with SU or insulin, Higher age, creatinine, triglycerides, HDL cholesterol, diabetes duration, lower systolic, diastolic blood pressure (although hypertension was almost significant), C-peptide, HbA1c and prior hypoglycemia (Saremi, Bahn, and Reaven 2016).

The DiaRegis study found patients with T2DM and vascular disease were at significantly increased risk of severe hypoglycemia even after adjusting for confounders. Interestingly, significantly more patients with vascular diseases were taking SUs and less were taking metformin (Gitt et al. 2012).

4.11.5 Chronic Kidney Disease and Hypoglycemia

Even in patients without diabetes, rates of hypoglycemia are greater in patients with CKD than those without CKD (3.46 vs 2.23 per 100 patient-months, respectively). In patients with diabetes, CKD doubles the risk relative to those with diabetes but no CKD (10.72 vs 5.33 per 100 patient months) (Moen et al. 2009).

4.11.6 Clinical Complexity / Frailty Status and Hypoglycemia

Using administrative data base claims from the US, n=31,542 noninsulin-treated patients with T2DM were evaluated on the association of clinical complexity and hypoglycemia. Patients with high clinical complexity showed a significant risk for hypoglycemia, irrespective of their glycemic control (Steinberg 2016).

An 11-year retrospective study of people with newly diagnosed diabetes from Taiwan evaluated diabetes progression using an adapted Diabetes Complication and Severity Index (aDCSI) to risk of severe hypoglycemia. Patients with a slow progression of diabetes severity had a HR of 1.10 CI 1.04-1.16) compared to those with a rapid progression HR 5.23 CI 5.00-5.47) (Kornelius et al. 2017).

73

4.11.7 Dementia and Hypoglycemia

A number of studies suggest dementia increases the risk of hypoglycemia (Biessels 2009). In a longitudinal study from 1980 -2007 of Kaiser Permanente of Northern California health claims, the number of hypoglycemia episodes requiring hospitalization were dose-dependently associated with dementia risk; one episode HR 1.26, two episodes HR 1.89 and three episodes HR 1.94 (Whitmer et al. 2009). Post hoc retrospective analysis of the ADVANCE trial suggested that severe but not mild cognitive dysfunction increased the risk of severe hypoglycemia. Assignment of treatment groups did not impact the association (de Galan et al. 2009). Finally, prospective analysis of the ACCORD trial found a poorer baseline DSST predicted a first episode of a severe hypoglycemia event. In addition, those with a greater decline in cognitive function over a 20-month period further increased the risk. Importantly, the authors found no association of dementia to intensive or standard treatment group (Punthakee et al. 2012).

The relationship between hypoglycemia and dementia is bi-directional; hypoglycemia increases the risk of dementia and dementia increases the risk of hypoglycemia (Mattishent and Loke 2016). This was also demonstrated in the prospective study of Health Aging and Body Composition Study which began in 1997. Older adults, n=783 with a modified mini-mental state examination score of 80 or higher, were followed for 12 years. Hypoglycemia increased the risk of dementia 2.1-fold and those who developed dementia had a 3.1-fold increased risk of experiencing a hypoglycemia event after adjusting for confounders and comorbidities. Several possible mechanisms may explain the bidirectional relationship, including brain damage caused by hypoglycemia and an inability to perceive symptoms in patients with cognitive impairment (Yaffe et al. 2013).

4.11.8 Depression and Hypoglycemia

A database analysis of 30 primary care physicians in the Washington area found depression increased risk of hypoglycemic events. Significance was maintained for this relationship even after adjusting for multiple demographic (age, sex, race, education, marital status) and clinical characteristics (diabetes duration, insulin use, rxrisk score, hypertension diagnosis, T1DM or T2DM complication score) (Katon et al. 2013).

74

4.11.9 Duration of Diabetes and Hypoglycemia

In the Global Hat Study, duration of diabetes did not impact rates of any hypoglycemia (Khunti et al. 2016). However, in a prospective observational study from New Mexico involving n=265 T2DM patients on insulin +/- SU and using SMBG, duration of diabetes correlated significantly with number of NSH events (Wendel et al. 2014).

4.11.10 Duration of Treatment and Hypoglycemia

A German primary care database of n=19,184 DPP4i and n=31,110 SU users demonstrated that severe hypoglycemia increased with increasing duration of treatment (Rathmann et al. 2013).

4.11.10.1 Duration of Insulin Treatment and Hypoglycemia

In patients with T2DM, duration of insulin treatment was significantly associated to any hypoglycemia, nocturnal and severe hypoglycemia. In contrast, in those with T1DM, duration of insulin treatment was associated with only severe and nocturnal hypoglycemia risk (Khunti et al. 2016).

4.11.11 Food Insecurity and Hypoglycemia

Patients with T2DM from San Francisco, California, Chicago and Illinois were interviewed to evaluate food insecurity and outcome of severe hypoglycemia. Of 711 patients, 46% were food insecure. Food insecurity was significantly associated with risk of 4 or more episodes of severe hypoglycemia. In fact, food insecurity was a better predictor than HbA1c, renal disease, income, age, race or education. Interestingly, patients with food insecurities had higher HbA1c levels. Food insecurity was associated with both hypoglycemia and hyperglycemia. Therefore, stricter glycemic control does not explain the association of food insecurity and hypoglycemia risk. The authors point out that food insecurity consists of varying intervals of adequate food supply, during which BG levels may be elevated, and food scarcity which may decrease BG levels (Seligman et al. 2012).

4.11.12 HbA1c Levels and Hypoglycemia

Risk of hypoglycemia is often a concern when considering treatment intensification of patients not at goal. It is often assumed that risk of hypoglycemia is greater in patients with a low

75

HbA1c. However, many studies have refuted this notion by finding hypoglycemia risk to be a greater risk in patients with a higher HbA1c. For instance, in the Supreme DM Study, a US retrospective cohort of 917,440 adults with T1DM and T2DM, severe hypoglycemia was found to be associated with higher HbA1c levels (Pathak, Schroeder, Seaquist, Zeng, Lafata, Thomas, Desai, Waitzfelder, Nichols, Lawrence, Karter, Steiner, Segal, O'Connor, et al. 2016). Healthcare records of n=5,974 patients with T2DM over the age of 75 was conducted in Cheshire UK. HbA1c levels and its correlation to hypoglycemia risk while on SU or insulin was evaluated. They also found risk of hypoglycemia to be highest in patients with the highest HbA1c category, suggesting erratic glycemic control maybe a risk factor predisposing patients to hypoglycemia (Heald et al. 2018).

Aside from baseline A1c, changes in glycemic control may have an impact on hypoglycemia risk. Hypoglycemia has always been more frequent in participants randomized to intensive vs less intensive glycemic control (Ray et al. 2009). However, the relationship of hypoglycemia risk in patients with treatment intensification vs level of glycemic control attained require clear separation. In patients whose glycemic control is improved, risk of hypoglycemia may be reduced. In ACCORD, risk of hypoglycemia requiring medical assistance increased with a 1% increased in the average HbA1c (including baseline HbA1c measurement) (Standard arm HR 1.76 CI 1.50 – 2.06 and intensive arm HR 1.15 CI 1.02 – 1.21). In addition, risk of hypoglycemia requiring medical assistance according to a 1% fall in HbA1c from baseline to 4- month visit was reduced by 28% and 14% in the intensive vs standard arms (Miller et al. 2010). ACCORD also demonstrated that risk of hypoglycemia is increased in patients who attempt but are unable to lower their blood glucose levels (Drake et al. 2016).

Findings questioning the existence of a relationship between HbA1c and hypoglycemia at all also exists. The Global Hat study, which surveyed over 27,000 insulin-treated patients with T1DM and T2DM did not find any significant association of HbA1c to the percentage of patients reporting hypoglycemia during a 4week prospective period using patient questionnaires and diaries (Khunti et al. 2016). Similarly, the Canadian Cohort of the HAT study also failed to show a correlation between HbA1c and any or severe hypoglycemic event (Aronson et al. 2018), as did a literature search conducted on frequency and risk factors of severe hypoglycemia in insulin-treated T2DM (Akram, Pedersen-Bjergaard, Borch-Johnsen, et al. 2006). In a New

76

Mexico prospective study involving n=265 T2DM patients on insulin +/- SU, baseline, 12-week nor 24-week HbA1c showed a significant correlation to frequency of NSH confirmed by SMBG. Age, sex, race also showed not association with number of events (Wendel et al. 2014).

The relationship of hypoglycemia risk with HbA1c may likely be U shaped. In the Diabetes and Aging study, among 9,094 survey respondents with T2DM in Northern California, 985 (10.8%) reported having experienced a severe hypoglycemic event. Hypoglycemia was found to be higher in patients with either near-normal or poor glycemic control. This was true even after adjusting for confounders such as age, sex, race diabetes duration, prior history of hypoglycemia, comorbidities, polypharmacy or category of diabetes medications.

(Lipska et al. 2013)

(Reproduced with permission)

A similar u-shaped trend of baseline HbA1c and hypoglycemia occurred in the SAVOR-TIMI 53 trial, with a clearer trend for major compared to any hypoglycemia (Cahn et al. 2016).

4.11.12.1 Hypoglycemia Severity and HbA1c

In the ORIGIN trial, the relationship of hypoglycemia risk to on-treatment HbA1c differed by treatment arm and severity of hypoglycemia. For instance, lower on treatment HbA1c increased the incidence of NSH. Higher on treatment HbA1c increased the risk of severe hypoglycemia in the standard arm more so than glargine arm. Further, the study demonstrated that on baseline HbA1c and on-treatment HbA1c should be considered as separate variables as they differ in their prediction of events. Higher baseline HbA1c was an independent predictor of NSH, yet, higher on treatment HbA1c was associated with less NSH. Independent predictors of NSH included 77

younger age, South Asian ethnicity, depression, lower BMI, higher LDL level, metformin use and higher baseline HbA1c. In contrast, independent predictors of severe hypoglycemia included older age, 12 years or less of education, hypertension, lower triglyceride and higher creatinine levels and greater impairment of cognitive function (Investigators 2014).

In the European CREDIT study, patients with T2DM starting insulin were evaluated on the predictors of hypoglycemia. A multivariable analysis documented symptomatic hypoglycemia had similar predictors to severe hypoglycemia. Being thin, active, taking SU (vs no oral agent), lower HbA1c and higher insulin doses at baseline significantly increased the risk of hypoglycemia. The authors concluded that unlike intensive vs less intensive trials, in which higher HbA1c may pose as a greater risk factor (Miller et al. 2010), real-world clinical practice suggests lower HbA1c is the concern (Home, Calvi-Gries, et al. 2018).

4.11.12.2 Class of AHA and HbA1c

The RECAP DM, a retrospective multicentered study from seven European and five Asian countries from 2006 to 2007, suggested that the association between hypoglycemia and HbA1c differs between users of SUs and those not using SUs. In SU users, (75% of cohort or n=3,278) the proportion of patients reporting hypoglycemia increased with lower HbA1c levels, yet in those not taking SUs (25% of cohort or n=1,121), the proportion of patients reporting hypoglycemia increased as HbA1c levels increased (Yu et al. 2016).

4.11.13 Glycemic Variability

Day-to-day variability of glycemic control has recently been shown in a post hoc analysis to be a significant predictor of any hypoglycemia, including nocturnal in patients with both T1DM and T2DM (Bailey et al. 2017).

In a continuous BG monitoring study of n=222 patients with T2DM from an outpatient unit in France, the mean amplitude of glycemic excursion (MAGE) but not HbA1c was found to be a significant predictor of asymptomatic hypoglycemia (Monnier et al. 2011). In fact, MAGE and duration of asymptomatic hypoglycemia have been found to be closely correlated after treatment intensification with either SU or (Wang, Lee, et al. 2016). For instance, in Poland, patients with T1DM and T2DM who underwent CGM, BG variability statistically and positively

78

correlated with duration of hypoglycemia (Czupryniak, Borkowska, and Szymanska-Garbacz 2017).

Baroflex sensitivity (BRS), a measure of CV autonomic dysfunction is impaired in patients with T2DM. Glycemic variability was recently shown to be related to BRS, independent of age, SBP and BMI. Further, a lower BRS score was found in patients with hypoglycemia, after adjusting for confounders (Matsutani et al. 2018).

4.11.14 Health Literacy

Poor health literacy was significantly associated with hypoglycemia risk in a survey of Californian patients with T2DM. The DISTANCE (The Diabetes Study of Northern California) study, evaluated n=14,357 treated adults with T2DM from the Kaiser Permanente Northern California (KPNC) health care delivery system. Adjusted and unadjusted odds ratios of hypoglycemia were significant with poor self-reported health literacy. Patients reporting problems learning, requiring help to read and not confident filling out forms were more likely to experience hypoglycemia at an increased odds ratio of 40%, 30% and 30%, respectively (Sarkar et al. 2010).

4.11.15 Seasonal Predictors

A prospective study from Germany captured all episodes of severe hypoglycemia between 2007 and 2014. Patients with T1DM had more frequent episodes of severe hypoglycemia on weekends in the warm seasons, likely attributed to behaviour changes. Seasonal variations, however, did not affect the frequency of hypoglycemia in patients with T2DM (Holstein et al. 2016).

Seasonality of severe hypoglycemia was also assessed in Japanese patients with diabetes and a history of severe hypoglycemia retrospectively between 2006 to 2012. Of 578 cases of severe hypoglycemia, 88 occurred in patients with T1DM, 317 in T2DM and 173 in non-DM group. Patients with T1DM experienced severe hypoglycemia significantly more frequently in the summer than in the winter. HbA1c was also highest in the winter and lowest in the summer in patients with T1DM. Occurrence of severe hypoglycemia and HbA1c levels did not differ amongst the season in patients with T2DM (Tsujimoto et al. 2014).

79

4.11.16 Sex

In an RCT where insulin naïve T2DM patients were given or NPH, similar HbA1c were achieved, yet rates of severe and severe nocturnal rates were higher in women than in men (Kautzky-Willer, Harreiter, and Pacini 2016). In the Canadian Cohort of the HAT study, of patients treated with insulin for at least 12 months, females were found to be a higher risk of NSH and NSNH, but not severe hypoglycemia (Aronson et al. 2018).

4.11.16.1 Sex-Gender Differences in Hypoglycemia from the CVOTs

In the SAVOR-TIMI 53 trial, men were significantly more likely to experience an episode of major but not any hypoglycemia in comparison to women (Cahn et al. 2016).

4.12 CONSEQUENCES OF HYPOGLYCEMIA

4.12.1 Attainment of HbA1c Target

Patients report deliberately allowing their HbA1c levels to rise to reduce their fear of hypoglycemia (Willis et al. 2013). A US claims study retrospectively studied n=28,371 patients with at least one prescription for a SU between 2008 and 2012. At 1-year post-index, 47.4% discontinued and 2.5% down titrated their SU on their own. Only 50.0% continued treatment as prescribed. Patients with post-index hypoglycemia had a significantly higher rate of discontinuation (HR=1.82 ss) and down titration (HR=4.25 ss) (Laires et al. 2016).

4.12.2 Increased Self-Monitoring

Patients experiencing an hypoglycemic episodes, even if not taking insulin, report more self BG testing than those without an episode (Malanda et al. 2012).

4.12.3 Impact on Adherence to Therapy

Patients who reported experiencing hypoglycemia also have significantly poorer adherence and treatment satisfaction. A Swedish cross-sectional study of n=430 consecutive T2DM patients on stable doses of metformin and SU for at least 6 months evaluated the impact of symptomatic hypoglycemia on medication adherence. They found that symptomatic hypoglycemia correlated with poor treatment satisfaction and reduced adherence, despite improvement in patients glycemic control (Walz et al. 2014). A 7 European country study (RECAP-DM) found that of 80

the 38% of patients experiencing hypoglycemia, lower treatment satisfaction and barriers to treatment adherence were reported (Alvarez Guisasola et al. 2008).

4.12.4 Dose Reduction

In a US Claims database, half the patients who initiated SU discontinued or down titrated by the end of one year. This retrospective US study assessed n=104,082 T2DM patients less than 65 years of age who had initiated SU treatment between 2008 and 2012. The probability of either discontinuation or down titration at 3, 6 and 12 months was 23.2%, 38.9%, and 52.3%, respectively. Hypoglycemia was found to be a significant risk factor for discontinuation (HR=1.78) and down titration (HR=2.79) (Iglay et al. 2016). Similarly, a cross-sectional study from 1201 general practices in Germany assessed treatment persistence, hypoglycemia and clinical outcomes in users of DPP4i’s and SU’s. They found DPP4i’s had lower treatment discontinuation and five-fold lower frequency of severe hypoglycemia by two years (0.18% with DPP4i vs 1.00% with SU). Furthermore, incidence of macrovascular events was significantly lower with DPP4i in comparison to SU (Rathmann et al. 2013).

4.12.5 Dosing Irregularities / Missed Doses

The GAPP2 study, a multinational online global Attitude of Patients and Physicians survey was conducted to assess real-world data on insulin dosing irregularities and hypoglycemia. The Canadian cohort consisted of n=156 patients who completed the questionnaire. A quarter of the patients (26%) reported adjusting their dose of insulin, the majority of them (60%) reported hypoglycemia as the reason. The authors concluded that HCPs need to be aware that hypoglycemia may frequently compromise diabetes management (Leiter, Boras, and Woo 2015).

4.12.6 Patient Fear

A literature review found 3 studies of patients with T2DM on oral agents evaluating fear of hypoglycemia and all 3 reported increased fear of an event after experiencing a hypoglycemia episode (Zhang et al. 2010). The Global Hat Study also demonstrated a strong association between any and nocturnal events with fear of hypoglycemia and suggesting those who experience an event, fear a future event (Khunti et al. 2016). A survey of n=626 patients with T2DM on the attributes of oral antihyperglycemic agents, found that fear of hypoglycemia was

81

more important than additional healthy life years or weight change (Muhlbacher and Bethge 2016).

The Canadian Cohort of GAPP2 evaluated n=156 insulin-treated patients and 202 HCPs who completed the survey. In the month prior to the survey, 26% reported a dosing irregularity (missed, mistimed, reduced basal insulin dose) and 60% identified hypoglycemia as the reason for the intentional irregularity. Ninety percent of physicians recommended a temporary reduction of insulin dose to manage hypoglycemia risk. Interestingly, patient concerns of hypoglycemia differed greatly with prescriber reported patient concerns, suggesting a need for improved communication (Leiter, Boras, and Woo 2015).

(Reprinted with permission)

In a similar questionnaire with n=202 patients with T1DM and n=133 patients with T2DM, more patients with T1DM feared a future event after a mild hypoglycemia than patients with T2DM. The opposite was true after a severe event. Although both groups feared severe hypoglycemia more than mild hypoglycemia, more patients with T2DM feared another event relative to patients with T1DM (Leiter et al. 2005).

82

4.12.7 Physician Fear

In GAPP (Global Attitude of Patients and Physicians in insulin Therapy), 1750 physicians (600 specialists, 650 primary care physicians) and 1530 insulin-treated patients (180 T1DM and 1350 T2DM) were surveyed over the internet (HCP) or phone (patients) from China, France, Japan, Germany, Spain, Turkey, UK and USA. A majority (87.6%) of physicians reported that many insulin-treated patients have inadequate glycemic control and 75.5% reported hypoglycemia as the limiting factor for not treating more aggressively (Peyrot et al. 2012).

GAPP2 was an online survey of physicians (n=1003) and T2DM patients (n=855) treated with basal insulin, with or without bolus insulin. While 75% of patients reported being worried about hypoglycemia, only 18% of physicians believed the patient worried (Brod et al. 2016). To assess physician response to hypoglycemic episodes, a US (New Mexico VA Health Care System) study involving 30 primary care providers prospectively evaluated hypoglycemia episodes in n=265 patients with T2DM on insulin +/- sulfonylurea. Physicians were more likely to reduce the dose of SU but not insulin with increased frequency of hypoglycemia episodes (Wendel et al. 2014).

4.12.8 Family Members Worry

The cross-sectional DAWN2 (Diabetes Attitudes, Wishes and Needs study) surveyed n=2057 family members of adults with diabetes from 17 countries across 4 continents. A majority (85%) of family members reported being very worried about the risk of hypoglycemia (Nefs et al. 2016). Similarly, a 63-item online questionnaire (InHypo-DMSOQ) was conducted to explore the burden of hypoglycemia on significant others of people living with diabetes in Canada. They found the majority (87%) of significant others would forgo other activities to help manage hypoglycemia. However, 56% reported difficulties in making such alterations to their schedules to assist. Sixty-two percent reported that they were involved in making decisions yet, 71% stated feeling inadequately informed on how to prevent or treat an event. The authors concluded that significant others feel a strong sense of altruism and HCP should involve them in educational training (Reichert et al. 2017).

83

4.12.9 Depression

A US retrospective analysis investigating the association of hypoglycemia and depression events was conducted of patients with T1DM or T2DM insurance claims from Thomson Reuters Market Scan between 2008 and 2009. Patients were eligible if they had not experienced hypoglycemia or depression during the baseline period (Jan 2008 to December 2008). Of the n=923,024 patients included, the odds of depression significantly increased in patients with hypoglycemia. Overall, 2.46% experienced a hypoglycemia event while 0.67% reported depression. With or without adjusting for covariates, those experiencing hypoglycemia had significantly higher odds of depression (OR 1.78 and 1.72, respectively) (Shao et al. 2013).

4.12.10 QoL

A five-country survey demonstrated that hypoglycemia, particularly nocturnal episodes have a pronounced effect on Health Related Quality of Life (HRQoL) (Evans et al. 2013). In a US survey of patients visiting a diabetes clinic, 23.8% of n=92 T1DM and 16.9% of n=326 T2DM patients reported have experienced a hypoglycemia episode in the last 6 months. Fear of hypoglycemia was higher in those experiencing severe hypoglycemia. Authors concluded that severe hypoglycemia negatively impacted HRQoL in patients with T2DM but not T1DM (McCoy et al. 2013).

4.12.10.1 QoL - Severity of Hypoglycemia

A US National Health Survey found a strong independent association of severe hypoglycemia and lower HRQoL. HRQoL decreased further with greater severity and frequency of hypoglycemia (Marrett et al. 2011). In corroboration with these findings, a survey of US patients with T2DM treated with an antihyperglycemic agent found severity of hypoglycemia was inversely and significantly associated with HRQoL and all SF-36 domains (Pawaskar et al. 2018). Even in non-insulin treated patients with T2DM on one or more agent, hypoglycemia was correlated with lower HRQoL (Williams, Pollack, and Dibonaventura 2011). In contrast, however, some studies have reported that hypoglycemia negatively impacts treatment satisfaction in patients treated with oral agents but not those on insulin (Zhang et al. 2010).

84

4.12.11 Motor Vehicle and Other Accidents

Of the 19 million people living with diabetes in the US, a large percentage of them seek or currently hold a driver’s licence. States and jurisdictions vary in the requirements and restrictions on drivers with a diabetes diagnosis and depend on the type of driving (American Diabetes, Lorber, et al. 2014).

In Canada, patients with diabetes who had reported to the Ontario Ministry of Transportation Medical Advisory Board between 2005 and 2007 were identified to examine the association between HbA1c and risk of a motor vehicle crash. HbA1c was significantly lower in those that had a crash compared to those that did not have a crash (7.4% vs 7.9%). For each reduction in HbA1c, there was a 26% increase in risk of a crash, after adjusting for confounders. Two variables, history of severe hypoglycemia (OR=4.07) and an older age of diabetes diagnosis (OR=1.29) were the strongest risk factors for a crash (Redelmeier, Kenshole, and Ray 2009).

In a US employer claims database, people with noninsulin-treated T2DM who had claims for hypoglycemia between 1998 – 2010 were evaluated for accident risk. Of n=5,582 claims for hypoglycemia and n=27,910 non-claims, hospitalization occurred in 5.5% of patients reporting hypoglycemia and 2.8% of non-hypoglycemia patients. Even after adjusting for potential confounders, hypoglycemia significantly increased the risk of any accident (HR =1.39), accidental fall (HR = 1.36) and motor vehicle accidents (HR = 1.82) (Signorovitch et al. 2013).

An online survey by n=1,569 UK drivers with T2DM reported that in the previous year, 62% experienced symptoms of mild or severe hypoglycemia. Following a hypoglycemia event, only 24% of insulin-secretagogue (SU or glinide) and 39% of insulin-treated drivers said they discontinued driving for the recommended 45 minutes. Moreover, 47% of patients taking non- insulin secretagogues reported experiencing symptoms of mild hypoglycemia and 9% reported severe hypoglycemia (Feher, Langerman, and Evans 2016).

4.12.12 Social Implications

The original DAWN study reported that 41% of adults with diabetes had poor psychosocial well- being. DAWN2 (Diabetes Attitudes, Wishes and Needs study), examined the psychosocial experiences of n=8,596 adults with T1DM (n=1,368) and T2DM (n=7,228) from 17 countries

85

across 4 continents. Two psychosocial themes emerged; first, anxiety/fear, worry about hypoglycemia, complications of diabetes, depression and negative mood/hopelessness and second, discrimination at work and public misunderstanding about diabetes (Stuckey et al. 2014).

4.12.13 Employment Implications

The major concern of diabetes in the workforce is hypoglycemia induced accidents, due to the use of insulin or SUs. Because hypoglycemia can lead to cognitive impairment, safety becomes of concern. A number of guidelines ban insulin treatment with certain types of employment, in particular the airline industry.

(Permission not required – Open Access)

It therefore becomes prudent to assess the extent of job safety sensitivity, in concordance with the severity of the hypoglycemic event. A matrix has been proposed to aid in employment decisions (Lee et al. 2011).

86

(Permission not required – Open Access)

4.12.14 Work Productivity and Burden of Non-Severe Hypoglycemia

A 20-minute survey of patients with self-reported diabetes was conducted in the US, UK, France and Germany to identify non-severe hypoglycemia (NSH) and its impact on workplace productivity. Of n=6,756 survey respondents, n=2,669 reported having NSH and 54% worked for pay. After excluding 27 respondents for unclassified work status or not remembering NSH events, n=1,404 respondents were included in the analysis. Approximately 50% had T1DM and 50% had T2DM. Per NSH event, loss in workplace productivity was estimated at $15.26 to $93.47 USD and represented 8.3 to 15.9 hours of lost time at work per month. If a NSH event occurred during work hours, 18.3% missed work for an average of 9.9 hours. When NSH events occurred outside work hours including nocturnal, 22.7% arrived late or missed a full day. Nocturnal NSH represented the greatest loss in work-place productivity, with an estimated 14.7 hours lost. Furthermore, 5.6 extra BG test strips and a 25% decrease in insulin dose of those using insulin was reported in the week following the NSH (Brod et al. 2011).

Of note, in the ORIGIN trial, unlike severe episodes of hypoglycemia, NSH (defined as symptoms and confirmation of ≤3 mmol/L) was not associated with an increased risk in death, CV death or arrhythmic death (Investigators et al. 2013).

4.12.15 Burden of Non-Severe Nocturnal Hypoglycemia

A survey of patients with T1DM and T2DM from nine countries (USA, UK, Germany, Canada, France, Italy, Spain, The Netherlands and Sweden) was conducted on the impact of non-severe nocturnal hypoglycemia (NSNH). Following an event, respondents reported an increase in glucose self-monitoring, a 3.4 hour delay in resuming normal functioning, a 16% decrease in insulin dose and poor-quality sleep the subsequent night with the majority reporting feeling tired or needing to take a nap (Brod et al. 2013a). The Nine Country Study also reported on the economic burden of NSNH. Following a NSNH event, loss of work productivity was estimated to be 3.3 to 7.5 hours of work time per event. In addition to the increased use and cost of test strip monitoring (approximately $87.10/year) and health care utilization, medical care required for falls or injuries was estimated at $2,111.30 per person per year (Brod et al. 2013b).

87

4.13 COMPLICATIONS OF HYPOGLYCEMIA

4.13.1 Mortality Risk

Observational data shows a clear link between hypoglycemia and increased risk of death. Over a four year period, elderly T2DM patients from Alberta who were hospitalized after a hypoglycemia episode had an increased risk of mortality (HR 2.55; 95% CI 2.25 - 2.88) in an independent, dose-dependent manner (Majumdar et al. 2013). Similarly, the UK Clinical Practice Research Datalink also demonstrated that in insulin-treated patients, hypoglycemia increases the risk of CV events and death, for an extensive period of time. In patients with T2DM, with and without pre-existing CVD, the risk for a CV event was HR 1.60 (95% CI 1.21 – 2.12) and HR 1.49 (1.23 – 1.82) and the risk of ACM was HR 1.74 (1.39 – 2.18) and HR 2.48 (2.21 – 2.79), respectively (Khunti et al. 2015). However, it is not clear if hypoglycemia is merely posing as a surrogate of a more complex vulnerable patient with diabetes and other risk factors and comorbidities. In the ADVANCE study, over a mean follow up period of 5 years, patients who experienced severe hypoglycemia were more likely to die from CV or any cause. They were also more likely to die from other non-CV causes such as respiratory, digestive and skin conditions (Zoungas et al. 2010).

88

(Reproduced with permission from Copyright Massachusetts Medical Society).

Similarly, the Hong Kong Diabetes Registry evaluated n=8,767 patients with T2DM between 1995 and 2007 and found those who had experienced severe hypoglycemia at baseline had a higher risk of all site cancer over the follow-up period (Kong et al. 2014). Finally, the Bezafibrate Infarction Prevention (BIP) study investigated the relationship between hypoglycemia and ACM in n=14,670 patients with CAD with and without diabetes. Over a mean 8-year follow up, hypoglycemia was associated with an increased risk of ACM and cancer.

89

The authors concluded that hypoglycemia may be a marker of increased risk of death (Fisman et al. 2016).

4.13.1.1 Mortality Risk - Intensive vs Standard Glycemic Control

An increased risk of hypoglycemia has been demonstrated in studies comparing more vs less intensive glucose control. However, despite a higher risk of hypoglycemia in the intensive arm, risk of death was lower in ADVANCE, ACCORD and ORIGIN.

(Hanefeld, Frier, and Pistrosch 2016).

(Reproduced with permission)

In ACCORD, however, although hypoglycemia was associated with increased risk of death irrespective of study arm, risk of death was lower in the intensive than the standard arm.

90

(Bonds et al. 2010).

(Reproduced with Permission)

Furthermore, the risk of death decreased in the intensively treated patients as the number of severe hypoglycemia episodes increased.

(Reproduced with Permission)

The authors concluded that symptomatic and severe hypoglycemia did not account for the increased mortality seen in the intensively treated patients (Seaquist et al. 2012).

Both uncontrolled high BG levels and hypoglycemic events are associated with worse outcomes. A retrospective analysis of prospectively collected patient data from Sweden corroborated the importance of HbA1c in the relationship between hypoglycemia and mortality. In n=713 patients with diabetes hospitalized with an acute coronary event revealed that after a 2 year follow up, mortality was significantly associated with highest vs lowest quartile of BG. The magnitude of risk was even greater than hypoglycemia during hospitalization, which was also a significant predictor of death, even after multivariable adjustments (Svensson et al. 2005).

91

4.13.1.2 Mortality Risk – Class of AHA

In ACCORD, irrespective of AHA class, the overall risk of hypoglycemia was significantly greater in the intensive vs standard treated groups. However, SU and metformin use, but not insulin use further increased the overall average risk (Miller et al. 2010). SUs, which were more frequently used in the standard arm of ORIGIN, may have increased the risk of death by blocking ATP-dependent potassium channels necessary for myocardial adaptation to ischemia (Pistrosch and Hanefeld 2015).

4.13.1.3 Mortality Risk – Severity of Episode

In the ACCORD study, patients who never experienced a hypoglycemic event experienced higher mortality rates in the intensively treated arm (1.2%/year) than in the standard arm (1.0%/year). However, in patients that had at least one previous severe event, defined as requiring any, medical, or non-medical assistance, mortality was lower in the intensively treated arm (2.8%/year vs 3.7%/year). Adjusted HR of mortality was 1.41 (CI 1.03-1.93) in the intensive arm vs 2.30 (CI 1.46-3.65) in the standard arm in patients experiencing at least one severe event vs no previous severe events. When severe events were further categorized as requiring medical assistance, the findings were even more robust. The risk of death in patients with hypoglycemia requiring medical attention in the intensive treated arm was 2.8%/year compared to 4.9%/year in the standard arm (adjusted HR 0.55 (CI 0.31-0.99)) (Bonds et al. 2010).

4.13.2 Biomarkers of Severe Hypoglycemia and Death (ACCORD)

A nested case-control study was conducted from the ACCORD study of those who died with at least one episode of severe hypoglycemia compared to those who did not die and did not experience any severe hypoglycemia episodes. Baseline insulin deficiency (fasting C-peptide ≤0.15 nmol/l) was compared after adjusting for glycemic intervention arm, age, race and BMI. After adjusting for confounders, insulin deficiency (OR 4.8) or baseline insulin use (OR 6.1) remained significantly elevated in those who experienced at least one episode of severe hypoglycemia and died. The authors provide several possible explanations including a blunted glucagon counterregulatory response in patients with insulin deficiency, increasing their risk of death (Chow et al. 2016).

92

4.13.3 Prediction of All-Cause Mortality in Patients with Hypoglycemia

The Multidimensional Prognostic Index (MPI) is a validated tool for predicting mortality. A Study of n=1,342 elderly Italian patients with T2DM concluded that advanced age, female gender, hospitalization for hypoglycemia were independently associated with a significant worse MPI score (Pilotto et al. 2014).

(Permission not required – open access)

In a Korean study, n=1428 patients with T2DM were followed for a median time of 10.4 years for CV or ACM. Presence of severe hypoglycemia was found to be the strongest predictor of both ACM and CV events. Severe hypoglycemia remained a significant predictor of both ACM and CV events even after adjusting for sex, age, diabetes duration, hypertension, CVD history, smoking, BMI, mean HbA1c, diabetic nephropathy, use of ACE/ARB, statin or ASA (Cha et al. 2016). This data, as well as the MACE data in the below section demonstrates the harms associated with hypoglycemia.

4.13.4 MACE (CV Death, Nonfatal MI, Stroke) Risk Following Hypoglycemia

In the ORIGIN trial, severe hypoglycemia increased the risk of death by 74%, CV death by 71% and arrhythmic death by 77%. Interestingly, however, the risk of dying from a hypoglycemic episode was greater in the standard arm compared to the glargine insulin arm. Despite the glargine arm experiencing significantly more episodes of severe hypoglycemia than the standard group (74% vs 26%, respectively), relationship of hypoglycemia and MACE (CV death, nonfatal

93

MI, nonfatal stroke or arrhythmic death) was higher in the standard arm even after propensity score matching (Investigators et al. 2013).

Multiple observational studies have evaluated the correlation between hypoglycemia and MACE. A retrospective analysis of n=860,845 patients with T2DM in a US health care claims suggested a 79% higher regression-adjusted odds of an acute CV event in outpatients with a hypoglycemia event. Confounders that were adjusted for included age, sex, geography, insurance type comorbidity scores, CV risk factors, diabetes complications, total baseline medical expenditures and prior acute CV event (Johnston et al. 2011). The Edinburgh T2DM Study, prospective cohort of n=1,066 patients aged 60-75 with T2DM, found the risk of a macrovascular event was significantly higher in patients who had experienced severe hypoglycemia. After adjusting for multiple CV risk factors, the association between severe hypoglycemia and macrovascular events remained significant (Bedenis et al. 2014).

A systematic review and meta-analysis of prospective and retrospective observational studies was conducted to assess the association between severe hypoglycemia and CVD and the role of comorbid illness as a confounder. A significant association of severe hypoglycemia and cardiovascular death in both prospective and retrospective study designs was found; prospective RR 2.67 and retrospective RR 1.93 (Goto et al. 2013).

(Permission not required – Open Access)

94

The same authors recently updated the meta-analysis and found a strong association of severe hypoglycemia and cvd in Japanese patients with T2DM (Goto et al. 2016).

4.13.4.1 Studies of Simultaneous Capillary Blood Glucose Monitoring and Holter Monitoring Evaluating Association of Hypoglycemia and Myocardial Infarction

CBGM studies identify hypoglycemic episodes that might not be symptomatic. In one such study from 2003, n=21 patients with insulin-treated T2DM and CAD were assessed while wearing simultaneous holter and continuous glucose monitors. Of 54 hypoglycemia episodes (from n=19 patients), ten episodes were associated with chest pain with four having associated electrocardiographic abnormalities. Further, no reports of ECG abnormalities or symptoms of chest pain were reported during normal BG range (Desouza et al. 2003).

(Reproduced with permission)

More recently, Chow et al. used continuous glucose monitoring and ambulatory ECG monitoring to study n=25 insulin-treated patients with T2DM and a history of CVD or risk factors. Hypoglycemia coincided with an increase in bradycardia and atrial ventricular ectopic activity (Chow et al. 2014).

95

(Reproduced with permission)

Cardiac autonomic dysregulation during hypoglycemia induced with the insulin clamp technique may be the cause differs in patients with diabetes vs those without. For example, over 60 minutes, despite hypoglycemia, in people with diabetes, the heart rate initially increased but then fell back to baseline levels. In patients without diabetes, the heart rate increased until the end of the clamp. Patients with diabetes also demonstrated greater repolarization abnormalities, despite similar epinephrine responses (Chow et al. 2017).

Severe hypoglycemia (<3.1 mmol/l) was also significantly associated with a higher number of severe ventricular arrhythmias in a study of n=95 insulin and SU-treated patients with T2DM and established CVD (Pistrosch et al. 2015). Similarly, in n=32 patients with T2DM and CVD from Germany wearing simultaneous CBG and ECG monitors, a strong association between severe hypoglycemia and severe ventricular arrhythmias were noted. Interestingly, patients treated with SU and or insulin were reported to have a high incidence of asymptomatic hypoglycemia (Stahn et al. 2014).

4.13.4.2 MACE Risk - Severity of Hypoglycemia

As illustrated in the ORIGIN trial, severe hypoglycemia, and not necessarily NSH, has been linked to MACE, even after adjusting risk with propensity score matching.

96

(Reproduced with permission)

There was no evidence that treatment allocation affected the relationship between NSH or NSNH and MACE outcomes, as it did with severe hypoglycemia.

(Reproduced with permission)

97

4.13.4.3 MACE Risk - Symptomatic vs Asymptomatic Hypoglycemia

In insulin-treated patients with T2DM (n=25) and either CVD or CV risk factors wearing simultaneous holter and continuous glucose monitors, only 3/34 episodes of hypoglycemia were symptomatic. Two of these three symptomatic episodes occurred during the day. Interestingly, no arrhythmias were noted in the 3 symptomatic episodes. Asymptomatic episodes had a lower heart rate and greater vagal tone compared to symptomatic ones (Chow et al. 2014).

4.13.4.4 MACE Risk - Nocturnal vs Daytime Hypoglycemia

In the ORIGIN trial, the CV outcomes attributable to nocturnal severe hypoglycemia were similar to daytime severe hypoglycemia (Investigators et al. 2013).

(Reproduced with permission)

Chow et al. demonstrated in T2DM patients wearing simultaneous holter and continuous glucose monitors that nocturnal hypoglycemia results in significantly more frequent arrhythmias. More specifically, bradycardia and atrial and ventricular ectopic counts were significantly greater during nocturnal hypoglycemia (Chow et al. 2014).

98

(Reproduced with permission)

In summary, both the ORIGIN and ACCORD trials demonstrated that severe hypoglycemia was associated with worse outcomes. More interestingly, when stratified by treatment arms, the standard arm in ORIGIN and the less intensive glucose lowering arm in ACCORD faired worse, for which a number of potential explanations have been proposed. Firstly, it remains plausible to state again that severe hypoglycemia may be merely serving as a marker of patients who are at higher CV risk. Secondly, higher use of sulfonylurea, which occurred in the standard arms of both trials could be blamed. As will be discussed in later sections (see 5.3), SU block ATP channels in order to stimulate pancreatic insulin secretion. However, ATP channels are located in other tissues including the heart. A number of studies, although observational in nature, have demonstrated that use of SU’s block ischemic preconditioning and are associated with higher risk of MACE and ACM. As evidence from other classes of antihyperglycemic agents continues to evolve, avoidance of hypoglycemia remains of critical importance for patients with diabetes.

4.13.5 Progression of Atherosclerosis

The VADT demonstrated that serious hypoglycemia was associated with the progression of CAD in the standard, but not intensively treated group (Saremi, Bahn, and Reaven 2016).

4.13.5.1 Coronary Artery Calcium (CAC)

Given the discordance in the literature between risk of CV events in patients in intensive treated vs standard glycemic control, a subset of n=197 patients from the VADT study was conducted. Patients who had baseline and follow-up computed tomography scans were used to evaluate coronary artery calcium progression and serious hypoglycemia. Serious hypoglycemia occurred in 74% of the intensive treated patients and 21% of the standard treated patients. Serious hypoglycemia was not associated with CAC progress in the overall cohort or in the intensively

99

treated patients. Interestingly, however, in the standard treated patients that experienced serious hypoglycemia event, a 50% CAC progression was observed. Results was maintained even after adjusting for baseline confounders including baseline CAC score. Furthermore, CAC progression was highest in the that experienced severe hypoglycemia and had a mean HbA1c greater than 7.5 mmol/L compared to ≤7.5 mmol/L (Saremi, Bahn, and Reaven 2016).

(Reproduced with permission)

100

(Reproduced with permission)

4.13.5.2 Carotid intima medial thickness (cIMT)

In a post hoc analysis of a Japanese study of n=274 patients with T2DM, those who experienced hypoglycemia demonstrated greater increases in cIMT than those without hypoglycemia. Further, greater frequency of hypoglycemia episodes positively correlated to further increases in cIMT. Although it may still be possible that hypoglycemia may be a marker of those with comorbid conditions, including atherosclerosis, significance was maintained even after adjusting for confounding factors between those with a history of hypoglycemia vs those without (Mita et al. 2017).

101

(Permission not required – Open Access)

4.13.5.3 Endothelial Dysfunction

In a small study of n=11 patients with T2DM and n=16 healthy subjects, the effects of equivalent sympathetic nervous system activity was evaluated during hypoglycemia on endothelial function. During hyperinsulinemic euglycemia, ICAM-1, VCAM-1, P-selectin, PAI-1, VEGF and endothelin-1 decreased in both healthy subjects and in patients with diabetes. During hypoglycemia of 2.9 mmol/L for 120 minutes, these parameters increased in both groups. Levels of epinephrine and norepinephrine levels were also similar during hypoglycemia in both healthy and patients with T2DM. However, despite milder hypoglycemia, levels of ICAM-1, VCAM-1, PAI-1, VEGF, and ET-1 responses were more elevated in patients with T2DM than in healthy controls. The authors concluded that for a milder level of hypoglycemia, despite similar SNS activation, patients with T2DM display a greater pro-inflammatory, pro-atherothrombotic and pro-coagulant states (Joy et al. 2016).

In a similar study by Wright et al., n=16 T1DM were compared to n=16 healthy controls during a hypoglycemia clamp of 2.5 mmol/L for 60 minutes to evaluate markers of inflammation, endothelial function and platelet activation. In both groups, hypoglycemia elicited an increase in vasoactive substances and epinephrine responses were similar elevated during hypoglycemia and reduced during euglycemia. In healthy subjects, increases in platelet activation (as assessed by platelet-monocyte activation and P-selectin) and CD40 expression occurred after hypoglycemia and was maximally elevated at 24 h. In patients with T1DM, CD40 expression peaked during hypoglycemia, and platelet monocyte activation also was significantly increased at 24h after

102

hypoglycemia. Authors conclude that hypoglycemia elicits release of potent vasoactive substances that could explain link of hypoglycemia to vascular injury (Wright et al. 2010).

4.13.6 Blood Coagulation Abnormalities

In n=16 healthy volunteers from the UK, hyper-, hypo- and euglycemic clamps were used to measure platelet aggregation. During periods of hypoglycemia, surrogates of increased coagulation such as platelet reactivity, monocyte and neutrophil platelet aggregates were observed (Iqbal et al. 2017).

4.13.7 Inflammation

In both healthy individuals and those with T1DM, hypoglycemia clamp of 2.9 mmol/L induced increases in inflammatory markers including PAI-1, VEGF, vascular adhesion molecules (VCAM, ICAM, E-selectin), IL-6 and markers of platelet activation (P-selectin) (Joy et al. 2015).

4.13.8 Cognitive Function

Hypoglycemia is known to impair executive functioning in patients with and without diabetes (Graveling, Deary, and Frier 2013). Impairment of cognitive function during hypoglycemia occurs prior to symptomatic awareness. Further, cognitive performance lags behind resolution of symptoms and restoration of glucose levels during the recovery phase of hypoglycemia (Evans et al. 2000). Neuroinflammatory mechanisms are thought to contribute to brain damage following cerebral ischemia in diabetes due to either hyper or hypoglycemia (Shukla et al. 2017).

103

(Permission not required – Open Access)

A cognitive function test was conducted at baseline and four-year follow up in n=831 patients aged 60-75 years of age with T2DM from The Edinburgh study. Hypoglycemia and history of hypoglycemia were associated with greater decline in cognition during follow up, even after adjustment of confounders (Feinkohl et al. 2014). In contrast, a subgroup analysis of the Freemantle Diabetes study, n=302 patients 70 years of age or older found no evidence that those with a history of severe hypoglycemia or prospectively reported hypoglycemia were at increased risk of cognitive impairment (Bruce et al. 2009).

4.13.9 Dementia

Reassuringly, results from ACCORD MIND study suggest that severe hypoglycemia was not associated with loss of TBV and or increase of AWM compared with T2DM patients without hypoglycemia (Zhang et al. 2014). Nevertheless, a retrospective study from the UK Clinical Practice Research datalink demonstrated that the odds of subsequent dementia was 27% higher after the occurrence of at least one hypoglycemia event. The risk increased to 50% in those with two or more episodes of hypoglycemia (Mehta et al. 2011).

104

4.13.10 Renal decline

In a 3-year cohort study of Japanese patients with T2DM experiencing severe hypoglycemia and with eGFR ≥60mL/min/1.73m², blood pressure surge during severe hypoglycemia independently correlated to an increase in the composite outcome of more than 15mL/min/1.73m² and initiation of chronic dialysis (Tsujimoto et al. 2016). In a similar study from Taiwan’s National Health Insurance Research Database, 1.3 million patients with T2DM were assessed for hypoglycemia events using inpatients, outpatients and emergency services between 2000 and 2010. CKD and those less than 20 years of age were excluded. During a four-year period, 1.7% or n=15,036 patients experienced one or more hypoglycemia event. Of N=906,368 eligible patients, n=15,036 were identified as matched controls of those without hypoglycemia. Patients with hypoglycemia were 77 times (HR=1.77) more likely to experience CKD. Furthermore, the risk of CKD increased with increasing number of episodes of hypoglycemia (Shih et al. 2015).

4.13.11 Falls and Fall-Related Fractures

A study from Taiwan’s National Health Insurance Research Database demonstrated that diabetes alone increases risk of falls and hypoglycemia further accentuates the risk. The difference was more pronounced in those <65 years of age (Lu et al. 2015).

(Reproduced with permission)

In fact, hypoglycemia can double the risk of falls over a 365-day period, as shown in a large US healthcare database of T2DM patients (n=21,613 patients with hypoglycemia matched with n=21,613 patients with non-hypoglycemia) (Kachroo et al. 2015). Further, an analysis from the US Medicare employee insurance claims between 2008 and 2009 found that patients with hypoglycemia had 70% higher adjusted odds of fall-related fractures. The relationship remained significant after multiple sensitivity analyses (Johnston et al. 2012).

105

Again, it may be that those with hypoglycemia are more vulnerable, as demonstrated in a retrospective study of older US veteran with T2DM between 2004 and 2010. Those with hypoglycemia (n=4,215) were compared on falls and fall related outcomes to those without hypoglycemia (n=4,215). The hypoglycemia group had significantly higher Charlson Comorbidity Index scores and higher rates of fall-related events, even after adjusting for sociodemographic, clinical, comorbidities, and medications use (Zhao et al. 2016).

4.14 COST

The Canadian Cohort of patients from the Global Hypoglycemia Assessment Tool (HAT) study, included n=183 and n=315 insulin-treated patients with T1DM and T2DM, respectively. Over 80% reported experiencing hypoglycemia in the preceding 4-months. A cost model estimated that over a 1-year period, the mean direct health care costs per patient were $1,777 per patient for hospitalizations and $204 per patient for additional appointments. Indirect costs per patient were also captured. Absenteeism from work was estimated at $500 per patient, late arrival at work at $187 per patient and leaving early from work at $128 per patient. In total, the direct and indirect cost of hypoglycemia was estimated at $125,932.

(O'Reilly et al. 2018)

(Reproduced with permission)

4.14.1 Cost of Severe Hypo, including Hospitalization

In a UK, Germany and France survey of n=1848 T1 and T2DM treated with insulin, total (direct and indirect) cost per severe hypoglycemia event was estimated at 3,000 Euros (Willis et al. 2013). The second largest health management organization in Israel also conducted a

106

retrospective analysis compare resource utilization (physician visits, hospitalizations and medications) 1 month pre- and 1 month post severe hypoglycemia event from 2005 to 2014. Of the n=3,691 patients with T2DM, 30% were treated with insulin and 70% with oral agents known to cause hypoglycemia. Health care costs and outpatient cost for patients with T2DM 1- month pre and 1-month post severe event increased dramatically.

(Goldstein et al. 2016)

(Permission not required – Open Access)

107

4.14.2 Cost of Non-Severe Hypoglycemia

Direct, indirect, and total annualized costs of hypoglycemia episodes in T2DM in the US is substantial. Authors cautioned that failure to account for indirect and direct costs of NSH greatly underestimates the overall cost (Foos et al. 2015).

(Reproduced with permission)

In another US study, rates of mild hypoglycemia were estimated using a mathematical model of patients with T2DM based on therapy, HbA1c levels, duration of diabetes, kidney function and BMI. Overall average rate of mild hypoglycemia was estimated at 2.2 events per person per year, and 4.9 if taking insulin. The economic impact of mild hypoglycemia was determined to be approximately equal to that of severe hypoglycemia, at roughly $900 million per year (Samuel et al. 2015).

4.15 PERFORMANCE MEASURES

A recent scoping review identified that although worldwide there exists 18 guidelines and 23 performance measure initiatives for diabetes, only 2 reports on performance measures specific to hypoglycemia, the National Institute of Health Excellence and the National Information Diabetes Service in the UK (Rodriguez-Gutierrez et al. 2016).

108

4.16 SUMMARY

In this chapter, we dug deep into the latest evidence on hypoglycemia, our outcome of interest. We learned that a single definition of hypoglycemia has been difficult to conceive, since BG thresholds of symptom perception and counterregulatory responses vary amongst individuals. In clinical trials of AHAs, the primary objective is efficacy-focused and hypoglycemia is captured as a secondary, tertiary or an adverse event outcome. In fact, many studies do not provide an explicit definition of hypoglycemia. When hypoglycemia is defined, BG thresholds, requirement of BG documentation or presence of symptoms vary. Further, study eligibility for those with previous history of an event or the statistical handling of patients with an event during the trial may also differ. This lack of a standardized hypoglycemia definition has been cited for restricting meta-analytical comparison of interventions on this outcome.

Since 2017, however, ADA has followed the recommendations from the International Hypoglycemia Study Group to define severe hypoglycemia as any episode requiring third party assistance, irrespective of BG values. Since 2013, ADA has defined any hypoglycemia as any abnormally low plasma glucose level that exposes an individual to harm. Following these recommendations, the primary outcome of our systematic review and meta-analysis is any hypoglycemia, irrespective of its definition and our secondary outcome of severe hypoglycemia is defined as requiring third party assistance or a BG value of <3.0 mmol/L.

Symptoms of hypoglycemia have been categorized into three categories; general malaise (headache and nausea), autonomic (such as hunger, sweating and tremors) and neuroglycopenic (such as dizziness, confusion, difficult speaking and vision changes). In normal individuals, when BG levels fall, the normal physiological response begins with the inhibition of endogenous insulin secretion, followed by glucagon, adrenaline and growth hormone release. In patients with newly diagnosed diabetes, these counterregulatory responses remain intact. However, in patients with a long duration of diabetes or frequent episodes of hypoglycemia, insulin levels decline and glucagon and sympathoadrenal responses are diminished. Such attenuation of a normally physiological response to low BG levels has been termed; ‘Hypoglycemia Associated Autonomic Failure’ or HAAF and can lead to a vicious cycle of more frequent episodes with lower thresholds of symptom perception and counterregulatory responses. This is evident is studies with elderly patients with diabetes where perception of hypoglycemia 109

symptoms is absent. Since episodes of hypoglycemia can occur undetected, continuous BG monitoring is the gold standard. However, a paucity of evidence is available with continuous BG monitoring and data is of short duration with poor patient adherence.

Incidence of hypoglycemia is greater in patients with type 1 than type 2 diabetes, although rates can be similar when duration of insulin use is accounted for. Hospitalization for hypoglycemia (ie a severe event) has been better documented than less severe forms, given the less subjective nature of the outcome. Epidemiological data on hypoglycemia in patients with type 2 diabetes is often conducted in insulin-treated individuals and estimates of hypoglycemia incidence in patients taking oral agents, without the use of SU are limited to clinical trial data. Temporal trends of hypoglycemia have also been confounded by the recent changing treatment paradigms. Although use of SU has drastically decreased in the last two decades, use of newer insulin regimens have increased. This has made hypoglycemia incidence with new AHA difficult to assess.

Risk factors for hypoglycemia, apart from SUs and insulin use, are extensive. It is often debated whether hypoglycemia is merely a bystander signifying a more complex patient. Patients experiencing hypoglycemia often co-present with vascular disease such as cardiovascular, renovascular and cerebrovascular and are thus at an increased risk of complications. In fact, some studies have demonstrated that patients who experience severe hypoglycemia are also at an increased risk of infection and cancers. Thus, whether hypoglycemia causes such negative outcomes or merely co-exists in patients with multiple comorbidities remains unclear. In any case, it remains clear that hypoglycemia is an undesired event and all efforts to circumvent it must be made.

Complications associated with hypoglycemia are extensive. Studies with simultaneous continuous BG and holter monitors have irrefutably demonstrated a correlation of hypoglycemia with electrocardiogram abnormalities, cardiac autonomic dysregulation and ventricular arrhythmias. Apart from the plethora of negative health complications associated with hypoglycemia, after experiencing a event, patients, family members and physicians worry of a subsequent event. This fear often leads to decreased treatment adherence and compliance. Subsequently, HbA1c targets are relaxed, placing patients at a higher risk of developing

110

diabetes-complications. Apart from the numerous health complications associated with hypoglycemia, undesirable consequences, such as motor vehicle accidents, falls and fractures can also increase after an event. To make matters worse, diabetes and hypoglycemia occur at a higher incidence in patients with low socioeconomic status, health literacy and food insecurity. Hence, avoidance of hypoglycemia is critical as it impacts the ability to achieve glycemic targets and complication-free survival.

The economic burden of hypoglycemia is frequently underestimated. Studies often capture hospitalization for hypoglycemia, given its less subjective nature. Less is known about the more frequent, nonsevere forms of hypoglycemia. Following an event, irrespective of severity, patients use of health care resources and self BG monitoring increase. A paucity of data on the indirect costs of hypoglycemia also exists, such as sick days or low work productivity. Hence, the total cost of hypoglycemia, both direct and indirect, is largely underappreciated.

Given the numerous complications, consequences and costs associated with hypoglycemia, it is clear that avoidance of hypoglycemia is key in the optimal management of diabetes. In order to attenuate the risk of hypoglycemia, improving our understanding of the risks associated with use of each class of AHA is necessary. For physicians, this could refine how treatment regimens are individualized. For patients and their family members, more precise estimates of hypoglycemia risk relative to placebo can help reduce the stress, worry and fear of an event. Policy makers may also be interested in how AHA classes compare to placebo, given the costs associated with this adverse outcome.

Antihyperglycemic Agents

In this chapter, we review the interventions currently indicated for patients with type 2 diabetes, including those of interest for our systematic review and meta-analysis. First, the glucose lowering mechanism of each class is examined. Next, an overview of the general efficacy and safety characteristics are provided, including any available data on hard outcomes such as all- cause mortality or major adverse cardiovascular outcomes. Finally, we assess what is known to date about the hypoglycemia risk of each class including mechanistic studies that evaluate hypoglycemia counterregulation.

111

5.1 METFORMIN

The therapeutic role for metformin extends well beyond diabetes. Described as the aspirin of the 21st century, metformin is currently being explored for its role in prediabetes, gestational diabetes, polycystic ovarian disease, preeclampsia and longevity (Romero et al. 2017) (ClinicalTrials.gov Identifier: NCT02432287).

5.1.1 Mechanism of Action

The first reports of guanides date back to the medieval times when extracts of French lilac (Galega officinalis) were used as herbal remedy. In 1957, metformin (a demethylated ) and phenphormin (a phenetylated biguanide) were introduced as a therapeutic treatment for patients with T2DM. However, in the 1970’s, phenphormin was withdrawn from most countries due to an increased risk of lactic acidosis. Metformin, continued to be used cautiously in most countries including Canada but was not FDA approved for use in the USA until 1995 (Stumvoll, Haring, and Matthaei 2007).

Considered an insulin-sensitizer, metformin reduces hyperglycemia, without altering plasma insulin concentrations. Metformin is believed to alter mitochondrial cellular respiration, resulting in inhibition of hepatic gluconeogenesis. Through its activation of AMPK (AMP - activated protein kinase), metformin interacts with complex I of the mitochondrial electron transport chain. In doing so, cellular concentration of ATP decreases while AMP concentrations increase. This favors AMP binding to adenylate cyclase and inhibits the actions of glucagon, thereby reducing hyperglycemia. This, however, represents a very simplistic picture of metformin’s multiple mechanisms of action which include AMPK-dependent and independent pathways (Rena, Hardie, and Pearson 2017).

5.1.2 UKPDS 34

Today, metformin is recommended as first-line in most diabetes guidelines (Qaseem et al. 2017; Lipscombe et al. 2018; Ryden et al. 2013). This recommendation is largely based on a subgroup analysis of n=342 obese and newly diagnosed T2DM patients from the UKPDS. Compared to the conventional arm of diet alone, metformin was used to intensively lower BG levels to a fasting BG below 6 mmol/L. After a median of 10.7 years, metformin was associated with

112

significant benefits including a 32% reduction for any diabetes-related endpoint, 42% reduction in diabetes-related death and a 36% reduction in ACM (Group 1998a).

5.1.3 Risk of All-Cause Mortality and MACE

Despite the promising results from UKPDS 34, a recent meta-analysis of 13 randomized studies failed to show a significant benefit with metformin on either risk of ACM, CV death, MI, stroke or peripheral vascular disease (Griffin, Leaver, and Irving 2017). In contrast, a larger meta- analysis of 53 studies which allowed for the inclusion of non-diabetics and observational studies, illustrated significant benefits in ACM in metformin vs non-metformin users (HR 0.72, CI 0.65 to 0.80), as well as relative to SU (HR 0.80, CI 0.66-0.97) and insulin users (HR 0.78, CI 0.73 to 0.83). Furthermore, metformin use was associated with less incidence of cancer and CVD (Campbell et al. 2017). Of course, observational studies may be at high risk of channelling bias, where metformin users may have been in better general health than say, users of insulin or SU.

Many still hold great promise for metformin. Similar to statins, metformin has demonstrated pleiotropic effects, beyond its glucose lowering. Recent evidence suggests, metformin protects against endothelial dysfunction (Nafisa et al. 2018).

113

(Reproduced with permission)

5.1.4 Efficacy

Measured as time to monotherapy failure, efficacy of metformin monotherapy was evaluated in n=4,360 patients recently diagnosed with T2DM in the ADOPT study. After a median of 4.0 years of treatment, the cumulative incidence of monotherapy failure at 5 years was 15% with , 21% with metformin and 34% with glyburide (Kahn et al. 2006). More recently, in a meta-analysis of placebo-controlled randomized trials, metformin monotherapy was associated with a 0.95% decrease in HbA1c at 3 months (Piera-Mardemootoo, Lambert, and Faillie 2018).

5.1.5 Safety & Tolerability

Based on metformin’s tendency to increase gastrointestinal side-effects, including diarrhea, a low starting dose and slow titration is recommended. Metformin has been associated with an increased incidence of vitamin B12 deficiency (Lipscombe et al. 2018)

5.2 INSULIN

Almost a century ago, a diagnosis of insulin-dependent diabetes mellitus (T1DM) meant an early death sentence. In 1922, at the University of Toronto Banting, Best Collip and Macleod were first to successfully isolate insulin from the pancreas of pigs and rabbits, prolonging the lives of millions with both T1DM and T2DM (Bliss 1982). Today, insulin remains a fundamental therapy for patients with T1DM and plays an important role for many patients with T2DM who will eventually require it .

5.2.1 Mechanism of Action

The inability of the pancreatic beta-cell to meet the insulin demands is a pillar of diabetes pathophysiology. Treatment with exogenous insulin simply aims to meet this need. Over the years, insulin manufacturers have strived to innovate a type of insulin that closely resembles that of normal insulin physiology (Brady 2017).

114

(Reproduced with permission)

Second generation analogues of insulin (insulin glargine 300 U/mL and ) promise longer duration allowing for once daily dosing and a lower risk of hypoglycemia, relative to first generation insulins (Mauricio and Hramiak 2018).

5.2.2 Risk of All-Cause Mortality and MACE

Interestingly only one randomized controlled trial has evaluated the CV safety of insulin as its primary objective. ORIGIN (Outcome Reduction with an Initial Glargine Intervention) randomized n=12,537 patients with T2D or prediabetes to insulin glargine or standard of care. Insulin glargine was associated with neutral rates of MACE (and cancer) after a 6.2-year median follow up (Gerstein et al. 2012). To note, at baseline, similar proportion of patients in the glargine arm (30.3%) vs standard of care (28.9%) were receiving SU treatment. By study end, however, SU use was significantly higher in the standard of care. 115

(Formal permission not required for use in thesis)

Retrospective studies from the UK General Practice Research database have suggested an increase risk of MACE endpoints and ACM, even after adjusting for confounding variables. It is not clear, however in non-randomized studies, what role channeling bias may have. For instance, insulin use may merely reflect a sicker patient (Holden et al. 2015; Currie et al. 2013).

5.2.3 Efficacy

An upper dose limit of insulin does not exist, however, up titration is limited by risk of hypoglycemia (Lipscombe et al. 2018).

5.2.4 Safety & Tolerability

Insulin is notoriously known to increase risk of weight gain and hypoglycemia. Further, because insulin requires subcutaneous injection, many patients view use of insulin as painful, inconvenient and a last resort (Mauricio and Hramiak 2018).

5.2.5 Risk of Hypoglycemia by Insulin Type

The DEVOTE study was a prespecified non-inferiority study of insulin degludec compared to basal insulin glargine on the primary endpoint of MACE and secondary adjudicated endpoint of severe hypoglycemia. The primary endpoint was met and degludec was numerically but not significantly better than glargine over 24 months. Despite similar glycemic efficacy, degludec had significantly lower rates of severe hypoglycemia than insulin glargine (Marso et al. 2017). The lower rate of hypoglycemia with insulin degludec is believed to be due to its longer half life and lower glycemic variability than basal insulin glargine. In a study of n=54 patients with T1DM on a 24-hour euglycemic clamp for 12 days, insulin degludec demonstrated four times

116

lower glucose day to day variability than insulin glargine (Heise et al. 2012). Similar results were observed in n=32 inpatients on oral agents randomized to liraglutide plus degludec (DEgLira) or basal bolus regimens. No differences were observed in FPG or PPG2h but significantly fewer episodes of confirmed hypoglycemia (Akiyama et al. 2017).

In a US retrospective study of older patients with T2DM (DELIVER 3) patients who switched from basal insulin to insulin Glargine 300 U/mL (GLA-300) or other insulins were evaluated. Those switching to GLA-300 were 57% less likely to have hypoglycemia at 6-month follow up, with similar glycemic control compared to other basal insulin (insulin glargine 100 U/mL, insulin determir or insulin degludec) (Zhou et al. 2017). Interestingly, another study from ADA 2017 (EDITION 3) demonstrated that Gla-300 offered greater efficacy and a lower risk of hypoglycemia than Gla-100 (Dailey III et al. 2017).

In a 24-week crossover study from the UK of patients with T1DM (SWITCH 1) or T2DM (SWITCH 2), IDeg OD was compared to IGlar U100 OD. Noninferiority of glycemic control was established but with significantly less hypoglycemia with IDeg vs IGlar U100, especially with respect to nocturnal episodes (Wysham et al. 2017).

Finally, a systematic review and meta-analysis comparing efficacy and safety of intensive insulin treatment demonstrated basal insulin to be associated with lower frequency of hypoglycemia, despite less efficacy, in comparison to prandial insulin or NPH (Pontiroli, Miele, and Morabito 2012). This was corroborated with a more recent patient level meta-analysis of insulin naïve patients with T2DM who showed a lower risk of hypoglycemia with insulin glargine vs Neutral Protamine Hagedorn insulin (NPH) (Owens et al. 2017).

5.2.6 Risk of Hypoglycemia vs Active Controls

In ACCORD, risk of hypoglycemia was higher in patients without insulin and with SU treatment (Miller et al. 2010).

117

(Reproduced with permission)

5.2.7 Risk of Hypoglycemia Co-administered with DPP4i

The ADDONIS study set out to answer which agent is a better add-on when initiating basal insulin asides from metformin. Patients (n=42) inadequately treated with insulin secretagogues (SU, meglitinides) remained on therapy or were switch to metformin and in a 1:1 ratio. The primary endpoint of HbA1c <7% and incidence of hypoglycemia (defined was symptomatic and SMBG <3.9 mmol/L) was achieved in significantly more patients in the vildagliptin arm than the IS arm. HbA1c was similar but with numerically less hypoglycemia. However, only the composite endpoint achieved statistical significance in this pilot study (Gautier et al. 2016).

118

5.2.8 Risk of CV Event after Hypoglycemia

The CREDIT study (CV Risk Evaluation in people with T2DM on Insulin) prospectively evaluated n=2,999 patients starting insulin therapy on CV outcomes. The study included centres from 12 countries in Canada, Europe and Japan. Although they found a significant relationship between HbA1c and CV events, HR 1.25 p<0.0001, driven by CV death and stroke, MI (HR 1.05 CI 0.83-1.32) did not show a strong association with glucose control. Interestingly, there was no relationship between severe/symptomatic hypoglycemia events and CV specific or ACM (Freemantle et al. 2016).

(Reproduced with permission)

5.2.9 Risk of All-Cause Mortality after Hypoglycemia in Insulin-treated Patients

A retrospective study from the US of hospitalized patients (not necessarily diagnosed with diabetes) with an acute MI found that hypoglycemia was associated with increased mortality. However, iatrogenic hypoglycemia caused by insulin use was not associated with an increased risk of death and only spontaneous events were significant (Kosiborod et al. 2009).

Nevertheless, in the Korean retrospective analysis of patients with T2DM, insulin use was a significant predictor of CV mortality (HR 3.21; 1.36-7.55) and ACM (HR 3.07; 2.01-4.69). However, given the comparator (patients with diabetes not taking insulin), it may merely be a reflection of a younger population with less comorbidities and underlying disease (Cha et al. 2016).

119

When analyzing a non-diabetic population, hypoglycemia is a marker of a vulnerable patient with comorbidities. A retrospective study revealed that drug associated hypoglycemia was not associated with increased risk of death in general wards, whereas spontaneous hypoglycemia was significantly associated in unadjusted or minimally adjusted models. After adjusting for comorbidities, hypoglycemia was no longer significantly associated with mortality, suggesting it is merely a marker of a higher risk patient (Boucai, Southern, and Zonszein 2011).

(Reproduced with permission)

A retrospective study from Italy questions the channel bias theory. An analysis of 126 episodes of severe hypoglycemia in patients with T2DM presenting to an emergency room in Italy found similar proportions were taking oral glucose lowering therapies compared to insulin. Moreover, those on oral medications were older, with more incidence of coma, duration of hypoglycemia and longer hospitalizations. Despite the worse presentation, mortality was not increased compared to those presenting to the ER on insulin (Fadini et al. 2009).

5.3 SULFONYLUREA

5.3.1 Mechanism of Action

SUs main site of action is the ATP-sensitive potassium channels (K-ATP) which are located on pancreatic beta-cell membrane. Insulin secretion is stimulated when theses channels are blocked. Of course, K-ATP channels are also located in other tissue types, including cardiac and smooth myocytes, endothelium and the brain, yet little is known about extra-pancreatic inhibition of K- ATP (Leonard et al. 2017).

120

5.3.2 Risk of All-Cause Mortality

Since randomization of intensive vs less intensive studies were not based on SU use, these studies cannot further our understanding of SU’s effect on CV or ACM. In ACCORD, both the intensive arm and the standard treatment arm had >65% SU use. In ADVANCE, although the intensive arm was randomly assigned to gliclazide use, the use of other SUs was permitted. In fact, >63% of the conventional arm at baseline began a SU and >57% ended the trial on a SU (Leonard et al. 2017).

5.3.3 Risk of CV Death

Many observational studies and meta-analysis suggest SU increase CV and ACM (Phung et al. 2013; Bain et al. 2017; Simpson et al. 2006; Simpson et al. 2015; Rathmann et al. 2013; Schramm et al. 2011). However, similar to the insulin literature, it is not clear what role channelling biases may play in these studies (Powell, Christiansen, and Miller 2018). A recent meta-regression analysis aimed to evaluate the bias but still found SU’s were associated with an increased risk of CV events and ACM. For instance, when the comparator was metformin, SU was associated with a 13% higher risk. Similarly, when the outcome was death, relative risk increased by 20%. Finally, when study design related issues, such as time-lag bias was considered, SU increased risk by 7%. The authors concluded, that even in studies with limited or no design biases, SUs were associated with an increased risk of CV and ACM (Azoulay and Suissa 2017).

(Reproduced with permission) 121

Some argue that ADVANCE provided evidence in a prospective randomized controlled trial and demonstrated the CV safety of SU’s. This is because, gliclazide was the antihyperglycemic agent used in 90.5% of participants in the intensively treated arm. However, by end of follow- up, the comparator or standard of control, had 57% use of ‘other SUs’ (Patel et al. 2008). Prospective CV studies, CAROLINA provides much needed guidance on the CV safety of an SU (glimepiride) in comparison to a DPP4i () (Marx et al. 2015). After 6 years, the study sponsors have announced that linagliptin met its non-inferiority MACE endpoint in comparison to glimepiride (https://www.boehringer-ingelheim.us/press-release/boehringer- ingelheim-and-lilly-announce-carolina-cardiovascular-outcome-trial ). This much awaited large long-term randomized controlled trial reassured many of the CV safety of SUs.

5.3.3.1 Ischemic Preconditioning

CV concerns relate to an abolishment of ischemic preconditioning (IPC) based on multiple animal and human studies. IPC, an innate self-protective mechanism requires K-ATP channels to open during periods of ischemia. By blocking myocardial K-ATP channels, SU may lead to larger infarct size during ischemia and affect cardiac excitability. Secondly, by blocking K-ATP channels, action potential duration is reduced, increasing the intracellular calcium load and thereby risk of post depolarization arrhythmias. The mechanism of SU’s on IPC are separate from those on arrhythmias. Despite the vast use of SUs for decades, the clinical significance has not been evaluated in a designated CV trial (Leonard et al. 2017).

5.3.4 Characteristics of Different SU’s

Glyburide has been the most extensively studied SU and has demonstrated a blocking of IPC. SUs may differ in their impact of IPC blocking as glimepiride and gliclazide have not demonstrated similar effects in comparison to glyburide, and little is known about glipizide’s impact on IPC. SU selectivity of K-ATP channels in pancreatic vs myocardial cells likely plays an important role. Highly selective (>100X) for pancreatic channels are , gliclazide, glipizide and , whereas, glimepiride and glyburide demonstrate modest (>10X) selectivity. Further, agents differ in their selectivity for mitochondrial vs sarcolemma K- ATP channels (Leonard et al. 2017).

122

5.3.5 Efficacy

According to Diabetes Canada, the glycemic lowering efficacy of SU’s is comparable to thiazolidinediones and DPP4i’s (Lipscombe et al. 2018)

5.3.5.1 SU Protective

SU’s are known to block ATP sensitive potassium channels, which are believed to regulate arrhythmias during hypoglycemia. K-ATP blocker (5mg/kg/hr), K-ATP opener diazoxide (DIAZ) 5mg/kg/hr or vehicle were provided to nondiabetic Sprague Dawley rats during hypo, hyper or euglycemic clamps. During second degree heart block, glibenclamide and vehicle were comparable, but DIAZ significantly increased second degree heart block 10-fold. Further, third degree heart block was reduced significantly by 22% in glibenclamide and increased by 19% in the DIAZ group compared to vehicle. Therefore, during severe hypoglycemia use of K-ATP blockade with SU’s may in fact reduce potentially fatal arrhythmias (Reno et al. 2017). This is reassuring for those on SU’s, but use of AHA with placebo-like risk of hypoglycemia would bolster patient and physician confidence even more.

5.3.5.2 SU Detrimental

Middleton et al. studied n=30 patients with T2DM treated with SU for 48 hours on simultaneous ECG and continuous glucose monitoring. They studied the effect of hypoglycemia (<3.5 mmol/L) for 20 minutes in comparison to euglycemia and hyperglycemia on ventricular repolarization (QTc) and QT dynamicity. 30% (9/30) experienced hypoglycemia. Sixty-seven percent of episodes were nocturnal and most (73%) were asymptomatic. QTC prolongation was seen in 5/9 of the patients and QT dynamicity was greater in those who experienced hypoglycemia compared to those that did not (Middleton et al. 2017).

5.3.6 Risk of Hypoglycemia - Gliclazide vs Glyburide vs Glipizide and Glimepiride

The risk of hypoglycemia varies among SUs and is generally considered to be related to potency and duration of action. However, despite glyburide and glimepiride having similar glycemic efficacy and duration of action, the frequency of severe hypoglycemia tends to be greater with glyburide, suggesting other important factors may be responsible. Similar to beta-cells,

123

pancreatic alpha cells also contain SUR linked to K-ATP sensitive potassium channels. In ducks, glucagon was suppressed by several SUs including gliclazide, glyburide, and tolbutamide. In humans, long term use of SU treatment with tolazamide, chlorpropamide, tolbutamide or has been reported to suppress postprandial glucagon response. Some studies have suggested that during hypoglycemia, glucagon is suppressed by tolbutamide and glyburide. However, in a recent investigation, neither glyburide or glimepiride affected plasma glucagon, epinephrine nor growth hormone response during hypoglycemia, although plasma cortisol levels were slightly reduced. Further, glyburide, but not glimepiride, causes persistent inappropriate insulin secretion during recovery from hypoglycemia. The reason for the discrepancy of SU glucagon and insulin response is not clear (Szoke et al. 2006).

5.3.7 Risk of Hypoglycemia vs Active Controls

In a prospective study from New Mexico of n=265 patients with T2DM, the frequency and rate of hypoglycemia was found to be higher with insulin than SUs (Wendel et al. 2014).

(Reproduced with permission)

5.3.8 Sex-Gender Differences in Hypoglycemia Risk

In a Japanese drug event monitoring project, female sex was associated with significantly more hypoglycemia symptoms associated with SU use compared to males. Further, risk factors for hypoglycemia unique to females were shorter duration of therapy and use of concomitant antihyperglycemic therapy (Kajiwara et al. 2015).

5.3.9 Risk of MACE after Hypoglycemia

Using deidentified nationwide electronic health records from the US between 2009 to 2014, CV outcomes associated with hypoglycemia among SU users was evaluated. Of n=143,635 patients,

124

the incidence of acute MI, congestive heart failure and stroke were positively associated with serious hypoglycemia (Nunes et al. 2017).

5.3.10 Healthcare Utilization SU vs other therapies

A decision tree evaluating the rates, costs, quality-adjusted life-years and incremental costs per quality-adjusted life year gained in association with hypoglycemia of varying severity was conducted using US Medicare and Canadian public health perspectives. For Canadians aged 65 years and older, DPP4i use were found to be more cost-effective. The annual costs of hypoglycemia with insulin or SUs was estimated at over 65 million dollars a year (Boulin, Diaby, and Tannenbaum 2016). Similarly, a retrospective analysis of the UK Clinical Practice Research Datalink found patients on metformin plus SU, compared to metformin with other oral agents, had higher rates of secondary healthcare utilization. Further, metformin plus SU patients had higher inpatient admissions for macrovascular complications and outpatient cardiology visits (Strongman et al. 2015). Finally, an Italian study demonstrated that although the total annual costs per patient were not significantly different among SU, thiazolidinediones or users, SU and thiazolidinedione treatment were associated with higher disease-related hospitalizations (Degli Esposti et al. 2014).

5.4 INCRETIN

5.4.1 The Incretin Effect in Healthy Subjects

In 1964, the incretin effect was described in which oral glucose elicited a higher insulin secretory response than intravenous glucose, despite similar levels of glycemia in healthy individuals. This observation was followed by the characterization of the two known incretin hormones, glucose-dependent insulinotropic polypeptide (GIP) in 1973 and glucagon-like peptide 1 (GLP1) in 1985 (Horowitz et al. 2016). GIP is produced by K cells located mainly in the duodenum and upper jejunum and GLP1 is produced by L cells which reside throughout the intestine and increase in presence towards the ileum and colon. Increases of GIP and GLP1 are only seen with oral glucose (Nauck and Meier 2016). It remains unresolved whether a GLP1 response to oral nutrients is derived from the upper gut where most nutrient absorption takes place, or the distal parts which would imply an indirect neural or hormonal mechanism of eliciting GLP1 release.

125

In healthy individuals, GIP is predominantly responsible for the incretin effect (Nauck and Meier 2016).

5.4.1.1 Glucagon in Healthy

Glucagon production by alpha cells is suppressed more strongly with iv than oral glucose. This may be explained by the fact that only with oral glucose, GIP and GLP2 are secreted, thereby increasing glucagon levels (Nauck and Meier 2016).

5.4.2 The Incretin Effect in Patients with T2DM

In T2DM, GLP1 is believed to be absent or reduced, yet the pancreas still remains responsive to GLP1 but not GIP. GIP is now gaining wider acceptance for its role for the reduced or absent incretin effect. Patients with T2DM lose the majority of their insulinotropic response to GIP. Studies with supra-physiological GIP administration have failed to show sustained stimulation of insulin. Since GIP is thought to be responsible for much of the incretin effect in healthy individuals, its lack of activity in patients with T2DM has the potential to explain the reduced incretin effect. Unfortunately, there are no GIP receptor antagonists available for clinical studies and the contribution of GIP to the incretin effect can only be estimated using exogenous GIP.

In contrast to GIP, GLP1 stimulates a considerable insulinotropic effect in patients with T2DM, with a linear response to dose and concentration. Presumably due to a reduction of functional B- cell mass, GLP1 induced insulin secretion in T2DM is reduced compared to healthy individuals. The ability of GLP1 to suppress glucagon is also retained in patients with T2DM (Nauck and Meier 2016).

5.4.2.1 Glucagon Effect in T2DM

In patients with T2DM, glucagon secretion is abnormally elevated in both the fasting and post prandial state, contributing to hyperglycemia. Postprandially, GIP and GLP1 are released and have opposing actions on glucagon. Where GLP1 suppresses glucagon secretion, GIP stimulates it. Measurement of glucagon secretion is difficult. In vivo glucagon secretion is measured by peripheral plasma concentrations using radioactive immunoassay. However, peripheral concentrations do not necessarily coincide with the alpha cell secretion rate since the liver is involved in glucagon’s clearance (Holst et al. 2011). 126

5.4.3 Incretin-Based Therapies for Patients with T2DM

The development of incretin therapies for T2DM stemmed from the notion that insulin secretory response to glucose load differed based on route of entry. Oral glucose loading demonstrated a greater insulinotropic response compared to parenteral glucose loading. Although GIP and GLP1 are both defined as incretin hormones in humans, only exogenous GLP1 has demonstrated insulinotropic effects. As a result, therapies that increase GLP1 levels through two paths emerged.

One path inhibits the catalytic activity of DPP4, preventing the degradation of endogenously released GLP1 and GIP. More specifically, DPP4 cleaves GLP1 (7-36) amide soon after its release from the enteroendocrine L cells at its N-terminal to its metabolite GLP1 (9-36). This led to the development of the DPP4 inhibitors, of which four are approved for use in Canada (, linagliptin, saxagliptin and sitagliptin). Alternatively, GLP1RA increase GLP1 levels to pharmacological concentrations. Direct agonists of the human GLP1 receptor use either modified nonmammalian peptides (such as , ) or recombinant human GLP1 (liraglutide, ) which are structurally resistant to DPP4 proteolysis were developed (Lovshin 2017).

5.5 GLUCAGON LIKE PEPTIDE-1 RECEPTOR AGONISTS

5.5.1 Mechanism of Action

One of the first studies to demonstrate the glucose-dependent effects of GLP1 was conducted in n=10 patients with T2DM with uncontrolled glucose, despite diet and SU usage. GLP1(7-36 amide), 1.2 pmol x kg-I x min- t or placebo was infused in the fasting state. Insulin and C- peptide increased significantly, glucagon declined, and plasma glucose reached normal fasting concentrations within 4h of GLP-1, but not with placebo. When normal plasma glucose concentrations were reached, insulin returned towards basal levels and plasma glucose concentrations remained stable despite the ongoing infusion of GLP-1 (7-36 amide). The glucose-dependent insulinotropic actions of GLP-1 (7-36 amide) was demonstrated (Nauck et al. 1993).

127

The first GLP1RA was discovered from the Gila monster. Exendin-4 shares 53% homology with native GLP1 and is resistant to DPP4 because of the penultimate NH2 terminal glycine instead of alanine. This allows for a longer half life of 26 minutes instead of 1-2 minutes for the intact biologically active GLP1 (Deacon 2004).

Liraglutide is a GLP1 analog with 97% homology to human native GLP1. Fatty acid chain acylation with liraglutide promotes binding to albumin, reducing access of DPP4 to the NH2 terminal and prevents renal elimination. As a result, liraglutide has a plasma half life of 12 hours in humans, allowing for once daily dosing (Deacon 2004). Interestingly, liraglutide has been shown to increase the endogenous GLP-1 response over time in patients with early T2DM (Kramer et al. 2017).

5.5.2 Risk of All-Cause Mortality

Recent meta-analysis demonstrated that GLP1RA significantly reduced ACM by 11% (RR=0.89, CI 0.81-0.99). A total of n=21,135 patients from 8 trials were included in the analysis (3 , 2 lixisenatide, 2 liraglutide and one semaglutide). Interestingly, there were no differences between groups with respect to CV death, nonfatal MI, nonfatal stroke or hospitalization for heart failure (Peterson and Barry 2017).

5.5.3 Risk of MACE

Four large CVOTs, designed to evaluate the CV safety of GLP1RA in patients with T2DM and CVD or CV risk factors have been published to date. Unlike the intensive vs less intensive trials, these FDA mandated studies aimed for routine glucose control in both arms, thus only evaluating the safety of the AHA molecule. ELIXA, a study conducted in patients with acute coronary syndrome using lixisenatide, and EXSCEL with exenatide, demonstrated neutral CV safety (Pfeffer et al. 2015; Holman et al. 2017). On the other hand, LEADER with liraglutide and SUSTAIN-6 with once weekly semaglutide, demonstrated a significant reduction on CV outcomes relative to standard of care (Marso, Daniels, Brown-Frandsen, Kristensen, Mann, Nauck, Nissen, Pocock, Poulter, Ravn, Steinberg, Stockner, Zinman, Bergenstal, and Buse 2016; Marso, Bain, et al. 2016). Based on these results, liraglutide is now recommended in patients with T2DM and CVD (Lipscombe et al. 2018). Semaglutide has not yet been approved for use in Canada. 128

Recombinant human GLP1 (ending in -glutide) may prove to have differing CV effects than the nonmammalian GLP1 receptor agonists (ending in -atide). This theory has recently emerged since the CVOT results of LEADER and SUSTAIN-6 demonstrated superiority with liraglutide and semaglutide, yet the findings of ELIXA and FREEDOM with lixisenatide and exenatide respectively, failed to demonstrate significance. Furthermore, in a larger animal model study of ischemic reperfusion injury, the porcine pig, liraglutide failed to demonstrate improvements in myocardial infarct size (Kristensen et al. 2009).

5.5.4 Efficacy

GLP1RA are considered powerful glucose lowering therapies, second only to insulin (Lipscombe et al. 2018).

5.5.5 Safety & Tolerability

Because GLP1RA require subcutaneous injection, patients may be reluctant to initiate therapy compared to oral agents. GI side-effects such as nausea, vomiting and diarrhea when initiating GLP1RA are common but transient and can be attenuated by slow up-titration. GLP1RA are contraindicated in patients with or with a family history of medullary thyroid cancer or multiple endocrine neoplasia syndrome type 2 (Lipscombe et al. 2018).

5.5.6 Hypoglycemia Counterregulation with GLP1 Infusion

GLP1 is released from the intestinal L-cells in response to orally ingested nutrients. GLP1 stimulates the pancreas to release insulin in a glucose-dependent manner, eliminating risk of hypoglycemia. GLP1 also inhibits food induced glucagon secretion, which is also glucose- dependent, but not hypoglycemia induced glucagon secretion. Since GLP1 is unlikely to alter the counterregulatory response to hypoglycemia, it is an ideal candidate for patients with diabetes (Deacon 2004).

In healthy volunteers, GLP1 suppresses glucagon during periods of euglycemia and hyperglycemia but not hypoglycemia (≤3.7 mmol/L). In response to hypoglycemia, GIP augments glucagon secretion (Malmgren and Ahren 2015). GLP1 does not impair the overall hypoglycemia counterregulation, although a small reduction in GH response was observed

129

during deeper stages of hypoglycemia. The mechanism is not fully understood although pituitary and hypothalamic actions of GLP1 are likely responsible (Nauck et al. 2002).

Hypoglycemia onset was compared in healthy (n=7) subjects and patients with T2DM (n=8), given subcutaneous injection of 1.5 nmol GLP1/kg body weight, representing the maximum tolerated dose. Fifteen minutes later, participants received bolus glucose to reach 15 nmol/l. They found GLP1 induced reactive hypoglycemia in healthy individuals but not in patients with T2DM. Five out of the 7 healthy subjects demonstrated clear symptoms of hypoglycemia. In patients with diabetes, the increase in insulin secretion in response to GLP1 accounted for only 50% of the AUC for C-peptide. The authors noted that the impaired insulin response to GLP1 was likely the main reason for a lack of hypoglycemia. Further, insulin resistance and higher glucagon levels in patients with diabetes may have also contributed to the lack of hypoglycemia to the high dose of GLP1 in comparison to healthy subjects (Vilsboll et al. 2001).

5.5.7 Hypoglycemia Counterregulation with GLP1RA

In a hypoglycemia clamp study designed to assess the glucose-dependent insulin secretion and overall counterregulatory response to hypoglycemia after intravenous infusion of exenatide in healthy volunteers, cortisol, epinephrine, norepinephrine and growth hormone were similar in comparison to placebo. Circulating plasma glucagon levels were augmented with exenatide vs placebo during hypoglycemia (contradictory to Nauck JCEM 2002). This study also reconfirmed that glucose-dependent insulinotropic actions of exenatide. When glycemic levels began to fall below 5 mmol/L, the insulin secretion rate drastically declined and was similar to the ISR of the placebo arm (Degn et al. 2004). Thus, GLP1’s pancreatic stimulation of insulin is glucose-dependent and wanes as BG levels approach normal range (Campbell 2011; Inkster, Zammitt, and Frier 2012).

In a randomized controlled trial of patients with T1DM, no difference was observed between liraglutide and placebo when added to insulin in glucagon changes during hypoglycemia, suggesting the counterregulatory response remains unaffected (Pieber et al. 2015). Similarly, in patients with T2DM, once-weekly semaglutide did not affect the glucagon counterregulatory response during hypoglycemia (average nadir reached of 2.9 mmol/L) in a double-blind cross- over study (Korsatko et al. 2018).

130

In patients with T2DM, a single 50mg dose of albiglutide did not impair the counterregulatory response to hypoglycemia (glucagon, epinephrine, norepinephrine, growth hormone and cortisol). Furthermore, albiglutide’s response to hypoglycemia appeared earlier than that of placebo, a response that had not been observed with other GLP1-R agonists, potentially due to the longer half life of albiglutide in comparison to liraglutide or exenatide. Glucagon levels remained low until plasma glucose was clammed at 4.0 mmol/L and mean glucagon levels were significantly greater in the albiglutide group during hypoglycemia level of 3.3 mmol/L. When glucose was clamped at more severe levels of 2.8 mmol/L, glucagon levels with placebo were greater than with albiglutide (Hompesch et al. 2015).

(Reproduced with permission)

Similar results were demonstrated in a study of lixisenatide vs placebo in patients with T2DM receiving insulin. Again, at a hypoglycemia level of 3.5 mmol/L, mean glucagon levels (and epinephrine) were significantly lower with lixisenatide treatment than placebo, however, at lower levels of 2.8 mmol/L, mean glucagon and epinephrine levels did not differ between the subjects, implying that increase in glucagon was more pronounced during lixisenatide treatment than with placebo (see below table 2). Given that glucagon levels were still reduced with lixisenatide at glucose levels of 3.5 mmol/L, there is a need to examine hypoglycemia counterregulation in T2DM patients given GLP1RA as add-on to basal insulin (Farngren, Persson, and Ahren 2016).

131

(Reproduced with permission)

5.5.8 Combination with Insulin

Risk of hypoglycemia with GLP1RA as add-on to insulin have provided mixed results depending on the comparator and use of background therapies. For instance, in one meta-analysis, significant reductions in HbA1c with lower symptomatic hypoglycemia was reported with GLP1RA combined with insulin treatment in comparison to intensified insulin treatment (Wysham, Lin, and Kuritzky 2017). However, a meta-analysis of the GetGoal clinical program (lixisenatide or placebo as add-on to basal insulin with or without additional oral agents) found better efficacy with lixisenatide, but with a greater incidence of hypoglycemia vs basal insulin alone. Although not separately evaluated, the increased incidence was skewed by the GetGoal L Asia study, in which patients were also on background SUs (Charbonnel et al. 2014). Other meta-analyses have demonstrated that GLP1RA use in combination with basal insulin provides significantly less hypoglycemia with similar glycemic efficacy compared to other injectable insulin regimens. When compared to placebo or placebo intensification trials, hypoglycemia was significantly increased with GLP1RA and basal insulin (Maiorino et al. 2017). GLP1RA and basal insulin combinations have also been associated with low glycemic variability in comparison insulin alone, which could explain the lower hypoglycemia risk (Bajaj et al. 2017)

132

(Reproduced with permission)

A meta-analysis of the efficacy and adverse effects of fixed ratio insulin with GLP1RA (DegLira or LixiLan) was conducted. In comparison to GLP1RA, fixed ratio combination therapy was associated with more hypoglycemia but less GI side effects (Gao, Cai, and Ji 2017). This again is an important reminder that risk of hypoglycemia is heavily dependent on the comparator and background therapies known to increase hypoglycemia.

5.5.9 Risk of Hypoglycemia in CVD

A meta-analysis of randomized trials was conducted with incretin-based therapies in patients with CV risk factors or established CVD. The six CVOTs (EXAMINE, SAVOR TIMI 53, TECOS, ELIXA, LEADER and SUSTAIN-6) published at the time made the inclusion criteria. CVOTs with GLP1RA demonstrated a neutral risk of any hypoglycemia and a lower risk of severe hypoglycemia, owing largely in part to the results of the LEADER trial. On the other hand, CVOTs with DPP4i demonstrated a significant increased risk in both any and severe hypoglycemia, owing largely in part to the SAVOR TIMI 53 trial (Zhang et al. 2017).

5.5.10 Risk of MACE after Hypoglycemia

A post hoc analysis of LEADER presented at ADA 2017 demonstrated that after severe hypoglycemia, the risk of MACE, CV death and ACM significantly increased in both treatment groups. Risk of an event was greater earlier (within 15 days) compared to later (within 60 days)

133

after the severe hypoglycemia episode. The authors emphasized the importance of minimizing risk of hypoglycemia in the management of patients with diabetes (Zinman et al. 2018).

5.6 DIPEPTIDYL PEPTIDASE IV INHIBITORS

5.6.1 Mechanism of Action of Glucose Lowering

Following their release, endogenous GLP1 and GIP are rapidly degraded by the ubiquitous enzyme dipeptidyl peptidase 4 (DPP4). DPP4i prevent the degradation of endogenously released GLP1 and markedly increase circulating intact (active) forms of GLP1 and GIP concentrations. The relative contribution of GLP1 mediated vs non-GLP1 mediated mechanisms by DPP4i to reduce glycemia has evolved. GLP1 is believed to be responsible for 50% of the glucose lowering effects of DPP4i, 42-47% of the insulinotropic effects and 67% of the glucagon suppressive effects (Horowitz et al. 2016).

5.6.2 Risk of MACE

Prior to the results of the DPP4i CVOTs that demonstrated a neutral effect on MACE and ACM, earlier pooled data of efficacy and safety trials suggested a potential benefit. For example, a systematic review and meta-analysis of 18 RCTs comparing DPP4i to other oral medications found the risk of MACE was significantly reduced (RR 0.48 (CI 0.31-0.75)) (Patil et al. 2012). Similarly, in a meta-analysis of RCTs comparing DPP4i to other glucose lowering drugs or placebo, DPP4i showed a 30% significant reduction in MACE (MH OR 0.71 (CI 0.59 -0.86)) (Monami et al. 2012). However, unlike the CVOTs, these early studies enrolled generally healthier patients without established CVD.

5.6.3 Efficacy

Glycemic lowering abilities of DPP4i are believed to be similar, although few head-to-head studies have compared within class differences. Diabetes Canada categorizes HbA1c lowering efficacy with DPP4i as comparable to SU’s and thiazolidinediones, but less than that of GLP1RA and SGLT2i (Lipscombe et al. 2018).

134

5.6.4 Safety and Tolerability

Diabetes Canada cautions against the use of saxagliptin in patients with heart failure (Lipscombe et al. 2018).

5.6.5 Risk of Pancreatitis

A number of systematic review and meta-analyses have evaluated the risk of pancreatitis in randomized controlled trials with DPP4i. When the three DPP4i CVOTs (SAVOR TIMI, TECOS and EXAMINE) are pooled, the risk of pancreatitis is significantly elevated (OR 1.79 (CI 1.13-2.82)), although the absolute risk is small (0.13%) (Tkac and Raz 2017). However, when these three CVOTs are excluded from the analysis, the risk of pancreatitis is no longer significantly elevated (OR 1.11 (CI 0.57-2.17)) (Roshanov and Dennis 2015). Based on this, the authors proposed that the increased risk of pancreatitis with DPP4i may only be a concern in patients with previous CVD due to the presence of low grade inflammation in this population (Sohani, Li, and Sun 2016). This notion was corroborated in a retrospective case-control study from Sweden where patients with CVD were found to be at a higher risk of acute pancreatitis (adjusted Odds Ratio 1.35 (CI 1.25-1.45)) (Sjöberg Bexelius et al. 2013). Nevertheless, use of DPP4i should be avoided in patients with a personal or family history of pancreatitis.

5.6.6 Hypoglycemia Counterregulation with DPP4i

DPP4i enhance GLP1 and GIP levels, while reducing glucagon levels post meal. Thus, the glucose lowering effects of DPP4i are not only GLP1 related (Deacon 2004). By preventing the degradation of GLP1 and GIP after a meal, increased glucose sensing is observed in islet, beta and alpha cells, which as a result, increases insulin and decreases glucagon secretion, respectively. In contrast to the suppressive effects of glucagon during food ingestion, DPP4i does not inhibit glucagon counterregulation to hypoglycemia. Therefore, DPP4i reduce glucagon levels when glucose levels are high via GLP1, yet, increase glucagon during hypoglycemia via GIP (Ahren and Foley 2016).

There are five main mechanisms of glucagon secretion by alpha cells:

1. Reduced beta-cell secretion elicits a reduction in intra islet insulin dependent inhibition of glucagon secretion. 135

2. Direct effect of low glucose

3. Activation of the autonomic nerves

4. stimulated release of adrenaline from the adrenals and

5. action of GIP

(Reproduced with permission)

5.6.6.1 Effects of GIP

Glucose-dependent insulinotropic polypeptide (GIP) has been considered as a bifunctional glucose-dependent regulator of glucagon and insulin secretion. In a study of healthy volunteers, during hypoglycemia, GIP infusion caused a greater glucagon response compared to saline. During euglycemia, GIP infusion still elicited a greater glucagon response in comparison to saline. During hyperglycemia, glucagon was suppressed to a similar extent between saline and GIP infusions, and furthermore GIP more than doubled the insulin secretion rate. As a result, the authors suggested GIP be considered a BG stabilizer based on its diverging effects of the main pancreatic glucoregulatory hormones; insulin and glucagon (Christensen et al. 2011).

136

In n=12 patients with T2DM, the same authors injected GIP to measure plasma concentrations of glucagon, glucose, insulin, C-peptide and intact GIP. Overall, GIP was found to oppose the action of insulin in lowering plasma glucose. However, again they demonstrated that GIP’s effect on insulin and glucagon is dependent on the existing plasma glucose level. During insulin induced hypoglycemia, GIP demonstrates a glucagonotropic effect. In contrast, during hyperglycemia, GIP increases glucose disposal by increasing insulin release. Finally, during fasting glycemia, GIP has minor effects on plasma glucose levels due to the offsetting effects of glucagon and insulin (Christensen et al. 2014).

(Reproduced with permission)

137

(Reproduced with permission)

Vildagliptin has been shown to protect mice from insulin-induced hypoglycemia through a GIP- glucagon counterregulatory axis. Increased levels of GIP with DPP4i contributed to glucagon’s counterregulatory response (Malmgren and Ahren 2015). Similarly, in patients with T2DM, vildagliptin enhanced alpha cell responsiveness to the suppressive effects of hyperglycemia and the stimulatory effects of hypoglycemia (Ahren et al. 2009). The effects of vildagliptin and sitagliptin on incretin hormones, glucose concentrations, insulin and glucagon secretion were found to be affected in a similar manner in patients with T2DM after a mixed meal (Baranov et al. 2016). Finally, a comparison of the counterregulatory response to hypoglycemia between the DPP4i linagliptin and the GLP1RA liraglutide was conducted in an exploratory trial in Japanese patients with T2DM. Hypoglycemic glucose clamp tests were conducted before and after a two- week period. Similar effects of glucagon, growth hormone, cortisol, epinephrine and norepinephrine were reported, demonstrating that the counterregulatory responses were not affected with either linagliptin or liraglutide (Yabe et al. 2017).

5.6.6.2 GIP over time

Despite the fact that intact GLP1 levels disappear with sustained treatment with the DPP4i sitagliptin, intact GIP responses continuously increased during the course of treatment. This may contribute to the sustained efficacy and improved glucagon counterregulation during

138

hypoglycemia over time with DPP4i (Aaboe et al. 2010). Further, although GLP1 and GIP share similar intracellular signalling pathways, there are marked differences between the two incretin hormones with respect to beta-cell function. GLP1 acts immediately to reduce glucotoxicity, which likely paves the way for the effects of GIP to be regained. This was demonstrated in a 12- week study with sitagliptin. Sitagliptin was shown to partially restore the diminished insulinotropic effect of GIP. At week 1, although sitagliptin increased active GIP levels compared to placebo, there were no additional improvements in insulin secretion. The effect of GIP, however improved by week 12 compared with week 1 only in the sitagliptin treated patients, despite no further changes in GIP concentrations. This suggests DPP4i may further improve the pancreatic responsiveness to GIP over time (Aaboe et al. 2015).

5.6.6.3 GIP response early phase vs late phase

In healthy subjects during elevated glucose levels, GIP augments early and late phase insulin responses and has no effect on glucagon release. In contrast, during periods of fasting or insulin induced hypoglycemia, GIP has glucagon releasing properties with no effect on phase insulin secretion. However, in patients with T2DM during hyperglycemia, the GIP-dependent amplification of late phase insulin secretion is impaired and contributes to the incretin defect in diabetes. This defect is likely attributable to a combination of the general impairment in beta- cell function and the quality of beta stimulus elicited by GIP. Here, during fasting glucose levels, GIP elicits only a minor, short-lived early-phase insulin response. During hyperglycemia, GIP infusion amplified the insulin secretion rate during the early and late phase insulin response. This is in contrast to GLP1, which elicits a robust late phase insulin response during hyperglycemia, but also during fasting glycemia (Christensen et al. 2014).

5.6.7 Risk of Hypoglycemia vs Active Controls

Using US MarketScan Commercial Claims (2007-2013), hypoglycemia as defined by ICD-9 codes was assessed after 12 months of SU usage in n=245,201 patients with diabetes. This was then applied to those initiating DPP4i (n=176,786) to predict events, as if they had been prescribed SUs, after adjusting for baseline characteristics. The observed rate per 100-person- years in SU users was 7.9 (CI 7.8-8.1). In DPP4i users, the rate was 4.0 (CI 3.9-4.1). Had the DPP4i users been prescribed SU instead, the predicted model suggested a rate of 7.4 (CI 7.2-7.5).

139

The authors concluded that the introduction of DPP4i over the last decade has helped avoid approximately 50% of hypoglycemia episodes (LIU et al. 2018).

5.6.8 Risk of Hypoglycemia Co-administered with Sulfonylurea vs Placebo

Salvo et al. conducted a meta-analysis of DPP4i therapy in patients taking SUs. They found a 52% increase in the risk of hypoglycemia when DPP4i were combined with SUs (Salvo et al. 2016). The analysis was criticized, however, for including the EXAMINE CVOT (weight 36%), which may have skewed the results. The concern was that Salvo’s meta-analysis was to compare to placebo and EXAMINE compared alogliptin to standard of care. Standard of care allows use of other (non-DPP4i) therapies such as SU’s or insulin (Tufan et al. 2016). This criticism highlights the issue with most efficacy-focused placebo-controlled studies. Placebo-controlled studies are often misinterpreted to mean that other background therapies are not permitted. In the case of hypoglycemia, this can and has drastically skewed risk estimates to date. We will revisit the hypoglycemia risk estimates of placebo-controlled studies in Chapter 7 on Scoping Review.

5.6.9 Risk of Hypoglycemia Co-administered with Insulin

DPP4 inhibition has been shown to improve beta-cell glucose sensitivity. Interestingly, sitagliptin and vildagliptin, but not alogliptin, increased glucose sensitivity of insulin secretion in patients on insulin therapy (Ahren and Foley 2016).

With respect to alpha cells, vildagliptin, sitagliptin, saxagliptin and alogliptin have shown to suppress inappropriate glucagon responses to mixed meal and oral glucose. DPP4i’s do not affect glucagon levels in healthy individuals, given their glucose-dependent mechanism of inhibitory action on glucagon (Ahren and Foley 2016). Therefore, DPP4i’s suppress inappropriate glucagon secretion during periods of hyperglycemia while enhancing the stimulatory effect of glucose during periods of hypoglycemia (Chen, Yu, et al. 2015).

5.6.10 Risk of Hypoglycemia Co-administered with Insulin vs Placebo

Due to its progressive nature, patients with T2DM will eventually require insulin therapy. Incretin and insulin combination therapy may offer a therapeutic advantage since insulin can

140

improve fasting glucose levels while DPP4i can inhibit postprandial hyperglycemia (Schweizer et al. 2013).

DPP4i improve glycemic control by increasing concentrations of GIP and GLP1, both of which act in a glucose-dependent manner to stimulate insulin secretion from the beta-cell. DPP4i have also shown to suppress glucagon secretion during periods of hyperglycemia. More importantly, during periods of hypoglycemia, GIP stimulate alpha cells to secrete glucagon. Thus, DPP4i may offer a lower risk of hypoglycemia when used in conjunction with insulin in comparison to insulin intensification or other insulin combination therapies for similar glycemic efficacy. In a recent meta-analysis, DPP4i demonstrated better glycemic control without inducing greater hypoglycemia risk. The comparator of this analysis was the combination of placebo or other anti-diabetic agents, making the benefit to risk profile difficult to interpret. In addition, insulin dose and rate of insulin titration vary amongst studies, further challenging interpretation of results (Chen, Yu, et al. 2015).

Nevertheless, promising results have been demonstrated with DPP4i and insulin combination therapies on hypoglycemia risk. In a pooled analysis of two 24-week RCTs of sitagliptin vs placebo as add-on to insulin, sitagliptin demonstrated similar HbA1C lowering but with less hypoglycemia. Interestingly, elderly individuals had similar or lower rates of hypoglycemia compared to younger individuals (Shankar, Long, et al. 2017). The addition of vildagliptin has also been shown to reduce glycemic variability during hyperglycemic episodes compared to up- titration of metformin in a study from Japan in patients with T2DM on multiple daily injections (MDI) (Yoshikawa et al. 2017).

5.6.11 Sex-Gender MACE Outcomes with DPP4i

In all three CVOTs of DPP4i reported thus far, no sex-gender specific differences were found for the primary MACE outcome (White et al. 2016; Scirica et al. 2013; Green et al. 2015).

5.6.12 Risk of Hypoglycemia in RCTs of CKD

In a 54-week randomized controlled trials of T2DM patients with moderate to severe kidney disease, sitagliptin had lower incidence of symptomatic hypoglycemia compared to glipizide, while providing similar HbA1c lowering (Arjona Ferreira, Marre, et al. 2013). In a second 54-

141

week study of T2DM patients with ESRD receiving dialysis, sitagliptin had numerically, but not significantly, less rates of symptomatic hypoglycemia compared to glipizide. For severe hypoglycemia, no patients in the sitagliptin arm and 7 patients in the SU arm reported an episode (Arjona Ferreira, Corry, et al. 2013). Finally, in a RCT of T2DM patients with CrCl <50ml/min, saxagliptin had similar rates of reported hypoglycemia compared to placebo (Nowicki et al. 2011).

5.6.13 Risk of Hypoglycemia in SRMA of CKD

As seen in the RCTs, risk of hypoglycemia in patients with T2DM on DPP4i and CKD depend on the comparator. In a meta-analysis of DPP4i trials conducted in patients with CKD (<60ml/min/1.73m²), hypoglycemia (although with variable definitions amongst the studies) was significantly increased in comparison to placebo, but neutral in comparison to active controls (Howse et al. 2016). In moderate to severe CKD including ESRD, DPP4i were associated with significantly less hypoglycemia when compared to glipizide and similar in comparison to placebo (Cheng et al. 2014). The results of a more recent SRMA of moderate to severe CKD including ESRD, demonstrated numerically lower hypoglycemia in comparison to glipizide (RR = 0.49) and numerically higher in comparison to placebo (RR = 1.19) (Singh-Franco, Harrington, and Tellez-Corrales 2016). Similar findings were found in a SRMA of T2DM patients with severe renal impairment (eGFR <30ml/min/1.73m²) or ESRD. DPP4i, when compared to placebo, resulted in a numerically increased risk of hypoglycemia (OR 1.43 (CI 0.89-2.30)) Alternatively, when DPP4i were compared to an SU (glipizide), risk of hypoglycemia was significantly lower (OR 0.55 (015, 1.99)) (Chen et al. 2016).

5.6.14 Risk of Hypoglycemia in CVD

A recent meta-analysis of incretin-based therapies in patients with CV risk factors or established CVD showed an increased risk of any and severe hypoglycemia with DPP4i, mainly driven by the results of SAVOR-TIMI 53. It should be noted that these trials were conducted with DPP4i vs placebo on top of standard therapy, and thus use of insulin, SUs and other non-incretin based therapies were permitted (Zhang et al. 2017; Abbas, Dehbi, and Ray 2016).

142

(Permission not required – Open Access)

One meta-analysis however, reported hypoglycemia but not severe hypoglycemia was significantly increased from the 3 CVOTs of DPP4i. In contrast, when the same authors evaluated hypoglycemia risk with DPP4i in a non-CV population, hypoglycemia risk was neutral and significantly decreased when compared to active controls (Xu et al. 2017).

The EXAMINE trial enrolled patients with T2DM and acute coronary syndrome and found a neutral effect on any and severe hypoglycemia compared to placebo. Nevertheless, a significant correlation was found between hypoglycemia and the primary MACE outcome. Moreover, hypoglycemia had a stronger correlation to MACE than MACE to hypoglycemia. It remains debated as to whether hypoglycemia causes subsequent MACE or whether patients experiencing hypoglycemia present with more comorbidities increasing their risk of MACE (Heller et al. 2017).

In SAVOR-TIMI 53, the risk of major and minor hypoglycemia was significantly increased, although hospitalization for hypoglycemia, the primary outcome measure of hypoglycemia in the trial, was neutral (Scirica et al. 2013). Post hoc analyses pointed to background SU use in saxagliptin patients experiencing hypoglycemia (Cahn et al. 2016). Yet, SU use increased more throughout the trial in the standard arm than in the saxagliptin arm. Perhaps, a combination of DPP4i with SU results in an even greater risk of hypoglycemia, than merely the sum of risk associated with each agent. It should be noted that background SU use was prevalent in all the CVOTs and SU and insulin use increased in the standard treatment arms more so than the new AHA being assessed. The use of diaries seen in SAVOR-TIMI 53 may have enhanced reporting

143

of hypoglycemia. However, both treatment arms received diaries and SU use was greater in the standard arm.

5.6.15 Risk of MACE after Hypo

A subgroup analysis from TECOS, demonstrated for the first time a bi-directional relationship between severe hypoglycemia events and CV outcomes. Severe hypoglycemia was rare and balanced between sitagliptin and placebo. Patient factors that increased severe hypoglycemia risk included older age, insulin treatment, women, non-white, longer diabetes duration, lower eGFR and lower weight. Patients that experienced severe hypoglycemia were at a significant increased risk of CV death, MI and ACM. Adjusted HR of MACE plus was 1.31 (CI 0.89-1.92) and of MACE was HR 1.39 (CI 0.94-2.06). Similarly, severe hypoglycemia occurred more frequently after a nonfatal MI, stroke or HHF (adjusted HR of MI 2.39 p<0.0001, stroke 1.50 p=0.2482, unstable angina 1.24 p=0.56 and HHF 2.80 p<0.0001). Since the risk of MACE was greater in patients that experienced a severe hypoglycemia event compared to the risk of severe hypoglycemia in patients with MACE, and although authors claim bidirectionality, hypoglycemia predicts MACE more so than MACE predicts hypoglycemia (Standl et al. 2018).

The SAVOR-TIMI 53 trial evaluated hospitalization for hypoglycemia as its primary hypoglycemia endpoint. Saxagliptin did not differ from placebo on this definition (HR 1.22 (CI 0.82-1.83, p=0.33)). However, significantly more patients in the saxagliptin arm reported at least one hypoglycemia event (15.3% vs 13.4%, p<0.001), major hypoglycemia (2.1% vs 1.7% p=0.047) and minor hypoglycemia (14.2% vs 12.5%, p=0.002) (Scirica et al. 2013). SAVOR TIMI 53 also conducted subgroup analyses and found that risk for any or major hypoglycemia occurred in those taking SUs, but not insulin.

144

(Reproduced with permission)

Further, rates increased in those on saxagliptin with a baseline HbA1c of ≤7.0%, but not in those with baseline HbA1c of >7.0%. In the overall population, multivariate analysis revealed that predictors of any hypoglycemia include; randomization to saxagliptin, long duration of diabetes, increased updated HbA1c, macroalbuminuria, moderate renal failure, SUR and insulin use. Predictors of major hypoglycemia were randomization to saxagliptin, advanced age, black race, reduced BMI, long duration of diabetes, declining renal failure, microalbuminuria and use of short acting insulin (Cahn et al. 2016).

In the overall SAVOR-TIMI 53 population, major hypoglycemia was associated with an increased risk of CV death, HF hospitalization, but not MI. Conversely, any hypoglycemia was not associated with CV death, HHF, or MI.

145

(Cavender et al. 2014)

(Reproduced with permission)

Based on the increased risk of HHF observed in the saxagliptin arm of SAVOR-TIMI 53, there was great interest to better understand the relationship of the increased major and any hypoglycemia with HHF. In a post hoc subgroup analysis of heart failure risk in SAVOR-TIMI 53, minor or major hypoglycemia was significantly increased (HR 1.16 p=0.001) in those without prior HF. In those with prior heart failure, hospitalization for hypoglycemia was significantly increased HR 2.55 p=0.004 (Scirica et al. 2014).

146

(Reproduced with permission)

5.7 SODIUM GLUCOSE COTRANSPORT 2 INHIBITORS

5.7.1 Mechanism of Action

The kidneys play an important role in gluconeogenesis, transportation and utilization of glucose from the circulation. They reabsorb glucose from the glomerular filtrate and regulating hormones associated with glucose metabolism. In healthy kidneys, approximately 160g/day of glucose is filtered. This represents 30% of the daily energy intake. To ensure the body does not lose energy in the urine, the proximal tubule reabsorbs glucose up to 450 g/day. The surplus is excreted in the urine when BG level surpasses this limit (Vallon and Thomson 2017).

The plasma membrane of living cells contains a lipid bilayer that does not allow for the free passage of highly polar glucose molecules. Therefore, membrane proteins are necessary for the movement of glucose into cells. Two classes of glucose transporters exist in the human body. The GLUT transporters passively facilitate glucose movement into cells along its chemical gradient into the intracellular space without consuming energy, whereas, sodium-glucose

147

cotransporters actively transport glucose against its concentration gradient and thus require energy. 14 different GLUT and 7 SGLT have so far been identified, both large membrane proteins. Located at the apical or luminal membrane of the proximal tubule of the nephron, SGLTs are a family of active sodium dependent membrane transport proteins that reabsorb 99% of the plasma glucose that filters through the glomerulus. SGLT couple sodium and glucose transport into the cell. Two SGLTs (SGLT1 and SGLT2) have been identified as playing an important role in the kidney. SGLT1 shares 58% homology with SGLT2 (Abdul-Ghani, Norton, and Defronzo 2011). SGLT2 and SGLT1 are located in the luminal membrane of the early (S1) and late (S3) segment of the proximal tubule, respectively. The low capacity, high affinity SGLT1 is responsible for most of the intestinal reabsorption and 10% of the renal glucose reabsorption. The high capacity, low affinity transporter, SGLT2 actively reabsorbs 90% of the filtered glucose. To facilitative glucose out of the cell alongside SGLT2 and SGLT1, are GLUT2 and GLUT1, located on the basolateral membrane of the S1 and S3 segments, respectively (Vallon and Thomson 2017).

In the first step, glucose is transported across the apical membrane by SGLTs, increasing the glucose concentration gradient. This gradient between the cell and the plasma drives the second stage; the net passive exit of glucose through GLUT2 across the basolateral membrane. Also located along the basolateral membrane is the Na/K pump which maintains the sodium gradient by pumping three Na ions out of the cell for every two potassium ions entering into the cell. The sodium gradient is critical for the sodium-glucose co-transport across the apical membrane (Ghezzi, Loo, and Wright 2018).

Most agents in this class selectively inhibit SGLT2 and reduce plasma glucose by augmenting glycosuria. The magnitude of glycosuria is dependent on the amount filtered by the kidneys, which is dependent on plasma glucose concentrations. Therefore, BG levels cannot be lowered below physiological levels and the risk of hypoglycemia is theoretically absent (Bakris et al. 2009). In patients with diabetes, the renal glucose reabsorption is increased compared to normal subjects. Furthermore, the renal threshold for glucose excretion (RTG) is also increased in patients with diabetes compared to healthy subjects. This is done via upregulation of the SGLT2 expression, increasing the reabsorption of sodium and glucose in the proximal renal tubules. By inhibiting SGLT2, reabsorption of glucose is inhibited, lowering plasma glucose levels. Age,

148

duration of diabetes, BMI and HOMA-IR have been shown to be independently be associated with high RTG (Yue et al. 2017) (Vallon and Thomson 2017).

(Reproduced with permission)

5.7.2 Risk of All-Cause Mortality

To date, three CVOTs, mandated by the FDA to demonstrate CV safety have presented results with SGLT2i. These three studies vary in size and the proportion of patients enrolled with established CVD vs CV risk factors, making comparisons challenging. Nevertheless, as it relates to death from any cause, only EMPAREG demonstrated a significant reduction (HR 0.68 (CI 0.57-0.82)) (Zinman et al. 2015), whereas CANVAS (HR 0.87 (CI 0.74-1.11)) (Neal et al. 2017) and DECLARE (HR 0.93 (CI 0.82-1.04)) (Wiviott et al. 2019) were neutral for this outcome. Systematic reviews and meta-analyses on ACM risk with SGL2i’s have mainly been driven by the results of the EMPAREG trial, where 99% of the participants had established CVD (Tang et al. 2016; Zelniker et al. 2019) . When meta-analyzing CVOTs vs non CVOT with SGLT2i, only the former is associated with a significant reduction in ACM (Monami, Dicembrini, and

149

Mannucci 2017). In a large observational study from national registries in the US, Norway, Denmark, Sweden, Germany and the UK, SGLT2i use (mostly dapagliflozin) was associated with a lower risk of HHF and death. Although many variables greatly differed between those prescribed an SGLT2i vs other glucose lowering drugs, significance was maintained after propensity score matching (Kosiborod et al. 2017).

5.7.2.1 Risk of MACE with SGLT2i

Both EMPAREG (HR 0.86 (CI 0.74-0.99)) (Zinman et al. 2015) and CANVAS (0.86 (CI 0.75- 0.97)) (Neal et al. 2017) demonstrated a significant reduction for the composite outcome of myocardial infarction, stroke and CV death. Results from the recent DECLARE study, although neutral overall for MACE (0.93 (CI 0.84-1.03)) (Wiviott et al. 2019) allowed for greater interpretation into the MACE risk reduction with SGLT2i. With n=10,186 patients at risk for CV events and n=6,974 patients with established CVD, subgroup analysis reaffirmed that the reductions of MACE, as well as the composite endpoint of heart failure and CV death occur primarily in patients with established atherosclerotic disease and not in those with CV risk factors (Zelniker et al. 2019).

5.7.2.2 Risk of Heart Failure Hospitalization

The CVOTs have demonstrated that SGLT2i exert their greatest benefit on reducing the risk of heart failure hospitalization (Zelniker et al. 2019), likely owing to the volume off-loading, natriuretic effect of the class. Given the reduction in heart failure hospitalization has consistently occurred both in patients with established CVD or CV risk factors, some have argued that this class should be recommended as first line therapy for patients with T2DM given their increased risk for declining renal function and heart failure (Verma, Juni, and Mazer 2019).

5.7.3 Efficacy

Diabetes Canada assigns similar glycemic lowering ability of SGLT2i to GLP1RA, given normal renal function (Lipscombe et al. 2018).

150

5.7.4 Safety and Tolerability

The SGLT2i are associated with a number of safety and tolerability issues, described in greater detail below.

5.7.4.1 Euglycemic Diabetic Ketoacidosis

Although eDKA is uncommon, the FDA has issued a warning about use of SGLT2i associated euglycemic diabetic ketoacidosis. Although SGLT2i causality has been debated, multiple biological mechanisms have been proposed and summarized by Taylor et al (Taylor, Blau, and Rother 2015).

(Reproduced with permission)

A cross-Canada expert panel have proposed a protocol as a preventative strategy to avoid SGLT2i-associated DKA (Goldenberg et al. 2016).

151

(Reproduced with permission)

5.7.4.2 Genital and Urinary Mycotic Infections

The most common side effect of SGLT2i are genital and urinary tract infections. The real-world incidence is difficult to estimate since patients may not attribute the infection to the medication if they are not briefed or may feel too embarrassed to mention to their health care providers. In a pooled analysis of studies aimed to evaluate the incidence of genital mycotic infections, reported an incidence of up to 14.7%. Many of the studies included in this analysis, however, had excluded patients at risk of an infection prior to randomization. Therefore, the true incidence in the real-world may indeed be much greater (Nyirjesy et al. 2012).

152

5.7.4.3 Volume Depletion

As indicated by Diabetes Canada, a number of medications need to be discontinued if a patient becomes ill, unable to ingest fluids or experiences a decline in renal function. The acronym SADMAN has now been changed to SADMANS to include SGLT2i (S-Sulfonylurea; A-Ace inhibitors; D-diuretics, direct renin inhibitors; M-metformin, A-angiotensin receptor blockers; N- non-steroidal anti-inflammatory; S-SGLT2i) guidelines.diabetes.ca/docs/CPG-2018-full-EN.pdf.

The results of EMPAREG has caught the attention of cardiologist, who for years have been desperately seeking to improve outcomes in patients with heart failure. A recent editorial by Cherney and Udell, however, has drawn caution to the use of SGLT2i in combination with diuretics due to concerns of volume depletion, especially in the frail elderly and in those with CKD. They have proposed a protocol for initiating SGLT2i for high risk patients taking concomitant diuretics (Cherney and Udell 2016).

(Reproduced with permission)

153

Caution is also recommended in patients consuming concomitant NSAIDs and SGLT2i due to concerns of acute kidney injury, more specifically hypoxic medullary injury (Heyman et al. 2017).

5.7.5 Hypoglycemia Counterregulation with SGLT2i

SGLT2i are known to increase glucagon concentrations and hepatic glucose production in patients with T2DM (Vallon and Thomson 2017). SGLT2i naturally stop working when the filtered glucose load falls below ~80 g/day. The low incidence of hypoglycemia with canagliflozin is thus expected, given that the renal glucose threshold (RTg) is above the threshold for hypoglycemia (≤3.9 mmol/L) (Stenlof et al. 2014). At less than 80g/day, the glucosuric load is handled by SGLT1. During inhibition of SGLT2, an increase in SGLT1 accounts for residual glucose reabsorption. The degree of glycosuria with SGLT2i depend on the degree of glycemia and GFR. The reduction in glucose levels are not dependent on insulin action or capacity of the beta-cells to secrete insulin. The liver reacts to the glucosuria by increasing glucose release, although the signal has yet to be identified (Ferrannini and Solini 2012). Therefore, SGLT2i are not expected to cause hypoglycemia, given that they decrease plasma glucose concentrations without augmenting insulin secretion and without inhibition of the counterregulatory response (Abdul-Ghani, Norton, and Defronzo 2011).

In an Austrian crossover study of n=17 metformin treated patients with T2DM, dapagliflozin did not impact glucagon levels during eu-, hyper-, or hypoglycemia. The addition of a DPP4i saxagliptin to dapagliflozin also did not impact glucagon secretion. Dapagliflozin, saxagliptin and the combination significantly reduced endogenous glucose production during hyperglycemia, but not eu- or hypoglycemia. Further, dapagliflozin (and saxagliptin) had no impact on lipolysis (measured as Ra: rate of appearance of glycerol) (Sach-Friedl et al. 2017).

5.7.6 Risk of Hypoglycemia vs Active Controls

Dapagliflozin demonstrated a ten-fold lower risk of hypoglycemia compared to glipizide over a 104-week period. For both treatment groups, the risk of hypoglycemia was lower in the second year of therapy than the first, however, it is not clear if rising HbA1c levels or less frequent study visits in the second year may have influenced these findings (Nauck, Del Prato, et al. 2014). The DECLARE study also demonstrated numerically less major hypoglycemia defined as requiring 154

external assistance (HR 0.68; 95% CI 0.49-0.95, p=0.02) compared to placebo on top of standard of care (Wiviott et al. 2019). In EMPAREG, any confirmed hypoglycemic events occurred in 27.8 vs 27.9 % of patients and those requiring assistance in 1.3 vs 1.5 % of patients with pooled vs placebo, both of which were neutral (Zinman et al. 2015). In CANVAS, the combined event rate of serious and nonserious hypoglycemia occurred in 50 and 46.4 per 1000 patient-years in canagliflozin and placebo treated patients, respectively (p=0.20) (Neal et al. 2017).

5.7.7 Risk of Hypoglycemia Co-administered with Sulfonylurea vs Placebo

Dapagliflozin, co-administered with glimepiride, significantly increased hypoglycemia rates in a dose dependent manner relative to placebo (Strojek et al. 2011).

5.7.8 Risk of Hypoglycemia Co-administered with Insulin

The beta-cell independent mechanism of glucose lowering with SGLT2i may allow for a lower risk of hypoglycemia. In addition, a decrease in the renal threshold for inhibition of glucose reabsorption or other insulin independent mechanisms may lower hypoglycemia risk and make for a good partner for insulin combination therapy. In a recent review of ideal partners for insulin combination therapy, authors concluded that SGLT2i and DPP4i should offer significant reductions in HbA1c without an increased risk of hypoglycemia (Singh and Singh 2016).

However, in a pooled analysis of RCTs with SGLT2i in patients with T2DM, dapagliflozin was found to double the risk of hypoglycemia vs placebo (OR 1.27; 1.05 -1.53). After conducting post hoc sensitivity analyses, the risk was entirely and exclusively explained by insulin co- administration. When the two insulin studies were excluded (Wilding et al. 2009; Wilding et al. 2012) the risk of hypoglycemia was not significantly elevated (OR 1.31; 0.93-1.86) (Musso et al. 2012). Empagliflozin 25 mg, and not 10mg, when used as add-on to basal insulin was also found to increase the risk of hypoglycemia in the first 18 weeks (21% on placebo, 20% empagliflozin 10mg, 28% empagliflozin 25mg). However, after physicians were allowed to titrate the dose of insulin, the rates of hypoglycemia were similar over the 78-week treatment period (Rosenstock, Jelaska, et al. 2015). Thus, although mechanistically, new AHA may suggest promise of lower

155

hypoglycemia risk, meta-analyses specifically designed to investigate hypoglycemia risk are still necessary as efficacy-focused studies have provided mixed results.

5.7.9 Risk of Hypoglycemia Co-administered with Incretin-based therapies

In a post hoc analysis of the CANVAS trial, documented hypoglycemia of ≤3.9 mmol/L occurred more frequently in patients taking canagliflozin with either DPP4i or GLP1RA. Unfortunately, statistical analyses were not conducted (Fulcher et al. 2016). Therefore, estimates of hypoglycemia risk when new AHA are combined with each other should be evaluated separately, as results may differ compared to use of AHA alone or in combination with metformin.

5.7.10 Risk of MACE after Hypoglycemia

In a pooled meta-analysis of Phase 2 and 3 RCTs of n=9,339 patients treated with dapagliflozin, those who experienced hypoglycemia did not have an increased risk of MACE. In fact, risk of MACE was numerically, but not significantly lower in those who experienced a hypoglycemia event (Sonesson et al. 2016).

(Permission not required – Open Access)

The opposite was shown in the SAVOR-TIMI 53 trial. Major hypoglycemia was associated with an 81-fold increase in CV death and an 80-fold increase in hospitalization for heart failure. Interestingly, there was no association between major hypoglycemia and MI.

156

(Permission not required – Open Access)

(Cavender et al. 2014)

5.8 SUMMARY

Metformin, commonly used as first line in Canada, US and Europe, is believed to increase insulin sensitivity and inhibit hepatic gluconeogenesis through the activation of AMP-activated protein kinase. When used as monotherapy, metformin provides an effective 0.95% decrease in HbA1c levels. Safety and tolerability issues mainly relate to gastrointestinal side effects. Hypoglycemia risk, when used as monotherapy is considered low.

The monumental discovery of insulin in 1920 at the University of Toronto prolonged the lives of patients with T1DM and T2DM around the world. Significant progress has since been made, improving the individualization of insulin treatment through a number of formulations and delivery methods. Given that a maximum dose limit does not exist, insulin provides the most effective HbA1c lowering. However, risk of hypoglycemia and weight gain is significant and requires careful consideration. Retrospective studies suggest insulin-treated patients are at an increased risk of all-cause mortality and cardiovascular events. However, despite efforts to adjust for confounders, patients requiring insulin may merely be reflecting a sicker population. ORIGIN, a large prospective study, determined insulin glargine was associated with neutral cardiovascular and cancer outcomes. Critics, however, have questioned the mostly SU use in the comparator arm. Nevertheless, insulin treatment is critical in patients with T1DM and eventually for patients with T2DM whose endogenous insulin levels become depleted.

157

SU’s stimulate endogenous insulin secretion by blocking ATP channels on the pancreatic beta- cell. They provide cheap, rapid and effective glucose lowering but are associated with weight gain and hypoglycemia risk. Similar to retrospective data with insulin, SUs have been associated with an increased risk of all-cause mortality and cardiovascular events. Given the ubiquitous nature of ATP channels, some fear that SU also block ischemic preconditioning, which protects the myocardium from subsequent and potentially fatal ischemia. However, recently, top line results of a long awaited cardiovascular safety study, CAROLINA, demonstrated that risk of cardiovascular events were non-inferior to a DPP4i, linagliptin.

The incretin-based therapies include the glucagon like peptide-1 receptors agonists and dipeptidyl peptidase-4 inhibitors. GLP1RA and DPP4i both stimulate pancreatic insulin secretion by increasing GLP1 levels to pharmacological and physiological doses, respectively. When used as monotherapy, the glucose-dependent action of GLP1 affords both classes with a negligible risk of hypoglycemia. Two subclasses of GLP1RA are available and approved for use as subcutaneous injections. GLP1RA ending in ‘atide’ use modified nonmammalian peptides whereas those ending in ‘utide' use recombinant human GLP1. Overall, GLP1RA are second to insulin in glycemic lowering efficacy. GLP1RA treatment requires dose titration to minimize GI side effects, which may be attributable to their effects on weight loss. GLP1RA are contraindicated in patients with a personal or family history of medullary thyroid cancer. Four large cardiovascular safety studies have been published with GLP1RA. ELIXA (with lixisenatide) and EXSCEL (with exenatide) were conducted in patients with acute coronary syndrome and demonstrated CV safety in comparison to standard of care. On the other hand, LEADER (with liraglutide) and SUSTAIN-6 (with once-weekly semaglutide) demonstrated superiority in MACE outcomes in patients with established cardiovascular disease, affording them priority recommendation in clinical practices guidelines for patients with T2DM and CVD.

DPP4i, also increase GLP1 levels but do so by inhibiting the DPP4 enzyme which degrades GLP1 (and GIP). Their glycemic lowering efficacy is considered comparable to sulfonylurea, according to Diabetes Canada. Overall, DPP4i are considered weight neutral and safe but must be used with caution in patients at risk or with a history of pancreatitis. Three cardiovascular safety studies and top line results of one have demonstrated the cardiovascular safety of this class. Only in one trial, SAVOR TIMI-53, an increased risk of heart failure hospitalization was

158

observed and thus must be used with caution in patients with or at risk of heart failure. By inhibiting the degradation of GIP, DPP4i may also offer an additional advantage during times of hypoglycemia. During times of hypoglycemia, GIP increases glucagon levels, acting as a BG stabilizer. This additional mechanism of hypoglycemia counterregulation may thus offer the DPP4i class with a unique advantage when it comes to hypoglycemia avoidance.

The sodium glucose cotransporter 2 inhibitors represent the newest class of AHA. By inhibiting renal glucose reabsorption, this class provides similar glucose lowering to that of DPP4i given the patient does not suffer from renal impairment. Risk of hypoglycemia is also considered negligible when used as monotherapy given that glucosuria only occurs during increased BG levels. By increasing glycosuria, patients on SGLT2i also experience weight loss but also an increase in genital mycotic and urinary tract infections. Safety considerations also include rare cases of ketoacidosis, lower extremity amputation, fractures and bladder cancer. Patients who are volume-compromised should avoid SGLT2i as use can lead to acute kidney injury. Three cardiovascular safety studies with differing baseline risk profiles have been published to date. They have demonstrated either a neutral effect on all-cause mortality (CANVAS and DECLARE) or a significant reduction (EMPAREG). Significant reductions in MACE endpoints was achieved in 2 of the 3 studies (CANVAS and EMPAREG), although DECLARE achieved a significant reduction when the subgroup of those with established CVD was evaluated. Overall, the most impressive outcome reductions with this class stems from their effect on heart failure hospitalization and renal endpoints.

Today’s era of AHA allows for CV safe reductions in hyperglycemia owing to their glucose- dependent mechanism of action and very low risk of severe hypoglycemia. With the results of recent CVOTs mandated by the FDA, the safety of new AHA has established confidence in patients and their physicians. In fact, cardiologists and nephrologists have shown increased interest in new AHA given their significant reductions in MACE and heart failure hospitalizations. Although guideline bodies have labelled hypoglycemia risk as negligible when used as monotherapy, diabetes is a progressive disease requiring the addition of multiple agents over time. When used together with insulin or SUs, these three class of AHA have clearly demonstrated an increased risk of hypoglycemia. When compared to insulin or sulfonylurea, new AHA have demonstrated a reduced risk of hypoglycemia. However, what remains unclear,

159

is how these agents compare to placebo on less severe events of hypoglycemia when background agents known to increase risk are eliminated. Given the interest of these AHA from multiple specialties, a more precise estimation of any hypoglycemia relative to placebo when used alone, with background metformin or in combination can help further personalize treatment strategies.

RATIONALE

Diabetes is one of the greatest health care challenges of our era. Attainment of glycemic control attenuates diabetes-related complications. Fear of hypoglycemia, however, limits target attainment. Reassurance of a low hypoglycemia risk can improve diabetes management and improve patient outcomes.

Three new classes of AHA for T2DM promise low to no hypoglycemia risk. Certainly, given the glucose dependent mechanism of action for each class, risk of severe hypoglycemia is minimal. However, a clear analysis which includes hypoglycemia of any severity has not been conducted. The low risk is widely accepted in relation to SU or insulin, but there are insufficient reports relative to placebo. In placebo-controlled studies, increases in hypoglycemia risk are often attributed to background SU and or insulin use and risk with AHA on their own or in combination has been difficult to evaluate. Each AHA class presents with its own unique mechanism of glucose-lowering and hypoglycemia counterregulation and may differ in risk of hypoglycemia relative to placebo. It is also not clear if risk of any hypoglycemia remains minimal when agents are combined with each other or simultaneously initiated.

6.1 DPP4i

6.1.1 Data from RCTs of DPP4i

A 4-year analysis of saxagliptin studies in which patients were treatment naïve or on metformin was conducted to evaluate long term safety. Reported hypoglycemia (ie signs and symptoms, without documented BG values) occurred in a similar frequency in the saxagliptin arm compared to placebo. However, there were more confirmed hypoglycemia events, defined as finger stick values ≤2.8 mmol/L (Rosenstock, Gross, et al. 2013).

160

Although no statistical analysis was conducted on this study either, numerically the rates of asymptomatic hypoglycemia was double in patients receiving initial combination therapy with metformin and alogliptin 12.5 (but not 25 mg) (Pratley, Fleck, and Wilson 2014).

6.2 GLP1RA

6.2.1 Data from RCTs of GLP1RA

Exenatide studied as monotherapy demonstrated no events of severe hypoglycemia, however, the incidence of mild to moderate hypoglycemia was 5-9%. Although these events were not severe, the study suggests that incidence of milder episodes may still exist with new AHA (Nelson et al. 2007). In addition, a five-year study of exenatide OW as add-on to other AHA demonstrated that although the risk of minor hypoglycemia was higher in those taking concomitant SU at baseline (25.3%) rates were not negligible in those not taking SU (8.2%) (Wysham et al. 2015).

Within class differences in hypoglycemia risk may also exist. In a head-to-head study of two GLP1RA, lixisenatide 20 µg OD demonstrated significantly less symptomatic hypoglycemia than exenatide 10µg BID (8 vs 48 events). No reports of severe hypoglycemia occurred in the study (Rosenstock, Raccah, et al. 2013).

Hypoglycemia risk may increase with GLP1RA dose. In a study comparing to metformin, incidence of total hypoglycemia was comparable (Dulaglutide 1.5 mg 12.3%, Dulaglutide 0.75 mg 11.1%, metformin 12.7%). However, rates for total hypoglycemia were 0.89, 0.47, and 0.29 events per patient year, respectively (Umpierrez et al. 2014).

Although severe episodes are unlikely to occur with new AHA’s, risk of milder episodes may differ. In a 52-week study comparing dulaglutide to sitagliptin, no events of severe hypoglycemia occurred. However, the incidence of total hypoglycemia at 52 weeks occurred in 10.2% dulaglutide 1.5mg, 5.3% dulaglutide 0.75mg and 4.8% sitagliptin. The one-year adjusted rates (events/patient/year) were 0.4 (1.6), 0.3 (2.6), and 1.1 (1.1), respectively (Nauck, Weinstock, et al. 2014).

161

6.2.2 Data from SRMA of GLP1RA

A recent systematic review and mixed treatment comparison of GLP1RA vs placebo, all agents except for albiglutide significantly increased the risk of hypoglycemia, ranging from an OR of 1.59; 1.10 – 2.31 for lixisenatide to OR 3.74; 1.51-2.24 for . Furthermore, the risk remained elevated after exclusion of studies with background insulin and or SU (Htike et al. 2017).

Table S3: Differences vs placebo for safety outcomes

Number of Number of Participants Participants Participants Studies Odds Ratio (95% CI) Participants Studies Odds Ratio (95% CI) Drug with events Drug with events

Hypoglycaemia Hypoglycaemia – Excluding background SU and Insulin TAS 30 398 3 3.74 (1.51,9.24) TAS 30 398 3 3.73 (1.61,8.66 ) DUL 122 1121 4 2.75 (1.51,5 .00 ) EBID 87 894 5 3.52 (1.94,6.41 ) EBID 409 2110 11 2.53 (1.74,3.67) DUL 72 882 3 2.59 (1.33,5.01) EQW 321 2242 7 1.88 (1.05,3.38) EQW - - - - LIR 149 1078 4 1.71 (1.14,2.57) LIR 31 478 2 1.62 (0.78,3.34 ) LIXI 307 2260 8 1.59 (1.10,2.31) LIXI 26 963 3 1.52 (0.64,3.59) ALB 42 404 1 1.33 (0.54,3.29) ALB - - - - PLA 293 3148 21 - PLA 36 1431 11 -

(Reproduced with permission)

6.2.3 Data from Network Meta-Analyses of GLP1RA

In a recent mixed treatment comparison, all GLP1RA increased risk of hypoglycemia compared with placebo except for albiglutide. Results were consistent even after excluding studies with background SU and Insulin (Htike et al. 2017).

6.3 SGLT2i

6.3.1 Data from RCTs of SGLT2i

Inagaki et al. conducted an RCT of canagliflozin monotherapy in Japanese patients. Again, although no statistical testing was conducted, symptomatic and asymptomatic hypoglycemia occurred in twice as many patients compared to placebo and without a dose dependent relationship (Inagaki, Kondo, et al. 2014).

Dose dependency on hypoglycemia risk may differ within the class of SGLT2i. showed double incidence of symptomatic hypoglycemia compared to placebo (placebo 1.3%, 5mg 1.3%, 15mg 2.6%) (Terra et al. 2017).

162

6.3.2 Data from SRMA of SGLT2i

In a SRMA conducted of RCTs with empagliflozin as add-on to metformin, the risk of hypoglycemia was neutral, but differed dependent on the comparator. In comparison to active agents, the risk of hypoglycemia with empagliflozin was numerically lower (52% and 57% non- significant reduction with 12.5mg and 25mg, respectively). However, risk was non-significantly elevated by 59% and 22% with empagliflozin 10mg and empagliflozin 25mg, respectively when compared to placebo (Zhong et al. 2016).

Pooled analyses of 7 placebo and active controlled studies of canagliflozin analyzed hypoglycemia risk based on AHA associated with hypoglycemia (ie insulin, SU, glinide) vs not. When studies of canagliflozin were pooled in which they were not on AHA associated with hypoglycemia, documented hypoglycemia occurred approximately twice as often (non- canagliflozin 2.9% vs canagliflozin 100mg 6.9% vs canagliflozin 300mg 4.1%). Similarly, severe hypoglycemia also occurred more frequently with canagliflozin 100mg (3 vs 1) compared to non-canagliflozin (Qiu et al. 2017).

Nevertheless, a pooled analysis of dapagliflozin revealed that increased risk of hypoglycemia was due to the studies of background SU or insulin. When all placebo-controlled trials were analyzed, the risk of hypoglycemia was double that of placebo, yet monotherapy or background metformin did not result in an increased risk (Ptaszynska et al. 2014).

6.4 Combination therapy

In an RCT of patients on linagliptin 5mg and metformin, randomized to empagliflozin or placebo, confirmed hypoglycemia of ≤3.9 mmol/L was numerically higher (3 vs 1 event) on the combination of DPP4i and SGLT2i than DPP4i and metformin alone (Softeland et al. 2017).

SCOPING REVIEW

A scoping review was first conducted to assess if the question of hypoglycemia risk with new AHA in comparison to placebo and without other background agents has previously been evaluated in a systematic review and meta-analysis. A search was conducted in English for

163

systematic reviews with metformin, DPP4i, GLP1RA and or SGLT2i in patients with type 2 diabetes. The Prisma flow diagram for the scoping review is available in Appendix I.

In short, PubMed search results retrieved 99 systematic reviews attempting to pool estimates with new AHA’s. Glycemic lowering efficacy was the primary objective in the majority of the studies. In fact, of the 99 studies, hypoglycemia was the primary objective or at least included in the title in only 8% (8/99), or 3 of the 59 systematic review and meta-analyses (Salvo et al. 2016; Kawalec, Mikrut, and Łopuch 2014; Yang et al. 2017; Phung et al. 2010; Liu et al. 2012; Andersen and Christensen 2016; Eriksson et al. 2016; Edridge et al. 2015). The 8% focus on adverse events is very much in line with the recent estimated 10% of systematic reviews focusing on adverse events described by the PRISMA harms group. This actually represents an improvement in reporting of harms in systematic reviews from the previous 5% from 1994 to 2010 (Zorzela et al. 2016).

Of the 99 studies retrieved, 59 studies were systematic reviews and meta-analyses. The 59 meta- analyses were evaluated to determine if background therapies of agents increasing risk were banned and if placebo was the comparator. Only 2 systematic review and meta analyses met this criteria (Liu et al. 2014; Kawalec, Mikrut, and Łopuch 2014). Liu et al. evaluated DPP4i monotherapy and background metformin, while Kawalec et al. assessed DPP4i or SGLT2i with metformin background therapy. The definition of hypoglycemia was not provided in Liu et al., and overall (ie any) hypoglycemia was pooled in Kawalec. Both studies found a neutral hypoglycemia signal. Similar analyses for GLP1RA were not found.

Given the findings of the scoping review, it was clear that an updated systematic review and meta-analyses focusing on the hypoglycemia risk with metformin and the 3 new classes of AHA’s was necessary in studies that did not permit use of insulin or SU as background therapy. With new AHA’s continuing to increase in popularity and as combination therapy gains wider acceptance, such data would be of clinical relevance and help improve diabetes management and patient outcomes.

164

METHODS

The protocol is registered with PROSPERO (CRD42018095458) and be found at http://www.crd.york.ac.uk/PROSPERO . The protocol for this systematic review and meta- analysis follows the 2015 Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) guidelines (Moher et al. 2015) and can be found in Appendix II. Given the safety-focus of this analysis, recommendations from the PRISMA Harms groups were also followed such as handling of zero event studies and search strategies (Zorzela et al. 2016).

8.1 OBJECTIVES

This systematic review and meta-analysis evaluates the risk of any and severe hypoglycemia specific to new AHA (DPP4i, GLP1RA and SGLT2) with or without metformin and alone or in combination with each other relative to placebo, by excluding studies allowing use of any other agents.

8.2 ELIGIBILITY CRITERIA

8.2.1 Study Design

Eligible studies for this systematic review and meta-analysis were double-blind, placebo- controlled randomized trials published in the English language, with a parallel design. Observational, cluster trials, cross-over, duplicate publications, extensions or sub-analyses of included studies were excluded.

8.2.1.1 Minimum Duration of Intervention

Studies with a minimum duration of therapy of 12 weeks were considered. Given the efficacy- focus of most randomized trials, a minimum duration of 12 weeks was chosen to reflect the minimum length of time for HbA1c changes to be detected.

8.2.2 Population

Adult participants with type 2 diabetes ≥18 years of age were included. Studies were excluded if they enrolled healthy volunteers, patients with type 1 diabetes, prediabetes, gestational diabetes, metabolic syndrome, or non-alcoholic steatohepatitis.

165

8.2.2.1 Diagnostic Criteria

Since the diagnostic criteria for type 2 diabetes has changed over the years, the diagnosis of diabetes was the established criteria valid at the start of the study.

8.2.3 Interventions

The interventions of interest included metformin, DPP4i’s GLP1RA’s or SGLT2i’s used as monotherapy, as add-on to metformin, as dual initiation or triple therapy irrespective of dose, dosing frequency, route of administration or local approval. Studies which allow open label or any other AHA such as alpha-glucosidase inhibitors, bile acid sequestrants, , insulin, meglitinides, SUs or thiazolidinediones as background therapy without indication of a wash-out period prior to randomization were excluded. Studies evaluating agents whose clinical trial programs have been terminated at time of search were not included. For instance, taspoglutide was excluded given termination of its clinical trial program in September 2010 (Rosenstock, Balas, et al. 2013)

8.2.4 Comparator

Only studies in which the comparator is a placebo pill were considered. Studies were excluded if the comparator was active therapy, standard of care or a lifestyle intervention without a placebo pill.

8.2.5 Outcomes

The primary outcome was any hypoglycemia, irrespective of severity, time of day or BG confirmation. The secondary outcome, severe hypoglycemia, was defined as requiring medical or third-party assistance, or with a confirmed BG value of <3.0 mmol/L or categorized as major. This threshold was selected based on recent recommendations aiming to standardize reporting of hypoglycemia in clinical trials (Group 2017).

8.2.5.1 Analysis Groups

An overall pooled estimate of all the included studies were not conducted since three classes of AHA as well as metformin were being evaluated, each with their own unique mechanism of action. Analyses were conducted by class of AHA and according to background therapy if two

166

or more studies were available per outcome. For instance, studies in which AHA including metformin were used as monotherapy were synthesized separately in comparison to AHA’s used with background metformin. Two agents simultaneously added to treatment naïve patients or to patients whose previous AHA were washed out were grouped separately. Triple therapy in which a third agent was added to dual therapy were also separately assessed. Finally, since metformin is not globally recommended as first-line, studies in which a second agents, including metformin was added to a non-metformin background were grouped together.

8.2.5.2 Sensitivity Analyses

A priori sensitivity analyses were conducted to investigate the influence of including studies with zero events in both arms using Risk Difference with 95% confidence intervals. This allowed a more comprehensive evaluation of all the available evidence. The robustness of study results was also evaluated by repeating the analysis with use of different effect sizes and statistical models.

8.2.6 Exclusion Criteria

Studies were excluded if they did not evaluate or provide information on hypoglycemia risk or are not randomized, parallel, placebo-controlled trials. Observational, cluster-randomized, or cross-over studies, as well as duplicate publications, extensions or sub-analyses of included studies were also excluded.

8.3 SEARCH STRATEGY

The databases of OVID Medline/Medline (Epub Ahead of Print, In-Process & Other Non- Indexed Citations 1946-Present) OVID EMBASE (Embase Classic + Embase 1947 onwards) and Cochrane Central Register of Controlled Trials (CENTRAL) were comprehensively searched. There were no restrictions on period of publication as metformin, one of the therapies of interest, has been around for many decades. Manual searches were performed from references of included studies and relevant systematic reviews. The search strategy for Medline can be found in Additional file 2. Grey literature, abstracts, conference meeting reports etc. were not included. The full search strategy for Medline OVID can be found in Appendix III.

167

The search strategy consisted of a validated string for RCTs (except for CENTRAL), T2DM and the AHA. AHA searches for metformin, DPP4i, GLP1RA and SGLT2i, included terms for the class of AHA, brand and scientific name. Since hypoglycemia is not expected to be the primary outcome of the studies, and may not be reported within the abstract, the term hypoglycemia was not searched. The PRISMA Harms group also suggests refraining from including harms within the search string (Zorzela et al. 2016).

For identifying RCTs, The Cochrane highly sensitive search strategy 6.4.d was used for MEDLINE (Higgins 2011). For EMBASE, a validated, specificity maximizing search string for RCTs was followed (Wong, Wilczynski, and Haynes 2006). For CENTRAL, the standard practice of selecting ‘Trials’ on the left panel was used.

8.4 STUDY SELECTION

Search results were uploaded into Endnote. After de-duplication, two reviewers (SK and LL) independently conducted title and abstract screening. If the citation met the eligibility criteria or if it was unclear, full text of the article was then investigated to further assess eligibility. Disagreements at any level of screening were first discussed amongst the two reviewers and if still unresolved, a third reviewer (ST) was consulted. A PRISMA flow diagram was used to show the process of study selection and reasons for exclusion.

8.4.1 Dealing with Duplicate and Companion Publications

For studies with multiple or companion publications, only the primary reference or the reference with the outcome of interest will be eligible for inclusion.

8.5 DATA EXTRACTION

Two reviewers (SK and PD) independently extracted key characteristics of included studies in a pre-piloted table. A customized extraction table included author, year, number of subjects, intervention, dose, duration of study, mean age, gender, ethnicity, duration of diabetes, use of rescue therapy, Hba1c at baseline, change in Hba1c vs baseline, change in Hba1c versus placebo, definition of hypoglycemia, hypoglycemia ascertainment and handling of patients with a history of or experiencing hypoglycemia during the study.

168

8.6 RISK OF BIAS

Studies were evaluated according to Cochrane’s Risk of bias table. Each domain per study was evaluated as low, high and unclear. Studies with low risk of bias will be considered unlikely to alter the results. Studies with high risk of bias will be considered as likely to weaken the confidence in the results. Studies with an unclear risk of bias will raise doubt in the confidence of the results. These 6 domains are defined below.

Table 1: Risk of Bias Classification

BIAS YES (low risk of bias) NO (high risk of bias) UNCLEAR Selection Bias Random sequence Computer generated, coin/dice toss, strata or By judgement of physician, odd Method of randomization not generation stratified by restricted or simple randomization, birth data, alternation, quasi- described with sufficient detail. Techniques to generate unrestricted randomization, blocked random, rotation. Example: “we randomly allocated” sequence randomization, (ie. future assignment cannot be or “using a randomized design” anticipated). Selection Bias Allocation concealment Central randomization by a third party, Open allocation sequence , Method of concealment not Techniques to assure sequentially numbered, sealed and opaque unsealed or non-opaque described or not described in sequence concealment envelopes/containers. envelopes, DOB, case record sufficient detail (ex. unclear if number. envelopes were sequentially numbered, opaque and sealed) LIKELY MAJORITY of STUDIES! Performance Blinding of participants Participants & study personnel blinded. Not blinded (open label) Insufficient information Bias and personnel LIKELY MAJORITY of STUDIES! Detection Blinding of outcome Outcome assessor b linded . Not blinded . In blinded placebo - Bias assessment For patient-reported outcomes, low risk of bias controlled trials, a suspected a/e (Differences in how if low risk of patient blinding. For outcome may more often be attributed to outcomes are criteria assessed from medical reports, low risk active treatment. LIKELY MAJORITY determined) of bias if a/e not suspected with treatment a/e of STUDIES! or a/e equally likely in either arm. Attrition Bias Incomplete outcome No missing data. Outcome data on all Exclusion /withdrawn Insufficient details on attrition or data, referring to randomized participants. /discontinuation due to adverse exclusion to permit judgement. attrition (drop out during Missing data and reasons for are balanced event or adverse effect greater in Reasons for missing data not study) or exclusions from amongst groups. A/E leading to discontinuation AHA than placebo arm, even if provided. analysis. Amount, similar or more favorable for AHA than placebo differential missingness is not ss. nature or handling of arm. In addition, similar or greater proportion Secondly, proportion completing outcome data. of AHA arm completed study. study greater in placebo arm. Reporting Selective reporting Protocol available . N o protocol but clear that Outco me(s) reported incompletely; LIKELY MAJORITY of STUDIES! Bias all prespecified outcomes are reported cannot be entered into meta-an. (UNLIKELY to find). Study fails to include results for a key outcome that would be expected for such study. Other Bias Risk of Confounding Due Rescue therapy was allowed. Hypoglycemia Rescue therapy was allowed. Use of rescue therapy not to Rescue Therapy data was not included after rescue initiation. Hypoglycemia data was included reported. after rescue initiation, thus at risk of bias.

169

8.6.1 Selection Bias: Random Sequence Generation

The first domain of selection bias was to evaluate the random sequence generation. Studies with clear description of how the sequence was generated will receive a low risk of bias. Studies with a low risk of bias, assignment to placebo or AHA cannot be anticipated by any means. As outlined in the above table, examples of studies with a high risk of bias of random sequence generation include physician judgement, birth date and rotation.

8.6.2 Selection Bias: Allocation Concealment

A low risk of bias is afforded to studies that assure concealment of the allocation schedule for example through the use of sealed and or opaque envelopes. Admittance of an allocation sequence that was not concealed, for example if posted in the lunch room would receive a high risk of bias.

8.6.3 Performance Bias: Blinding of Participants and Personnel

Inclusion of open-label studies should not be found, as this would violate our study protocol. Given our criteria for blinded studies, all included studies will be double-checked to ensure blinding occurred and was maintained for the duration of our outcome assessment.

8.6.4 Detection Bias: Blinding of Outcome Assessment

Although blinding will be ensured in the included studies, our primary objective of evaluating hypoglycemia, an adverse event, in comparison to placebo, will result in a high risk of detection bias in all included studies.

8.6.5 Attrition Bias: Incomplete Outcome Data

Missing outcome data on randomized patients will be evaluated. Studies without any drop outs, although likely rare, will be considered as low risk of bias, if hypoglycemia is evaluated in all randomized patients. Patient disposition diagrams will be reviewed to compare number of drop outs due to adverse events. In the event of an imbalance, for example, a 3-fold greater number of drop outs due to adverse event, with consideration of the denominator size swith AHA compared to placebo, will be considered at high risk of attrition bias.

170

8.6.6 Reporting Bias: Selective Reporting of Outcomes

Protocols of included studies will be reviewed, where available, to ensure hypoglycemia outcomes as planned are reported. Should study protocols not be available, a low risk of bias will still be assigned if both any and severe hypoglycemia are reported. Alternatively, if only one form of hypoglycemia is presented, without any mention of the severity of the episodes resulting in an inability to enter data into the primary and secondary outcome of our meta- analysis, an unclear risk of bias will be designated. High risk of reporting bias will be assigned if outcomes are indicated in the protocol or methods but not reported.

8.6.7 Other Bias: Confounding by Rescue Medication

Rescue medication provided to patients during the trial may increase the risk of hypoglycemia. A high risk of bias will be assigned to studies that include hypoglycemia outcomes in patients receiving rescue medication. Studies that exclude rescued patients from safety or hypoglycemia data will be assigned a low risk of bias.

8.7 DATA SYNTHESIS

The principal summary measure used is Risk Ratio with 95% confidence intervals for the primary endpoint of number of patients reporting an event. REVMAN 5.3 was used to meta- analyze the data if the outcome is available in ≥2 studies per subgroup. Since heterogeneity is anticipated between studies, Random-Effects Model (REM) was used. Mantel-Haenszel was used to compute Risk Ratio and Risk Difference for the primary analysis of studies reporting incidence of hypoglycemia and for the sensitivity analysis of including studies with zero hypoglycemia in either arm, respectively.

8.8 STATISTICAL ANALYSIS

8.8.1 STATISTICAL METHOD

Rates of hypoglycemia with new AHA’s, especially of the severe form is anticipated to be low. As seen with the 2007 Nissen meta-analysis of MI risk with rosiglitazone, meta-analyses of rare adverse events can be widely criticized (Nissen and Wolski 2007, 2010). Which statistical

171

method to use and how to handles studies with zero events in both arms is debated and the nuances are described below.

8.8.1.1 Peto Odds Ratio

The Cochrane book recommends use of Peto OR for rare events (Higgins 2011). In a simulation study of OR, Peto methods were found to provide the least biased results when a true treatment effect existed (Cheng et al. 2016). In contrast, Prochaska et al. recommends Peto OR be avoided for rare events (Prochaska and Hilton 2012).

In another study, 12 methods for pooling rare events were compared. They found that the Inverse Variance and DerSimonian and Laird Odds Ratio and Risk Difference (RD) and the MH Odds Ratio with a 0.5 zero-cell correction gave the most biased results. The authors suggested that the Peto Odds Ratio for events where the rates are below one percent gave the least biased, most powerful and best confidence coverage. However, the key to Peto is to ensure balance between the number of participants in the treatment and control arms. Otherwise, MH OR without a zero- correction factor performed with less bias (Bradburn et al. 2007).

In this analysis, a balance between the number of participants between the arms is not anticipated. Dose-ranging studies will be eligible for inclusion and the various doses will be combined as a single pairwise comparison. Given the potential imbalance of participants between the arms, Peto Odds Ratio will not be considered.

8.8.1.2 Inverse Variance

Inverse Variance and DerSimonian and Laird methods should also be avoided for rare events, especially when using random effects (Bradburn et al. 2007). When event rates are low, the standard error used in this effect estimate is considered poor.

8.8.1.3 Mantel Haenszel (MH)

Cochrane recommends use of Mantel Haenszel when event rates of outcomes are anticipated to be low. It is believed MH has better statistical properties than the Inverse Variance method for low event rates because its weighting scheme is dependent on the type of effect measure (RR, OR, RD) used. For rare or sparse events, where one or both arms contain zero events, use of a

172

correction factor with MH is believed to give the least biased results (Sweeting, Sutton, and Lambert 2004).

8.8.2 EFFECT MEASURE 8.8.2.1 Odds Ratio (OR)

The Odds Ratio (OR) are the odds that an outcome will occur given a particular exposure, compared to the odds of the outcome occurring in the absence of that exposure: OR=a/b / c/d. Similar to Risk Ratio (RR), OR is more likely to maintain consistency with differing baseline risk and used more frequently in studies compared to Risk Difference (RD). However, knowledge translation of OR is more difficult and frequently misinterpreted as RR. Use of OR in meta-analyses may be due to convenience more so than appropriateness (Morton et al. 2008).

When using OR, decisions to include studies in which both arms have zero events (BA0E) or not depend on the magnitude of the true treatment effect. If a small treatment effect is anticipated, including BA0E introduces little bias, decreased means square error and narrows the 95% CI. Excluding BA0E when a true effect exists, however, leads to a smaller bias in OR calculations (Cheng et al. 2016).

8.8.2.2 Risk Ratio (aka Relative Risk) (RR)

Relative Risk or Risk Ratio (RR) is the probability of an event occurring in an exposed group to the probability of the event occurring in a comparison or non-exposed group: RR = a/(a+b) / c/(c+d). Relative risk, as opposed to odds ratio, is recommended for simple binary variables. Deeks et al. recommended use of RR over OR based on a sample of 114 meta analyses. They found higher events observed in trials with higher baseline risk, corresponding more with the RR model (Deeks 2002). Odds Ratio can still be considered if an increase in outcome is expected. RR is commonly reported in studies and meta-analyses, more likely to maintain consistency with differing baseline risks and easier to interpret (Morton et al. 2008).

Since our baseline risks will differ and the outcome of interest will be dichotomous, ie patient with hypoglycemia, risk ratio will be used.

173

8.8.2.3 Risk Difference (RD)

Typically, meta-analyses using OR or RR include studies with zero events in either the control or treatment arm, but not both. Meta analyses using RD include studies with zero events in either or both arms. Risk difference may appear to be more advantageous for studies with zero events, however, Bradburn et al. demonstrated with simulation studies that RD yield wide confidence intervals especially when dealing with rare events and thus have poor statistical power (Higgins 2011). Further, although risk difference can address zero event studies and be easily converted to NNT and NNH, it should not be used if baseline risks differ between treatment arms or incidence of event rates differ over time (Morton et al. 2008). The decision to include or exclude zero events impacts risk difference more than OR or RR. Inclusion of zero events can negate an otherwise statistically significant finding if zero events are excluded (Friedrich, Adhikari, and Beyene 2007).

Despite the limitations of using RD, studies with zero events in both arms is expected and therefore RD will be used in the sensitivity analyses when including these trials.

8.8.3 ANALYSIS METHOD (FEM or REM)

To determine the appropriate analysis method, a common practice is to assess heterogeneity using Cochrane’s Q statistic. However, the model to use should be based on the nature of the studies and not on the test for heterogeneity (Shuster and Walker 2016). Since these studies are not conducted under universal conditions, random effect method (REM) is recommended. REM is considered more appropriate because studies included in meta-analyses are rarely identical when comparing the population or interventions (Rao et al. 2017).

However, using REM in meta-analyses of many small studies may find statistical significance, whereas using fixed effects method (FEM) or even a single large study may show neutrality. This is because in REM, small studies are weighted more equally given the large between-trial heterogeneity than in FEM. To overcome this, we will examine all outcomes primarily using REM, but will report if using FEM significantly alters results (Higgins 2011).

174

8.8.4 BOTH ARM ZERO EVENTS (BA0E)

Consensus is lacking on whether to include BA0E. Those against its inclusion claim it lacks information. However, these studies are important to demonstrate low and equal event rates. For rare events, enormous sample sizes are needed to detect treatment effects. By elimination studies with zero events, the bias is further exaggerated (Bhaumik et al. 2012).

Studies with zero events in both arms should be included to summarize all available data. However, in doing so, the treatment effect estimate is moved closer to nil, decreasing the confidence interval and between study heterogeneity and negating a possibly significant finding. The difference may not seem obvious when using RR or OR, however, RD demonstrate more substantial findings between inclusion and exclusion of BA0E. Excluding BA0E studies can increase pooled effects. Regardless of effect measure chosen (RR, OR or RD), zero total event trials should be included (Friedrich, Adhikari, and Beyene 2007).

In this meta-analysis, sensitivity analyses were conducted to investigate the influence of including studies with zero events in both arms using Risk Difference with 95% confidence intervals. In doing so, a more comprehensive evaluation of all the available evidence was achieved. The robustness of study results was affirmed by repeating the analysis with use of different effect sizes and statistical models.

8.8.4.1 Continuity Corrections

Most meta-analytic software, including RevMan, yield a computational error when studies with zero-cell counts are included. Hence, zero counts are automatically filled with a fixed value known as a continuity correction (typically 0.5), if one arm is zero, but not both (Higgins 2011). However, if the effect is zero, the would induce a bias.

Sweeting et al. however, discovered that adding a correction proportional to the reciprocal of the size of the comparator arm to perform better than the standard 0.5 (Sweeting, Sutton, and Lambert 2004). Newer statistical pooling methods have been studied and determined to provide better estimates than MH (Bhaumik et al. 2012). Ignoring BA0E studies or using continuity corrections have also been criticized. Beta binomial models (BBIN) were found to outperform continuity corrections (Kuss 2015).

175

8.8.4.2 How have other SRMA’s of RCTS of rare events addressed BA0E?

Since this is not the first SRMA of RCTs evaluating hypoglycemia risk, here’s how other analysis have handled BA0E:

Salvo et al. used Risk Ratio with MH using FEM, and if heterogeneity was found between estimates, REM. (Salvo et al. 2016)

In Edridge et al., if a study reported zero hypoglycemia events, 0.001 was entered in the meta- analysis to avoid the study from being excluded. (Edridge et al. 2015)

Andersen NMA Hypoglycemia when added on to SU: Hypoglycemia analyzed as a dichotomous endpoint and reported as OR. Bayesian model. (Andersen and Christensen 2016)

Schopman 2014 MA w SUs: When none of the patients experienced hypo, 0.5 was added to the number of hypoglycemia and non-hypoglycemia events so that the logit transformation could be performed using OR in REVMAN. (Schopman et al. 2014)

8.8.4.3 Lessons from Nissen’s 2007 SRMA of MI risk with Rosiglitazone

In 2007, Nissen and Wolski conducted a SRMA to evaluate the risk of MI and CV death with rosiglitazone. Peto Odds Ratio was used. Since this approach did not allow for 0 cell counts, studies in which patients did not experience a CV event in either arm were excluded. In total 42 studies were pooled. Their analysis found that rosiglitazone significantly increased the risk of an MI (OR 1.43, CI 1.03 to 1.98, p=0.03). The analysis received much criticism, even to this day. In 2010, Nissen re-conducted the study, this time, allowing for studies with zero events to be included. The number of included studies increased to 56. MH OR using FEM was used, and the OR maintained significance for elevated MI with rosiglitazone (Nissen and Wolski 2010). Although the updated meta-analysis did not change the results, it highlighted the importance of considering all the available evidence, not just ones with events.

176

8.8.5 Unit of Analysis

Studies reporting number or percent of patients with hypoglycemia were included. Only data from the initial study phase where double blinding is maintained was eligible. A single pair-wise comparison was used for dose-ranging studies.

8.8.5.1 Number of Participants vs Number of Events

The primary and secondary hypoglycemic outcome will be number of patients experiencing a hypoglycemic event. Authors of studies which only provide count data, or number of episodes, and do not also provide number of patients with experiencing events, were to be contacted. To avoid the common pitfall of incorporating count data as dichotomous data, unit of analysis was assessed for each included study.

8.8.5.2 Single Dose-Ranging Study

Studies with multiple arms examining various doses of a single AHA were combined into a single comparison arm.

8.8.5.3 Single Study Contributing to Multiple Comparisons

When faced with a study that contains multiple interventions that can contribute to multiple comparison, the Cochrane Handbook (section 16.5.4) warns against double-counting, as this creates a unit of analysis error. They offer several recommendations, including combining groups to create a single pair-wise comparison. In the case of dose-ranging studies, where multiple arms of the same intervention were being evaluated, this was done. However, when interventions differ, combing is not applicable. The second suggestion is selecting a pair of interventions and excluding others, however, this approach would omit valuable data. For this analysis, the third suggestion was adopted, which was to split the shared intervention, which in studies, was the placebo arm. Thus, depending on whether the study contributed to two or three comparison, the placebo numerator and denominator was divided into 2 or 3 (Higgins 2011).

177

8.8.6 Assessment of Heterogeneity

Chi² test and I² were used to conduct tests of statistical heterogeneity. Irrespective of heterogeneity, random effects model was used to pool estimates. Statistical significance was achieved if P <0.05.

8.8.7 Assessment of Reporting Bias

If ≥10 studies are available per outcome, a funnel plot will be used to determine publication bias. Several factors may account for funnel plot asymmetry including poor methodology and true heterogeneity.

8.8.8 Quality of Evidence

The overall quality for each outcome will be presented according to the Grading of Recommendations Assessment Development and Evaluation (GRADE) approach (Atkins et al. 2004). The summary of the evidence will be presented in the “Summary of findings” table as described in the Cochrane Handbook for assessing risk of bias in randomized trial (Higgins et al. 2011).

RESULTS 9.1 SEARCH RESULTS

9.1.1 Prisma Flow

Of 16,141 records identified from our searches, 141 studies were included in 10 intervention groups vs placebo, of which 124 were unique studies (Agarwal, Jindal, and Sapakal 2018; Ahren et al. 2013; Amin et al. 2015; Aschner et al. 2006; Bailey et al. 2010; Bailey et al. 2012; Barnett et al. 2012; Barzilai et al. 2011; Bolinder et al. 2012; Bolli et al. 2014; Bosi et al. 2007; Charbonnel et al. 2006; Chen, Ning, et al. 2015; Chiasson and Naditch 2001; Dagogo-Jack et al. 2018; Davies et al. 2017; DeFronzo and Goodman 1995; DeFronzo et al. 2005; DeFronzo et al. 2008; DeFronzo et al. 2009; Dejager et al. 2007; Ferrannini et al. 2010; Ferrannini et al. 2013; Fonseca et al. 2012; Fonseca et al. 2013; Forst et al. 2010; Goldstein et al. 2007; Gantz et al. 2017; Goodman, Thurston, and Penman 2009; Grunberger et al. 2012; Haak et al. 2012; Hanefeld et al. 2007; Haring et al. 2014; Home, Shankar, et al. 2018; Hong et al. 2016; Ikeda et

178

al. 2015; Inagaki et al. 2013; Inagaki, Kondo, et al. 2014; Inagaki, Onouchi, et al. 2014; Inagaki et al. 2015; Iwamoto et al. 2010; Jabbour et al. 2014; Ji et al. 2014; Ji et al. 2016; Ji et al. 2017; Jung et al. 2015; Kadowaki and Kondo 2013; Kadowaki et al. 2013; Kadowaki et al. 2014; Kadowaki et al. 2017; Kadowaki et al. 2018; Kaku et al. 2013; Kaku, Kiyosue, et al. 2014; Kaku, Watada, et al. 2014; Kashiwagi et al. 2014; Kashiwagi et al. 2015; Kawamori et al. 2012; Kikuchi et al. 2009; Kim et al. 2007; Kumar et al. 2014; Lu et al. 2016; Ludvik et al. 2018; Madsbad et al. 2004; Mathieu et al. 2015; Matthaei et al. 2015; Miller et al. 2018; Mohan et al. 2009; Moretto et al. 2008; Nauck et al. 2009; Nauck, Petrie, et al. 2016; Nauck, Stewart, et al. 2016; Nonaka et al. 2008; Odawara, Hamada, and Suzuki 2014; Odawara et al. 2015; Pan, Yang, et al. 2012; Pan, Xing, et al. 2012; Pan et al. 2017; Park et al. 2017; Pi-Sunyer et al. 2007; Pratley et al. 2006; Pratley, Fleck, and Wilson 2014; Qiu, Capuano, and Meininger 2014; Ratner, Rosenstock, and Boka 2010; Raz et al. 2006; Raz et al. 2008; Rhee et al. 2010; Ristic et al. 2005; Rodbard et al. 2016; Roden et al. 2013; Rosenstock, Sankoh, and List 2008; Rosenstock, Aguilar-Salinas, et al. 2009; Rosenstock, Reusch, et al. 2009; Rosenstock et al. 2012; Rosenstock, Seman, et al. 2013; Rosenstock et al. 2011; Rosenstock, Cefalu, et al. 2015; Ross et al. 2012; Ross et al. 2015; Scherbaum et al. 2008; Schumm-Draeger et al. 2015; Scott et al. 2007; Scott et al. 2008; Seino et al. 2008; Seino, Inagaki, et al. 2014; Seino, Sasaki, Fukatsu, Sakai, et al. 2014; Seino, Sasaki, Fukatsu, Ubukata, et al. 2014; Shankar, Inzucchi, et al. 2017; Sheu et al. 2015; Softeland et al. 2017; Sorli et al. 2017; Stenlof et al. 2013; Taskinen et al. 2011; Terra et al. 2011; Terra et al. 2017; Vilsboll et al. 2007; Wang, Yang, et al. 2016; White et al. 2014; Wilding et al. 2013; Wu et al. 2015; Yang et al. 2011; Yang et al. 2012; Yang et al. 2013; Yang et al. 2015; Yang et al. 2016).

The Prisma Flow Diagram can be found in Appendix IV.

9.1.1.1 Kappa Estimate

The Kappa estimate was 0.80, which according to Cochrane’s handbook section 7.2.6 reflects excellent agreement (Higgins 2011).

Table 2: Kappa Estimate

Review Author 2

179

Include Exclude Unsure Total

Review Include a=84 b=30 c=0 I1=114

Author Exclude d=10 e=8,850 f=0 E1=8,860

1 Unsure g=0 h=0 i=0 U1=0

Total I2=94 E2=8,880 U2=0 K=8,974

Where; Kappa= Po-PE/1-PE

And Po=a+e+i/K and Pe=(I1xI2)+(E1xE2)+(U1xU2)/K²

From our data;

Po=84+8,850+0/8974 = 0.9955

PE=(114x94)+(8,860x8,880)+0/80,532,676

PE=10,716+78,676,800/80,532,676

PE=78,687,516/80,532,676

PE=0.97709

And,

Kappa=Po-PE/1-PE

Kappa=0.9955-0.97709/1-0.97709

Kappa=0.01841/0.02291=0.8036

9.1.1.2 Studies with Multiple Comparisons

Twelve studies contained double-blind hypoglycemia data that could contribute to multiple comparison groups; 7 to two comparisons, and 5 to three comparison groups. Of these, only one

180

study had separate placebo arms for each intervention (Pan et al. 2017). The remaining 11 studies shared a placebo group, whose numerator and denominator were divided according to the number of comparisons applicable (ie 2 or 3). Below table provides a summary of these studies.

Table 3: Studies with Multiple Comparison

Study Comparison 1 Comparison 2 Comparison 3 Amin 2015 DPP4i +Met SGLT2i + Met Fonseca 2013 Met Mono SGLT2i Mono Goldstein 2007 Met Mono DPP4i Mono Dual Initiation Haak 2012 Met Mono DPP4i Mono Dual Initiation Ji 2016 Met Mono DPP4i Mono Dual Initiation Ji 2017 Met Mono DPP4i Mono Dual Initiation Pan 2017 (placebo arm per intervention) DPP4i Mono DPP4i +Met Pratley 2014 Met Mono DPP4i Mono Dual Initiation Roden 2013 SGLT2i Mono DPP4i Mono Rosenstock 2012 DPP4i +Met SGLT2i + Met Inagaki 2015 DPP4i Mono DPP4i Mono Gantz 2017 DPP4i Mono DPP4i Mono

In 4 of the remaining 11 studies, zero patients experienced hypoglycemia in the shared placebo arm (Amin et al. 2015, Fonseca et al. 2013, Gantz et al. 2017 and Inagaki et al 2015).

For the remaining 7 studies in which the numerator of the shared placebo could not be split evenly, the studies were listed in chronological then alphabetical order to balance distribution of the unpaired patient with hypoglycemia. Distribution of the leftover placebo patient was added in a pattern, beginning within the comparison of Placebo vs Dual Initiation, then Placebo vs DPP4i, Placebo vs SGLT2i and Placebo vs Metformin.

Table 4: Distribution of Single Hypoglycemia Count in Shared Placebo Arm

Leftover placebo patient split pbo in 3: Comparison 1 Comparison 2 Comparison 3 added to: Goldstein 2007 1 2 8 pbo vs dual Haak 2012 1 2 8 pbo vs dpp4 Pratley 2014 1 2 8 pbo vs met Ji 2016 1 2 8 pbo vs dual Ji 2017 1 2 8 pbo vs dpp4 split pbo in 2: Rosenstock 2012 3 7 pbo vs sglt2 Roden 2013 2 6 pbo vs dpp4i 181

Multiple Comparisons studies are also captured in the Characteristics of Included Studies found in

182

Appendix VI.

9.1.2 Excluded Studies (that appeared at first to be eligible)

Appendix V contains a list of studies which at first appeared eligible, but upon further inspection, failed to meet the inclusion criteria.

9.2 CHARACTERISTICS OF INCLUDED STUDIES Characteristics of included studies are summarized in the below text and tables. Individual study details and can be found in

1

Appendix VI .

9.2.1 Study Duration

Most studies were 12 or 24 weeks in duration with the exception of a few 52 weeks studies.

9.2.2 Unit of Analysis All included studies reported percent of patients reporting hypoglycemia. Few studies reported number of episodes and those that did, also provided percent of the population with events. Unit of analysis for each study can be found in

1

Appendix VI.

9.2.3 Incomplete Outcome Data The primary hypoglycemia outcome in all included studies were conducted on an intent-to-treat basis and none of the data included were based on a per protocol analysis. The details for each included study can be found under “Safety Analysis” found in

1

Appendix VI.

9.2.4 Funding

All studies except for one (Wu et al. 2015) were funded by the manufacturing pharmaceutical company.

9.3 CHARACTERISTIC OF THE STUDY POPULATION

9.3.1 Number of Participants

Total number of participants included in the analysis is n=45,843. Of the 10 comparison groups, DPP4i monotherapy contributed most number of studies (44 or 37 unique) with the most number of participants (n=13,072). Second is the DPP4i + metformin group with 23 studies (22 unique) and with n=8,218. The smallest comparison group was the non-metformin background dual therapy with 3 studies and n=578 participants.

9.3.2 Mean Age

The mean age of most studies ranged from 50-60 years of age. DPP4i monotherapy group had the widest age range of 48.9 to 72.1.

9.3.3 Gender

The overall range of percent male in the included studies was 25.8 to 86.0. GLP1RA with Metformin (25.8 to 77.0) and SGLT2i with Metformin (28.3 to 74.5) included most percent females and the Non-Metformin background Dual Initiation group included most % of males (71.3 to 83.1).

9.3.4 Race/Ethnicity

Unlike Canada and USA, metformin is not recommended first line in Japan for patients with T2DM. This may explain why all 3 studies within the non-metformin background dual therapy group were of Asian origin. Most of the participants in the SGLT2i monotherapy group were of Asian origin (11 of 17 studies). Close to half of the studies in the DPP4i monotherapy group were conducted in Asian patients (20 studies) and 3 were conducted in Indian patients.

1

Similarly, of the 23 studies included in DPP4i with met, 7 studies enrolled Asian participants. For the remaining studies, most were of Caucasian origin.

9.3.5 Duration of Diabetes

Triple therapy arm included patients with the longest diabetes duration (5.64 to 10.1 years). For the 3 classes of AHA’s, the monotherapy group had less years of diabetes duration than the respective group with metformin background.

9.3.6 Baseline HbA1c

Dual initiation group had the highest baseline HbA1c range (8.21 to 9.0) and DPP4i monotherapy group had the lowest range (6.7 to 9.0).

9.3.7 Change in HbA1c

Likely related to the longer duration of diabetes and declining beta-cell function, the smallest change in HbA1c compared to placebo was seen in the triple therapy group (-0.35 to -0.89). The largest change in HbA1c was seen in the dual initiation group -1.08 to -2.07.

Table 5: Summary of Characteristics of Included Studies

Comparator Studies Participants % Ethnicity Mean Diabetes Study Baseline Change Male Age Duration Duration A1c in A1c in weeks Vs pbo (Range) (Range) (Range) (Range) (Range)

1. MET MONO 8 2,121 38 to 2 studies 52.2 to 1.0 to 7.5 12 to 36 7.84 to -0.7 to - 73.5 Asian 60.2 9.7 1.3

2 DPP4i MONO 45 13,072 36.4 to 20 studies 48.9 to 0.5 to 8.6 12 to 26 6.7 to 9.0 -0.14 to 86.0 in Asians, 72.1 -1.2 3 in Indians

3 DPP4i w MET 23 8,218 41.5 to 7 studies 51.6 to 1.9 to 9.4 12 to 52 7.69 to -0.31 to 73.7 in Asians 61.8 9.3 -1.1

4 GLP1RA 9 2,210 31.4 to 2 studies 52.0 to 1.0 to 12 to 52 7.1 to -0.38 to MONO 84.61 in Asians 60.0 8.87 8.54 -1.85

2

5 GLP1RA w 8 3,342 25.8 to 50.4 to 1.8 to 7.7 12 to 26 7.46 to -0.1 to - MET 77 58.9 8.6 2.1

6 SGLT2i 17 6,150 41.3 to 11 studies 49.9 to 0.25 to 12 to 52 7.46 to -0.35 to MONO 81.8 in Asians 60.4 7.8 weeks 8.45 -1.31

7 SGLT2i w 15 5,845 28.3 to 3 studies 52.7 to 4.2 to 12 to 24 7.16 to -0.171 to MET 74.5 in Asians 60.8 8.05 weeks 8.4 -1.30

8 DUAL 6 1,788 43.2 to 2 studies 52.4 to 1.1 to 6.8 24 to 26 8.21 to -1.08 to INITIATION 69.7 in Asians 56.4 weeks 9.0 -2.07

9 TRIPLE 7 2,519 43.7 to 54.3 to 5.64 to 24 to 26 7.86 to -0.35 to THERAPY 64.5 59.7 10.1 weeks 8.5 -0.89

10 NON-MET 3 578 71.3 to All 3 54.1 to 6.5 to 14 to 24 7.87 to -0.88 to Background 83.1 Asian 58.4 8.34 weeks 8.18 -0.94 Dual Therapy

TOTAL 141 45,843

9.4 HETEROGENEITY OF HYPOGLYCEMIA

9.4.1 Definition of Hypoglycemia The heterogeneity of hypoglycemia definitions was an anticipated barrier of this meta-analysis. Thus, our primary outcome of any hypoglycemia was irrespective of the definition, BG confirmation, threshold, time of day (daytime or nocturnal), or the presence or absence of symptoms. The majority of included studies did provide a definition (90/141 or 64%) of hypoglycemia. Table 6 below provides a condensed summary of the definitions. For the definition of hypoglycemia per included study, please refer to

3

Appendix VI.

9.4.2 Definition of Severe Hypoglycemia

Just under half of the included studies, however, reported a definition for severe hypoglycemia (65/141 or 46%). Requiring assistance (whether medical or third party) without a need for BG documentation was observed in the majority of the 65 studies providing a definition for severe hypoglycemia (42/65 or 65%). Requiring assistance with BG confirmations were noted in the remaining studies with BG thresholds ranging from <3.89 to <2.8 mmol/L.

Our protocol defined severe hypoglycemia with a threshold of <3.0 or requiring assistance (either third party or medical, including hospitalization). As outlined in the table below, some studies defined any hypoglycemia with a BG documentation of ≤2.8 mmol/L and were captured in the meta-analysis as severe. Alternatively, 11 studies set a threshold for severe hypoglycemia at ≤3.1 mmol/L and were thus captured in our analysis as any hypoglycemia. In 4 of the 11 studies, hypoglycemia did not occur.

Table 6: Heterogeneity of Hypoglycemia – Definitions

1

Number of Number of Criteria for Severe No of studies Defining Any studies studies Hypoglycemia Hypoglycemia as ≤3.1 mmol? providing a providing a Definition Definition of (RA=Requiring (captured as any) of Severe assistance) Hypoglyce Hypoglycemia

mia

MET MONO 4/8 4/4 4/4 RA (Fonseca et al. 2013 0/4 coma requiring (n=8 studies) hospitalization)

DPP4i MONO 25/45 15/25 12/15 RA 4/15

(n=45) 3/15 “confirmed” as BG ≤2.8 Dejager 2007 (3/472; 0/160), mmol/L Pi-Sunyer 2007- no hypo,

(Pan 2012, Rosenstock 2008, Pratley 2006 (1/70; 0/28), Rosenstock 2009) Scherbaum 2008 (0/156; 1/150)

DPP4i w MET 15/23 13/15 8/13 RA 4/13

(n=23) 1/13 RA and BG <3.9 Bosi 2007 (2/362; 1/182)

1/13 RA and BG <3.3 Goodman 2009 (2/248; 0/122)

3/13 “confirmed” as BG ≤2.8 Odawara 2014-no hypo mmol/L Pan 2012 vilda (1/294; 0/144)

GLP1RA 7/9 6/7 2/6 RA 1/6 MONO 1/6 RA & BG <3.0 mmol/L Seino 2008 – no hypo (n=9) 1/6 RA & BG <2.9 mmol/L

1/6 BG <3.1 mmol/L (Sorli 2017; no hypo)

1/6 BG <2.8 mmol/L (“minor” Madsbad 2004)

GLP1RA w 6/8 5/6 2/5 RA and BG <2.0 mmol/L 1/5 MET 3/5 RA Nauck 2016 sema (n=8) (12/270; 2/46)

2

SGLT2i 11/17 6/11 3/6 RA and BG <3.0 mmo/L 0/6 MONO 3/6 RA (Fonseca coma w (n=17) hosp)

SGLT2i w 10/15 6/10 4/6 RA and BG <3.0 mmol/L 0/6 MET 1/6 RA (n=15) 1/6 <3.0 mmol/L

DUAL 4/6 4/4 3/4 Requiring Assistance 0/4 INITIATION 1/4 Requiring Assistance and (n=6) BG <3.89 mmol/L

TRIPLE 6/7 4/6 2/4 Requiring Assistance 0/4 THERAPY 2/4 Requiring Assistance or (n=7) BG <3.0 mmol/L

NON-MET 2/3 2/2 2/2 Requiring Assistance 1/2 Background Dual Therapy Odawara 2015 -no hypo

(n=3)

90/141 65/90 42/65 = RA only 11/65

9.4.3 Exclusion of Patients Experiencing or With a History of Hypoglycemia

Although investigators of randomized controlled trials can use their discretional judgement to enroll patients, we investigated whether the study explicitly stated exclusion of patients with a history of hypoglycemia (usually of the severe form) during the screening or throughout the trial. Seven included studies indicated excluding patients either prior to randomization or during the study period.

3

9.4.4 Rescue Medication

Given the placebo comparison of our analysis, we also investigated the role of rescue medication within the studies. Approximately half of the studies (48% (68/141) allowed for rescue therapy. Criteria for rescue therapy were mentioned in some, but not others. Hyperglycemic duration, BG values, and recommendation or not of specific agents also varied amongst studies.

Just under half of the included studies (64/141 or 45%) did not report information on use of rescue medication, making it difficult to interpret whether or not it was allowed. More importantly, heterogeneity existed in statistical analysis of patients receiving rescue therapy. Only a few studies (9/141 or 6%) indicated that rescue therapy was not allowed, and if patients required additional glycemic lowering, they were withdrawn from the study.

9.4.5 Inclusion of Hypoglycemia Reports After Rescue Medication

Most of the studies that allowed rescue therapy for hyperglycemia, did not include safety data after patients received rescue therapy (53/68 or 78%). Data observed post rescue were largely treated as missing. Last observation carried forward method was used in most studies to handle data observed prior to rescue therapy administration. Only one study, Nauck 2016 (GLP1 Mono), admittedly changed hypoglycemia data post hoc to only include data pre-rescue.

In the remaining 22% (15/68) of studies allowing for rescue therapy, hypoglycemia data post rescue was included. Three of these fifteen studies included safety data post rescue only if hypoglycemia was severe. Of these three, in Terra 2017 (SGLT2i mono) and Miller 2018 (Dual initiation), 2 patients reported severe hypoglycemia in the AHA arm and none in the placebo arm. In Dagogo-Jack (Triple Therapy), no severe hypoglycemia was observed in the AHA arm and one in the placebo group.

Inclusion of data after the initiation of rescue therapy further adds to the heterogeneity of hypoglycemia reported in the literature. In two studies, they indicated that hypoglycemia occurred after the initiation of rescue therapy and were therefore not included in our meta- analysis. In Sorli 2017 (GLP1 Mono), two patients in the placebo arm experienced hypoglycemia after the initiation of rescue therapy. In Aschner 2006, one instance of hypoglycemia also occurred after rescue and was therefore subtracted. As most studies did not

4

specify the association of rescue therapy to hypoglycemic events, it remains unclear how handling of these patients may have biased results. Although rescue therapy was permissible for control of hyperglycemia, most of the studies including hypoglycemia post rescue had patients reporting events in the AHA arm and not in the placebo arm. In fact, only two studies allowing for hypoglycemia data post rescue observed more hypoglycemia in the placebo than the AHA arm (Taskinenen 2011 and Bolinder 2012).

Table 7: Heterogeneity of Hypoglycemia - Rescue Therapy

5

Exclusion of Rescue Of those studies allowing rescue Patients with therapy, how many included safety data events? Permitted? post rescue therapy?

MET MONO 0/8 5/8 yes Ji 2017 Rescue Confounding

(n=8 studies) 3/8 not reported

DPP4i MONO 0/45 19/45 yes Ji 2017 Rescue Confounding

(n=45) 23/45 not reported

3/45 not allowed (Hone 2016, Inagaki 2014, Jung 2015)

DPP4i w MET 0/23 9/23 yes Taskinen 2011 Rescue Confounding

(n=23) 10/23 not reported

4/23 not allowed (Amin 2015, Nauck 2009, Odawara 2014, Yang 2011)

GLP1RA MONO 0/9 4/9 yes Grunberger 2012 Rescue Confounding

(n=9) 5/9 not reported

GLP1RA w MET 2/8 3/8 yes Ahren 2013 Rescue Confounding

(n=8) 5/8 not reported

SGLT2i MONO 1/17 8/17 yes Bailey 2012 Rescue Confounding

(n=17) 9/17 not reported Ferrannini 2010 Rescue Confounding

Ji 2014 Rescue Confounding

Terra 2017 Rescue Confounding only if severe

SGLT2i w MET 3/15 7/15 yes Bailey 2010 Rescue Confounding

(n=15) 7/15 not reported Bolinder 2012 Rescue Confounding

1/15 not allowed Yang 2016 Rescue Confounding

6

DUAL INITIATION 0/6 6/6 yes Ji 2017 Rescue confounding

(n=6) (not hypo Miller 2018 Rescue confounding only if specific but at severe discretion of investigator)

TRIPLE THERAPY 1/7 7/7 yes Dagogo-Jack 2018 Included post rescue safety data only if severe hypoglycemia (n=7) occurred .

NON-MET 0/3 2/3 Not no Background Dual Therapy reported

(n=3) 1/3 not allowed

Total 7/141 68/141 – Allowed rescue 53/68 - Did not include post rescue data

Excluded 9/141- Did not allow 12/68 - Included post rescue safety data patients with rescue or a history of 3/68 – Included post rescue safety data hypoglycemia 64/141- Did not indicate only if severe hypoglycemia occurred 9.5 RISK OF BIAS

Risk of bias for all included studies can be found in Appendix VII for any hypoglycemia and Appendix VIII for severe hypoglycemia.

9.5.1 Selection Bias: Random Sequence Generation

As anticipated, given our study eligibility, none of the studies had a high risk of bias for random sequence generation. Included studies were designated as having a low risk of bias whether or not randomization method was described.

9.5.2 Selection Bias: Allocation Concealment

Similarly, allocation concealment was designated as an unclear risk for the majority of included studies and low bias for the remaining studies. None of the studies included were found to have a high risk of allocation concealment bias.

7

9.5.3 Performance Bias: Blinding of Participants and Personnel

Given our study protocol, all included studies were blinded and received a low risk of bias.

9.5.4 Detection Bias: Blinding of Outcome Assessment

Detection bias was considered high risk for all included studies, given the harms outcome in a placebo-controlled study. Hypoglycemia may have been more likely to be attributed to the treatment arm given the placebo comparator. Further, investigators may have been more likely to discontinue participants in the placebo arm for unsatisfactory therapeutic effect or at their discretion for any concern.

9.5.5 Attrition Bias: Incomplete Outcome Data

Most studies had comparable loss of patients throughout the study due to adverse events and were designated as low risk of bias. Very few studies provided incomplete details. The remaining studies were assigned a high-risk bias due to the imbalance of study discontinuation due to adverse events or for indication participant discontinuation after experiencing a hypoglycemic episode.

9.5.6 Reporting Bias: Selective Reporting of Outcomes

Majority of included studies provided information on hypoglycemia outcomes and thus were designated as low risk of reporting bias. Studies were designated as high risk of bias for providing a range of hypoglycemia events, for not describing the severity or for only reporting hypoglycemia findings during the extension phase.

9.5.7 Other Bias: Confounding by Rescue Medication

As described in greater detail in the earlier sections, use of rescue medication was permitted in almost half of the included studies. Since most of these studies did not incorporate safety or hypoglycemia data post rescue, they were designated as low risk of bias. Studies received a high risk of bias if rescue medication was permitted and included in the safety outcome.

8

9.6 HYPOGLYCEMIA OUTCOMES

9.6.1 Primary Outcome: Any Hypoglycemia

We found no evidence to suggest an increased risk for any hypoglycemia with any of the comparators. Below is the summary forest plot for risk of any hypoglycemia. Forest Plots with individual studies per comparator can be found in Appendix VII.

Figure 1: Summary Forest Plot of Any Hypoglycemia

AHA Placebo (n/N) (n/N) MONOTHERAPY

Metformin 53/1512 12/442 RR 1.36 (95% CI 0.71 - 2.63)

DPP4i 151/7567 38/2978 RR 1.07 (95% CI 0.76 - 1.51)

GLP1RA 43/1020 8/414 RR 1.77 (95%CI 0.85 - 3.67)

SGLT2i 78/4531 17/1505 RR 1.29 (95% CI 0.79 - 2.09)

METFORMIN BACKGROUND

DPP4i 87/4853 45/2426 RR 0.88 (95% CI 0.61 - 1.27)

GLP1RA 163/2618 23/734 RR 1.30 (95% CI 0.79 - 2.14)

SGLT2i 70/4069 20/1144 RR 0.92 (95% CI 0.56 - 1.53)

Dual Therapy Inittiation 60/1477 4/311 RR 2.12 (95% CI 0.89 - 5.05)

Third AHA added to Dual RR 1.12 (95% CI 0.69 - 1.83) Therapy 45/1459 25/1057

0.0625 0.125 0.25 0.5 1 2 4 8 Favours AHA Favours Placebo Metformin Background 9

9.6.2 Secondary Outcome: Severe Hypoglycemia

As anticipated, we found no evidence to suggest an increase in severe hypoglycemia with any comparators. Below are the summary forest plots of the two comparators in which studies reported incidence of severe hypoglycemia. Forest plots containing individual studies of severe hypoglycemia can be found in Appendix VIII.

Figure 2: Summary Forest Plot of Severe Hypoglycemia

AHA Placebo (n/N) (n/N)

DPP4i Added to Metformin RR 0.62 (95% CI 0.14 - 2.67) Background 4/839 4/466

Third Agent Added to Dual Background 2/814 1/403 RR 0.72 (95% CI 0.11 - 4.53)

0.0625 0.125 0.25 0.5 1 2 4 8

Favours AHA Favours Placebo

9.6.3 Sensitivity Analysis: Inclusion of Studies with Zero Any Hypoglycemia

Only 19.1% of the included studies did not find any hypoglycemia. The number of patients reporting any hypoglycemia was low and slightly higher with overall AHA therapy compared to placebo, 750/33,132 (2.26%) compared to 192/12844 (1.49%), respectively. Two comparisons, GLP1RA + metformin (RD 0.02 (CI 0.01-0.03)) and Dual Therapy Initiation (RD 0.02 (CI 0.01- 0.04)), resulted in a significantly higher risk difference in comparison to placebo. Although statistically significant, the clinical significance of a 2% higher actual difference with these two groups relative to placebo is not known.

10

Below is a summary of the findings for any hypoglycemia using risk difference per comparator when studies with no hypoglycemia were considered. Individual studies per comparison can be found in Appendix IX for any hypoglycemia.

Table 8: Summary of Studies with Zero Any Hypoglycemia

Comparison Number Number of Number of Risk Difference of AHA patients placebo (95% CI) Studies with Any patients with with hypoglycemia / Any Zero Number of hypoglycemia / Any AHA Patients Number of Hypo / Placebo Total Patients Number of Studies

Met Mono 1/8 53/1581 12/477 0.01 (-0.00 – 0.03)

DPP4i Mono 12/45 151/9640 38/3845 0.0 (-0.00 – 0.01)

DPP4i + Met 5/23 87/5476 45/2742 -0.00 (-0.01 – 0.00)

GLP1RA 3/9 43/1581 8/629 0.01 (-0.00 – 0.02) Mono

GLP1RA + 0/8 163/2618 23/734 0.02 (0.01 – 0.03) Met

SGLT2i 0/17 78/4531 17/1505 0.01 (0.00 – 0.01) Mono

11

SGLT2i + 3/15 70/4507 20/1343 0.00 (-0.00 – 0.01) Met

Dual 0/6 60/1477 4/311 0.02 (0.01 – 0.04) Initiation

Triple 0/7 45/1459 25/1057 0.01 (-0.01 – 0.02) Therapy

Non-Met 3/3 0/262 0/201 0.00 (-0.02 – 0.02) Background Dual Therapy

Total 27/141 750/33,132 192/12,844

9.6.4 Sensitivity Analysis: Inclusion of Studies with Zero Severe Hypoglycemia

On the other hand, the majority of studies (127/139 or 91.4%) did not report any incidence of severe hypoglycemia. Further, in those studies that did report severe hypoglycemia, the number of patients were low and similar between the AHA and placebo arms, 17/32,719 or 0.052% and 7/12,539 or 0.056%, respectively. Below is a summary of the findings for severe hypoglycemia when studies with zero incidence were considered. Forest plots of individual studies per comparisons can be found in Appendix X for sever hypoglycemia.

Table 9: Summary of Studies with Zero Severe Hypoglycemia

Comparison Number Number of Number of Risk Difference of AHA patients placebo (95% CI) Studies with Severe patients with with hypoglycemia / Severe

12

Zero Number of hypoglycemia / Severe AHA Patients Number of Hypo / Placebo Total Patients Number of Studies

Met Mono 7/8 1/1581 0/477 0.00 (-0.01 – 0.01)

DPP4i Mono 44/44 0/9452 0/3766 0.00 (-0.00 – 0.00)

DPP4i + Met 20/23 4/5476 4/2742 -0.00 (-0.00 – 0.00)

GLP1RA 8/9 1/1581 0/629 0.00 (-0.01 – 0.01) Mono

GLP1RA + 7/8 2/2618 0/734 0.00 (-0.00 – 0.01) Met

SGLT2i 16/17 2/4531 0/1505 0.00 (-0.00 – 0.00) Mono

SGLT2i + 14/15 3/4507 2/1343 -0.00 (-0.00 – 0.00) Met

Dual 5/6 2/1477 0/311 0.00 (-0.01 – 0.01) Initiation

Triple 3/6 2/1234 1/831 0.00 (-0.01 – 0.01) Therapy

Non-Met 3/3 0/262 0/201 0.00 (-0.02 – 0.002) Background

13

Dual Therapy

TOTAL 127/139 17/32,719 7/12,539

9.6.5 Post Hoc Analysis: Impact on Results if Defining Severe Hypoglycemia as ≤3.1 mmol/L

Given the current lack of standardization for BG thresholds of severe hypoglycemia, we conducted a post hoc analysis of the results if severe hypoglycemia was defined as ≤3.1 mmol/L, instead of <3.0 mmol/L. Eleven studies defined severe hypoglycemia as a threshold of ≤3.1 mmol/L and are depicted in Table 6. Of these 11 studies, no hypoglycemia occurred in 4 studies. Our results, presented in Appendix XI, did not change when severe hypoglycemia was defined as ≤3.1 mmol/L.

9.7 STATISTICAL HETEROGENEITY

Statistical heterogeneity is inevitable in meta-analyses. Below the results and interpretation of Tau², Chi² and I² are reviewed. The results of all statistical tests for heterogeneity can be found at the bottom of every forest plot in Appendix VII for any hypoglycemia and in Appendix VIII for severe hypoglycemia.

9.7.1 Tau²

Tau² is used for random effect model meta-analyses to describe the extent of between-study variance (Deeks et al. 2008). All analyses except for risk ratio of any hypoglycemia for GLP1 + metformin resulted in a Tau² value of 0, demonstrating overall very low between study variance. The Tau² of GLP11RA + metformin was low at 0.07 but nevertheless taken into consideration for the SoF table.

9.7.2 Chi²

Chi² is a formal test of the consistency of study results. The overlap of confidence intervals of individual studies in a forest plot is used to determine whether the heterogeneity is due to chance

14

alone. A low p-value or a large Chi² suggests presence of true heterogeneity, beyond that of chance alone (Higgins 2011). It is recommended to interpret Chi² value with caution, since small sample sizes or a low number of studies can have low power to detect heterogeneity. On the other hand, too many studies can have high power to detect heterogeneity. In all our analyses, the Chi² p-value was greater than 0.05, rejecting evidence of heterogeneity, and suggesting statistical homogeneity. In other words, the true treatment effect is similar to that observed in the studies included in our meta-analyses.

9.7.3 I²

I² represents the percentage of variation between sample estimates that is due to true, between study heterogeneity (Deeks et al. 2008). Most (36/40) analyses conducted resulted in a I² value of 0, further corroborating that the variability in effect size estimates was due to sampling error within the studies, and not due to true heterogeneity between studies. Further, the highest I² value was found for risk difference of any hypoglycemia with GLP1RA monotherapy at 28%. I² values below 40% are generally considered to be unimportant (Higgins 2011). Only three other analyses contained non-zero I² values; GLP1RA + metformin any hypoglycemia relative risk (I² =12%), GLP1RA + metformin any hypoglycemia risk difference (I² =6%) and SGLT2i monotherapy any hypoglycemia risk difference (I² = 4%). Although a p-value of 0.10 is sometime considered for tests of heterogeneity (Higgins 2011), our results contained overall low I² values and a p-value of 0.10 would not have changed the results .

9.7.4 Overall Effect (Z)

The z statistics incorporates the weighted average effect size. A significant p-value for the z-test was observed for the risk difference of any hypoglycemia for GLP1RA + met, SGLT2i monotherapy and Dual therapy initiation, suggesting that the effect size is different from zero, rejecting the null hypothesis.

9.8 PUBLICATION BIAS

As recommended by the Cochrane Handbook, funnel plot asymmetry was visually assessed for outcomes that contained 10 or more studies. Further, given the dichotomous outcomes used, statistical tests such as Egger’s were not conducted (Higgins 2011). Upon visual inspection of

15

the 4 funnel plots containing 10 or more studies, no evidence of asymmetry was noted, suggesting publication bias was an unlikely factor.

Figure 3: Funnel Plot of DPP4i Monotherapy - Any Hypoglycemia

Figure 4: Funnel Plot of DPP4i + Metformin - Any Hypoglycemia

16

Figure 5: Funnel Plot of SGLT2i Monotherapy - Any Hypoglycemia

Figure 6: Funnel Plot of SGLT2i + Metformin - Any Hypoglycemia

17

9.9 SUMMARY OF FINDINGS

The summary of findings (SoF) for the primary and secondary outcomes can be found in Appendix XII. SoF presents relative effects and not absolute effects. Thus, sensitivity analyses of any and severe hypoglycemia using Risk Difference were not included.

The outcome of severe hypoglycemia was assessed as high certainty for all comparisons. Severe hypoglycemia was extremely rare and many studies reported zero events. Further, rates were similar between placebo and the treatment arm.

For the certainty assessment of any hypoglycemia, a single down grade was applied to all comparisons for serious risk of bias. As mentioned earlier, we could not rule out that detection bias did not occur in a placebo-controlled study assessing an adverse safety outcome.

A second down grade, resulting in low certainty was applied to metformin monotherapy, GLP1RA monotherapy and GLP1RA with metformin therapy, dual therapy initiation and triple therapy for attrition bias or imbalances in the drop out rates between treatment and placebo arms. The remaining categories of biases, including inconsistency, imprecision, indirectness and publication bias was considered not serious for any outcome or comparison.

18

The GRADE working group advises no rate down for indirectness from RCTs unless compelled by a biological basis that the population within the included trials would significantly differ from that of the real world (Guyatt et al. 2011). Although our findings were extracted form randomized trials where patients at risk could be excluded prior or during the conduct of the study, indirectness was not considered a major threat to the relative risk estimates.

9.10 DISCUSSION

In this systematic review and meta-analysis, we found no evidence that relative to placebo, new AHA increase risk of any or severe hypoglycemia when used as monotherapy, with background metformin, as dual initiation or triple combination therapy. Although we anticipated no increase in risk of severe hypoglycemia with the three new AHA given their glucose-dependent mechanism of action, this analysis was important to evaluate the patient-important outcome of less severe episodes. By including studies in which zero reports of hypoglycemia occurred, we were able to provide even greater reassurance on the placebo-like risk for any and severe hypoglycemia.

Attainment of glycemic control in patients with diabetes is critical in attenuating risk of developing diabetes complications. However, the rate limiting step in achieving a lower BG is the fear of hypoglycemia, for both patients and their health care providers. Unlike SUs and insulin, DPP4i, GLP1RA and SGLT2i offer reduction of BG with the promise of low to no risk of hypoglycemia. However, evidence to support this claim is extracted from efficacy-based studies and their respective meta-analyses in which hypoglycemia risk is a secondary, tertiary or exploratory endpoint. Use of background therapies, such as SUs or insulin often increase risk estimates, making it difficult to assess risk with AHA alone. Given the importance of hypoglycemia avoidance to patients and their physicians, this meta-analysis was specifically designed to evaluate hypoglycemia risk as the primary objective in studies where background agents known to increase the risk were not permitted.

Previously published meta-analyses that compare to placebo and exclude studies which allow use of other therapies are limited. Two such studies of similar criteria were found for DPP4i’s

19

(Kawalec and Liu et al) and one for SGLT2i (Kawalec et al.). In both instances, risk of hypoglycemia was found to be similar to placebo. Similar meta-analyses with GLP1RA in which hypoglycemia is compared to placebo and does not allow use of background therapies were not found, as they allowed background use of other agents and compared to active controls.

Eligible studies included in our meta-analysis was greatest with DPP4i followed by SGLT2i and GLP1RA. This may be attributed to the fact that DPP4i were first to receive approval for use in patients with T2DM and have thus been studied more extensively. Moreover, since most GLP1RA are administered as subcutaneous injections, they are often reserved for use after failure to achieve glycemic targets with oral agents. Given our study inclusion criteria, that disallowed studies with use of background SUs, insulin and TZDs for example, it is plausible that more GLP1RA studies did not fit our inclusion criteria. This is changing, however, since many diabetes guideline bodies now recommend use of SGLT2i and GLP1RA with proven evidence to be used earlier in the paradigm of diabetes management, owing to the significant reductions of cardiovascular outcomes in patients with diabetes and CVD.

9.10.1 Limitations

Our meta-analysis has several limitations. Given that we only included randomized controlled trials, our results may not be generalizable to the real-world. Randomized trials select relatively ‘healthier’ patients by excluding those with major comorbidities, who are often the same individuals at risk of experiencing hypoglycemia. And although we reported on studies excluding patients at risk of hypoglycemia prior to randomization or during the study, patients could still have been withdrawn from the study for any reason or concern at the investigator’s discretion, as indicated in most studies.

Other post randomization biases may have occurred given our harms outcome and placebo comparator. As a result, detection bias was rated high for all included trials. In placebo- controlled trials, an adverse event may more likely be attributable to the treatment arm. Also given the placebo comparator, investigators may more likely withdraw patients for unsatisfactory therapeutic effect, or more likely to offer placebo patients rescue medication. Again, although

20

we attempted to identify studies permitting rescue therapy and how patients initiating rescue therapy were statistically handled, many studies did not provide information on whether rescue medication was permitted, making bias ranking difficult. Such biases post-randomization may have led to a different placebo population completing the study than the one who began the study. As a result, the certainty in our findings within the SoF table was downgraded to moderate for serious risk of bias for all comparisons for the outcome of any hypoglycemia.

In this large analysis, we pooled all AHA within a class into one estimate. However, molecule- specific differences have been reported and thus it remains possible that each individual agent is associated with a differing hypoglycemia risk. We included only published trials and did not search the grey literature, in which more unfavorable data may be less likely to be published. All but one study was sponsored by the manufacturing pharmaceutical company. However, a recent Cochrane review found studies sponsored by industry were more likely to have positive efficacy, but not safety findings (Lundh et al. 2018). In addition, we restricted our study inclusion criteria to only those published in the English language. Although a number of studies included in our analysis were conducted in Asian populations, studies published in other languages may be associated with differing environments and patient populations in which the safety evidence also differs.

A major limitation of this analysis is the heterogeneity of hypoglycemia definitions in efficacy- focused studies, a limiting factor that has discouraged others from conducting similar pooled estimates. However, our primary endpoint of any hypoglycemia was irrespective of the definition used within each study. Our definition for the secondary endpoint of severe hypoglycemia included the need for third party assistance or a BG measurement of ≤3.0 mmol/L. We observed that most of our included studies defined severe hypoglycemia as requiring assistance or thresholds ≤3.0 mmol/L. Further, despite only a few studies defining severe hypoglycemia at ≤3.1 mmol/L, we conducted a post hoc sensitivity analysis to explore the impact of defining severe hypoglycemia at this higher threshold and found that our results did not differ.

21

9.11 CONCLUSION

Our meta-analysis provides convincing reassurance that risk of any and severe hypoglycemia with new AHA used alone, with metformin or in combination is comparable to placebo. In addition, our study offers several important insights. This is the first study to evaluate hypoglycemia risk relative to placebo with new AHA in which background agents known to increase risk are excluded. This is important since relative to insulin and sulfonylurea, therapies notorious for increasing risk of hypoglycemia, new AHA offer a clinically important advantage. However, relative to placebo, risk of hypoglycemia with new AHA in previous RCTs and meta- analyses have been difficult to discern given the use of SU and/or insulin. In addition to the quantitative pooled estimate of hypoglycemia risk, our study provides a qualitative description of heterogeneous definitions of hypoglycemia and its assessment used within the clinical literature.

9.11.1 Future Research

Hypoglycemia is rarely the primary objective of clinical research, yet it is a key barrier to optimal diabetes management. More focus on this important patient outcome is clearly necessary. International collaborative efforts to standardize hypoglycemia classifications and definitions used in clinical research must continue. Moreover, guidance on how hypoglycemia risk in participants is screened, monitored, documented, reported and statistically handled require guidance.

As the first known attempt to pool and describe the evidence of hypoglycemia risk with new AHA on their own, additional research from observational studies, grey literature and studies without a language restriction is necessary to corroborate our findings. Risk of severe hypoglycemia with new AHA is clearly not a concern, however, less severe episodes do occur and need to be captured in clinical trials evaluating diabetes therapies. Additional research on the association of patient characteristics (such as advanced age, diabetes duration) and glycemic control (such as baseline HbA1c and change in HbA1c) could also further advance our understanding of hypoglycemia risk with new therapies.

New continuous monitoring devices will play a tremendous role in advancing hypoglycemia research. It is believed that much of mild to moderate hypoglycemia is asymptomatic and not perceived by the patient especially those with advanced age. These new advances in continuous 22

blood glucose monitoring can assure that all events are captured even if not perceived by the patients. Limited studies report on nocturnal hypoglycemia, despite the dangers of such an event. As device manufacturers strive to improve patient comfort and adherence, more insight into the risk of nocturnal hypoglycemia with various therapies and their combinations can provide additional safety evaluations.

IMPLICATIONS

10.1.1 TO HCP

The placebo-like risk of any hypoglycemia with new AHA and metformin can reassure physicians to strive for better glycemic control in more patients with diabetes. Whether used alone, on top of metformin or as triple therapy, risk of any hypoglycemia was similar to placebo for all comparators. Results are even more reassuring for severe hypoglycemia as number of patients reporting events was found to be exceedingly rare.

The results of this analysis also reaffirm diabetes guideline categorization of the rare risk with new AHA and reaffirm Diabetes Canada categorization of low risk.

10.1.2 TO PATIENT

Patients with diabetes who have experienced a hypoglycemic episode fear its reoccurrence. As described in the earlier chapters, patients with events are less likely to achieve glycemic targets and adhere to therapy. This study can help convince them of the placebo-like risk associated with metformin, DPP4i, GLP1RA and SGLT2i used alone or in combination with each other.

Beyond improving adherence and likelihood to achieve glycemic targets and thus reduce complications, elimination of hypoglycemia can improve the quality of life of patients with diabetes. Further, patients taking new AHA are at less risk of incurring the multiple complications and consequences associated with hypoglycemia, including depression, motor vehicle accidents and employment implications.

10.1.3 TO SYSTEM

Hypoglycemia incurs significant direct and indirect health care costs. Costs associated with severe hypoglycemia is well described in the literature. Cost analyses often capture 23

hospitalization for hypoglycemia (using hospital codes) and evaluate the increased use of health care resources and increased self monitoring post event. Less studied, is the impact of less severe hypoglycemic events and the associated indirect costs such as loss of work productivity or time off work after experiencing an episode. By averting hypoglycemia, whether severe or mild, government spending on diabetes-related costs can be improved.

10.1.3.1 TEST STRIPS

In 2013, the Canadian government introduced limits to the number of glucose tests strips reimbursed for patients with diabetes according to the type of therapy (non-hypoglycemia- causing OHA, hypoglycemia causing OHA, insulin or no therapy) provided. At $1 a strip, this policy has made a significant impact a reducing diabetes-associated costs. However, two-years after the introduction of this policy, patients taking non-hypoglycemia causing therapies exceeded quantity limits more than any other group (Gomes et al. 2016). Although, for some patients, especially when newly diagnosed, self monitoring can increase disease awareness management and engagement, the need to test multiple times a day in many others on new AHA can be eliminated. Results of this meta-analysis can be communicated to reassure patients, physicians and policy maker of the placebo like risk with new AHA. By further strengthening these policy limits and recommendations for those on new AHA, even greater savings can be achieved.

24

References

Aaboe, K., S. Akram, C. F. Deacon, J. J. Holst, S. Madsbad, and T. Krarup. 2015. 'Restoration of the insulinotropic effect of glucose-dependent insulinotropic polypeptide contributes to the antidiabetic effect of dipeptidyl peptidase-4 inhibitors', Diabetes Obes Metab , 17: 74- 81.

Aaboe, K., F. K. Knop, T. Vilsboll, C. F. Deacon, J. J. Holst, S. Madsbad, and T. Krarup. 2010. 'Twelve weeks treatment with the DPP-4 inhibitor, sitagliptin, prevents degradation of peptide YY and improves glucose and non-glucose induced insulin secretion in patients with type 2 diabetes mellitus', Diabetes Obes Metab , 12: 323-33.

Abbas, A. S., H. M. Dehbi, and K. K. Ray. 2016. 'Cardiovascular and non-cardiovascular safety of dipeptidyl peptidase-4 inhibition: a meta-analysis of randomized controlled cardiovascular outcome trials', Diabetes Obes Metab , 18: 295-9.

Abdul-Ghani, M. A., L. Norton, and R. A. Defronzo. 2011. 'Role of sodium-glucose cotransporter 2 (SGLT 2) inhibitors in the treatment of type 2 diabetes', Endocr Rev , 32: 515-31.

Abdul-Ghani, M. A., D. Tripathy, and R. A. DeFronzo. 2006. 'Contributions of beta-cell dysfunction and insulin resistance to the pathogenesis of impaired glucose tolerance and impaired fasting glucose', Diabetes Care , 29: 1130-9.

Action to Control Cardiovascular Risk in Diabetes Study, Group, H. C. Gerstein, M. E. Miller, R. P. Byington, D. C. Goff, Jr., J. T. Bigger, J. B. Buse, W. C. Cushman, S. Genuth, F. Ismail-Beigi, R. H. Grimm, Jr., J. L. Probstfield, D. G. Simons-Morton, and W. T. Friedewald. 2008. 'Effects of intensive glucose lowering in type 2 diabetes', N Engl J Med , 358: 2545-59.

Agarwal, P., C. Jindal, and V. Sapakal. 2018. 'Efficacy and Safety of in Indian Patients with Inadequately Controlled Type 2 Diabetes Mellitus: A Randomized, Double- blind Study', Indian J Endocrinol Metab , 22: 41-46.

Ahren, B., and J. E. Foley. 2016. 'Improved glucose regulation in type 2 diabetic patients with DPP-4 inhibitors: focus on alpha and beta cell function and lipid metabolism', Diabetologia , 59: 907-17.

Ahren, B., A. Leguizamo Dimas, P. Miossec, S. Saubadu, and R. Aronson. 2013. 'Efficacy and safety of lixisenatide once-daily morning or evening injections in type 2 diabetes inadequately controlled on metformin (GetGoal-M)', Diabetes Care , 36: 2543-50.

Ahren, B., A. Schweizer, S. Dejager, B. E. Dunning, P. M. Nilsson, M. Persson, and J. E. Foley. 2009. 'Vildagliptin enhances islet responsiveness to both hyper- and hypoglycemia in patients with type 2 diabetes', J Clin Endocrinol Metab , 94: 1236-43.

25

Akeroyd, J. M., E. A. Suarez, B. Bartali, G. R. Chiu, M. Yang, A. V. Schwartz, and A. B. Araujo. 2014. 'Differences in skeletal and non-skeletal factors in a diverse sample of men with and without type 2 diabetes mellitus', J Diabetes Complications , 28: 679-83.

Akiyama, Tomoaki, Masayo Yamada, Sho Katsuragawa, Sakiko Terui, Taichi Minami, and Yasuo Terauchi. 2017. "The Effect of Combination Therapy of Degludec and Liraglutide in Hospitalized Patients with Uncontrolled Type 2 Diabetes (T2DM)." In Diabetes , A295-A95. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Akram, K., U. Pedersen-Bjergaard, K. Borch-Johnsen, and B. Thorsteinsson. 2006. 'Frequency and risk factors of severe hypoglycemia in insulin-treated type 2 diabetes: a literature survey', J Diabetes Complications , 20: 402-8.

Akram, K., U. Pedersen-Bjergaard, B. Carstensen, K. Borch-Johnsen, and B. Thorsteinsson. 2006. 'Frequency and risk factors of severe hypoglycaemia in insulin-treated Type 2 diabetes: a cross-sectional survey', Diabet Med , 23: 750-6.

Alberti, K. George M. M., Paul Zimmet, and Jonathan Shaw. 2005. 'The metabolic syndrome—a new worldwide definition', The Lancet , 366: 1059-62.

Ali, M. K., K. M. Bullard, J. B. Saaddine, C. C. Cowie, G. Imperatore, and E. W. Gregg. 2013. 'Achievement of goals in U.S. diabetes care, 1999-2010', N Engl J Med , 368: 1613-24.

Alsahli, M., and J. E. Gerich. 2013. 'Hypoglycemia', Endocrinol Metab Clin North Am , 42: 657- 76.

Alvarez Guisasola, F., S. Tofe Povedano, G. Krishnarajah, R. Lyu, P. Mavros, and D. Yin. 2008. 'Hypoglycaemic symptoms, treatment satisfaction, adherence and their associations with glycaemic goal in patients with type 2 diabetes mellitus: findings from the Real-Life Effectiveness and Care Patterns of Diabetes Management (RECAP-DM) Study', Diabetes Obes Metab , 10 Suppl 1: 25-32.

American Diabetes, Association, J. E. Anderson, M. A. Greene, J. W. Griffin, Jr., D. B. Kohrman, D. Lorber, C. D. Saudek, D. Schatz, and L. Siminerio. 2014. 'Diabetes and employment', Diabetes Care , 37 Suppl 1: S112-7.

American Diabetes, Association, D. Lorber, J. Anderson, S. Arent, D. J. Cox, B. M. Frier, M. A. Greene, J. Griffin, Jr., G. Gross, K. Hathaway, I. Hirsch, D. B. Kohrman, D. G. Marrero, T. J. Songer, and A. L. Yatvin. 2014. 'Diabetes and driving', Diabetes Care , 37 Suppl 1: S97-103.

Amin, N. B., X. Wang, S. M. Jain, D. S. Lee, G. Nucci, and J. M. Rusnak. 2015. 'Dose-ranging efficacy and safety study of ertugliflozin, a sodium-glucose co-transporter 2 inhibitor, in patients with type 2 diabetes on a background of metformin', Diabetes Obes Metab , 17: 591-8.

26

Andersen, S. E., and M. Christensen. 2016. 'Hypoglycaemia when adding sulphonylurea to metformin: a systematic review and network meta-analysis', Br J Clin Pharmacol, 82: 1291-302.

Arjona Ferreira, J. C., D. Corry, C. E. Mogensen, L. Sloan, L. Xu, G. T. Golm, E. J. Gonzalez, M. J. Davies, K. D. Kaufman, and B. J. Goldstein. 2013. 'Efficacy and safety of sitagliptin in patients with type 2 diabetes and ESRD receiving dialysis: a 54-week randomized trial', Am J Kidney Dis , 61: 579-87.

Arjona Ferreira, J. C., M. Marre, N. Barzilai, H. Guo, G. T. Golm, C. M. Sisk, K. D. Kaufman, and B. J. Goldstein. 2013. 'Efficacy and safety of sitagliptin versus glipizide in patients with type 2 diabetes and moderate-to-severe chronic renal insufficiency', Diabetes Care , 36: 1067-73.

Aronson, R., R. Goldenberg, D. Boras, R. Skovgaard, and H. Bajaj. 2018. 'The Canadian Hypoglycemia Assessment Tool Program: Insights Into Rates and Implications of Hypoglycemia From an Observational Study', Can J Diabetes , 42: 11-17.

Aschner, P., M. S. Kipnes, J. K. Lunceford, M. Sanchez, C. Mickel, and D. E. Williams-Herman. 2006. 'Effect of the dipeptidyl peptidase-4 inhibitor sitagliptin as monotherapy on glycemic control in patients with type 2 diabetes', Diabetes Care , 29: 2632-7.

Association, American Diabetes. 2005. 'Defining and reporting hypoglycemia in diabetes: a report from the American Diabetes Association Workgroup on Hypoglycemia', Diabetes Care , 28: 1245-49.

Association, American Diabetes. 2017. 'Standards of medical care in diabetes—2017 abridged for primary care providers', Clinical Diabetes , 35: 5-26.

Association, American Diabetes. 2018. 'Economic Costs of Diabetes in the US in 2017', Diabetes Care , 41: 917.

Association, American Diabetes. 2019. '6. Glycemic Targets: Standards of Medical Care in Diabetes-2019', Diabetes Care , 42: S61.

Association, Canadian Diabetes. 2014. "An economic tsunami: the cost of diabetes in Canada. December 2009." In.

Atkins, D., D. Best, P. A. Briss, M. Eccles, Y. Falck-Ytter, S. Flottorp, G. H. Guyatt, R. T. Harbour, M. C. Haugh, D. Henry, S. Hill, R. Jaeschke, G. Leng, A. Liberati, N. Magrini, J. Mason, P. Middleton, J. Mrukowicz, D. O'Connell, A. D. Oxman, B. Phillips, H. J. Schunemann, T. Edejer, H. Varonen, G. E. Vist, J. W. Williams, Jr., and S. Zaza. 2004. 'Grading quality of evidence and strength of recommendations', BMJ , 328: 1490.

Atlas, IDF Diabetes. 2016. "International Diabetes Federation 7th Edition, 2015." In.: Nov.

27

Azoulay, L., and S. Suissa. 2017. 'Sulfonylureas and the Risks of Cardiovascular Events and Death: A Methodological Meta-Regression Analysis of the Observational Studies', Diabetes Care , 40: 706-14.

Bailey, C. J., J. L. Gross, A. Pieters, A. Bastien, and J. F. List. 2010. 'Effect of dapagliflozin in patients with type 2 diabetes who have inadequate glycaemic control with metformin: a randomised, double-blind, placebo-controlled trial', Lancet , 375: 2223-33.

Bailey, C. J., N. Iqbal, C. T'Joen, and J. F. List. 2012. 'Dapagliflozin monotherapy in drug-naive patients with diabetes: a randomized-controlled trial of low-dose range', Diabetes Obes Metab , 14: 951-9.

Bailey, Timothy S, Anuj Bhargava, J Hans De Vries, Gregg Gerety, Janusz Gumprecht, Wendy S Lane, Carol H Wysham, Britta Anker Bak, Elise Hachmann-Nielsen, and Athena Philis-Tsimikas. 2017. "Day-to-Day Variability of Fasting Self-Measured Plasma Glucose (SMPG) Correlates with Risk of Hypoglycemia in Adults with Type 1 (T1D) or Type 2 Diabetes (T2D)." In Diabetes , A104-A05. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Bain, S., E. Druyts, C. Balijepalli, C. A. Baxter, C. J. Currie, R. Das, R. Donnelly, K. Khunti, H. Langerman, P. Leigh, G. Siliman, K. Thorlund, K. Toor, J. Vora, and E. J. Mills. 2017. 'Cardiovascular events and all-cause mortality associated with sulphonylureas compared with other antihyperglycaemic drugs: A Bayesian meta-analysis of survival data', Diabetes Obes Metab , 19: 329-35.

Bajaj, H. S., K. Venn, C. Ye, A. Patrick, S. Kalra, H. Khandwala, N. Aslam, D. Twum-Barima, and R. Aronson. 2017. 'Lowest Glucose Variability and Hypoglycemia Are Observed With the Combination of a GLP-1 Receptor Agonist and Basal Insulin (VARIATION Study)', Diabetes Care , 40: 194-200.

Bakris, G. L., V. A. Fonseca, K. Sharma, and E. M. Wright. 2009. 'Renal sodium-glucose transport: role in diabetes mellitus and potential clinical implications', Kidney Int , 75: 1272-7.

Balijepalli, C., E. Druyts, G. Siliman, M. Joffres, K. Thorlund, and E. J. Mills. 2017. 'Hypoglycemia: a review of definitions used in clinical trials evaluating antihyperglycemic drugs for diabetes', Clin Epidemiol , 9: 291-96.

Baranov, O., M. Kahle, C. F. Deacon, J. J. Holst, and M. A. Nauck. 2016. 'Feedback suppression of meal-induced glucagon-like peptide-1 (GLP-1) secretion mediated through elevations in intact GLP-1 caused by dipeptidyl peptidase-4 inhibition: a randomized, prospective comparison of sitagliptin and vildagliptin treatment', Diabetes Obes Metab , 18: 1100-09.

Barlow, P., M. McKee, S. Basu, and D. Stuckler. 2017. 'Impact of the North American Free Trade Agreement on high-fructose corn syrup supply in Canada: a natural experiment using synthetic control methods', CMAJ , 189: E881-E87.

28

Barnett, A. H., S. Patel, R. Harper, R. Toorawa, S. Thiemann, M. von Eynatten, and H. J. Woerle. 2012. 'Linagliptin monotherapy in type 2 diabetes patients for whom metformin is inappropriate: an 18-week randomized, double-blind, placebo-controlled phase III trial with a 34-week active-controlled extension', Diabetes Obes Metab , 14: 1145-54.

Barzilai, N., H. Guo, E. M. Mahoney, S. Caporossi, G. T. Golm, R. B. Langdon, D. Williams- Herman, K. D. Kaufman, J. M. Amatruda, B. J. Goldstein, and H. Steinberg. 2011. 'Efficacy and tolerability of sitagliptin monotherapy in elderly patients with type 2 diabetes: a randomized, double-blind, placebo-controlled trial', Curr Med Res Opin , 27: 1049-58.

Bedenis, R., A. H. Price, C. M. Robertson, J. R. Morling, B. M. Frier, M. W. Strachan, and J. F. Price. 2014. 'Association between severe hypoglycemia, adverse macrovascular events, and inflammation in the Edinburgh Type 2 Diabetes Study', Diabetes Care , 37: 3301-8.

Belalcazar, L. M., and C. M. Ballantyne. 2017. 'Looking Back at Look AHEAD Through the Lens of Recent Diabetes Outcome Trials', Circulation , 135: 720-23.

Bellou, V., L. Belbasis, I. Tzoulaki, and E. Evangelou. 2018. 'Risk factors for type 2 diabetes mellitus: An exposure-wide umbrella review of meta-analyses', PLoS One , 13: e0194127.

Bhaumik, D. K., A. Amatya, S. L. Normand, J. Greenhouse, E. Kaizar, B. Neelon, and R. D. Gibbons. 2012. 'Meta-Analysis of Rare Binary Adverse Event Data', J Am Stat Assoc , 107: 555-67.

Biessels, G. J. 2009. 'Diabetes: hypoglycemia and dementia in type 2 diabetes: chick or egg?', Nat Rev Endocrinol , 5: 532-4.

Biessels, Geert Jan, Mark W. J. Strachan, Frank L. J. Visseren, L. Jaap Kappelle, and Rachel A. Whitmer. 2014. 'Dementia and cognitive decline in type 2 diabetes and prediabetic stages: towards targeted interventions', The Lancet Diabetes & Endocrinology , 2: 246-55.

Birge, S. J. 2008. 'Osteoporotic fractures: a brain or bone disease?', Curr Osteoporos Rep , 6: 57- 61.

Bliss, M. 1982. 'Banting's, Best's, and Collip's accounts of the discovery of insulin', Bull Hist Med , 56: 554-68.

Bohn, B., W. Kerner, J. Seufert, H. P. Kempe, P. M. Jehle, F. Best, M. Fuchtenbusch, A. Knauerhase, M. Hofer, J. Rosenbauer, R. W. Holl, and D. P. V. initiative. 2016. 'Trend of antihyperglycaemic therapy and glycaemic control in 184,864 adults with type 1 or 2 diabetes between 2002 and 2014: Analysis of real-life data from the DPV registry from Germany and Austria', Diabetes Res Clin Pract , 115: 31-8.

Bolinder, J., O. Ljunggren, J. Kullberg, L. Johansson, J. Wilding, A. M. Langkilde, J. Sugg, and S. Parikh. 2012. 'Effects of dapagliflozin on body weight, total fat mass, and regional adipose tissue distribution in patients with type 2 diabetes mellitus with inadequate glycemic control on metformin', J Clin Endocrinol Metab , 97: 1020-31. 29

Bolli, G. B., P. De Feo, S. De Cosmo, G. Perriello, M. M. Ventura, M. M. Benedetti, F. Santeusanio, J. E. Gerich, and P. Brunetti. 1984. 'A reliable and reproducible test for adequate glucose counterregulation in type I diabetes mellitus', Diabetes , 33: 732-7.

Bolli, G. B., M. Munteanu, S. Dotsenko, E. Niemoeller, G. Boka, Y. Wu, and M. Hanefeld. 2014. 'Efficacy and safety of lixisenatide once daily vs. placebo in people with Type 2 diabetes insufficiently controlled on metformin (GetGoal-F1)', Diabet Med , 31: 176-84.

Bonds, D. E., M. E. Miller, R. M. Bergenstal, J. B. Buse, R. P. Byington, J. A. Cutler, R. J. Dudl, F. Ismail-Beigi, A. R. Kimel, B. Hoogwerf, K. R. Horowitz, P. J. Savage, E. R. Seaquist, D. L. Simmons, W. I. Sivitz, J. M. Speril-Hillen, and M. E. Sweeney. 2010. 'The association between symptomatic, severe hypoglycaemia and mortality in type 2 diabetes: retrospective epidemiological analysis of the ACCORD study', BMJ , 340: b4909.

Booth, G. L., J. E. Hux, J. Fang, and B. T. Chan. 2005. 'Time trends and geographic disparities in acute complications of diabetes in Ontario, Canada', Diabetes Care , 28: 1045-50.

Bosi, E., R. P. Camisasca, C. Collober, E. Rochotte, and A. J. Garber. 2007. 'Effects of vildagliptin on glucose control over 24 weeks in patients with type 2 diabetes inadequately controlled with metformin', Diabetes Care , 30: 890-5.

Boucai, L., W. N. Southern, and J. Zonszein. 2011. 'Hypoglycemia-associated mortality is not drug-associated but linked to comorbidities', Am J Med , 124: 1028-35.

Boulin, M., V. Diaby, and C. Tannenbaum. 2016. 'Preventing Unnecessary Costs of Drug- Induced Hypoglycemia in Older Adults with Type 2 Diabetes in the United States and Canada', PLoS One , 11: e0162951.

Bradburn, M. J., J. J. Deeks, J. A. Berlin, and A. Russell Localio. 2007. 'Much ado about nothing: a comparison of the performance of meta-analytical methods with rare events', Stat Med , 26: 53-77.

Brady, V. J. 2017. 'Insulin Therapy: The Old, the New and the Novel-An Overview', Nurs Clin North Am , 52: 539-52.

Braga, M. F., A. Casanova, H. Teoh, H. C. Gerstein, D. H. Fitchett, G. Honos, P. A. McFarlane, E. Ur, J. F. Yale, A. Langer, S. G. Goodman, L. A. Leiter, and Investigators Diabetes Registry to Improve Vascular Events. 2012. 'Poor achievement of guidelines- recommended targets in type 2 diabetes: findings from a contemporary prospective cohort study', Int J Clin Pract , 66: 457-64.

Bremer, J. P., K. Jauch-Chara, M. Hallschmid, S. Schmid, and B. Schultes. 2009. 'Hypoglycemia unawareness in older compared with middle-aged patients with type 2 diabetes', Diabetes Care , 32: 1513-7.

30

Brod, M., T. Christensen, T. L. Thomsen, and D. M. Bushnell. 2011. 'The impact of non-severe hypoglycemic events on work productivity and diabetes management', Value Health , 14: 665-71.

Brod, M., G. Galstyan, A. G. Unnikrishnan, I. Harman-Boehm, V. Prusty, F. Lavalle, M. McGill, A. Murphy, and F. Puchulu. 2016. 'Self-Treated Hypoglycemia in Type 2 Diabetes Mellitus: Results from the Second Wave of an International Cross-Sectional Survey', Diabetes Ther , 7: 279-93.

Brod, M., M. Wolden, T. Christensen, and D. M. Bushnell. 2013a. 'A nine country study of the burden of non-severe nocturnal hypoglycaemic events on diabetes management and daily function', Diabetes Obes Metab , 15: 546-57.

Brod, M., M. Wolden, T. Christensen, and D. M. Bushnell. 2013b. 'Understanding the economic burden of nonsevere nocturnal hypoglycemic events: impact on work productivity, disease management, and resource utilization', Value Health , 16: 1140-9.

Broz, J., M. Brabec, D. Janickova Zdarska, Z. Fedakova, L. Hoskovcova, J. Y. You, V. Donicova, P. Hlado, D. Rahelic, M. Kvapil, and J. Polak. 2015. 'Fear of driving license withdrawal in patients with insulin-treated diabetes mellitus negatively influences their decision to report severe hypoglycemic events to physicians', Patient Prefer Adherence , 9: 1367-70.

Bruce, D. G., W. A. Davis, G. P. Casey, R. M. Clarnette, S. G. Brown, I. G. Jacobs, O. P. Almeida, and T. M. Davis. 2009. 'Severe hypoglycaemia and cognitive impairment in older patients with diabetes: the Fremantle Diabetes Study', Diabetologia , 52: 1808-15.

Budnitz, D. S., M. C. Lovegrove, N. Shehab, and C. L. Richards. 2011. 'Emergency hospitalizations for adverse drug events in older Americans', N Engl J Med , 365: 2002- 12.

Cahn, A., I. Raz, O. Mosenzon, G. Leibowitz, I. Yanuv, A. Rozenberg, N. Iqbal, B. Hirshberg, M. Sjostrand, C. Stahre, K. Im, E. Kanevsky, B. M. Scirica, D. L. Bhatt, and E. Braunwald. 2016. 'Predisposing Factors for Any and Major Hypoglycemia With Saxagliptin Versus Placebo and Overall: Analysis From the SAVOR-TIMI 53 Trial', Diabetes Care , 39: 1329-37.

Campbell, J. M., S. M. Bellman, M. D. Stephenson, and K. Lisy. 2017. 'Metformin reduces all- cause mortality and diseases of ageing independent of its effect on diabetes control: A systematic review and meta-analysis', Ageing Res Rev , 40: 31-44.

Campesi, I., F. Franconi, G. Seghieri, and M. Meloni. 2017. 'Sex-gender-related therapeutic approaches for cardiovascular complications associated with diabetes', Pharmacol Res , 119: 195-207.

Cavender, M, B Scirica, E Braunwald, I Raz, O Mozenson, K Im, A Umez-Eronini, B Hirshberg, PG Steg, and D Bhatt. 2014. "Major hypoglycemia is associated with cardiovascular death and hospitalization for heart failure: findings from SAVOR-TIMI 53." In European 31

Heart Journal , 342-42. OXFORD UNIV PRESS GREAT CLARENDON ST, OXFORD OX2 6DP, ENGLAND.

Cha, S. A., J. S. Yun, T. S. Lim, S. Hwang, E. J. Yim, K. H. Song, K. D. Yoo, Y. M. Park, Y. B. Ahn, and S. H. Ko. 2016. 'Severe Hypoglycemia and Cardiovascular or All-Cause Mortality in Patients with Type 2 Diabetes', Diabetes Metab J , 40: 202-10.

Charbonnel, B., M. Bertolini, F. J. Tinahones, M. P. Domingo, and M. Davies. 2014. 'Lixisenatide plus basal insulin in patients with type 2 diabetes mellitus: a meta-analysis', J Diabetes Complications , 28: 880-6.

Charbonnel, B., A. Karasik, J. Liu, M. Wu, and G. Meininger. 2006. 'Efficacy and safety of the dipeptidyl peptidase-4 inhibitor sitagliptin added to ongoing metformin therapy in patients with type 2 diabetes inadequately controlled with metformin alone', Diabetes Care , 29: 2638-43.

Chatterjee, S., A. Sharma, E. Lichstein, and D. Mukherjee. 2013. 'Intensive glucose control in diabetics with an acute myocardial infarction does not improve mortality and increases risk of hypoglycemia-a meta-regression analysis', Curr Vasc Pharmacol , 11: 100-4.

Chen, C., Q. Yu, S. Zhang, P. Yang, and C. Y. Wang. 2015. 'Assessing the efficacy and safety of combined DPP-4 inhibitor and insulin treatment in patients with type 2 diabetes: a meta- analysis', Int J Clin Exp Pathol , 8: 14141-50.

Chen, M., Y. Liu, J. Jin, and Q. He. 2016. 'The efficacy and safety of dipeptidyl peptidase-4 inhibitors for treatment of type 2 diabetes mellitus patients with severe renal impairment: a meta-analysis', Ren Fail , 38: 581-7.

Chen, Y. J., C. C. Yang, L. C. Huang, L. Chen, and C. M. Hwu. 2015. 'Increasing trend in emergency department visits for hypoglycemia from patients with type 2 diabetes mellitus in Taiwan', Prim Care Diabetes , 9: 490-6.

Chen, Y., G. Ning, C. Wang, Y. Gong, S. Patel, C. Zhang, T. Izumoto, H. J. Woerle, and W. Wang. 2015. 'Efficacy and safety of linagliptin monotherapy in Asian patients with inadequately controlled type 2 diabetes mellitus: A multinational, 24-week, randomized, clinical trial', J Diabetes Investig , 6: 692-8.

Cheng, D., Y. Fei, Y. Liu, J. Li, Y. Chen, X. Wang, and N. Wang. 2014. 'Efficacy and safety of dipeptidyl peptidase-4 inhibitors in type 2 diabetes mellitus patients with moderate to severe renal impairment: a systematic review and meta-analysis', PLoS One , 9: e111543.

Cheng, J., E. Pullenayegum, J. K. Marshall, A. Iorio, and L. Thabane. 2016. 'Impact of including or excluding both-armed zero-event studies on using standard meta-analysis methods for rare event outcome: a simulation study', BMJ Open , 6: e010983.

Cherney, D. Z., and J. A. Udell. 2016. 'Use of Sodium Glucose Cotransporter 2 Inhibitors in the Hands of Cardiologists: With Great Power Comes Great Responsibility', Circulation , 134: 1915-17. 32

Chiasson, J. L., and L. Naditch. 2001. 'The synergistic effect of plus metformin combination therapy in the treatment of type 2 diabetes', Diabetes Care , 24: 989-94.

Cho, N. H., J. E. Shaw, S. Karuranga, Y. Huang, J. D. da Rocha Fernandes, A. W. Ohlrogge, and B. Malanda. 2018. 'IDF Diabetes Atlas: Global estimates of diabetes prevalence for 2017 and projections for 2045', Diabetes Res Clin Pract , 138: 271-81.

Chow, E., A. Bernjak, E. Walkinshaw, A. Lubina-Solomon, J. Freeman, I. A. Macdonald, P. J. Sheridan, and S. R. Heller. 2017. 'Cardiac Autonomic Regulation and Repolarization During Acute Experimental Hypoglycemia in Type 2 Diabetes', Diabetes , 66: 1322-33.

Chow, E., A. Bernjak, S. Williams, R. A. Fawdry, S. Hibbert, J. Freeman, P. J. Sheridan, and S. R. Heller. 2014. 'Risk of cardiac arrhythmias during hypoglycemia in patients with type 2 diabetes and cardiovascular risk', Diabetes , 63: 1738-47.

Chow, L. S., H. Chen, M. E. Miller, S. M. Marcovina, and E. R. Seaquist. 2016. 'Biomarkers associated with severe hypoglycaemia and death in ACCORD', Diabet Med , 33: 1076-83.

Christensen, M. B., S. Calanna, J. J. Holst, T. Vilsboll, and F. K. Knop. 2014. 'Glucose- dependent insulinotropic polypeptide: blood glucose stabilizing effects in patients with type 2 diabetes', J Clin Endocrinol Metab , 99: E418-26.

Christensen, M., L. Vedtofte, J. J. Holst, T. Vilsboll, and F. K. Knop. 2011. 'Glucose-dependent insulinotropic polypeptide: a bifunctional glucose-dependent regulator of glucagon and insulin secretion in humans', Diabetes , 60: 3103-9.

Clemens, K. K., S. Shariff, K. Liu, I. Hramiak, J. L. Mahon, E. McArthur, and A. X. Garg. 2015. 'Trends in Antihyperglycemic Medication Prescriptions and Hypoglycemia in Older Adults: 2002-2013', PLoS One , 10: e0137596.

Cranston, I., J. Lomas, A. Maran, I. Macdonald, and S. A. Amiel. 1994. 'Restoration of hypoglycaemia awareness in patients with long-duration insulin-dependent diabetes', Lancet , 344: 283-7.

Crowshoe, L., D. Dannenbaum, M. Green, R. Henderson, M. N. Hayward, and E. Toth. 2018. 'Type 2 Diabetes and Indigenous Peoples', Can J Diabetes , 42 Suppl 1: S296-s306.

Cryer, P. E. 1992. 'Iatrogenic hypoglycemia as a cause of hypoglycemia-associated autonomic failure in IDDM. A vicious cycle', Diabetes , 41: 255-60.

Cryer, P. E. 2005. 'Mechanisms of hypoglycemia-associated autonomic failure and its component syndromes in diabetes', Diabetes , 54: 3592-601.

Cryer, P. E. 2013. 'Hypoglycemia-associated autonomic failure in diabetes', Handb Clin Neurol , 117: 295-307.

Cryer, P. E., L. Axelrod, A. B. Grossman, S. R. Heller, V. M. Montori, E. R. Seaquist, F. J. Service, and Society Endocrine. 2009. 'Evaluation and management of adult

33

hypoglycemic disorders: an Endocrine Society Clinical Practice Guideline', J Clin Endocrinol Metab , 94: 709-28.

Cukierman, T., H. C. Gerstein, and J. D. Williamson. 2005. 'Cognitive decline and dementia in diabetes--systematic overview of prospective observational studies', Diabetologia , 48: 2460-9.

Currie, C. J., S. E. Holden, S. Jenkins-Jones, C. L. Morgan, B. Voss, S. N. Rajpathak, B. Alemayehu, J. R. Peters, and S. S. Engel. 2018. 'Impact of differing glucose-lowering regimens on the pattern of association between glucose control and survival', Diabetes Obes Metab , 20: 821-30.

Currie, C. J., C. D. Poole, M. Evans, J. R. Peters, and C. L. Morgan. 2013. 'Mortality and other important diabetes-related outcomes with insulin vs other antihyperglycemic therapies in type 2 diabetes', J Clin Endocrinol Metab , 98: 668-77.

Czupryniak, Leszek, Anna Borkowska, and Elektra Szymanska-Garbacz. 2017. "Higher Glycemic Variability Is Associated with Increased Risk of Hypoglycemia in Well or Poorly Controlled Type 1 or Type 2 Diabetes Patients." In Diabetes , A103-A04. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Dagogo-Jack, S., J. Liu, R. Eldor, G. Amorin, J. Johnson, D. Hille, Y. Liao, S. Huyck, G. Golm, S. G. Terra, J. P. Mancuso, S. S. Engel, and B. Lauring. 2018. 'Efficacy and safety of the addition of ertugliflozin in patients with type 2 diabetes mellitus inadequately controlled with metformin and sitagliptin: The VERTIS SITA2 placebo-controlled randomized study', Diabetes Obes Metab , 20: 530-40.

Dagogo-Jack, S., C. Rattarasarn, and P. E. Cryer. 1994. 'Reversal of hypoglycemia unawareness, but not defective glucose counterregulation, in IDDM', Diabetes , 43: 1426-34.

Dailey III, George E, Timothy Reid, John White, Jason Chao, Fang Liz Zhou, Sachin Paranjape, and Paulos Berhanu. 2017. "Improved Glycemic Control and Lower Hypoglycemia Risk with Reduced Prior Oral Antidiabetes Drug (OAD) Therapy in Patients (Pts) with T2D Treated with Insulin Glargine 300 U/mL (Gla-300)." In Diabetes , A251-A51. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Danaei, Goodarz, Mariel M. Finucane, Yuan Lu, Gitanjali M. Singh, Melanie J. Cowan, Christopher J. Paciorek, John K. Lin, Farshad Farzadfar, Young-Ho Khang, Gretchen A. Stevens, Mayuree Rao, Mohammed K. Ali, Leanne M. Riley, Carolyn A. Robinson, and Majid Ezzati. 2011. 'National, regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980: systematic analysis of health examination surveys and epidemiological studies with 370 country-years and 2·7 million participants', The Lancet, 378: 31-40.

Davies, M., T. R. Pieber, M. L. Hartoft-Nielsen, O. K. H. Hansen, S. Jabbour, and J. Rosenstock. 2017. 'Effect of Oral Semaglutide Compared With Placebo and Subcutaneous 34

Semaglutide on Glycemic Control in Patients With Type 2 Diabetes: A Randomized Clinical Trial', JAMA , 318: 1460-70. de Galan, B. E., S. Zoungas, J. Chalmers, C. Anderson, C. Dufouil, A. Pillai, M. Cooper, D. E. Grobbee, M. Hackett, P. Hamet, S. R. Heller, L. Lisheng, S. MacMahon, G. Mancia, B. Neal, C. Y. Pan, A. Patel, N. Poulter, F. Travert, M. Woodward, and Advance Collaborative group. 2009. 'Cognitive function and risks of cardiovascular disease and hypoglycaemia in patients with type 2 diabetes: the Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation (ADVANCE) trial', Diabetologia , 52: 2328-36.

Deacon, C. F. 2004. 'Therapeutic strategies based on glucagon-like peptide 1', Diabetes , 53: 2181-9.

Deeks, J. J. 2002. 'Issues in the selection of a summary statistic for meta-analysis of clinical trials with binary outcomes', Stat Med , 21: 1575-600.

DeFronzo, R. A. 1988. 'Lilly lecture 1987. The triumvirate: beta-cell, muscle, liver. A collusion responsible for NIDDM', Diabetes , 37: 667-87.

Defronzo, R. A. 2009. 'Banting Lecture. From the triumvirate to the ominous octet: a new paradigm for the treatment of type 2 diabetes mellitus', Diabetes , 58: 773-95.

DeFronzo, R. A., P. R. Fleck, C. A. Wilson, and Q. Mekki. 2008. 'Efficacy and safety of the dipeptidyl peptidase-4 inhibitor alogliptin in patients with type 2 diabetes and inadequate glycemic control: a randomized, double-blind, placebo-controlled study', Diabetes Care , 31: 2315-7.

DeFronzo, R. A., and A. M. Goodman. 1995. 'Efficacy of metformin in patients with non- insulin-dependent diabetes mellitus. The Multicenter Metformin Study Group', N Engl J Med , 333: 541-9.

DeFronzo, R. A., M. N. Hissa, A. J. Garber, J. Luiz Gross, R. Yuyan Duan, S. Ravichandran, and R. S. Chen. 2009. 'The efficacy and safety of saxagliptin when added to metformin therapy in patients with inadequately controlled type 2 diabetes with metformin alone', Diabetes Care , 32: 1649-55.

DeFronzo, R. A., R. E. Ratner, J. Han, D. D. Kim, M. S. Fineman, and A. D. Baron. 2005. 'Effects of exenatide (exendin-4) on glycemic control and weight over 30 weeks in metformin-treated patients with type 2 diabetes', Diabetes Care , 28: 1092-100.

Degli Esposti, L., S. Saragoni, S. Buda, and E. Degli Esposti. 2014. 'Clinical outcomes and health care costs combining metformin with sitagliptin or sulphonylureas or thiazolidinediones in uncontrolled type 2 diabetes patients', Clinicoecon Outcomes Res , 6: 463-72.

35

Dejager, S., S. Razac, J. E. Foley, and A. Schweizer. 2007. 'Vildagliptin in drug-naive patients with type 2 diabetes: a 24-week, double-blind, randomized, placebo-controlled, multiple- dose study', Horm Metab Res , 39: 218-23.

Desouza, C., H. Salazar, B. Cheong, J. Murgo, and V. Fonseca. 2003. 'Association of hypoglycemia and cardiac ischemia: a study based on continuous monitoring', Diabetes Care , 26: 1485-9.

Diamond, M. P., T. Jones, S. Caprio, L. Hallarman, M. C. Diamond, M. Addabbo, W. V. Tamborlane, and R. S. Sherwin. 1993. 'Gender influences counterregulatory hormone responses to hypoglycemia', Metabolism , 42: 1568-72.

Donnelly, L. A., A. D. Morris, B. M. Frier, J. D. Ellis, P. T. Donnan, R. Durrant, M. M. Band, G. Reekie, and G. P. Leese. 2005. 'Frequency and predictors of hypoglycaemia in Type 1 and insulin-treated Type 2 diabetes: a population-based study', Diabet Med , 22: 749-55.

Drake, T. C., F. C. Hsu, D. Hire, S. H. Chen, R. M. Cohen, R. McDuffie, E. Nylen, P. O'Connor, S. Rehman, and E. R. Seaquist. 2016. 'Factors associated with failure to achieve a glycated haemoglobin target of <8.0% in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial', Diabetes Obes Metab , 18: 92-5.

Duckworth, W., C. Abraira, T. Moritz, D. Reda, N. Emanuele, P. D. Reaven, F. J. Zieve, J. Marks, S. N. Davis, R. Hayward, S. R. Warren, S. Goldman, M. McCarren, M. E. Vitek, W. G. Henderson, and G. D. Huang. 2009. 'Glucose control and vascular complications in veterans with type 2 diabetes', N Engl J Med , 360: 129-39.

Edridge, C. L., A. J. Dunkley, D. H. Bodicoat, T. C. Rose, L. J. Gray, M. J. Davies, and K. Khunti. 2015. 'Prevalence and Incidence of Hypoglycaemia in 532,542 People with Type 2 Diabetes on Oral Therapies and Insulin: A Systematic Review and Meta-Analysis of Population Based Studies', PLoS One , 10: e0126427.

Elliott, L., C. Fidler, A. Ditchfield, and T. Stissing. 2016. 'Hypoglycemia Event Rates: A Comparison Between Real-World Data and Randomized Controlled Trial Populations in Insulin-Treated Diabetes', Diabetes Ther , 7: 45-60.

Eriksson, J. W., J. Bodegard, D. Nathanson, M. Thuresson, T. Nystrom, and A. Norhammar. 2016. 'Sulphonylurea compared to DPP-4 inhibitors in combination with metformin carries increased risk of severe hypoglycemia, cardiovascular events, and all-cause mortality', Diabetes Res Clin Pract , 117: 39-47.

Evans, M., K. Khunti, M. Mamdani, C. B. Galbo-Jorgensen, J. Gundgaard, M. Bogelund, and S. Harris. 2013. 'Health-related quality of life associated with daytime and nocturnal hypoglycaemic events: a time trade-off survey in five countries', Health Qual Life Outcomes , 11: 90.

Evans, M. L., A. Pernet, J. Lomas, J. Jones, and S. A. Amiel. 2000. 'Delay in onset of awareness of acute hypoglycemia and of restoration of cognitive performance during recovery', Diabetes Care , 23: 893-7. 36

Fadini, G. P., M. Rigato, A. Tiengo, and A. Avogaro. 2009. 'Characteristics and mortality of type 2 diabetic patients hospitalized for severe iatrogenic hypoglycemia', Diabetes Res Clin Pract , 84: 267-72.

Fanelli, C. G., L. Epifano, A. M. Rambotti, S. Pampanelli, A. Di Vincenzo, F. Modarelli, M. Lepore, B. Annibale, M. Ciofetta, P. Bottini, and et al. 1993. 'Meticulous prevention of hypoglycemia normalizes the glycemic thresholds and magnitude of most of neuroendocrine responses to, symptoms of, and cognitive function during hypoglycemia in intensively treated patients with short-term IDDM', Diabetes , 42: 1683-9.

Farngren, J., M. Persson, and B. Ahren. 2016. 'Effect of the GLP-1 Receptor Agonist Lixisenatide on Counterregulatory Responses to Hypoglycemia in Subjects With Insulin- Treated Type 2 Diabetes', Diabetes Care , 39: 242-9.

Feher, M. D., H. Langerman, and M. Evans. 2016. 'Hypoglycemia, diabetes therapies and driving categories in type 2 diabetes', Curr Med Res Opin , 32: 1005-12.

Feinkohl, I., P. P. Aung, M. Keller, C. M. Robertson, J. R. Morling, S. McLachlan, I. J. Deary, B. M. Frier, M. W. Strachan, J. F. Price, and Investigators Edinburgh Type 2 Diabetes Study. 2014. 'Severe hypoglycemia and cognitive decline in older people with type 2 diabetes: the Edinburgh type 2 diabetes study', Diabetes Care , 37: 507-15.

Ferrannini, E., S. J. Ramos, A. Salsali, W. Tang, and J. F. List. 2010. 'Dapagliflozin monotherapy in type 2 diabetic patients with inadequate glycemic control by diet and exercise: a randomized, double-blind, placebo-controlled, phase 3 trial', Diabetes Care , 33: 2217-24.

Ferrannini, E., L. Seman, E. Seewaldt-Becker, S. Hantel, S. Pinnetti, and H. J. Woerle. 2013. 'A Phase IIb, randomized, placebo-controlled study of the SGLT2 inhibitor empagliflozin in patients with type 2 diabetes', Diabetes Obes Metab , 15: 721-8.

Ferrannini, E., and A. Solini. 2012. 'SGLT2 inhibition in diabetes mellitus: rationale and clinical prospects', Nat Rev Endocrinol , 8: 495-502.

Fisman, Enrique Z., Michael Motro, Alexander Tenenbaum, Jonathan Leor, Valentina Boyko, Lori Mandelzweig, Yaniv Sherer, Yehuda Adler, and Solomon Behar. 2016. 'Is hypoglycaemia a marker for increased long-term mortality risk in patients with coronary artery disease? An 8-year follow-up', European Journal of Cardiovascular Prevention & Rehabilitation , 11: 135-43.

Fonseca, V. A., R. Alvarado-Ruiz, D. Raccah, G. Boka, P. Miossec, and J. E. Gerich. 2012. 'Efficacy and safety of the once-daily GLP-1 receptor agonist lixisenatide in monotherapy: a randomized, double-blind, placebo-controlled trial in patients with type 2 diabetes (GetGoal-Mono)', Diabetes Care , 35: 1225-31.

Fonseca, V. A., E. Ferrannini, J. P. Wilding, W. Wilpshaar, P. Dhanjal, G. Ball, and S. Klasen. 2013. 'Active- and placebo-controlled dose-finding study to assess the efficacy, safety,

37

and tolerability of multiple doses of in patients with type 2 diabetes mellitus', J Diabetes Complications , 27: 268-73.

Foos, V., N. Varol, B. H. Curtis, K. S. Boye, D. Grant, J. L. Palmer, and P. McEwan. 2015. 'Economic impact of severe and non-severe hypoglycemia in patients with Type 1 and Type 2 diabetes in the United States', J Med Econ , 18: 420-32.

Forst, T., B. Uhlig-Laske, A. Ring, U. Graefe-Mody, C. Friedrich, K. Herbach, H. J. Woerle, and K. A. Dugi. 2010. 'Linagliptin (BI 1356), a potent and selective DPP-4 inhibitor, is safe and efficacious in combination with metformin in patients with inadequately controlled Type 2 diabetes', Diabet Med , 27: 1409-19.

Freemantle, N., N. Danchin, F. Calvi-Gries, M. Vincent, and P. D. Home. 2016. 'Relationship of glycaemic control and hypoglycaemic episodes to 4-year cardiovascular outcomes in people with type 2 diabetes starting insulin', Diabetes Obes Metab , 18: 152-8.

Friedrich, J. O., N. K. Adhikari, and J. Beyene. 2007. 'Inclusion of zero total event trials in meta- analyses maintains analytic consistency and incorporates all available data', BMC Med Res Methodol , 7: 5.

Frier, B. M. 2014. 'Hypoglycaemia in diabetes mellitus: epidemiology and clinical implications', Nat Rev Endocrinol , 10: 711-22.

Fu, H., W. Xie, B. Curtis, and D. Schuster. 2014. 'Identifying factors associated with hypoglycemia-related hospitalizations among elderly patients with T2DM in the US: a novel approach using influential variable analysis', Curr Med Res Opin , 30: 1787-93.

Fulcher, G., D. R. Matthews, V. Perkovic, D. de Zeeuw, K. W. Mahaffey, C. Mathieu, V. Woo, C. Wysham, G. Capuano, M. Desai, W. Shaw, F. Vercruysse, G. Meininger, B. Neal, and Canvas trial collaborative group. 2016. 'Efficacy and safety of canagliflozin when used in conjunction with incretin-mimetic therapy in patients with type 2 diabetes', Diabetes Obes Metab , 18: 82-91.

Gantz, I., T. Okamoto, Y. Ito, K. Okuyama, E. A. O'Neill, K. D. Kaufman, S. S. Engel, and E. Lai. 2017. 'A randomized, placebo- and sitagliptin-controlled trial of the safety and efficacy of , a once-weekly dipeptidyl peptidase-4 inhibitor, in Japanese patients with type 2 diabetes', Diabetes Obes Metab , 19: 1602-09.

Gao, Xueying, Xiaoling Cai, and Linong Ji. 2017. "Meta-Analysis of the Efficacy and Adverse Effects of a Fixed-Ratio Combination of Basal Insulin and GLP-1RA in T2DM Patients." In Diabetes , A304-A04. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Gautier, J. F., P. Monguillon, O. Verier-Mine, P. Valensi, B. Fiquet, S. Dejager, and B. Charbonnel. 2016. 'Which oral antidiabetic drug to combine with metformin to minimize the risk of hypoglycemia when initiating basal insulin?: A randomized controlled trial of a DPP4 inhibitor versus insulin secretagogues', Diabetes Res Clin Pract , 116: 26-8.

38

Geddes, J., J. E. Schopman, N. N. Zammitt, and B. M. Frier. 2008. 'Prevalence of impaired awareness of hypoglycaemia in adults with Type 1 diabetes', Diabet Med , 25: 501-4.

Geiss, L. S., J. Wang, Y. J. Cheng, T. J. Thompson, L. Barker, Y. Li, A. L. Albright, and E. W. Gregg. 2014. 'Prevalence and incidence trends for diagnosed diabetes among adults aged 20 to 79 years, United States, 1980-2012', JAMA , 312: 1218-26.

Geng, Y., J. C. Lo, L. Brickner, and N. P. Gordon. 2017. 'Racial-Ethnic Differences in Fall Prevalence among Older Women: A Cross-Sectional Survey Study', BMC Geriatr , 17: 65.

George, E., N. Harris, C. Bedford, I. A. Macdonald, C. A. Hardisty, and S. R. Heller. 1995. 'Prolonged but partial impairment of the hypoglycaemic physiological response following short-term hypoglycaemia in normal subjects', Diabetologia , 38: 1183-90.

Gerich, J. E., M. Langlois, C. Noacco, J. H. Karam, and P. H. Forsham. 1973. 'Lack of glucagon response to hypoglycemia in diabetes: evidence for an intrinsic pancreatic alpha cell defect', Science , 182: 171-3.

Gerstein, H. C. 2013. 'Do lifestyle changes reduce serious outcomes in diabetes?', N Engl J Med , 369: 189-90.

Gerstein, H. C., J. Bosch, G. R. Dagenais, R. Diaz, H. Jung, A. P. Maggioni, J. Pogue, J. Probstfield, A. Ramachandran, M. C. Riddle, L. E. Ryden, and S. Yusuf. 2012. 'Basal insulin and cardiovascular and other outcomes in dysglycemia', N Engl J Med , 367: 319- 28.

Gerstein, H. C., and G. H. Werstuck. 2013. 'Dysglycaemia, vasculopenia, and the chronic consequences of diabetes', Lancet Diabetes Endocrinol , 1: 71-8.

Gerstein, Hertzel C., Michael E. Miller, Faramarz Ismail-Beigi, Joe Largay, Charlotte McDonald, Heather A. Lochnan, and Gillian L. Booth. 2014. 'Effects of intensive glycaemic control on ischaemic heart disease: analysis of data from the randomised, controlled ACCORD trial', The Lancet , 384: 1936-41.

Ghezzi, C., D. D. F. Loo, and E. M. Wright. 2018. 'Physiology of renal glucose handling via SGLT1, SGLT2 and GLUT2', Diabetologia , 61: 2087-97.

Ginde, A. A., J. A. Espinola, and C. A. Camargo, Jr. 2008. 'Trends and disparities in U.S. emergency department visits for hypoglycemia, 1993-2005', Diabetes Care , 31: 511-3.

Gitt, A. K., P. Bramlage, C. Binz, M. Krekler, T. Plate, E. Deeg, and D. Tschope. 2012. 'Hypoglycaemia is more frequent in type 2 diabetic patients with co-morbid vascular disease: an analysis of the DiaRegis registry', Eur J Prev Cardiol , 19: 765-72.

Goldenberg, R. M., L. D. Berard, A. Y. Cheng, J. D. Gilbert, S. Verma, V. C. Woo, and J. F. Yale. 2016. 'SGLT2 Inhibitor-associated Diabetic Ketoacidosis: Clinical Review and Recommendations for Prevention and Diagnosis', Clin Ther , 38: 2654-64 e1.

39

Goldstein, B. J., M. N. Feinglos, J. K. Lunceford, J. Johnson, and D. E. Williams-Herman. 2007. 'Effect of initial combination therapy with sitagliptin, a dipeptidyl peptidase-4 inhibitor, and metformin on glycemic control in patients with type 2 diabetes', Diabetes Care , 30: 1979-87.

Goldstein, D., G. Chodick, V. Shalev, B. L. Thorsted, L. Elliott, and A. Karasik. 2016. 'Use of Healthcare Services Following Severe Hypoglycemia in Patients with Diabetes: Analysis of Real-World Data', Diabetes Ther , 7: 295-308.

Gomes, T., D. Martins, M. Tadrous, J. M. Paterson, B. R. Shah, D. N. Juurlink, S. Singh, and M. M. Mamdani. 2016. 'Self-Monitoring of Blood Glucose Levels: Evaluating the Impact of a Policy of Quantity Limits on Test-Strip Use and Costs', Can J Diabetes , 40: 431-35.

Goodman, M., H. Thurston, and J. Penman. 2009. 'Efficacy and tolerability of vildagliptin in patients with type 2 diabetes inadequately controlled with metformin monotherapy', Horm Metab Res , 41: 368-73.

Goto, A., O. A. Arah, M. Goto, Y. Terauchi, and M. Noda. 2013. 'Severe hypoglycaemia and cardiovascular disease: systematic review and meta-analysis with bias analysis', BMJ , 347: f4533.

Goto, A., M. Goto, Y. Terauchi, N. Yamaguchi, and M. Noda. 2016. 'Association Between Severe Hypoglycemia and Cardiovascular Disease Risk in Japanese Patients With Type 2 Diabetes', J Am Heart Assoc , 5: e002875.

Graveling, A. J., I. J. Deary, and B. M. Frier. 2013. 'Acute hypoglycemia impairs executive cognitive function in adults with and without type 1 diabetes', Diabetes Care , 36: 3240-6.

Green, J. B., M. A. Bethel, P. W. Armstrong, J. B. Buse, S. S. Engel, J. Garg, R. Josse, K. D. Kaufman, J. Koglin, S. Korn, J. M. Lachin, D. K. McGuire, M. J. Pencina, E. Standl, P. P. Stein, S. Suryawanshi, F. Van de Werf, E. D. Peterson, R. R. Holman, and Tecos Study Group. 2015. 'Effect of Sitagliptin on Cardiovascular Outcomes in Type 2 Diabetes', N Engl J Med , 373: 232-42.

Greenway, F. L. 2016. 'Severe hypoglycemia in the Look AHEAD Trial', J Diabetes Complications , 30: 935-43.

Griffin, S. J., J. K. Leaver, and G. J. Irving. 2017. 'Impact of metformin on cardiovascular disease: a meta-analysis of randomised trials among people with type 2 diabetes', Diabetologia , 60: 1620-29.

Group, International Diabetes Federation Guideline Development. 2014. 'Global guideline for type 2 diabetes', Diabetes Research and Clinical Practice , 104: 1.

Group, International Hypoglycaemia Study. 2017. 'Glucose concentrations of less than 3.0 mmol/L (54 mg/dL) should be reported in clinical trials: a joint position statement of the American Diabetes Association and the European Association for the Study of Diabetes', Diabetes Care , 40: 155-57. 40

Group, UK Prospective Diabetes Study. 1998a. 'Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34)', The Lancet , 352: 854-65.

Group, UK Prospective Diabetes Study. 1998b. 'Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33)', The Lancet , 352: 837-53.

Grunberger, G., A. Chang, G. Garcia Soria, F. T. Botros, R. Bsharat, and Z. Milicevic. 2012. 'Monotherapy with the once-weekly GLP-1 analogue dulaglutide for 12 weeks in patients with Type 2 diabetes: dose-dependent effects on glycaemic control in a randomized, double-blind, placebo-controlled study', Diabet Med , 29: 1260-7.

Guyatt, G. H., A. D. Oxman, R. Kunz, J. Woodcock, J. Brozek, M. Helfand, P. Alonso-Coello, Y. Falck-Ytter, R. Jaeschke, G. Vist, E. A. Akl, P. N. Post, S. Norris, J. Meerpohl, V. K. Shukla, M. Nasser, and H. J. Schunemann. 2011. 'GRADE guidelines: 8. Rating the quality of evidence--indirectness', J Clin Epidemiol , 64: 1303-10.

Haak, T., T. Meinicke, R. Jones, S. Weber, M. von Eynatten, and H. J. Woerle. 2012. 'Initial combination of linagliptin and metformin improves glycaemic control in type 2 diabetes: a randomized, double-blind, placebo-controlled study', Diabetes Obes Metab , 14: 565-74.

Halimi, S. 2015. 'Severe hypoglycaemia the "tip of the iceberg": an underestimated risk in both type 1 and type 2 diabetic patients', Diabetes Metab , 41: 105-6.

Hamilton, E., W. A. Davis, D. G. Bruce, and T. M. Davis. 2017. 'Influence of Premature Mortality on the Link Between Type 2 Diabetes and Hip Fracture: The Fremantle Diabetes Study', J Clin Endocrinol Metab , 102: 551-59.

Hanefeld, M., B. M. Frier, and F. Pistrosch. 2016. 'Hypoglycemia and Cardiovascular Risk: Is There a Major Link?', Diabetes Care , 39 Suppl 2: S205-9.

Hanefeld, M., G. A. Herman, M. Wu, C. Mickel, M. Sanchez, and P. P. Stein. 2007. 'Once-daily sitagliptin, a dipeptidyl peptidase-4 inhibitor, for the treatment of patients with type 2 diabetes', Curr Med Res Opin , 23: 1329-39.

Haring, H. U., L. Merker, E. Seewaldt-Becker, M. Weimer, T. Meinicke, U. C. Broedl, and H. J. Woerle. 2014. 'Empagliflozin as add-on to metformin in patients with type 2 diabetes: a 24-week, randomized, double-blind, placebo-controlled trial', Diabetes Care , 37: 1650-9.

Harris, S. B., J. M. Ekoe, Y. Zdanowicz, and S. Webster-Bogaert. 2005. 'Glycemic control and morbidity in the Canadian primary care setting (results of the diabetes in Canada evaluation study)', Diabetes Res Clin Pract , 70: 90-7.

Hay, L. C., E. G. Wilmshurst, and G. Fulcher. 2003. 'Unrecognized hypo- and hyperglycemia in well-controlled patients with type 2 diabetes mellitus: the results of continuous glucose monitoring', Diabetes Technol Ther , 5: 19-26.

41

Heald, A. H., S. G. Anderson, G. J. Cortes, V. Cholokova, M. Narajos, A. Khan, G. Donnahey, and M. Livingston. 2018. 'Hypoglycaemia in the over 75s: Understanding the predisposing factors in type 2 diabetes (T2DM)', Prim Care Diabetes , 12: 133-38.

Heinemann, L., G. Freckmann, D. Ehrmann, G. Faber-Heinemann, S. Guerra, D. Waldenmaier, and N. Hermanns. 2018. 'Real-time continuous glucose monitoring in adults with type 1 diabetes and impaired hypoglycaemia awareness or severe hypoglycaemia treated with multiple daily insulin injections (HypoDE): a multicentre, randomised controlled trial', Lancet , 391: 1367-77.

Heise, T., L. Hermanski, L. Nosek, A. Feldman, S. Rasmussen, and H. Haahr. 2012. 'Insulin degludec: four times lower pharmacodynamic variability than insulin glargine under steady-state conditions in type 1 diabetes', Diabetes Obes Metab , 14: 859-64.

Heller, S. R., R. M. Bergenstal, W. B. White, S. Kupfer, G. L. Bakris, W. C. Cushman, C. R. Mehta, S. E. Nissen, C. A. Wilson, F. Zannad, Y. Liu, N. M. Gourlie, C. P. Cannon, and Examine Investigators. 2017. 'Relationship of glycated haemoglobin and reported hypoglycaemia to cardiovascular outcomes in patients with type 2 diabetes and recent acute coronary syndrome events: The EXAMINE trial', Diabetes Obes Metab , 19: 664- 71.

Hepburn, D. A., K. M. MacLeod, A. C. Pell, I. J. Scougal, and B. M. Frier. 1993. 'Frequency and symptoms of hypoglycaemia experienced by patients with type 2 diabetes treated with insulin', Diabet Med , 10: 231-7.

Heyman, S. N., M. Khamaisi, S. Rosen, C. Rosenberger, and Z. Abassi. 2017. 'Potential Hypoxic Renal Injury in Patients With Diabetes on SGLT2 Inhibitors: Caution Regarding Concomitant Use of NSAIDs and Iodinated Contrast Media', Diabetes Care , 40: e40-e41.

Heymsfield, S. B., and T. A. Wadden. 2017. 'Mechanisms, Pathophysiology, and Management of Obesity', N Engl J Med , 376: 254-66.

Higgins, J. P., D. G. Altman, P. C. Gotzsche, P. Juni, D. Moher, A. D. Oxman, J. Savovic, K. F. Schulz, L. Weeks, and J. A. Sterne. 2011. 'The Cochrane Collaboration's tool for assessing risk of bias in randomised trials', BMJ , 343: d5928.

Higgins, JPT, Green S. 2011. 'Cochrane Handbook of Systematic Reviews of Interventions ', The Cochrane Collaboration , version 5.1.0.

Holden, S. E., S. Jenkins-Jones, C. L. Morgan, G. Schernthaner, and C. J. Currie. 2015. 'Glucose-lowering with exogenous insulin monotherapy in type 2 diabetes: dose association with all-cause mortality, cardiovascular events and cancer', Diabetes Obes Metab , 17: 350-62.

Holman, R. R., M. A. Bethel, R. J. Mentz, V. P. Thompson, Y. Lokhnygina, J. B. Buse, J. C. Chan, J. Choi, S. M. Gustavson, N. Iqbal, A. P. Maggioni, S. P. Marso, P. Ohman, N. J. Pagidipati, N. Poulter, A. Ramachandran, B. Zinman, and A. F. Hernandez. 2017. 'Effects

42

of Once-Weekly Exenatide on Cardiovascular Outcomes in Type 2 Diabetes', N Engl J Med , 377: 1228-39.

Holman, R. R., S. K. Paul, M. A. Bethel, D. R. Matthews, and H. A. Neil. 2008. '10-year follow- up of intensive glucose control in type 2 diabetes', N Engl J Med , 359: 1577-89.

Holst, J. J., M. Christensen, A. Lund, J. de Heer, B. Svendsen, U. Kielgast, and F. K. Knop. 2011. 'Regulation of glucagon secretion by incretins', Diabetes Obes Metab , 13 Suppl 1: 89-94.

Holstein, A., T. Wohland, O. M. Patzer, F. Trachte, P. Kovacs, and J. D. Holstein. 2016. 'Accumulation of severe hypoglycemia at weekends and in warm seasons in patients with type 1 diabetes but not with type 2 diabetes', J Diabetes Complications , 30: 1308-14.

Home, P., F. Calvi-Gries, L. Blonde, V. Pilorget, J. Berlingieri, and N. Freemantle. 2018. 'Clinical correlates of hypoglycaemia over 4 years in people with type 2 diabetes starting insulin: An analysis from the CREDIT study', Diabetes Obes Metab , 20: 921-29.

Home, P., R. R. Shankar, I. Gantz, C. Iredale, E. A. O'Neill, L. Jain, A. Pong, S. Suryawanshi, S. S. Engel, K. D. Kaufman, and E. Lai. 2018. 'A randomized, double-blind trial evaluating the efficacy and safety of monotherapy with the once-weekly dipeptidyl peptidase-4 inhibitor omarigliptin in people with type 2 diabetes', Diabetes Res Clin Pract , 138: 253- 61.

Hompesch, M., A. Jones-Leone, M. C. Carr, J. Matthews, H. Zhi, M. Young, L. Morrow, and R. R. Reinhardt. 2015. 'Albiglutide does not impair the counter-regulatory hormone response to hypoglycaemia: a randomized, double-blind, placebo-controlled, stepped glucose clamp study in subjects with type 2 diabetes mellitus', Diabetes Obes Metab , 17: 82-90.

Hong, S., C. Y. Park, K. A. Han, C. H. Chung, B. J. Ku, H. C. Jang, C. W. Ahn, M. K. Lee, M. K. Moon, H. S. Son, C. B. Lee, Y. W. Cho, and S. W. Park. 2016. 'Efficacy and safety of teneligliptin, a novel dipeptidyl peptidase-4 inhibitor, in Korean patients with type 2 diabetes mellitus: a 24-week multicentre, randomized, double-blind, placebo-controlled phase III trial', Diabetes Obes Metab , 18: 528-32.

Horowitz, M., T. Wu, A. M. Deane, K. L. Jones, and C. K. Rayner. 2016. 'DPP-4 Inhibition and the Known Unknown', Diabetes , 65: 2124-6.

Howse, P. M., L. N. Chibrikova, L. K. Twells, B. J. Barrett, and J. M. Gamble. 2016. 'Safety and Efficacy of Incretin-Based Therapies in Patients With Type 2 Diabetes Mellitus and CKD: A Systematic Review and Meta-analysis', Am J Kidney Dis , 68: 733-42.

Htike, Z. Z., F. Zaccardi, D. Papamargaritis, D. R. Webb, K. Khunti, and M. J. Davies. 2017. 'Efficacy and safety of glucagon-like peptide-1 receptor agonists in type 2 diabetes: A systematic review and mixed-treatment comparison analysis', Diabetes Obes Metab , 19: 524-36.

43

Iglay, K., Y. Qiu, C. P. Steve Fan, Z. Li, J. Tang, and P. Laires. 2016. 'Risk factors associated with treatment discontinuation and down-titration in type 2 diabetes patients treated with sulfonylureas', Curr Med Res Opin , 32: 1567-75.

Ikeda, S., Y. Takano, O. Cynshi, R. Tanaka, A. D. Christ, V. Boerlin, U. Beyer, A. Beck, C. Ciorciaro, M. Meyer, and T. Kadowaki. 2015. 'A novel and selective sodium-glucose cotransporter-2 inhibitor, , improves glycaemic control and lowers body weight in patients with type 2 diabetes mellitus', Diabetes Obes Metab , 17: 984-93.

Imran, S. A., G. Agarwal, H. S. Bajaj, and S. Ross. 2018. 'Targets for Glycemic Control', Can J Diabetes , 42 Suppl 1: S42-s46.

Inagaki, N., K. Kondo, T. Yoshinari, N. Maruyama, Y. Susuta, and H. Kuki. 2013. 'Efficacy and safety of canagliflozin in Japanese patients with type 2 diabetes: a randomized, double- blind, placebo-controlled, 12-week study', Diabetes Obes Metab , 15: 1136-45.

Inagaki, N., K. Kondo, T. Yoshinari, N. Takahashi, Y. Susuta, and H. Kuki. 2014. 'Efficacy and safety of canagliflozin monotherapy in Japanese patients with type 2 diabetes inadequately controlled with diet and exercise: a 24-week, randomized, double-blind, placebo-controlled, Phase III study', Expert Opin Pharmacother , 15: 1501-15.

Inagaki, N., H. Onouchi, H. Maezawa, S. Kuroda, and K. Kaku. 2015. 'Once-weekly versus daily alogliptin in Japanese patients with type 2 diabetes: a randomised, double- blind, phase 3, non-inferiority study', Lancet Diabetes Endocrinol , 3: 191-7.

Inagaki, N., H. Onouchi, H. Sano, N. Funao, S. Kuroda, and K. Kaku. 2014. 'SYR-472, a novel once-weekly dipeptidyl peptidase-4 (DPP-4) inhibitor, in type 2 diabetes mellitus: a phase 2, randomised, double-blind, placebo-controlled trial', Lancet Diabetes Endocrinol , 2: 125-32.

International Hypoglycaemia Study, Group. 2017. 'Glucose Concentrations of Less Than 3.0 mmol/L (54 mg/dL) Should Be Reported in Clinical Trials: A Joint Position Statement of the American Diabetes Association and the European Association for the Study of Diabetes', Diabetes Care , 40: 155-57.

Investigators, ORIGIN Trial. 2014. 'Predictors of nonsevere and severe hypoglycemia during glucose-lowering treatment with insulin glargine or standard drugs in the ORIGIN trial', Diabetes Care: DC_141329.

Investigators, Origin Trial, L. G. Mellbin, L. Ryden, M. C. Riddle, J. Probstfield, J. Rosenstock, R. Diaz, S. Yusuf, and H. C. Gerstein. 2013. 'Does hypoglycaemia increase the risk of cardiovascular events? A report from the ORIGIN trial', Eur Heart J , 34: 3137-44.

Iqbal, A., and S. Heller. 2016. 'Managing hypoglycaemia', Best Pract Res Clin Endocrinol Metab , 30: 413-30.

Iqbal, Ahmed, Mark Thomas, Peter Novodvorsky, Alan Bernjak, Lewis Birch, Danielle Lambert, Linda Kay, Fiona Wright, Lynne Prince, and Sheila Francis. 2017. "Effect of 44

Hypoglycemia on Platelet Reactivity and Platelet-Leukocyte Kinetics." In Diabetes , A41- A41. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Ito, H., E. Tsugami, S. Ando, A. Imai, S. Matsumoto, T. Omoto, M. Shinozaki, S. Nishio, M. Abe, S. Antoku, M. Mifune, and M. Togane. 2016. 'Secular Trends in the Clinical Characteristics of Type 2 Diabetic Patients With Severe Hypoglycemia Between 2008 and 2013', J Clin Med Res , 8: 710-4.

Iwamoto, Y., T. Taniguchi, K. Nonaka, T. Okamoto, K. Okuyama, J. C. Arjona Ferreira, and J. Amatruda. 2010. 'Dose-ranging efficacy of sitagliptin, a dipeptidyl peptidase-4 inhibitor, in Japanese patients with type 2 diabetes mellitus', Endocr J , 57: 383-94.

Jaacks, L. M., K. R. Siegel, U. P. Gujral, and K. M. Narayan. 2016. 'Type 2 diabetes: A 21st century epidemic', Best Pract Res Clin Endocrinol Metab , 30: 331-43.

Jabbour, S. A., E. Hardy, J. Sugg, and S. Parikh. 2014. 'Dapagliflozin is effective as add-on therapy to sitagliptin with or without metformin: a 24-week, multicenter, randomized, double-blind, placebo-controlled study', Diabetes Care , 37: 740-50.

Janghorbani, M., R. M. Van Dam, W. C. Willett, and F. B. Hu. 2007. 'Systematic review of type 1 and type 2 diabetes mellitus and risk of fracture', Am J Epidemiol , 166: 495-505.

Ji, L., P. Han, X. Wang, J. Liu, S. Zheng, Y. M. Jou, E. A. O'Neill, G. T. Golm, S. S. Engel, K. D. Kaufman, and R. R. Shankar. 2016. 'Randomized clinical trial of the safety and efficacy of sitagliptin and metformin co-administered to Chinese patients with type 2 diabetes mellitus', J Diabetes Investig , 7: 727-36.

Ji, L., L. Li, J. Kuang, T. Yang, D. J. Kim, A. A. Kadir, C. N. Huang, and D. Lee. 2017. 'Efficacy and safety of fixed-dose combination therapy, alogliptin plus metformin, in Asian patients with type 2 diabetes: A phase 3 trial', Diabetes Obes Metab , 19: 754-58.

Ji, L., J. Ma, H. Li, T. A. Mansfield, L. T'Joen C, N. Iqbal, A. Ptaszynska, and J. F. List. 2014. 'Dapagliflozin as monotherapy in drug-naive Asian patients with type 2 diabetes mellitus: a randomized, blinded, prospective phase III study', Clin Ther , 36: 84-100.e9.

Johnston, S. S., C. Conner, M. Aagren, K. Ruiz, and J. Bouchard. 2012. 'Association between hypoglycaemic events and fall-related fractures in Medicare-covered patients with type 2 diabetes', Diabetes Obes Metab , 14: 634-43.

Johnston, S. S., C. Conner, M. Aagren, D. M. Smith, J. Bouchard, and J. Brett. 2011. 'Evidence linking hypoglycemic events to an increased risk of acute cardiovascular events in patients with type 2 diabetes', Diabetes Care , 34: 1164-70.

Jones, T. W., P. Porter, R. S. Sherwin, E. A. Davis, P. O'Leary, F. Frazer, G. Byrne, S. Stick, and W. V. Tamborlane. 1998. 'Decreased epinephrine responses to hypoglycemia during sleep', N Engl J Med , 338: 1657-62.

45

Joy, N. G., M. Mikeladze, L. M. Younk, D. B. Tate, and S. N. Davis. 2016. 'Effects of equivalent sympathetic activation during hypoglycemia on endothelial function and pro- atherothrombotic balance in healthy individuals and obese standard treated type 2 diabetes', Metabolism , 65: 1695-705.

Joy, N. G., D. B. Tate, L. M. Younk, and S. N. Davis. 2015. 'Effects of Acute and Antecedent Hypoglycemia on Endothelial Function and Markers of Atherothrombotic Balance in Healthy Humans', Diabetes , 64: 2571-80.

Jung, C. H., C. Y. Park, K. J. Ahn, N. H. Kim, H. C. Jang, M. K. Lee, J. Y. Park, C. H. Chung, K. W. Min, Y. A. Sung, J. H. Park, S. J. Kim, H. J. Lee, and S. W. Park. 2015. 'A randomized, double-blind, placebo-controlled, phase II clinical trial to investigate the efficacy and safety of oral DA-1229 in patients with type 2 diabetes mellitus who have inadequate glycaemic control with diet and exercise', Diabetes Metab Res Rev , 31: 295- 306.

Kachroo, S., H. Kawabata, S. Colilla, L. Shi, Y. Zhao, J. Mukherjee, U. Iloeje, and V. Fonseca. 2015. 'Association between hypoglycemia and fall-related events in type 2 diabetes mellitus: analysis of a U.S. commercial database', J Manag Care Spec Pharm , 21: 243- 53.

Kadowaki, T., M. Haneda, N. Inagaki, Y. Terauchi, A. Taniguchi, K. Koiwai, H. Rattunde, H. J. Woerle, and U. C. Broedl. 2014. 'Empagliflozin monotherapy in Japanese patients with type 2 diabetes mellitus: a randomized, 12-week, double-blind, placebo-controlled, phase II trial', Adv Ther , 31: 621-38.

Kadowaki, T., N. Inagaki, K. Kondo, K. Nishimura, G. Kaneko, N. Maruyama, N. Nakanishi, M. Gouda, H. Iijima, and Y. Watanabe. 2018. 'Efficacy and safety of teneligliptin added to canagliflozin monotherapy in Japanese patients with type 2 diabetes mellitus: A multicentre, randomized, double-blind, placebo-controlled, parallel-group comparative study', Diabetes Obes Metab , 20: 453-57.

Kadowaki, T., N. Inagaki, K. Kondo, K. Nishimura, G. Kaneko, N. Maruyama, N. Nakanishi, H. Iijima, Y. Watanabe, and M. Gouda. 2017. 'Efficacy and safety of canagliflozin as add-on therapy to teneligliptin in Japanese patients with type 2 diabetes mellitus: Results of a 24- week, randomized, double-blind, placebo-controlled trial', Diabetes Obes Metab , 19: 874-82.

Kadowaki, T., and K. Kondo. 2013. 'Efficacy, safety and dose-response relationship of teneligliptin, a dipeptidyl peptidase-4 inhibitor, in Japanese patients with type 2 diabetes mellitus', Diabetes Obes Metab , 15: 810-8.

Kadowaki, T., N. Tajima, M. Odawara, M. Nishii, T. Taniguchi, and J. C. Ferreira. 2013. 'Addition of sitagliptin to ongoing metformin monotherapy improves glycemic control in Japanese patients with type 2 diabetes over 52 weeks', J Diabetes Investig , 4: 174-81.

Kahn, S. E., S. M. Haffner, M. A. Heise, W. H. Herman, R. R. Holman, N. P. Jones, B. G. Kravitz, J. M. Lachin, M. C. O'Neill, B. Zinman, and G. Viberti. 2006. 'Glycemic 46

durability of rosiglitazone, metformin, or glyburide monotherapy', N Engl J Med , 355: 2427-43.

Kajiwara, A., A. Kita, J. Saruwatari, K. Oniki, K. Morita, M. Yamamura, M. Murase, H. Koda, S. Hirota, T. Ishizuka, and K. Nakagawa. 2015. 'Higher risk of sulfonylurea-associated hypoglycemic symptoms in women with type 2 diabetes mellitus', Clin Drug Investig , 35: 593-600.

Kaku, K., S. Inoue, O. Matsuoka, A. Kiyosue, H. Azuma, N. Hayashi, T. Tokudome, A. M. Langkilde, and S. Parikh. 2013. 'Efficacy and safety of dapagliflozin as a monotherapy for type 2 diabetes mellitus in Japanese patients with inadequate glycaemic control: a phase II multicentre, randomized, double-blind, placebo-controlled trial', Diabetes Obes Metab , 15: 432-40.

Kaku, K., A. Kiyosue, S. Inoue, N. Ueda, T. Tokudome, J. Yang, and A. M. Langkilde. 2014. 'Efficacy and safety of dapagliflozin monotherapy in Japanese patients with type 2 diabetes inadequately controlled by diet and exercise', Diabetes Obes Metab , 16: 1102- 10.

Kaku, K., H. Watada, Y. Iwamoto, K. Utsunomiya, Y. Terauchi, K. Tobe, Y. Tanizawa, E. Araki, M. Ueda, H. Suganami, and D. Watanabe. 2014. 'Efficacy and safety of monotherapy with the novel sodium/glucose cotransporter-2 inhibitor tofogliflozin in Japanese patients with type 2 diabetes mellitus: a combined Phase 2 and 3 randomized, placebo-controlled, double-blind, parallel-group comparative study', Cardiovasc Diabetol , 13: 65.

Kannel, W. B., and D. L. McGee. 1979. 'Diabetes and cardiovascular disease. The Framingham study', JAMA , 241: 2035-8.

Karter, A. J., K. J. Lipska, P. J. O'Connor, J. Y. Liu, H. H. Moffet, E. B. Schroeder, J. M. Lawrence, G. A. Nichols, K. M. Newton, R. D. Pathak, J. Desai, B. Waitzfelder, M. G. Butler, A. Thomas, and J. F. Steiner. 2017. 'High rates of severe hypoglycemia among African American patients with diabetes: the surveillance, prevention, and Management of Diabetes Mellitus (SUPREME-DM) network', J Diabetes Complications , 31: 869-73.

Kashiwagi, A., K. Kazuta, K. Goto, S. Yoshida, E. Ueyama, and A. Utsuno. 2015. 'Ipragliflozin in combination with metformin for the treatment of Japanese patients with type 2 diabetes: ILLUMINATE, a randomized, double-blind, placebo-controlled study', Diabetes Obes Metab , 17: 304-8.

Kashiwagi, A., K. Kazuta, S. Yoshida, and I. Nagase. 2014. 'Randomized, placebo-controlled, double-blind glycemic control trial of novel sodium-dependent glucose cotransporter 2 inhibitor ipragliflozin in Japanese patients with type 2 diabetes mellitus', J Diabetes Investig , 5: 382-91.

Katon, W. J., B. A. Young, J. Russo, E. H. Lin, P. Ciechanowski, E. J. Ludman, and M. R. Von Korff. 2013. 'Association of depression with increased risk of severe hypoglycemic episodes in patients with diabetes', Ann Fam Med , 11: 245-50. 47

Kautzky-Willer, A., J. Harreiter, and G. Pacini. 2016. 'Sex and Gender Differences in Risk, Pathophysiology and Complications of Type 2 Diabetes Mellitus', Endocr Rev , 37: 278- 316.

Kawalec, Paweł, Alicja Mikrut, and Sylwia Łopuch. 2014. 'The safety of dipeptidyl peptidase-4 (DPP-4) inhibitors or sodium-glucose cotransporter 2 (SGLT-2) inhibitors added to metformin background therapy in patients with type 2 diabetes mellitus: a systematic review and meta-analysis', Diabetes/Metabolism Research and Reviews , 30: 269-83.

Kawamori, R., N. Inagaki, E. Araki, H. Watada, N. Hayashi, Y. Horie, A. Sarashina, Y. Gong, M. von Eynatten, H. J. Woerle, and K. A. Dugi. 2012. 'Linagliptin monotherapy provides superior glycaemic control versus placebo or with comparable safety in Japanese patients with type 2 diabetes: a randomized, placebo and active comparator- controlled, double-blind study', Diabetes Obes Metab , 14: 348-57.

Kenny, C. 2014. 'When hypoglycemia is not obvious: diagnosing and treating under-recognized and undisclosed hypoglycemia', Prim Care Diabetes , 8: 3-11.

Khalangot, M., G. Hu, M. Tronko, V. Kravchenko, and V. Guryanov. 2009. 'Gender risk of nonfatal stroke in type 2 diabetic patients differs depending on the type of treatment', J Womens Health (Larchmt) , 18: 97-103.

Khan, T. A., and J. L. Sievenpiper. 2016. 'Controversies about sugars: results from systematic reviews and meta-analyses on obesity, cardiometabolic disease and diabetes', Eur J Nutr , 55: 25-43.

Khunti, K., S. Alsifri, R. Aronson, M. Cigrovski Berkovic, C. Enters-Weijnen, T. Forsen, G. Galstyan, P. Geelhoed-Duijvestijn, M. Goldfracht, H. Gydesen, R. Kapur, N. Lalic, B. Ludvik, E. Moberg, U. Pedersen-Bjergaard, and A. Ramachandran. 2016. 'Rates and predictors of hypoglycaemia in 27 585 people from 24 countries with insulin-treated type 1 and type 2 diabetes: the global HAT study', Diabetes Obes Metab , 18: 907-15.

Khunti, K., M. Davies, A. Majeed, B. L. Thorsted, M. L. Wolden, and S. K. Paul. 2015. 'Hypoglycemia and risk of cardiovascular disease and all-cause mortality in insulin- treated people with type 1 and type 2 diabetes: a cohort study', Diabetes Care , 38: 316- 22.

Kikuchi, M., N. Abe, M. Kato, S. Terao, N. Mimori, and H. Tachibana. 2009. 'Vildagliptin dose- dependently improves glycemic control in Japanese patients with type 2 diabetes mellitus', Diabetes Res Clin Pract , 83: 233-40.

Kim, D., L. MacConell, D. Zhuang, P. A. Kothare, M. Trautmann, M. Fineman, and K. Taylor. 2007. 'Effects of once-weekly dosing of a long-acting release formulation of exenatide on glucose control and body weight in subjects with type 2 diabetes', Diabetes Care , 30: 1487-93.

48

Kim, J. T., T. J. Oh, Y. A. Lee, J. H. Bae, H. J. Kim, H. S. Jung, Y. M. Cho, K. S. Park, S. Lim, H. C. Jang, and H. K. Lee. 2011. 'Increasing trend in the number of severe hypoglycemia patients in Korea', Diabetes Metab J , 35: 166-72.

Kong, A. P., X. Yang, A. Luk, R. C. Ma, W. Y. So, R. Ozaki, R. Ting, K. Cheung, C. S. Ho, M. H. Chan, C. C. Chow, and J. C. Chan. 2014. 'Severe hypoglycemia identifies vulnerable patients with type 2 diabetes at risk for premature death and all-site cancer: the Hong Kong diabetes registry', Diabetes Care , 37: 1024-31.

Kornelius, Edy, Yi Sun Yang, Chiung Huei Peng, Shih Chang Lo, Yung Rung Lai, Jeng Yuan Chiou, and Chien Ning Huang. 2017. "Progress of Diabetes Severity Associated with Higher Risk of Severe Hypoglycemia." In Diabetes , A106-A06. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Kosiborod, M., M. A. Cavender, A. Z. Fu, J. P. Wilding, K. Khunti, R. W. Holl, A. Norhammar, K. I. Birkeland, M. E. Jorgensen, M. Thuresson, N. Arya, J. Bodegard, N. Hammar, P. Fenici, Cvd-Real Investigators, and Group Study. 2017. 'Lower Risk of Heart Failure and Death in Patients Initiated on Sodium-Glucose Cotransporter-2 Inhibitors Versus Other Glucose-Lowering Drugs: The CVD-REAL Study (Comparative Effectiveness of Cardiovascular Outcomes in New Users of Sodium-Glucose Cotransporter-2 Inhibitors)', Circulation , 136: 249-59.

Kosiborod, M., S. E. Inzucchi, A. Goyal, H. M. Krumholz, F. A. Masoudi, L. Xiao, and J. A. Spertus. 2009. 'Relationship between spontaneous and iatrogenic hypoglycemia and mortality in patients hospitalized with acute myocardial infarction', JAMA , 301: 1556-64.

Kramer, C. K., B. Zinman, H. Choi, P. W. Connelly, and R. Retnakaran. 2017. 'Chronic liraglutide therapy induces an enhanced endogenous glucagon-like peptide-1 secretory response in early type 2 diabetes', Diabetes Obes Metab , 19: 744-48.

Kristensen, J., U. M. Mortensen, M. Schmidt, P. H. Nielsen, T. T. Nielsen, and M. Maeng. 2009. 'Lack of cardioprotection from subcutaneously and preischemic administered liraglutide in a closed chest porcine ischemia reperfusion model', BMC Cardiovasc Disord , 9: 31.

Kumar, KM Prasanna, Sunil M Jain, Conrad Tou, and Kajs-Marie Schützer. 2014. 'Saxagliptin as initial therapy in treatment-naive Indian adults with type 2 diabetes mellitus inadequately controlled with diet and exercise alone: a randomized, double-blind, placebo-controlled, phase IIIb clinical study', International Journal of Diabetes in Developing Countries , 34: 201-09.

Kuss, O. 2015. 'Statistical methods for meta-analyses including information from studies without any events-add nothing to nothing and succeed nevertheless', Stat Med , 34: 1097-116.

Laires, P. A., J. Conceicao, F. Araujo, J. Dores, C. Silva, L. Radican, and A. Nogueira. 2016. 'The cost of managing severe hypoglycemic episodes in Type 2 diabetic patients', Expert Rev Pharmacoecon Outcomes Res , 16: 315-25.

49

Lawton, J., D. Rankin, J. Elliott, S. R. Heller, H. A. Rogers, N. De Zoysa, S. Amiel, and U. K. Nihr Dafne Study Group. 2014. 'Experiences, views, and support needs of family members of people with hypoglycemia unawareness: interview study', Diabetes Care , 37: 109-15.

Le Feuvre, C., S. Jacqueminet, and O. Barthelemy. 2011. 'Myocardial ischemia: a silent epidemic in Type 2 diabetes patients', Future Cardiol , 7: 183-90.

Lee, Alexandra K, A Richey Sharrett, Beverly G Windham, Elbert S Huang, Karen Bandeen- Roche, and Elizabeth Selvin. 2017. "Is Frailty a Risk Factor for Severe Hypoglycemia in Older Adults? The Atherosclerosis Risk in Communities (ARIC) Study." In Diabetes , A104-A04. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Lee, C., L. Joseph, A. Colosimo, and K. Dasgupta. 2012. 'Mortality in diabetes compared with previous cardiovascular disease: a gender-specific meta-analysis', Diabetes Metab , 38: 420-7.

Lee, S. M., D. Koh, W. K. Chui, and C. F. Sum. 2011. 'Diabetes management and hypoglycemia in safety sensitive jobs', Saf Health Work , 2: 9-16.

Leiter, L. A., L. Berard, C. K. Bowering, A. Y. Cheng, K. G. Dawson, J. M. Ekoe, C. Fournier, L. Goldin, S. B. Harris, P. Lin, T. Ransom, M. Tan, H. Teoh, R. T. Tsuyuki, D. Whitham, V. Woo, J. F. Yale, and A. Langer. 2013. 'Type 2 diabetes mellitus management in Canada: is it improving?', Can J Diabetes , 37: 82-9.

Leiter, L. A., D. Boras, and V. C. Woo. 2015. 'Dosing irregularities and self-treated hypoglycemia in type 2 diabetes: results from the Canadian Cohort of an International Survey of Patients and Healthcare Professionals', Can J Diabetes , 39 Suppl 4: 19-25.

Leiter, Lawrence, Jean-Francois Yale, J. L. Chiasson, Sharmen Harris, P. Kleinstiver, and L. Sauriol. 2005. Assessment of the impact of fear of hypoglycemic episodes on glycemic and hypoglycemic management .

Leonard, C. E., S. Hennessy, X. Han, D. S. Siscovick, J. H. Flory, and R. Deo. 2017. 'Pro- and Antiarrhythmic Actions of Sulfonylureas: Mechanistic and Clinical Evidence', Trends Endocrinol Metab , 28: 561-86.

Lipscombe, L., G. Booth, S. Butalia, K. Dasgupta, D. T. Eurich, R. Goldenberg, N. Khan, L. MacCallum, B. R. Shah, and S. Simpson. 2018. 'Pharmacologic Glycemic Management of Type 2 Diabetes in Adults', Can J Diabetes , 42 Suppl 1: S88-s103.

Lipska, K. J., J. S. Ross, Y. Wang, S. E. Inzucchi, K. Minges, A. J. Karter, E. S. Huang, M. M. Desai, T. M. Gill, and H. M. Krumholz. 2014. 'National trends in US hospital admissions for hyperglycemia and hypoglycemia among Medicare beneficiaries, 1999 to 2011', JAMA Intern Med , 174: 1116-24.

50

Lipska, K. J., E. M. Warton, E. S. Huang, H. H. Moffet, S. E. Inzucchi, H. M. Krumholz, and A. J. Karter. 2013. 'HbA1c and risk of severe hypoglycemia in type 2 diabetes: the Diabetes and Aging Study', Diabetes Care , 36: 3535-42.

Lipska, K. J., X. Yao, J. Herrin, R. G. McCoy, J. S. Ross, M. A. Steinman, S. E. Inzucchi, T. M. Gill, H. M. Krumholz, and N. D. Shah. 2017. 'Trends in Drug Utilization, Glycemic Control, and Rates of Severe Hypoglycemia, 2006-2013', Diabetes Care , 40: 468-75.

LIU, JINAN, YUEXIN TANG, HAKIMA HANNACHI, SAMUEL S ENGEL, and SWAPNIL RAJPATHAK. 2018. "Impact of Switching from Sulfonylureas to Dipeptidyl Peptidase-4 Inhibitors on Hypoglycemia Burden in the United States—A Predictive Modeling Approach." In.: Am Diabetes Assoc.

Liu, S. C., Y. K. Tu, M. N. Chien, and K. L. Chien. 2012. 'Effect of antidiabetic agents added to metformin on glycaemic control, hypoglycaemia and weight change in patients with type 2 diabetes: a network meta-analysis', Diabetes Obes Metab , 14: 810-20.

Liu, X., Q. Xiao, L. Zhang, Q. Yang, X. Liu, L. Xu, and W. Cheng. 2014. 'The long-term efficacy and safety of DPP-IV inhibitors monotherapy and in combination with metformin in 18,980 patients with type-2 diabetes mellitus--a meta-analysis', Pharmacoepidemiol Drug Saf , 23: 687-98.

Lu, C. H., K. W. Min, L. M. Chuang, S. Kokubo, S. Yoshida, and B. S. Cha. 2016. 'Efficacy, safety, and tolerability of ipragliflozin in Asian patients with type 2 diabetes mellitus and inadequate glycemic control with metformin: Results of a phase 3 randomized, placebo- controlled, double-blind, multicenter trial', J Diabetes Investig , 7: 366-73.

Lu, C. L., P. C. Hsu, H. N. Shen, Y. H. Chang, H. F. Chen, and C. Y. Li. 2015. 'Association Between History of Severe Hypoglycemia and Risk of Falls in Younger and Older Patients With Diabetes', Medicine (Baltimore) , 94: e1339.

Ludvik, B., J. P. Frias, F. J. Tinahones, J. Wainstein, H. Jiang, K. E. Robertson, L. E. Garcia- Perez, D. B. Woodward, and Z. Milicevic. 2018. 'Dulaglutide as add-on therapy to SGLT2 inhibitors in patients with inadequately controlled type 2 diabetes (AWARD-10): a 24-week, randomised, double-blind, placebo-controlled trial', Lancet Diabetes Endocrinol , 6: 370-81.

Lundh, A., J. Lexchin, B. Mintzes, J. B. Schroll, and L. Bero. 2018. 'Industry sponsorship and research outcome: systematic review with meta-analysis', Intensive Care Med , 44: 1603- 12.

Luo, J., S. J. Jacober, M. J. Prince, M. A. Carey, and Y. Qu. 2013. 'The effect of adjusting for baseline hypoglycemia when analyzing hypoglycemia data: a systematic analysis of 15 diabetes clinical trials', Diabetes Technol Ther , 15: 654-61.

Madsbad, S., O. Schmitz, J. Ranstam, G. Jakobsen, and D. R. Matthews. 2004. 'Improved glycemic control with no weight increase in patients with type 2 diabetes after once-daily

51

treatment with the long-acting glucagon-like peptide 1 analog liraglutide (NN2211): a 12- week, double-blind, randomized, controlled trial', Diabetes Care , 27: 1335-42.

Maiorino, M. I., P. Chiodini, G. Bellastella, A. Capuano, K. Esposito, and D. Giugliano. 2017. 'Insulin and Glucagon-Like Peptide 1 Receptor Agonist Combination Therapy in Type 2 Diabetes: A Systematic Review and Meta-analysis of Randomized Controlled Trials', Diabetes Care , 40: 614-24.

Majumdar, S. R., B. R. Hemmelgarn, M. Lin, K. McBrien, B. J. Manns, and M. Tonelli. 2013. 'Hypoglycemia associated with hospitalization and adverse events in older people: population-based cohort study', Diabetes Care , 36: 3585-90.

Malanda, U. L., L. M. Welschen, Riphagen, II, J. M. Dekker, G. Nijpels, and S. D. Bot. 2012. 'Self-monitoring of blood glucose in patients with type 2 diabetes mellitus who are not using insulin', Cochrane Database Syst Rev , 1: Cd005060.

Malmgren, S., and B. Ahren. 2015. 'DPP-4 inhibition contributes to the prevention of hypoglycaemia through a GIP-glucagon counterregulatory axis in mice', Diabetologia , 58: 1091-9.

Marrett, E., L. Radican, M. J. Davies, and Q. Zhang. 2011. 'Assessment of severity and frequency of self-reported hypoglycemia on quality of life in patients with type 2 diabetes treated with oral antihyperglycemic agents: A survey study', BMC Res Notes , 4: 251.

Marso, S. P., S. C. Bain, A. Consoli, F. G. Eliaschewitz, E. Jodar, L. A. Leiter, I. Lingvay, J. Rosenstock, J. Seufert, M. L. Warren, V. Woo, O. Hansen, A. G. Holst, J. Pettersson, T. Vilsboll, and Sustain- Investigators. 2016. 'Semaglutide and Cardiovascular Outcomes in Patients with Type 2 Diabetes', N Engl J Med , 375: 1834-44.

Marso, S. P., G. H. Daniels, K. Brown-Frandsen, P. Kristensen, J. F. Mann, M. A. Nauck, S. E. Nissen, S. Pocock, N. R. Poulter, L. S. Ravn, W. M. Steinberg, M. Stockner, B. Zinman, R. M. Bergenstal, and J. B. Buse. 2016. 'Liraglutide and Cardiovascular Outcomes in Type 2 Diabetes', N Engl J Med , 375: 311-22.

Marso, S. P., G. H. Daniels, K. Brown-Frandsen, P. Kristensen, J. F. Mann, M. A. Nauck, S. E. Nissen, S. Pocock, N. R. Poulter, L. S. Ravn, W. M. Steinberg, M. Stockner, B. Zinman, R. M. Bergenstal, J. B. Buse, Leader Steering Committee, and Leader Trial Investigators. 2016. 'Liraglutide and Cardiovascular Outcomes in Type 2 Diabetes', N Engl J Med , 375: 311-22.

Marso, S. P., D. K. McGuire, B. Zinman, N. R. Poulter, S. S. Emerson, T. R. Pieber, R. E. Pratley, P. M. Haahr, M. Lange, K. Brown-Frandsen, A. Moses, S. Skibsted, K. Kvist, and J. B. Buse. 2017. 'Efficacy and Safety of Degludec versus Glargine in Type 2 Diabetes', N Engl J Med , 377: 723-32.

Martin-Timon, I., and F. J. Del Canizo-Gomez. 2015. 'Mechanisms of hypoglycemia unawareness and implications in diabetic patients', World J Diabetes , 6: 912-26.

52

Marx, N., J. Rosenstock, S. E. Kahn, B. Zinman, J. J. Kastelein, J. M. Lachin, M. A. Espeland, E. Bluhmki, M. Mattheus, B. Ryckaert, S. Patel, O. E. Johansen, and H. J. Woerle. 2015. 'Design and baseline characteristics of the CARdiovascular Outcome Trial of LINAgliptin Versus Glimepiride in Type 2 Diabetes (CAROLINA(R))', Diab Vasc Dis Res , 12: 164-74.

Mathieu, C., A. E. Ranetti, D. Li, E. Ekholm, W. Cook, B. Hirshberg, H. Chen, L. Hansen, and N. Iqbal. 2015. 'Randomized, Double-Blind, Phase 3 Trial of Triple Therapy With Dapagliflozin Add-on to Saxagliptin Plus Metformin in Type 2 Diabetes', Diabetes Care , 38: 2009-17.

Matsutani, D., M. Sakamoto, S. Minato, Y. Kayama, N. Takeda, R. Horiuchi, and K. Utsunomiya. 2018. 'Visit-to-visit HbA1c variability is inversely related to baroreflex sensitivity independently of HbA1c value in type 2 diabetes', Cardiovasc Diabetol , 17: 100.

Matthaei, S., D. Catrinoiu, A. Celinski, E. Ekholm, W. Cook, B. Hirshberg, H. Chen, N. Iqbal, and L. Hansen. 2015. 'Randomized, Double-Blind Trial of Triple Therapy With Saxagliptin Add-on to Dapagliflozin Plus Metformin in Patients With Type 2 Diabetes', Diabetes Care , 38: 2018-24.

Mattishent, K., and Y. K. Loke. 2016. 'Bi-directional interaction between hypoglycaemia and cognitive impairment in elderly patients treated with glucose-lowering agents: a systematic review and meta-analysis', Diabetes Obes Metab , 18: 135-41.

Mauricio, D., and I. Hramiak. 2018. 'Second-Generation Insulin Analogues - a Review of Recent Real-World Data and Forthcoming Head-to-Head Comparisons', Eur Endocrinol , 14: 2-9.

Mauricio, D., L. Meneghini, J. Seufert, L. Liao, H. Wang, L. Tong, A. Cali, P. Stella, P. Carita, and K. Khunti. 2017. 'Glycaemic control and hypoglycaemia burden in patients with type 2 diabetes initiating basal insulin in Europe and the USA', Diabetes Obes Metab , 19: 1155-64.

McCoy, R. G., J. Herrin, K. J. Lipska, and N. D. Shah. 2018. 'Recurrent hospitalizations for severe hypoglycemia and hyperglycemia among U.S. adults with diabetes', J Diabetes Complications , 32: 693-701.

McCoy, R. G., H. K. Van Houten, J. Y. Ziegenfuss, N. D. Shah, R. A. Wermers, and S. A. Smith. 2013. 'Self-report of hypoglycemia and health-related quality of life in patients with type 1 and type 2 diabetes', Endocr Pract , 19: 792-9.

McCrimmon, R. J., and R. S. Sherwin. 2010. 'Hypoglycemia in type 1 diabetes', Diabetes , 59: 2333-9.

McGowan, K., W. Thomas, and A. Moran. 2002. 'Spurious reporting of nocturnal hypoglycemia by CGMS in patients with tightly controlled type 1 diabetes', Diabetes Care , 25: 1499- 503.

53

Mehta, R. L., M. J. Davies, S. Ali, N. A. Taub, M. A. Stone, R. Baker, P. G. McNally, I. G. Lawrence, and K. Khunti. 2011. 'Association of cardiac and non-cardiac chronic disease comorbidity on glycaemic control in a multi-ethnic population with type 1 and type 2 diabetes', Postgrad Med J , 87: 763-8.

Meneilly, G. S., E. Cheung, and H. Tuokko. 1994. 'Counterregulatory hormone responses to hypoglycemia in the elderly patient with diabetes', Diabetes , 43: 403-10.

Middleton, T. L., J. Wong, L. Molyneaux, B. A. Brooks, D. K. Yue, S. M. Twigg, and T. Wu. 2017. 'Cardiac Effects of Sulfonylurea-Related Hypoglycemia', Diabetes Care , 40: 663- 70.

Miller, M. E., D. E. Bonds, H. C. Gerstein, E. R. Seaquist, R. M. Bergenstal, J. Calles-Escandon, R. D. Childress, T. E. Craven, R. M. Cuddihy, G. Dailey, M. N. Feinglos, F. Ismail- Beigi, J. F. Largay, P. J. O'Connor, T. Paul, P. J. Savage, U. K. Schubart, A. Sood, S. Genuth, and Accord Investigators. 2010. 'The effects of baseline characteristics, glycaemia treatment approach, and glycated haemoglobin concentration on the risk of severe hypoglycaemia: post hoc epidemiological analysis of the ACCORD study', BMJ , 340: b5444.

Miller, M. E., J. D. Williamson, H. C. Gerstein, R. P. Byington, W. C. Cushman, H. N. Ginsberg, W. T. Ambrosius, L. Lovato, and W. B. Applegate. 2014. 'Effects of randomization to intensive glucose control on adverse events, cardiovascular disease, and mortality in older versus younger adults in the ACCORD Trial', Diabetes Care, 37: 634-43.

Miller, S., T. Krumins, H. Zhou, S. Huyck, J. Johnson, G. Golm, S. G. Terra, J. P. Mancuso, S. S. Engel, and B. Lauring. 2018. 'Ertugliflozin and Sitagliptin Co-initiation in Patients with Type 2 Diabetes: The VERTIS SITA Randomized Study', Diabetes Ther , 9: 253-68.

Mita, T., N. Katakami, T. Shiraiwa, H. Yoshii, N. Kuribayashi, T. Osonoi, H. Kaneto, K. Kosugi, Y. Umayahara, M. Gosho, I. Shimomura, and H. Watada. 2017. 'Relationship between frequency of hypoglycemic episodes and changes in carotid atherosclerosis in insulin- treated patients with type 2 diabetes mellitus', Sci Rep , 7: 39965.

Moen, M. F., M. Zhan, V. D. Hsu, L. D. Walker, L. M. Einhorn, S. L. Seliger, and J. C. Fink. 2009. 'Frequency of hypoglycemia and its significance in chronic kidney disease', Clin J Am Soc Nephrol , 4: 1121-7.

Mohan, V., W. Yang, H. Y. Son, L. Xu, L. Noble, R. B. Langdon, J. M. Amatruda, P. P. Stein, and K. D. Kaufman. 2009. 'Efficacy and safety of sitagliptin in the treatment of patients with type 2 diabetes in China, India, and Korea', Diabetes Res Clin Pract , 83: 106-16.

Moher, D., L. Shamseer, M. Clarke, D. Ghersi, A. Liberati, M. Petticrew, P. Shekelle, and L. A. Stewart. 2015. 'Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement', Syst Rev , 4: 1.

54

Monami, M., I. Dicembrini, and E. Mannucci. 2017. 'Effects of SGLT-2 inhibitors on mortality and cardiovascular events: a comprehensive meta-analysis of randomized controlled trials', Acta Diabetol , 54: 19-36.

Monami, M., C. Lamanna, C. M. Desideri, and E. Mannucci. 2012. 'DPP-4 inhibitors and lipids: systematic review and meta-analysis', Adv Ther , 29: 14-25.

Monnier, L., A. Wojtusciszyn, C. Colette, and D. Owens. 2011. 'The contribution of glucose variability to asymptomatic hypoglycemia in persons with type 2 diabetes', Diabetes Technol Ther , 13: 813-8.

Moretto, T. J., D. R. Milton, T. D. Ridge, L. A. Macconell, T. Okerson, A. M. Wolka, and R. G. Brodows. 2008. 'Efficacy and tolerability of exenatide monotherapy over 24 weeks in antidiabetic drug-naive patients with type 2 diabetes: a randomized, double-blind, placebo-controlled, parallel-group study', Clin Ther , 30: 1448-60.

Morton, SC, MH Murad, E O’Connor, CS Lee, M Booth, BW Vandermeer, JM Snowden, KE D’Anci, R Fu, and G Gartlehner. 2008. 'Quantitative Synthesis—An Update'.

Mozaffarian, D. 2016. 'Dietary and Policy Priorities for Cardiovascular Disease, Diabetes, and Obesity: A Comprehensive Review', Circulation , 133: 187-225.

Muhlbacher, A., and S. Bethge. 2016. 'What matters in type 2 diabetes mellitus oral treatment? A discrete choice experiment to evaluate patient preferences', Eur J Health Econ , 17: 1125- 40.

Muller, N., T. Lehmann, B. Gerste, J. B. Adler, C. Kloos, M. Hartmann, G. Kramer, N. Kuniss, and U. A. Muller. 2017. 'Increase in the incidence of severe hypoglycaemia in people with Type 2 diabetes in spite of new drugs: analysis based on health insurance data from Germany', Diabet Med , 34: 1212-18.

Musso, G., R. Gambino, M. Cassader, and G. Pagano. 2012. 'A novel approach to control hyperglycemia in type 2 diabetes: sodium glucose co-transport (SGLT) inhibitors: systematic review and meta-analysis of randomized trials', Ann Med , 44: 375-93.

Nafisa, A., S. G. Gray, Y. Cao, T. Wang, S. Xu, F. H. Wattoo, M. Barras, N. Cohen, D. Kamato, and P. J. Little. 2018. 'Endothelial function and dysfunction: impact of metformin', Pharmacol Ther .

Namba, M., T. Iwakura, R. Nishimura, K. Akazawa, M. Matsuhisa, Y. Atsumi, J. Satoh, and T. Yamauchi. 2018. 'The current status of treatment-related severe hypoglycemia in Japanese patients with diabetes mellitus: A report from the committee on a survey of severe hypoglycemia in the Japan Diabetes Society', J Diabetes Investig .

Nanditha, A., R. C. Ma, A. Ramachandran, C. Snehalatha, J. C. Chan, K. S. Chia, J. E. Shaw, and P. Z. Zimmet. 2016. 'Diabetes in Asia and the Pacific: Implications for the Global Epidemic', Diabetes Care , 39: 472-85.

55

Napoli, N., M. Chandran, D. D. Pierroz, B. Abrahamsen, A. V. Schwartz, S. L. Ferrari, I. O. F. Bone, and Group Diabetes Working. 2017. 'Mechanisms of diabetes mellitus-induced bone fragility', Nat Rev Endocrinol , 13: 208-19.

Nauck, M. A., S. Del Prato, S. Duran-Garcia, K. Rohwedder, A. M. Langkilde, J. Sugg, and S. J. Parikh. 2014. 'Durability of glycaemic efficacy over 2 years with dapagliflozin versus glipizide as add-on therapies in patients whose type 2 diabetes mellitus is inadequately controlled with metformin', Diabetes Obes Metab , 16: 1111-20.

Nauck, M. A., G. C. Ellis, P. R. Fleck, C. A. Wilson, and Q. Mekki. 2009. 'Efficacy and safety of adding the dipeptidyl peptidase-4 inhibitor alogliptin to metformin therapy in patients with type 2 diabetes inadequately controlled with metformin monotherapy: a multicentre, randomised, double-blind, placebo-controlled study', Int J Clin Pract , 63: 46-55.

Nauck, M. A., M. M. Heimesaat, K. Behle, J. J. Holst, M. S. Nauck, R. Ritzel, M. Hufner, and W. H. Schmiegel. 2002. 'Effects of glucagon-like peptide 1 on counterregulatory hormone responses, cognitive functions, and insulin secretion during hyperinsulinemic, stepped hypoglycemic clamp experiments in healthy volunteers', J Clin Endocrinol Metab , 87: 1239-46.

Nauck, M. A., J. R. Petrie, G. Sesti, E. Mannucci, J. P. Courreges, M. L. Lindegaard, C. B. Jensen, and S. L. Atkin. 2016. 'A Phase 2, Randomized, Dose-Finding Study of the Novel Once-Weekly Human GLP-1 Analog, Semaglutide, Compared With Placebo and Open- Label Liraglutide in Patients With Type 2 Diabetes', Diabetes Care , 39: 231-41.

Nauck, M. A., M. W. Stewart, C. Perkins, A. Jones-Leone, F. Yang, C. Perry, R. R. Reinhardt, and M. Rendell. 2016. 'Efficacy and safety of once-weekly GLP-1 receptor agonist albiglutide (HARMONY 2): 52 week primary endpoint results from a randomised, placebo-controlled trial in patients with type 2 diabetes mellitus inadequately controlled with diet and exercise', Diabetologia , 59: 266-74.

Nauck, M., R. S. Weinstock, G. E. Umpierrez, B. Guerci, Z. Skrivanek, and Z. Milicevic. 2014. 'Efficacy and safety of dulaglutide versus sitagliptin after 52 weeks in type 2 diabetes in a randomized controlled trial (AWARD-5)', Diabetes Care , 37: 2149-58.

Nauck, Michael A., and Juris J. Meier. 2016. 'The incretin effect in healthy individuals and those with type 2 diabetes: physiology, pathophysiology, and response to therapeutic interventions', The Lancet Diabetes & Endocrinology , 4: 525-36.

Neal, B., V. Perkovic, K. W. Mahaffey, D. de Zeeuw, G. Fulcher, N. Erondu, W. Shaw, G. Law, M. Desai, D. R. Matthews, and Canvas Program Collaborative Group. 2017. 'Canagliflozin and Cardiovascular and Renal Events in Type 2 Diabetes', N Engl J Med , 377: 644-57.

Nefs, G., F. Pouwer, R. I. Holt, S. Skovlund, N. Hermanns, A. Nicolucci, and M. Peyrot. 2016. 'Correlates and outcomes of worries about hypoglycemia in family members of adults with diabetes: The second Diabetes Attitudes, Wishes and Needs (DAWN2) study', J Psychosom Res , 89: 69-77. 56

Nelson, P., T. Poon, X. Guan, C. Schnabel, M. Wintle, and M. Fineman. 2007. 'The incretin mimetic exenatide as a monotherapy in patients with type 2 diabetes', Diabetes Technol Ther , 9: 317-26.

Nesto, R. W., R. T. Phillips, K. G. Kett, T. Hill, E. Perper, E. Young, and O. S. Leland, Jr. 1988. 'Angina and exertional myocardial ischemia in diabetic and nondiabetic patients: assessment by exercise thallium scintigraphy', Ann Intern Med , 108: 170-5.

Nissen, S. E., and K. Wolski. 2007. 'Effect of rosiglitazone on the risk of myocardial infarction and death from cardiovascular causes', N Engl J Med , 356: 2457-71.

Nissen, S. E., and K. Wolski. 2010. 'Rosiglitazone revisited: an updated meta-analysis of risk for myocardial infarction and cardiovascular mortality', Arch Intern Med , 170: 1191-201.

Nonaka, K., T. Kakikawa, A. Sato, K. Okuyama, G. Fujimoto, N. Kato, H. Suzuki, Y. Hirayama, T. Ahmed, M. J. Davies, and P. P. Stein. 2008. 'Efficacy and safety of sitagliptin monotherapy in Japanese patients with type 2 diabetes', Diabetes Res Clin Pract , 79: 291- 8.

Nowicki, M., I. Rychlik, H. Haller, M. Warren, L. Suchower, I. Gause-Nilsson, and K. M. Schutzer. 2011. 'Long-term treatment with the dipeptidyl peptidase-4 inhibitor saxagliptin in patients with type 2 diabetes mellitus and renal impairment: a randomised controlled 52-week efficacy and safety study', Int J Clin Pract , 65: 1230-9.

Nunes, A. P., K. Iglay, L. Radican, S. S. Engel, J. Yang, M. C. Doherty, and D. D. Dore. 2017. 'Hypoglycaemia seriousness and weight gain as determinants of cardiovascular disease outcomes among sulfonylurea users', Diabetes Obes Metab .

Nunes, A. P., J. Yang, L. Radican, S. S. Engel, K. Kurtyka, K. Tunceli, S. Yu, K. Iglay, M. C. Doherty, and D. D. Dore. 2016. 'Assessing occurrence of hypoglycemia and its severity from electronic health records of patients with type 2 diabetes mellitus', Diabetes Res Clin Pract , 121: 192-203.

Nyirjesy, P., Y. Zhao, K. Ways, and K. Usiskin. 2012. 'Evaluation of vulvovaginal symptoms and Candida colonization in women with type 2 diabetes mellitus treated with canagliflozin, a sodium glucose co-transporter 2 inhibitor', Curr Med Res Opin , 28: 1173- 8.

O'Reilly, D. J., N. Burke, J. E. Tarride, J. Hahn, and L. Nurkanovic. 2018. 'Direct Health-Care Costs and Productivity Costs Associated With Hypoglycemia in Adults With Type 1 and Type 2 Diabetes Mellitus That Participated in the Canadian Hypoglycemia Assessment Tool Program', Can J Diabetes .

Odawara, M., I. Hamada, and M. Suzuki. 2014. 'Efficacy and Safety of Vildagliptin as Add-on to Metformin in Japanese Patients with Type 2 Diabetes Mellitus', Diabetes Ther , 5: 169-81.

Odawara, M., M. Yoshiki, M. Sano, I. Hamada, V. Lukashevich, and W. Kothny. 2015. 'Efficacy and safety of a single-pill combination of vildagliptin and metformin in Japanese patients 57

with type 2 diabetes mellitus: a randomized, double-blind, placebo-controlled trial', Diabetes Ther , 6: 17-27.

Ostenson, C. G., P. Geelhoed-Duijvestijn, J. Lahtela, R. Weitgasser, M. Markert Jensen, and U. Pedersen-Bjergaard. 2014. 'Self-reported non-severe hypoglycaemic events in Europe', Diabet Med , 31: 92-101.

Owens, D. R., L. Traylor, P. Mullins, and W. Landgraf. 2017. 'Patient-level meta-analysis of efficacy and hypoglycaemia in people with type 2 diabetes initiating insulin glargine 100U/mL or neutral protamine Hagedorn insulin analysed according to concomitant oral antidiabetes therapy', Diabetes Res Clin Pract , 124: 57-65.

Pan, C., P. Han, Q. Ji, C. Li, J. Lu, J. Yang, W. Li, J. Zeng, A. T. Hsieh, and J. Chan. 2017. 'Efficacy and safety of alogliptin in patients with type 2 diabetes mellitus: A multicentre randomized double-blind placebo-controlled Phase 3 study in mainland China, Taiwan, and Hong Kong', J Diabetes , 9: 386-95.

Pan, C., X. Xing, P. Han, S. Zheng, J. Ma, J. Liu, X. Lv, J. Lu, and G. Bader. 2012. 'Efficacy and tolerability of vildagliptin as add-on therapy to metformin in Chinese patients with type 2 diabetes mellitus', Diabetes Obes Metab , 14: 737-44.

Pan, C. Y., W. Yang, C. Tou, I. Gause-Nilsson, and J. Zhao. 2012. 'Efficacy and safety of saxagliptin in drug-naive Asian patients with type 2 diabetes mellitus: a randomized controlled trial', Diabetes Metab Res Rev , 28: 268-75.

Pantalone, K. M., T. M. Hobbs, B. J. Wells, S. X. Kong, M. W. Kattan, J. Bouchard, K. M. Chagin, C. Yu, B. Sakurada, A. Milinovich, W. Weng, J. M. Bauman, and R. S. Zimmerman. 2016. 'Changes in Characteristics and Treatment Patterns of Patients with Newly Diagnosed Type 2 Diabetes in a Large United States Integrated Health System between 2008 and 2013', Clin Med Insights Endocrinol Diabetes , 9: 23-30.

Park, J., S. W. Park, K. H. Yoon, S. R. Kim, K. J. Ahn, J. H. Lee, J. O. Mok, C. H. Chung, K. A. Han, G. P. Koh, J. G. Kang, C. B. Lee, S. H. Kim, N. Y. Kwon, and D. M. Kim. 2017. 'Efficacy and safety of monotherapy in patients with type 2 diabetes and moderately elevated glycated haemoglobin levels after diet and exercise', Diabetes Obes Metab , 19: 1681-87.

Park, S. W., B. H. Goodpaster, E. S. Strotmeyer, N. de Rekeneire, T. B. Harris, A. V. Schwartz, F. A. Tylavsky, and A. B. Newman. 2006. 'Decreased muscle strength and quality in older adults with type 2 diabetes: the health, aging, and body composition study', Diabetes , 55: 1813-8.

Patel, A., S. MacMahon, J. Chalmers, B. Neal, L. Billot, M. Woodward, M. Marre, M. Cooper, P. Glasziou, D. Grobbee, P. Hamet, S. Harrap, S. Heller, L. Liu, G. Mancia, C. E. Mogensen, C. Pan, N. Poulter, A. Rodgers, B. Williams, S. Bompoint, B. E. de Galan, R. Joshi, and F. Travert. 2008. 'Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes', N Engl J Med , 358: 2560-72.

58

Pathak, R. D., E. B. Schroeder, E. R. Seaquist, C. Zeng, J. E. Lafata, A. Thomas, J. Desai, B. Waitzfelder, G. A. Nichols, J. M. Lawrence, A. J. Karter, J. F. Steiner, J. Segal, and P. J. O'Connor. 2016. 'Severe Hypoglycemia Requiring Medical Intervention in a Large Cohort of Adults With Diabetes Receiving Care in U.S. Integrated Health Care Delivery Systems: 2005-2011', Diabetes Care , 39: 363-70.

Pathak, R. D., E. B. Schroeder, E. R. Seaquist, C. Zeng, J. E. Lafata, A. Thomas, J. Desai, B. Waitzfelder, G. A. Nichols, J. M. Lawrence, A. J. Karter, J. F. Steiner, J. Segal, P. J. O'Connor, and Supreme-Dm Study Group. 2016. 'Severe Hypoglycemia Requiring Medical Intervention in a Large Cohort of Adults With Diabetes Receiving Care in U.S. Integrated Health Care Delivery Systems: 2005-2011', Diabetes Care , 39: 363-70.

Patil, H. R., F. J. Al Badarin, H. A. Al Shami, S. K. Bhatti, C. J. Lavie, D. S. Bell, and J. H. O'Keefe. 2012. 'Meta-analysis of effect of dipeptidyl peptidase-4 inhibitors on cardiovascular risk in type 2 diabetes mellitus', Am J Cardiol , 110: 826-33.

Pawaskar, M., K. Iglay, E. A. Witt, S. S. Engel, and S. Rajpathak. 2018. 'Impact of the severity of hypoglycemia on health - Related quality of life, productivity, resource use, and costs among US patients with type 2 diabetes', J Diabetes Complications , 32: 451-57.

Pelletier, C, S Dai, KC Roberts, and A Bienek. 2012. 'Report summary Diabetes in Canada: facts and figures from a public health perspective', Chronic diseases and injuries in Canada , 33.

Peng, C., M. C. Bind, E. Colicino, I. Kloog, H. M. Byun, L. Cantone, L. Trevisi, J. Zhong, K. Brennan, A. E. Dereix, P. S. Vokonas, B. A. Coull, J. D. Schwartz, and A. A. Baccarelli. 2016. 'Particulate Air Pollution and Fasting Blood Glucose in Nondiabetic Individuals: Associations and Epigenetic Mediation in the Normative Aging Study, 2000-2011', Environ Health Perspect , 124: 1715-21.

Peterson, S., and A. Barry. 2017. 'Effect of glucagon-like peptide-1 receptor agonists on all- cause mortality and cardiovascular outcomes: A meta-analysis', Curr Diabetes Rev .

Peyrot, M., A. H. Barnett, L. F. Meneghini, and P. M. Schumm-Draeger. 2012. 'Insulin adherence behaviours and barriers in the multinational Global Attitudes of Patients and Physicians in Insulin Therapy study', Diabet Med , 29: 682-9.

Pfeffer, M. A., B. Claggett, R. Diaz, K. Dickstein, H. C. Gerstein, L. V. Kober, F. C. Lawson, L. Ping, X. Wei, E. F. Lewis, A. P. Maggioni, J. J. McMurray, J. L. Probstfield, M. C. Riddle, S. D. Solomon, and J. C. Tardif. 2015. 'Lixisenatide in Patients with Type 2 Diabetes and Acute Coronary Syndrome', N Engl J Med , 373: 2247-57.

Phung, O. J., J. M. Scholle, M. Talwar, and C. I. Coleman. 2010. 'Effect of noninsulin antidiabetic drugs added to metformin therapy on glycemic control, weight gain, and hypoglycemia in type 2 diabetes', JAMA , 303: 1410-8.

59

Phung, O. J., E. Schwartzman, R. W. Allen, S. S. Engel, and S. N. Rajpathak. 2013. 'Sulphonylureas and risk of cardiovascular disease: systematic review and meta-analysis', Diabet Med , 30: 1160-71.

Pi-Sunyer, F. X., A. Schweizer, D. Mills, and S. Dejager. 2007. 'Efficacy and tolerability of vildagliptin monotherapy in drug-naive patients with type 2 diabetes', Diabetes Res Clin Pract , 76: 132-8.

Piera-Mardemootoo, C., P. Lambert, and J. L. Faillie. 2018. 'Efficacy of metformin on glycemic control and weight in drug-naive type 2 diabetes mellitus patients: A systematic review and meta-analysis of placebo-controlled randomized trials', Therapie .

Pilotto, A., M. Noale, S. Maggi, F. Addante, A. Tiengo, P. C. Perin, G. Rengo, and G. Crepaldi. 2014. 'Hypoglycemia is independently associated with multidimensional impairment in elderly diabetic patients', Biomed Res Int , 2014: 906103.

Pistrosch, F., X. Ganz, S. R. Bornstein, A. L. Birkenfeld, E. Henkel, and M. Hanefeld. 2015. 'Risk of and risk factors for hypoglycemia and associated arrhythmias in patients with type 2 diabetes and cardiovascular disease: a cohort study under real-world conditions', Acta Diabetol , 52: 889-95.

Pistrosch, F., and M. Hanefeld. 2015. 'Hypoglycemia and Cardiovascular Disease: Lessons from Outcome Studies', Curr Diab Rep , 15: 117.

Pontiroli, A. E., L. Miele, and A. Morabito. 2012. 'Metabolic control and risk of hypoglycaemia during the first year of intensive insulin treatment in type 2 diabetes: systematic review and meta-analysis', Diabetes Obes Metab , 14: 433-46.

Powell, W. R., C. L. Christiansen, and D. R. Miller. 2018. 'Meta-Analysis of Sulfonylurea Therapy on Long-Term Risk of Mortality and Cardiovascular Events Compared to Other Oral Glucose-Lowering Treatments', Diabetes Ther , 9: 1431-40.

Pramming, S., B. Thorsteinsson, I. Bendtson, and C. Binder. 1991. 'Symptomatic hypoglycaemia in 411 type 1 diabetic patients', Diabet Med , 8: 217-22.

Pratley, R. E., P. Fleck, and C. Wilson. 2014. 'Efficacy and safety of initial combination therapy with alogliptin plus metformin versus either as monotherapy in drug-naive patients with type 2 diabetes: a randomized, double-blind, 6-month study', Diabetes Obes Metab , 16: 613-21.

Pratley, RE, S Jauffret-Kamel, E Galbreath, and D Holmes. 2006. 'Twelve-week monotherapy with the DPP-4 inhibitor vildagliptin improves glycemic control in subjects with type 2 diabetes', Hormone and metabolic research , 38: 423-28.

Prochaska, J. J., and J. F. Hilton. 2012. 'Risk of cardiovascular serious adverse events associated with varenicline use for tobacco cessation: systematic review and meta-analysis', BMJ , 344: e2856.

60

Ptaszynska, A., K. M. Johnsson, S. J. Parikh, T. W. de Bruin, A. M. Apanovitch, and J. F. List. 2014. 'Safety profile of dapagliflozin for type 2 diabetes: pooled analysis of clinical studies for overall safety and rare events', Drug Saf , 37: 815-29.

Puente, E. C., J. Silverstein, A. J. Bree, D. R. Musikantow, D. F. Wozniak, S. Maloney, D. Daphna-Iken, and S. J. Fisher. 2010. 'Recurrent moderate hypoglycemia ameliorates brain damage and cognitive dysfunction induced by severe hypoglycemia', Diabetes , 59: 1055-62.

Punthakee, Z., R. Goldenberg, and P. Katz. 2018. 'Definition, Classification and Diagnosis of Diabetes, Prediabetes and Metabolic Syndrome', Can J Diabetes , 42 Suppl 1: S10-s15.

Punthakee, Z., M. E. Miller, L. J. Launer, J. D. Williamson, R. M. Lazar, T. Cukierman-Yaffee, E. R. Seaquist, F. Ismail-Beigi, M. D. Sullivan, L. C. Lovato, R. M. Bergenstal, and H. C. Gerstein. 2012. 'Poor cognitive function and risk of severe hypoglycemia in type 2 diabetes: post hoc epidemiologic analysis of the ACCORD trial', Diabetes Care , 35: 787- 93.

Punthakee, Z., G. H. Werstuck, and H. C. Gerstein. 2007. 'Diabetes and cardiovascular disease: explaining the relationship', Rev Cardiovasc Med , 8: 145-53.

Qaseem, A., M. J. Barry, L. L. Humphrey, M. A. Forciea, and Physicians Clinical Guidelines Committee of the American College of. 2017. 'Oral Pharmacologic Treatment of Type 2 Diabetes Mellitus: A Clinical Practice Guideline Update From the American College of Physicians', Ann Intern Med , 166: 279-90.

Qiu, R., D. Balis, J. Xie, M. J. Davies, M. Desai, and G. Meininger. 2017. 'Longer-term safety and tolerability of canagliflozin in patients with type 2 diabetes: a pooled analysis', Curr Med Res Opin , 33: 553-62.

Qiu, R., G. Capuano, and G. Meininger. 2014. 'Efficacy and safety of twice-daily treatment with canagliflozin, a sodium glucose co-transporter 2 inhibitor, added on to metformin monotherapy in patients with type 2 diabetes mellitus', J Clin Transl Endocrinol , 1: 54- 60.

Rana, O. A., C. D. Byrne, and K. Greaves. 2014. 'Intensive glucose control and hypoglycaemia: a new cardiovascular risk factor?', Heart , 100: 21-7.

Rao, G., F. Lopez-Jimenez, J. Boyd, F. D'Amico, N. H. Durant, M. A. Hlatky, G. Howard, K. Kirley, C. Masi, T. M. Powell-Wiley, A. E. Solomonides, C. P. West, J. Wessel, Lifestyle American Heart Association Council on, Health Cardiometabolic, Cardiovascular Council on, Nursing Stroke, Surgery Council on Cardiovascular, Anesthesia, Cardiology Council on Clinical, Genomics Council on Functional, Biology Translational, and Council Stroke. 2017. 'Methodological Standards for Meta-Analyses and Qualitative Systematic Reviews of Cardiac Prevention and Treatment Studies: A Scientific Statement From the American Heart Association', Circulation , 136: e172-e94.

61

Rathmann, W., K. Kostev, J. B. Gruenberger, M. Dworak, G. Bader, and G. Giani. 2013. 'Treatment persistence, hypoglycaemia and clinical outcomes in type 2 diabetes patients with dipeptidyl peptidase-4 inhibitors and sulphonylureas: a primary care database analysis', Diabetes Obes Metab , 15: 55-61.

Ratner, R. E., J. Rosenstock, and G. Boka. 2010. 'Dose-dependent effects of the once-daily GLP- 1 receptor agonist lixisenatide in patients with Type 2 diabetes inadequately controlled with metformin: a randomized, double-blind, placebo-controlled trial', Diabet Med , 27: 1024-32.

Ratzki-Leewing, A., S. B. Harris, S. Mequanint, S. M. Reichert, J. Belle Brown, J. E. Black, and B. L. Ryan. 2018. 'Real-world crude incidence of hypoglycemia in adults with diabetes: Results of the InHypo-DM Study, Canada', BMJ Open Diabetes Res Care , 6: e000503.

Ray, K. K., S. R. Seshasai, S. Wijesuriya, R. Sivakumaran, S. Nethercott, D. Preiss, S. Erqou, and N. Sattar. 2009. 'Effect of intensive control of glucose on cardiovascular outcomes and death in patients with diabetes mellitus: a meta-analysis of randomised controlled trials', Lancet , 373: 1765-72.

Raz, I., Y. Chen, M. Wu, S. Hussain, K. D. Kaufman, J. M. Amatruda, R. B. Langdon, P. P. Stein, and M. Alba. 2008. 'Efficacy and safety of sitagliptin added to ongoing metformin therapy in patients with type 2 diabetes', Curr Med Res Opin , 24: 537-50.

Raz, I., M. Hanefeld, L. Xu, C. Caria, D. Williams-Herman, and H. Khatami. 2006. 'Efficacy and safety of the dipeptidyl peptidase-4 inhibitor sitagliptin as monotherapy in patients with type 2 diabetes mellitus', Diabetologia , 49: 2564-71.

Reaven, P. D., T. E. Moritz, D. C. Schwenke, R. J. Anderson, M. Criqui, R. Detrano, N. Emanuele, M. Kayshap, J. Marks, S. Mudaliar, R. Harsha Rao, J. H. Shah, S. Goldman, D. J. Reda, M. McCarren, C. Abraira, W. Duckworth, and Trial Veterans Affairs Diabetes. 2009. 'Intensive glucose-lowering therapy reduces cardiovascular disease events in veterans affairs diabetes trial participants with lower calcified coronary atherosclerosis', Diabetes , 58: 2642-8.

Redelmeier, D. A., A. B. Kenshole, and J. G. Ray. 2009. 'Motor vehicle crashes in diabetic patients with tight glycemic control: a population-based case control analysis', PLoS Med , 6: e1000192.

Reichert, Sonja M, Alexandria Ratzki-Leewing, Bridget L Ryan, Selam Mequanint, Susan Webster-Bogaert, Judith B Brown, and Stewart B Harris. 2017. "Hypoglycemia Management through the Eyes of the Significant Other: Highlights from the InHypo-DM Study (Canada)." In Diabetes , A106-A06. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Rena, G., D. G. Hardie, and E. R. Pearson. 2017. 'The mechanisms of action of metformin', Diabetologia , 60: 1577-85.

62

Reno, C. M., D. Daphna-Iken, Y. S. Chen, J. VanderWeele, K. Jethi, and S. J. Fisher. 2013. 'Severe hypoglycemia-induced lethal cardiac arrhythmias are mediated by sympathoadrenal activation', Diabetes , 62: 3570-81.

Reno, Candace M, Justin Bayles, Allie Skinner, and Simon J Fisher. 2017. "KATP Channel Opening Increases Severe Hypoglycemia-Induced Cardiac Arrhythmias." In Diabetes , A101-A01. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Rhee, E. J., W. Y. Lee, K. H. Yoon, S. J. Yoo, I. K. Lee, S. H. Baik, Y. K. Kim, M. K. Lee, K. S. Park, J. Y. Park, B. S. Cha, H. W. Lee, K. W. Min, H. Y. Bae, M. J. Kim, J. A. Kim, D. K. Kim, and S. W. Kim. 2010. 'A multicenter, randomized, placebo-controlled, double- blind phase II trial evaluating the optimal dose, efficacy and safety of LC 15-0444 in patients with type 2 diabetes', Diabetes Obes Metab , 12: 1113-9.

Riddle, M. C. 2010. 'Effects of intensive glucose lowering in the management of patients with type 2 diabetes mellitus in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial', Circulation , 122: 844-6.

Riddle, M. C., W. T. Ambrosius, D. J. Brillon, J. B. Buse, R. P. Byington, R. M. Cohen, D. C. Goff, Jr., S. Malozowski, K. L. Margolis, J. L. Probstfield, A. Schnall, E. R. Seaquist, and Investigators Action to Control Cardiovascular Risk in Diabetes. 2010. 'Epidemiologic relationships between A1C and all-cause mortality during a median 3.4- year follow-up of glycemic treatment in the ACCORD trial', Diabetes Care , 33: 983-90.

Riddle, M. C., H. C. Gerstein, R. R. Holman, S. E. Inzucchi, B. Zinman, S. Zoungas, and W. T. Cefalu. 2018. 'A1C Targets Should Be Personalized to Maximize Benefits While Limiting Risks', Diabetes Care , 41: 1121-24.

Ristic, S., S. Byiers, J. Foley, and D. Holmes. 2005. 'Improved glycaemic control with dipeptidyl peptidase-4 inhibition in patients with type 2 diabetes: vildagliptin (LAF237) dose response', Diabetes Obes Metab , 7: 692-8.

Rodbard, H. W., J. Seufert, N. Aggarwal, A. Cao, A. Fung, M. Pfeifer, and M. Alba. 2016. 'Efficacy and safety of titrated canagliflozin in patients with type 2 diabetes mellitus inadequately controlled on metformin and sitagliptin', Diabetes Obes Metab , 18: 812-9.

Roden, M., J. Weng, J. Eilbracht, B. Delafont, G. Kim, H. J. Woerle, and U. C. Broedl. 2013. 'Empagliflozin monotherapy with sitagliptin as an active comparator in patients with type 2 diabetes: a randomised, double-blind, placebo-controlled, phase 3 trial', Lancet Diabetes Endocrinol , 1: 208-19.

Rodriguez-Gutierrez, R., N. S. Ospina, R. G. McCoy, K. J. Lipska, N. D. Shah, V. M. Montori, and Group Hypoglycemia as a Quality Measure in Diabetes Study. 2016. 'Inclusion of Hypoglycemia in Clinical Practice Guidelines and Performance Measures in the Care of Patients With Diabetes', JAMA Intern Med , 176: 1714-16.

63

Romero, R., O. Erez, M. Huttemann, E. Maymon, B. Panaitescu, A. Conde-Agudelo, P. Pacora, B. H. Yoon, and L. I. Grossman. 2017. 'Metformin, the aspirin of the 21st century: its role in gestational diabetes mellitus, prevention of preeclampsia and cancer, and the promotion of longevity', Am J Obstet Gynecol , 217: 282-302.

Rosella, L. C., M. Lebenbaum, T. Fitzpatrick, D. O'Reilly, J. Wang, G. L. Booth, T. A. Stukel, and W. P. Wodchis. 2016. 'Impact of diabetes on healthcare costs in a population-based cohort: a cost analysis', Diabet Med , 33: 395-403.

Rosenstock, J., N. Aggarwal, D. Polidori, Y. Zhao, D. Arbit, K. Usiskin, G. Capuano, and W. Canovatchel. 2012. 'Dose-ranging effects of canagliflozin, a sodium-glucose cotransporter 2 inhibitor, as add-on to metformin in subjects with type 2 diabetes', Diabetes Care , 35: 1232-8.

Rosenstock, J., C. Aguilar-Salinas, E. Klein, S. Nepal, J. List, and R. Chen. 2009. 'Effect of saxagliptin monotherapy in treatment-naive patients with type 2 diabetes', Curr Med Res Opin , 25: 2401-11.

Rosenstock, J., B. Balas, B. Charbonnel, G. B. Bolli, M. Boldrin, R. Ratner, and R. Balena. 2013. '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 , 36: 498-504.

Rosenstock, J., W. T. Cefalu, P. Lapuerta, B. Zambrowicz, I. Ogbaa, P. Banks, and A. Sands. 2015. 'Greater dose-ranging effects on A1C levels than on glucosuria with LX4211, a dual inhibitor of SGLT1 and SGLT2, in patients with type 2 diabetes on metformin monotherapy', Diabetes Care , 38: 431-8.

Rosenstock, J., J. L. Gross, C. Aguilar-Salinas, M. Hissa, N. Berglind, S. Ravichandran, and D. Fleming. 2013. 'Long-term 4-year safety of saxagliptin in drug-naive and metformin- treated patients with Type 2 diabetes', Diabet Med , 30: 1472-6.

Rosenstock, J., A. Jelaska, C. Zeller, G. Kim, U. C. Broedl, H. J. Woerle, and Empa-Reg Basaltm trial investigators. 2015. 'Impact of empagliflozin added on to basal insulin in type 2 diabetes inadequately controlled on basal insulin: a 78-week randomized, double- blind, placebo-controlled trial', Diabetes Obes Metab , 17: 936-48.

Rosenstock, J., A. J. Lewin, P. Norwood, V. Somayaji, T. T. Nguyen, J. G. Teeter, S. L. Johnson, H. Dai, and S. G. Terra. 2011. 'Efficacy and safety of the dipeptidyl peptidase-4 inhibitor PF-734200 added to metformin in Type 2 diabetes', Diabet Med , 28: 464-9.

Rosenstock, J., D. Raccah, L. Koranyi, L. Maffei, G. Boka, P. Miossec, and J. E. Gerich. 2013. 'Efficacy and safety of lixisenatide once daily versus exenatide twice daily in type 2 diabetes inadequately controlled on metformin: a 24-week, randomized, open-label, active-controlled study (GetGoal-X)', Diabetes Care , 36: 2945-51.

Rosenstock, J., J. Reusch, M. Bush, F. Yang, and M. Stewart. 2009. 'Potential of albiglutide, a long-acting GLP-1 receptor agonist, in type 2 diabetes: a randomized controlled trial exploring weekly, biweekly, and monthly dosing', Diabetes Care , 32: 1880-6. 64

Rosenstock, J., S. Sankoh, and J. F. List. 2008. 'Glucose-lowering activity of the dipeptidyl peptidase-4 inhibitor saxagliptin in drug-naive patients with type 2 diabetes', Diabetes Obes Metab , 10: 376-86.

Rosenstock, J., L. J. Seman, A. Jelaska, S. Hantel, S. Pinnetti, T. Hach, and H. J. Woerle. 2013. 'Efficacy and safety of empagliflozin, a sodium glucose cotransporter 2 (SGLT2) inhibitor, as add-on to metformin in type 2 diabetes with mild hyperglycaemia', Diabetes Obes Metab , 15: 1154-60.

Roshanov, P. S., and B. B. Dennis. 2015. 'Incretin-based therapies are associated with acute pancreatitis: Meta-analysis of large randomized controlled trials', Diabetes Res Clin Pract , 110: e13-7.

Ross, S. A., E. Rafeiro, T. Meinicke, R. Toorawa, S. Weber-Born, and H. J. Woerle. 2012. 'Efficacy and safety of linagliptin 2.5 mg twice daily versus 5 mg once daily in patients with type 2 diabetes inadequately controlled on metformin: a randomised, double-blind, placebo-controlled trial', Curr Med Res Opin , 28: 1465-74.

Ross, S., C. Thamer, J. Cescutti, T. Meinicke, H. J. Woerle, and U. C. Broedl. 2015. 'Efficacy and safety of empagliflozin twice daily versus once daily in patients with type 2 diabetes inadequately controlled on metformin: a 16-week, randomized, placebo-controlled trial', Diabetes Obes Metab , 17: 699-702.

Ryden, L., P. J. Grant, S. D. Anker, C. Berne, F. Cosentino, N. Danchin, C. Deaton, J. Escaned, H. P. Hammes, H. Huikuri, M. Marre, N. Marx, L. Mellbin, J. Ostergren, C. Patrono, P. Seferovic, M. S. Uva, M. R. Taskinen, M. Tendera, J. Tuomilehto, P. Valensi, J. L. Zamorano, J. L. Zamorano, S. Achenbach, H. Baumgartner, J. J. Bax, H. Bueno, V. Dean, C. Deaton, C. Erol, R. Fagard, R. Ferrari, D. Hasdai, A. W. Hoes, P. Kirchhof, J. Knuuti, P. Kolh, P. Lancellotti, A. Linhart, P. Nihoyannopoulos, M. F. Piepoli, P. Ponikowski, P. A. Sirnes, J. L. Tamargo, M. Tendera, A. Torbicki, W. Wijns, S. Windecker, G. De Backer, P. A. Sirnes, E. A. Ezquerra, A. Avogaro, L. Badimon, E. Baranova, H. Baumgartner, J. Betteridge, A. Ceriello, R. Fagard, C. Funck-Brentano, D. C. Gulba, D. Hasdai, A. W. Hoes, J. K. Kjekshus, J. Knuuti, P. Kolh, E. Lev, C. Mueller, L. Neyses, P. M. Nilsson, J. Perk, P. Ponikowski, Z. Reiner, N. Sattar, V. Schachinger, A. Scheen, H. Schirmer, A. Stromberg, S. Sudzhaeva, J. L. Tamargo, M. Viigimaa, C. Vlachopoulos, and R. G. Xuereb. 2013. 'ESC Guidelines on diabetes, pre-diabetes, and cardiovascular diseases developed in collaboration with the EASD: the Task Force on diabetes, pre-diabetes, and cardiovascular diseases of the European Society of Cardiology (ESC) and developed in collaboration with the European Association for the Study of Diabetes (EASD)', Eur Heart J , 34: 3035-87.

Sach-Friedl, Stefanie, Thomas Augustin, Christoph Magnes, Eva Ekardt, Anita Eberl, Sophie Narath, Martina Brunner, Stefan Korsatko, Eva Svehlikova, and Gerlies Treiber. 2017. "Effect of SGLT2i, DPP-4i, and the Combination of SGLT2i+ DPP-4i on Glucagon, Endogenous Glucose Production (EGP), and Lipolysis in Patients with Type 2 Diabetes (T2DM)." In Diabetes , A312-A12. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

65

Salvo, F., N. Moore, M. Arnaud, P. Robinson, E. Raschi, F. De Ponti, B. Begaud, and A. Pariente. 2016. 'Addition of dipeptidyl peptidase-4 inhibitors to sulphonylureas and risk of hypoglycaemia: systematic review and meta-analysis', BMJ , 353: i2231.

Samuel, S., D. Goswami, K. Shum, K. S. Boye, B. Rengarajan, B. Curtis, and S. E. Curtis. 2015. 'A model of mild hypoglycemia', Curr Med Res Opin , 31: 633-41.

Saremi, A., G. D. Bahn, and P. D. Reaven. 2016. 'A Link Between Hypoglycemia and Progression of Atherosclerosis in the Veterans Affairs Diabetes Trial (VADT)', Diabetes Care , 39: 448-54.

Sarkar, U., A. J. Karter, J. Y. Liu, H. H. Moffet, N. E. Adler, and D. Schillinger. 2010. 'Hypoglycemia is more common among type 2 diabetes patients with limited health literacy: the Diabetes Study of Northern California (DISTANCE)', J Gen Intern Med , 25: 962-8.

Scherbaum, W. A., A. Schweizer, A. Mari, P. M. Nilsson, G. Lalanne, S. Jauffret, and J. E. Foley. 2008. 'Efficacy and tolerability of vildagliptin in drug-naive patients with type 2 diabetes and mild hyperglycaemia*', Diabetes Obes Metab , 10: 675-82.

Schloot, N. C., A. Haupt, M. Schutt, K. Badenhoop, M. Laimer, C. Nicolay, M. Reaney, K. Fink, and R. W. Holl. 2016. 'Risk of severe hypoglycemia in sulfonylurea-treated patients from diabetes centers in Germany/Austria: How big is the problem? Which patients are at risk?', Diabetes Metab Res Rev , 32: 316-24.

Schopman, J. E., J. Geddes, and B. M. Frier. 2010. 'Prevalence of impaired awareness of hypoglycaemia and frequency of hypoglycaemia in insulin-treated type 2 diabetes', Diabetes Res Clin Pract , 87: 64-8.

Schopman, J. E., A. C. Simon, S. J. Hoefnagel, J. B. Hoekstra, R. J. Scholten, and F. Holleman. 2014. 'The incidence of mild and severe hypoglycaemia in patients with type 2 diabetes mellitus treated with sulfonylureas: a systematic review and meta-analysis', Diabetes Metab Res Rev , 30: 11-22.

Schramm, T. K., G. H. Gislason, A. Vaag, J. N. Rasmussen, F. Folke, M. L. Hansen, E. L. Fosbol, L. Kober, M. L. Norgaard, M. Madsen, P. R. Hansen, and C. Torp-Pedersen. 2011. 'Mortality and cardiovascular risk associated with different insulin secretagogues compared with metformin in type 2 diabetes, with or without a previous myocardial infarction: a nationwide study', Eur Heart J , 32: 1900-8.

Schumm-Draeger, P. M., L. Burgess, L. Koranyi, V. Hruba, J. E. Hamer-Maansson, and T. W. de Bruin. 2015. 'Twice-daily dapagliflozin co-administered with metformin in type 2 diabetes: a 16-week randomized, placebo-controlled clinical trial', Diabetes Obes Metab , 17: 42-51.

Schwartz, A. V., T. A. Hillier, D. E. Sellmeyer, H. E. Resnick, E. Gregg, K. E. Ensrud, P. J. Schreiner, K. L. Margolis, J. A. Cauley, M. C. Nevitt, D. M. Black, and S. R. Cummings.

66

2002. 'Older women with diabetes have a higher risk of falls: a prospective study', Diabetes Care , 25: 1749-54.

Schweizer, A., J. E. Foley, W. Kothny, and B. Ahren. 2013. 'Clinical evidence and mechanistic basis for vildagliptin's effect in combination with insulin', Vasc Health Risk Manag , 9: 57-64.

Scirica, B. M., D. L. Bhatt, E. Braunwald, P. G. Steg, J. Davidson, B. Hirshberg, P. Ohman, R. Frederich, S. D. Wiviott, E. B. Hoffman, M. A. Cavender, J. A. Udell, N. R. Desai, O. Mosenzon, D. K. McGuire, K. K. Ray, L. A. Leiter, I. Raz, Savor-Timi Steering Committee, and Investigators. 2013. 'Saxagliptin and cardiovascular outcomes in patients with type 2 diabetes mellitus', N Engl J Med , 369: 1317-26.

Scirica, B. M., E. Braunwald, I. Raz, M. A. Cavender, D. A. Morrow, P. Jarolim, J. A. Udell, O. Mosenzon, K. Im, A. A. Umez-Eronini, P. S. Pollack, B. Hirshberg, R. Frederich, B. S. Lewis, D. K. McGuire, J. Davidson, P. G. Steg, D. L. Bhatt, Savor-Timi Steering Committee, and Investigators*. 2014. 'Heart failure, saxagliptin, and diabetes mellitus: observations from the SAVOR-TIMI 53 randomized trial', Circulation , 130: 1579-88.

Scott, R., T. Loeys, M. J. Davies, and S. S. Engel. 2008. 'Efficacy and safety of sitagliptin when added to ongoing metformin therapy in patients with type 2 diabetes', Diabetes Obes Metab , 10: 959-69.

Scott, R., M. Wu, M. Sanchez, and P. Stein. 2007. 'Efficacy and tolerability of the dipeptidyl peptidase-4 inhibitor sitagliptin as monotherapy over 12 weeks in patients with type 2 diabetes', Int J Clin Pract , 61: 171-80.

Seaquist, E. R., J. Anderson, B. Childs, P. Cryer, S. Dagogo-Jack, L. Fish, S. R. Heller, H. Rodriguez, J. Rosenzweig, and R. Vigersky. 2013. 'Hypoglycemia and diabetes: a report of a workgroup of the American Diabetes Association and the Endocrine Society', Diabetes Care , 36: 1384-95.

Seaquist, E. R., M. E. Miller, D. E. Bonds, M. Feinglos, D. C. Goff, Jr., K. Peterson, and P. Senior. 2012. 'The impact of frequent and unrecognized hypoglycemia on mortality in the ACCORD study', Diabetes Care , 35: 409-14.

Segel, S. A., D. S. Paramore, and P. E. Cryer. 2002. 'Hypoglycemia-associated autonomic failure in advanced type 2 diabetes', Diabetes , 51: 724-33.

Seino, Y., N. Inagaki, H. Miyahara, I. Okuda, M. Bush, J. Ye, M. C. Holland, S. Johnson, E. Lewis, and H. Nakajima. 2014. 'A randomized dose-finding study demonstrating the efficacy and tolerability of albiglutide in Japanese patients with type 2 diabetes mellitus', Curr Med Res Opin , 30: 1095-106.

Seino, Y., M. F. Rasmussen, M. Zdravkovic, and K. Kaku. 2008. 'Dose-dependent improvement in glycemia with once-daily liraglutide without hypoglycemia or weight gain: A double- blind, randomized, controlled trial in Japanese patients with type 2 diabetes', Diabetes Res Clin Pract , 81: 161-8. 67

Seino, Y., T. Sasaki, A. Fukatsu, S. Sakai, and Y. Samukawa. 2014. 'Efficacy and safety of monotherapy in Japanese patients with type 2 diabetes mellitus: a 12-week, randomized, placebo-controlled, phase II study', Curr Med Res Opin , 30: 1219-30.

Seino, Y., T. Sasaki, A. Fukatsu, M. Ubukata, S. Sakai, and Y. Samukawa. 2014. 'Dose-finding study of luseogliflozin in Japanese patients with type 2 diabetes mellitus: a 12-week, randomized, double-blind, placebo-controlled, phase II study', Curr Med Res Opin , 30: 1231-44.

Seligman, H. K., E. A. Jacobs, A. Lopez, J. Tschann, and A. Fernandez. 2012. 'Food insecurity and glycemic control among low-income patients with type 2 diabetes', Diabetes Care , 35: 233-8.

Shaefer, C., D. Hinnen, and C. Sadler. 2016. 'Hypoglycemia and diabetes: increased need for awareness', Curr Med Res Opin , 32: 1479-86.

Shai, I., R. Jiang, J. E. Manson, M. J. Stampfer, W. C. Willett, G. A. Colditz, and F. B. Hu. 2006. 'Ethnicity, obesity, and risk of type 2 diabetes in women: a 20-year follow-up study', Diabetes Care , 29: 1585-90.

Shankar, R Ravi, Jianmin Long, Lei Xu, Edward A O'Neill, and Samuel S Engel. 2017. "Impact of Sitagliptin on Hypoglycemia in Elderly Subjects with T2DM Treated with Insulin." In Diabetes , A585-A86. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Shankar, R. R., S. E. Inzucchi, V. Scarabello, I. Gantz, K. D. Kaufman, E. Lai, P. Ceesay, S. Suryawanshi, and S. S. Engel. 2017. 'A randomized clinical trial evaluating the efficacy and safety of the once-weekly dipeptidyl peptidase-4 inhibitor omarigliptin in patients with type 2 diabetes inadequately controlled on metformin monotherapy', Curr Med Res Opin , 33: 1853-60.

Shao, W., R. Ahmad, N. Khutoryansky, M. Aagren, and J. Bouchard. 2013. 'Evidence supporting an association between hypoglycemic events and depression', Curr Med Res Opin , 29: 1609-15.

Sheu, W. H., I. Gantz, M. Chen, S. Suryawanshi, A. Mirza, B. J. Goldstein, K. D. Kaufman, and S. S. Engel. 2015. 'Safety and Efficacy of Omarigliptin (MK-3102), a Novel Once- Weekly DPP-4 Inhibitor for the Treatment of Patients With Type 2 Diabetes', Diabetes Care , 38: 2106-14.

Shih, C. J., Y. L. Wu, Y. H. Lo, S. C. Kuo, D. C. Tarng, C. C. Lin, S. M. Ou, and Y. T. Chen. 2015. 'Association of hypoglycemia with incident chronic kidney disease in patients with type 2 diabetes: a nationwide population-based study', Medicine (Baltimore) , 94: e771.

Shukla, V., A. K. Shakya, M. A. Perez-Pinzon, and K. R. Dave. 2017. 'Cerebral ischemic damage in diabetes: an inflammatory perspective', J Neuroinflammation , 14: 21.

68

Shuster, J. J., and M. A. Walker. 2016. 'Low-event-rate meta-analyses of clinical trials: implementing good practices', Stat Med , 35: 2467-78.

Signorovitch, J. E., D. Macaulay, M. Diener, Y. Yan, E. Q. Wu, J. B. Gruenberger, and B. M. Frier. 2013. 'Hypoglycaemia and accident risk in people with type 2 diabetes mellitus treated with non-insulin antidiabetes drugs', Diabetes Obes Metab , 15: 335-41.

Simpson, S. H., S. R. Majumdar, R. T. Tsuyuki, D. T. Eurich, and J. A. Johnson. 2006. 'Dose- response relation between sulfonylurea drugs and mortality in type 2 diabetes mellitus: a population-based cohort study', CMAJ , 174: 169-74.

Simpson, Scot H., Jayson Lee, Sabina Choi, Ben Vandermeer, Ahmed S. Abdelmoneim, and Travis R. Featherstone. 2015. 'Mortality risk among sulfonylureas: a systematic review and network meta-analysis', The Lancet Diabetes & Endocrinology , 3: 43-51.

Singh-Franco, D., C. Harrington, and E. Tellez-Corrales. 2016. 'An updated systematic review and meta-analysis on the efficacy and tolerability of dipeptidyl peptidase-4 inhibitors in patients with type 2 diabetes with moderate to severe chronic kidney disease', SAGE Open Med , 4: 2050312116659090.

Singh, A. K., and R. Singh. 2016. 'Dipeptidyl peptidase-4 inhibitors or sodium glucose co- transporter-2 inhibitors as an add-on to insulin therapy: A comparative review', Indian J Endocrinol Metab , 20: 32-42.

Sjöberg Bexelius, Tomas, Rickard Ljung, Fredrik Mattsson, and Jesper Lagergren. 2013. Sa1477 Cardiovascular Disease and Risk of Acute Pancreatitis in a Population-Based Study .

Softeland, E., J. J. Meier, B. Vangen, R. Toorawa, M. Maldonado-Lutomirsky, and U. C. Broedl. 2017. 'Empagliflozin as Add-on Therapy in Patients With Type 2 Diabetes Inadequately Controlled With Linagliptin and Metformin: A 24-Week Randomized, Double-Blind, Parallel-Group Trial', Diabetes Care , 40: 201-09.

Sohani, Z. N., L. Li, and X. Sun. 2016. 'Incretin-based therapy: Is the risk of pancreatitis driven by cardiovascular disease?', Diabetes Res Clin Pract , 117: 28-31.

Sonesson, C., P. A. Johansson, E. Johnsson, and I. Gause-Nilsson. 2016. 'Cardiovascular effects of dapagliflozin in patients with type 2 diabetes and different risk categories: a meta- analysis', Cardiovasc Diabetol , 15: 37.

Sorli, C., S. I. Harashima, G. M. Tsoukas, J. Unger, J. D. Karsbol, T. Hansen, and S. C. Bain. 2017. 'Efficacy and safety of once-weekly semaglutide monotherapy versus placebo in patients with type 2 diabetes (SUSTAIN 1): a double-blind, randomised, placebo- controlled, parallel-group, multinational, multicentre phase 3a trial', Lancet Diabetes Endocrinol , 5: 251-60.

Spyer, G., A. T. Hattersley, I. A. MacDonald, S. Amiel, and K. M. MacLeod. 2000. 'Hypoglycaemic counter-regulation at normal blood glucose concentrations in patients with well controlled type-2 diabetes', Lancet , 356: 1970-4. 69

Sreenan, S., M. Andersen, B. L. Thorsted, M. L. Wolden, and M. Evans. 2014. 'Increased Risk of Severe Hypoglycemic Events with Increasing Frequency of Non-severe Hypoglycemic Events in Patients with Type 1 and Type 2 Diabetes', Diabetes Ther , 5: 447-58.

Stahn, A., F. Pistrosch, X. Ganz, M. Teige, C. Koehler, S. Bornstein, and M. Hanefeld. 2014. 'Relationship between hypoglycemic episodes and ventricular arrhythmias in patients with type 2 diabetes and cardiovascular diseases: silent hypoglycemias and silent arrhythmias', Diabetes Care , 37: 516-20.

Standl, E., S. R. Stevens, P. W. Armstrong, J. B. Buse, J. C. N. Chan, J. B. Green, J. M. Lachin, A. Scheen, F. Travert, F. Van de Werf, E. D. Peterson, and R. R. Holman. 2018. 'Increased Risk of Severe Hypoglycemic Events Before and After Cardiovascular Outcomes in TECOS Suggests an At-Risk Type 2 Diabetes Frail Patient Phenotype', Diabetes Care , 41: 596-603.

Steinberg, D. I. 2016. 'Intensive glucose-lowering in type 2 diabetes was linked to hypoglycemia in high-complexity patients', Ann Intern Med , 165: Jc33.

Stenlof, K., W. T. Cefalu, K. A. Kim, M. Alba, K. Usiskin, C. Tong, W. Canovatchel, and G. Meininger. 2013. 'Efficacy and safety of canagliflozin monotherapy in subjects with type 2 diabetes mellitus inadequately controlled with diet and exercise', Diabetes Obes Metab , 15: 372-82.

Stenlof, K., W. T. Cefalu, K. A. Kim, E. Jodar, M. Alba, R. Edwards, C. Tong, W. Canovatchel, and G. Meininger. 2014. 'Long-term efficacy and safety of canagliflozin monotherapy in patients with type 2 diabetes inadequately controlled with diet and exercise: findings from the 52-week CANTATA-M study', Curr Med Res Opin , 30: 163-75.

Strojek, K., K. H. Yoon, V. Hruba, M. Elze, A. M. Langkilde, and S. Parikh. 2011. 'Effect of dapagliflozin in patients with type 2 diabetes who have inadequate glycaemic control with glimepiride: a randomized, 24-week, double-blind, placebo-controlled trial', Diabetes Obes Metab , 13: 928-38.

Strongman, H., K. D'Oca, H. Langerman, and R. Das. 2015. 'Comparison of diabetes-associated secondary healthcare utilization between alternative oral antihyperglycaemic dual therapy combinations with metformin in patients with type 2 diabetes: an observational cohort study', Diabetes Obes Metab , 17: 573-80.

Stuckey, H. L., C. B. Mullan-Jensen, G. Reach, K. Kovacs Burns, N. Piana, M. Vallis, J. Wens, I. Willaing, S. E. Skovlund, and M. Peyrot. 2014. 'Personal accounts of the negative and adaptive psychosocial experiences of people with diabetes in the second Diabetes Attitudes, Wishes and Needs (DAWN2) study', Diabetes Care , 37: 2466-74.

Stumvoll, M., H. U. Haring, and S. Matthaei. 2007. 'Metformin', Endocr Res , 32: 39-57.

Svensson, A. M., D. K. McGuire, P. Abrahamsson, and M. Dellborg. 2005. 'Association between hyper- and hypoglycaemia and 2 year all-cause mortality risk in diabetic patients with acute coronary events', Eur Heart J , 26: 1255-61. 70

Sweeting, M. J., A. J. Sutton, and P. C. Lambert. 2004. 'What to add to nothing? Use and avoidance of continuity corrections in meta-analysis of sparse data', Stat Med , 23: 1351- 75.

Szoke, E., N. R. Gosmanov, J. C. Sinkin, A. Nihalani, A. B. Fender, P. E. Cryer, C. Meyer, and J. E. Gerich. 2006. 'Effects of glimepiride and glyburide on glucose counterregulation and recovery from hypoglycemia', Metabolism , 55: 78-83.

Tang, H., Z. Fang, T. Wang, W. Cui, S. Zhai, and Y. Song. 2016. 'Meta-Analysis of Effects of Sodium-Glucose Cotransporter 2 Inhibitors on Cardiovascular Outcomes and All-Cause Mortality Among Patients With Type 2 Diabetes Mellitus', Am J Cardiol , 118: 1774-80.

Tang, Yuexin, Jinan Liu, Swapnil Rajpathak, Samuel S Engel, and Hakima Hannachi. 2017. "Temporal Changes in Inpatient and Outpatient Hypoglycemia among Patients Treated with Sulfonylureas or Dipeptidyl Peptidase-4 Inhibitors in the US." In Diabetes , A586- A86. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Taskinen, M. R., J. Rosenstock, I. Tamminen, R. Kubiak, S. Patel, K. A. Dugi, and H. J. Woerle. 2011. 'Safety and efficacy of linagliptin as add-on therapy to metformin in patients with type 2 diabetes: a randomized, double-blind, placebo-controlled study', Diabetes Obes Metab , 13: 65-74.

Taylor, S. I., J. E. Blau, and K. I. Rother. 2015. 'SGLT2 Inhibitors May Predispose to Ketoacidosis', J Clin Endocrinol Metab , 100: 2849-52.

Terra, S. G., K. Focht, M. Davies, J. Frias, G. Derosa, A. Darekar, G. Golm, J. Johnson, D. Saur, B. Lauring, and S. Dagogo-Jack. 2017. 'Phase III, efficacy and safety study of ertugliflozin monotherapy in people with type 2 diabetes mellitus inadequately controlled with diet and exercise alone', Diabetes Obes Metab , 19: 721-28.

Terra, SG, V Somayaji, S Schwartz, AJ Lewin, JG Teeter, H Dai, TT Nguyen, and RA Calle. 2011. 'A dose-ranging study of the DPP-IV inhibitor PF-734200 added to metformin in subjects with type 2 diabetes', Experimental and clinical endocrinology & diabetes , 119: 401-07.

Thibault, V., M. Belanger, E. LeBlanc, L. Babin, S. Halpine, B. Greene, and M. Mancuso. 2016. 'Factors that could explain the increasing prevalence of type 2 diabetes among adults in a Canadian province: a critical review and analysis', Diabetol Metab Syndr , 8: 71.

Tkac, I., and I. Raz. 2017. 'Combined Analysis of Three Large Interventional Trials With Gliptins Indicates Increased Incidence of Acute Pancreatitis in Patients With Type 2 Diabetes', Diabetes Care , 40: 284-86.

Tschope, D., P. Bramlage, C. Binz, M. Krekler, T. Plate, E. Deeg, and A. K. Gitt. 2011. 'Antidiabetic pharmacotherapy and anamnestic hypoglycemia in a large cohort of type 2 diabetic patients--an analysis of the DiaRegis registry', Cardiovasc Diabetol , 10: 66.

71

Tschope, D., P. Bramlage, S. Schneider, and A. K. Gitt. 2016. 'Incidence, characteristics and impact of hypoglycaemia in patients receiving intensified treatment for inadequately controlled type 2 diabetes mellitus', Diab Vasc Dis Res , 13: 2-12.

Tseng, C. L., O. Soroka, and L. M. Pogach. 2018. 'An expanded prevention quality diabetes composite: Quantifying the burden of preventable hospitalizations for older adults with diabetes', J Diabetes Complications , 32: 458-64.

Tsujimoto, T., R. Yamamoto-Honda, H. Kajio, M. Kishimoto, H. Noto, R. Hachiya, A. Kimura, M. Kakei, and M. Noda. 2014. 'Seasonal variations of severe hypoglycemia in patients with type 1 diabetes mellitus, type 2 diabetes mellitus, and non-diabetes mellitus: clinical analysis of 578 hypoglycemia cases', Medicine (Baltimore) , 93: e148.

Tsujimoto, T., R. Yamamoto-Honda, H. Kajio, M. Kishimoto, H. Noto, R. Hachiya, A. Kimura, M. Kakei, and M. Noda. 2016. 'Accelerated decline of renal function in type 2 diabetes following severe hypoglycemia', J Diabetes Complications , 30: 681-5.

Tu, J. V., A. Chu, M. R. Rezai, H. Guo, L. C. Maclagan, P. C. Austin, G. L. Booth, D. G. Manuel, M. Chiu, D. T. Ko, D. S. Lee, B. R. Shah, L. R. Donovan, Q. Z. Sohail, and D. A. Alter. 2015. 'The Incidence of Major Cardiovascular Events in Immigrants to Ontario, Canada: The CANHEART Immigrant Study', Circulation .

Tufan, F., O. Soyluk, G. Bahat, and M. A. Karan. 2016. 'Under-representation of frail older people in meta-analysis of dipeptidyl peptidase-4 inhibitors and hypoglycaemia', BMJ , 353: i3183.

Tuligenga, R. H. 2015. 'Intensive glycaemic control and cognitive decline in patients with type 2 diabetes: a meta-analysis', Endocr Connect , 4: R16-24.

Tuomi, Tiinamaija, Nicola Santoro, Sonia Caprio, Mengyin Cai, Jianping Weng, and Leif Groop. 2014. 'The many faces of diabetes: a disease with increasing heterogeneity', The Lancet , 383: 1084-94.

Umpierrez, G., S. Tofe Povedano, F. Perez Manghi, L. Shurzinske, and V. Pechtner. 2014. 'Efficacy and safety of dulaglutide monotherapy versus metformin in type 2 diabetes in a randomized controlled trial (AWARD-3)', Diabetes Care , 37: 2168-76.

Vallon, V., and S. C. Thomson. 2017. 'Targeting renal glucose reabsorption to treat hyperglycaemia: the pleiotropic effects of SGLT2 inhibition', Diabetologia , 60: 215-25. van Dijk, P., A. Bouma, G. W. Landman, K. H. Groenier, H. Bilo, N. Kleefstra, and K. J. van Hateren. 2017. 'Hypoglycemia in Frail Elderly Patients With Type 2 Diabetes Mellitus Treated With Sulfonylurea', J Diabetes Sci Technol , 11: 438-39.

Verma, S., P. Juni, and C. D. Mazer. 2019. 'Pump, pipes, and filter: do SGLT2 inhibitors cover it all?', Lancet , 393: 3-5.

72

Vilsboll, T., T. Krarup, S. Madsbad, and J. J. Holst. 2001. 'No reactive hypoglycaemia in Type 2 diabetic patients after subcutaneous administration of GLP-1 and intravenous glucose', Diabet Med , 18: 144-9.

Vilsboll, T., M. Zdravkovic, T. Le-Thi, T. Krarup, O. Schmitz, J. P. Courreges, R. Verhoeven, I. Buganova, and S. Madsbad. 2007. 'Liraglutide, a long-acting human glucagon-like peptide-1 analog, given as monotherapy significantly improves glycemic control and lowers body weight without risk of hypoglycemia in patients with type 2 diabetes', Diabetes Care , 30: 1608-10. von Birgelen, C., M. M. Kok, N. Sattar, P. Zocca, C. Doelman, G. D. Kant, M. M. Lowik, L. C. van der Heijden, H. Sen, K. G. van Houwelingen, M. G. Stoel, J. H. W. Louwerenburg, M. Hartmann, Fhaf de Man, G. C. M. Linssen, C. J. M. Doggen, and K. Tandjung. 2018. '"Silent" Diabetes and Clinical Outcome After Treatment With Contemporary Drug- Eluting Stents: The BIO-RESORT Silent Diabetes Study', JACC Cardiovasc Interv , 11: 448-59.

Wackers, F. J., L. H. Young, S. E. Inzucchi, D. A. Chyun, J. A. Davey, E. J. Barrett, R. Taillefer, S. D. Wittlin, G. V. Heller, N. Filipchuk, S. Engel, R. E. Ratner, and A. E. Iskandrian. 2004. 'Detection of silent myocardial ischemia in asymptomatic diabetic subjects: the DIAD study', Diabetes Care , 27: 1954-61.

Walz, L., B. Pettersson, U. Rosenqvist, A. Deleskog, G. Journath, and P. Wandell. 2014. 'Impact of symptomatic hypoglycemia on medication adherence, patient satisfaction with treatment, and glycemic control in patients with type 2 diabetes', Patient Prefer Adherence , 8: 593-601.

Wang, J. S., I. T. Lee, W. J. Lee, S. D. Lin, S. L. Su, S. T. Tu, Y. H. Tseng, S. Y. Lin, and W. H. Sheu. 2016. 'Glycemic excursions are positively associated with changes in duration of asymptomatic hypoglycemia after treatment intensification in patients with type 2 diabetes', Diabetes Res Clin Pract , 113: 108-15.

Wang, W., J. Yang, G. Yang, Y. Gong, S. Patel, C. Zhang, T. Izumoto, and G. Ning. 2016. 'Efficacy and safety of linagliptin in Asian patients with type 2 diabetes mellitus inadequately controlled by metformin: A multinational 24-week, randomized clinical trial', J Diabetes , 8: 229-37.

Wang, Z., and M. Liu. 2016. 'Life years lost associated with diabetes: An individually matched cohort study using the U.S. National Health Interview Survey data', Diabetes Res Clin Pract , 118: 69-76.

Wendel, C. S., G. G. Fotieo, J. H. Shah, J. Felicetta, B. H. Curtis, and G. H. Murata. 2014. 'Incidence of non-severe hypoglycaemia and intensity of treatment among veterans with type 2 diabetes in the U.S.A.: a prospective observational study', Diabet Med , 31: 1524- 31.

Weng, W., Y. Liang, E. S. Kimball, T. Hobbs, S. X. Kong, B. Sakurada, and J. Bouchard. 2016. 'Decreasing incidence of type 2 diabetes mellitus in the United States, 2007-2012: 73

Epidemiologic findings from a large US claims database', Diabetes Res Clin Pract , 117: 111-8.

White, J. L., P. Buchanan, J. Li, and R. Frederich. 2014. 'A randomized controlled trial of the efficacy and safety of twice-daily saxagliptin plus metformin combination therapy in patients with type 2 diabetes and inadequate glycemic control on metformin monotherapy', BMC Endocr Disord , 14: 17.

White, W. B., S. Kupfer, F. Zannad, C. R. Mehta, C. A. Wilson, L. Lei, G. L. Bakris, S. E. Nissen, W. C. Cushman, S. R. Heller, R. M. Bergenstal, P. R. Fleck, C. P. Cannon, and Examine Investigators. 2016. 'Cardiovascular Mortality in Patients With Type 2 Diabetes and Recent Acute Coronary Syndromes From the EXAMINE Trial', Diabetes Care , 39: 1267-73.

Whitmer, R. A., A. J. Karter, K. Yaffe, C. P. Quesenberry, Jr., and J. V. Selby. 2009. 'Hypoglycemic episodes and risk of dementia in older patients with type 2 diabetes mellitus', JAMA , 301: 1565-72.

Wilding, J. P., E. Ferrannini, V. A. Fonseca, W. Wilpshaar, P. Dhanjal, and A. Houzer. 2013. 'Efficacy and safety of ipragliflozin in patients with type 2 diabetes inadequately controlled on metformin: a dose-finding study', Diabetes Obes Metab , 15: 403-9.

Wilding, J. P., P. Norwood, C. T'Joen, A. Bastien, J. F. List, and F. T. Fiedorek. 2009. 'A study of dapagliflozin in patients with type 2 diabetes receiving high doses of insulin plus insulin sensitizers: applicability of a novel insulin-independent treatment', Diabetes Care , 32: 1656-62.

Wilding, J. P., V. Woo, N. G. Soler, A. Pahor, J. Sugg, K. Rohwedder, and S. Parikh. 2012. 'Long-term efficacy of dapagliflozin in patients with type 2 diabetes mellitus receiving high doses of insulin: a randomized trial', Ann Intern Med , 156: 405-15.

Williams, S. A., M. F. Pollack, and M. Dibonaventura. 2011. 'Effects of hypoglycemia on health- related quality of life, treatment satisfaction and healthcare resource utilization in patients with type 2 diabetes mellitus', Diabetes Res Clin Pract , 91: 363-70.

Willis, W. D., J. I. Diago-Cabezudo, A. Madec-Hily, and A. Aslam. 2013. 'Medical resource use, disturbance of daily life and burden of hypoglycemia in insulin-treated patients with diabetes: results from a European online survey', Expert Rev Pharmacoecon Outcomes Res , 13: 123-30.

Wing, R. R., M. A. Espeland, J. M. Clark, H. P. Hazuda, W. C. Knowler, H. J. Pownall, J. Unick, T. Wadden, and L. Wagenknecht. 2016. 'Association of Weight Loss Maintenance and Weight Regain on 4-Year Changes in CVD Risk Factors: the Action for Health in Diabetes (Look AHEAD) Clinical Trial', Diabetes Care , 39: 1345-55.

Wiviott, S. D., I. Raz, M. P. Bonaca, O. Mosenzon, E. T. Kato, A. Cahn, M. G. Silverman, T. A. Zelniker, J. F. Kuder, S. A. Murphy, D. L. Bhatt, L. A. Leiter, D. K. McGuire, J. P. H. Wilding, C. T. Ruff, I. A. M. Gause-Nilsson, M. Fredriksson, P. A. Johansson, A. M. 74

Langkilde, and M. S. Sabatine. 2019. 'Dapagliflozin and Cardiovascular Outcomes in Type 2 Diabetes', N Engl J Med , 380: 347-57.

Wong, M. G., V. Perkovic, J. Chalmers, M. Woodward, Q. Li, M. E. Cooper, P. Hamet, S. Harrap, S. Heller, S. MacMahon, G. Mancia, M. Marre, D. Matthews, B. Neal, N. Poulter, A. Rodgers, B. Williams, and S. Zoungas. 2016. 'Long-term Benefits of Intensive Glucose Control for Preventing End-Stage Kidney Disease: ADVANCE-ON', Diabetes Care , 39: 694-700.

Wong, S. S., N. L. Wilczynski, and R. B. Haynes. 2006. 'Developing optimal search strategies for detecting clinically sound treatment studies in EMBASE', J Med Libr Assoc , 94: 41-7.

Wright, R. J., D. E. Newby, D. Stirling, C. A. Ludlam, I. A. Macdonald, and B. M. Frier. 2010. 'Effects of acute insulin-induced hypoglycemia on indices of inflammation: putative mechanism for aggravating vascular disease in diabetes', Diabetes Care , 33: 1591-7.

Wu, W., Y. Li, X. Chen, D. Lin, S. Xiang, F. Shen, and X. Gu. 2015. 'Effect of Linagliptin on Glycemic Control in Chinese Patients with Newly-Diagnosed, Drug-Naive Type 2 Diabetes Mellitus: A Randomized Controlled Trial', Med Sci Monit , 21: 2678-84.

Wysham, C. H., J. Lin, and L. Kuritzky. 2017. 'Safety and efficacy of a glucagon-like peptide-1 receptor agonist added to basal insulin therapy versus basal insulin with or without a rapid-acting insulin in patients with type 2 diabetes: results of a meta-analysis', Postgrad Med , 129: 436-45.

Wysham, C. H., L. A. MacConell, D. G. Maggs, M. Zhou, P. S. Griffin, and M. E. Trautmann. 2015. 'Five-year efficacy and safety data of exenatide once weekly: long-term results from the DURATION-1 randomized clinical trial', Mayo Clin Proc , 90: 356-65.

Wysham, Carol H, Janusz Gumprecht, Wendy S Lane, Lone Norgard Troelsen, Deniz Tutkunkardas, and Simon Heller. 2017. "Insulin degludec (IDeg) shows consistent risk reductions across hypoglycemia definitions vs. insulin glargine U100 (IGlar) in the SWITCH 1 and 2 trials." In Diabetes , A252-A52. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Xu, S., X. Zhang, L. Tang, F. Zhang, and N. Tong. 2017. 'Cardiovascular effects of dipeptidyl peptidase-4 inhibitor in diabetic patients with and without established cardiovascular disease: a meta-analysis and systematic review', Postgrad Med , 129: 205-15.

Yabe, D., T. Eto, M. Shiramoto, S. Irie, K. Murotani, Y. Seino, H. Kuwata, T. Kurose, S. Seino, B. Ahren, and Y. Seino. 2017. 'Effects of DPP-4 inhibitor linagliptin and GLP-1 receptor agonist liraglutide on physiological response to hypoglycaemia in Japanese subjects with type 2 diabetes: A randomized, open-label, 2-arm parallel comparative, exploratory trial', Diabetes Obes Metab , 19: 442-47.

Yaffe, K., C. M. Falvey, N. Hamilton, T. B. Harris, E. M. Simonsick, E. S. Strotmeyer, R. I. Shorr, A. Metti, A. V. Schwartz, and A. B. C. Study Health. 2013. 'Association between

75

hypoglycemia and dementia in a biracial cohort of older adults with diabetes mellitus', JAMA Intern Med , 173: 1300-6.

Yale, J. F., B. Paty, and P. A. Senior. 2018. 'Hypoglycemia', Can J Diabetes , 42 Suppl 1: S104- s08.

Yang, H. K., K. W. Min, S. W. Park, C. H. Chung, K. S. Park, S. H. Choi, K. H. Song, D. M. Kim, M. K. Lee, Y. A. Sung, S. H. Baik, I. J. Kim, B. S. Cha, J. H. Park, Y. B. Ahn, I. K. Lee, S. J. Yoo, J. Kim, B. Park Ie, T. S. Park, and K. H. Yoon. 2015. 'A randomized, placebo-controlled, double-blind, phase 3 trial to evaluate the efficacy and safety of in drug-naive patients with type 2 diabetes', Endocr J , 62: 449-62.

Yang, S. J., K. W. Min, S. K. Gupta, J. Y. Park, V. K. Shivane, S. U. Pitale, P. K. Agarwal, A. Sosale, P. Gandhi, M. Dharmalingam, V. Mohan, U. Mahesh, D. M. Kim, Y. S. Kim, J. A. Kim, P. K. Kim, and S. H. Baik. 2013. 'A multicentre, multinational, randomized, placebo-controlled, double-blind, phase 3 trial to evaluate the efficacy and safety of (LC15-0444) in patients with type 2 diabetes', Diabetes Obes Metab , 15: 410-6.

Yang, W., Y. Guan, Y. Shentu, Z. Li, A. O. Johnson-Levonas, S. S. Engel, K. D. Kaufman, B. J. Goldstein, and M. Alba. 2012. 'The addition of sitagliptin to ongoing metformin therapy significantly improves glycemic control in Chinese patients with type 2 diabetes', J Diabetes , 4: 227-37.

Yang, W., P. Han, K. W. Min, B. Wang, T. Mansfield, C. T'Joen, N. Iqbal, E. Johnsson, and A. Ptaszynska. 2016. 'Efficacy and safety of dapagliflozin in Asian patients with type 2 diabetes after metformin failure: A randomized controlled trial', J Diabetes , 8: 796-808.

Yang, W., C. Y. Pan, C. Tou, J. Zhao, and I. Gause-Nilsson. 2011. 'Efficacy and safety of saxagliptin added to metformin in Asian people with type 2 diabetes mellitus: a randomized controlled trial', Diabetes Res Clin Pract , 94: 217-24.

Yang, Y., S. Chen, H. Pan, Y. Zou, B. Wang, G. Wang, and H. Zhu. 2017. 'Safety and efficiency of SGLT2 inhibitor combining with insulin in subjects with diabetes: Systematic review and meta-analysis of randomized controlled trials', Medicine (Baltimore) , 96: e6944.

Yau, R. K., E. S. Strotmeyer, H. E. Resnick, D. E. Sellmeyer, K. R. Feingold, J. A. Cauley, E. Vittinghoff, N. De Rekeneire, T. B. Harris, M. C. Nevitt, S. R. Cummings, R. I. Shorr, and A. V. Schwartz. 2013. 'Diabetes and risk of hospitalized fall injury among older adults', Diabetes Care , 36: 3985-91.

Yokomoto-Umakoshi, M., I. Kanazawa, S. Kondo, and T. Sugimoto. 2017. 'Association between the risk of falls and osteoporotic fractures in patients with type 2 diabetes mellitus', Endocr J , 64: 727-34.

Yoshikawa, Fukumi, Hiroshi Uchino, Tomoko Nagashima, Ken Kanazawa, Fumika Shigiyama, Shuki Usui, Masahiko Miyagi, Naoki Kumashiro, Hiroshi Yoshino, and Yasuyo Ando. 2017. "Glycemic Variability in Metformin/DPP-4 Inhibitor Compared with the High- 76

Dose Metformin in Patients with Multiple-Daily-Insulin-Treated Type 2 Diabetes: A Prospective, Randomized, Controlled Trial." In Diabetes , A235-A35. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Young, T. K., and B. A. Roche. 1990. 'Factors associated with clinical gallbladder disease in a Canadian Indian population', Clin Invest Med , 13: 55-9.

Yu, S., A. Z. Fu, S. S. Engel, R. R. Shankar, and L. Radican. 2016. 'Association between hypoglycemia risk and hemoglobin A1C in patients with type 2 diabetes mellitus', Curr Med Res Opin , 32: 1409-16.

Yue, X. D., J. Y. Wang, X. R. Zhang, J. H. Yang, C. Y. Shan, M. Y. Zheng, H. Z. Ren, Y. Zhang, S. H. Yang, Z. H. Guo, B. Chang, and B. C. Chang. 2017. 'Characteristics and Impact Factors of Renal Threshold for Glucose Excretion in Patients with Type 2 Diabetes Mellitus', J Korean Med Sci , 32: 621-27.

Yun, J. S., and S. H. Ko. 2015. 'Avoiding or coping with severe hypoglycemia in patients with type 2 diabetes', Korean J Intern Med , 30: 6-16.

Zaccardi, F., M. J. Davies, N. N. Dhalwani, D. R. Webb, G. Housley, D. Shaw, J. W. Hatton, and K. Khunti. 2016. 'Trends in hospital admissions for hypoglycaemia in England: a retrospective, observational study', Lancet Diabetes Endocrinol , 4: 677-85.

Zammitt, N. N., and B. M. Frier. 2005. 'Hypoglycemia in type 2 diabetes: pathophysiology, frequency, and effects of different treatment modalities', Diabetes Care , 28: 2948-61.

Zelniker, T. A., S. D. Wiviott, I. Raz, K. Im, E. L. Goodrich, M. P. Bonaca, O. Mosenzon, E. T. Kato, A. Cahn, R. H. M. Furtado, D. L. Bhatt, L. A. Leiter, D. K. McGuire, J. P. H. Wilding, and M. S. Sabatine. 2019. 'SGLT2 inhibitors for primary and secondary prevention of cardiovascular and renal outcomes in type 2 diabetes: a systematic review and meta-analysis of cardiovascular outcome trials', Lancet , 393: 31-39.

Zhang, Y., H. Wieffer, R. Modha, B. Balar, M. Pollack, and G. Krishnarajah. 2010. The Burden of hypoglycemia in type 2 diabetes: A systematic review of patient and economic perspectives .

Zhang, Z., X. Chen, P. Lu, J. Zhang, Y. Xu, W. He, M. Li, S. Zhang, J. Jia, S. Shao, J. Xie, Y. Yang, and X. Yu. 2017. 'Incretin-based agents in type 2 diabetic patients at cardiovascular risk: compare the effect of GLP-1 agonists and DPP-4 inhibitors on cardiovascular and pancreatic outcomes', Cardiovasc Diabetol , 16: 31.

Zhang, Z., J. Lovato, H. Battapady, C. Davatzikos, H. C. Gerstein, F. Ismail-Beigi, L. J. Launer, A. Murray, Z. Punthakee, A. A. Tirado, J. Williamson, R. N. Bryan, and M. E. Miller. 2014. 'Effect of hypoglycemia on brain structure in people with type 2 diabetes: epidemiological analysis of the ACCORD-MIND MRI trial', Diabetes Care , 37: 3279-85.

77

Zhao, Y., S. Kachroo, H. Kawabata, S. Colilla, J. Mukherjee, V. Fonseca, U. Iloeje, and L. Shi. 2016. 'Association between Hypoglycemia and Fall-Related Fractures and Health Care Utilization in Older Veterans with Type 2 Diabetes', Endocr Pract , 22: 196-204.

Zhong, X., D. Lai, Y. Ye, X. Yang, B. Yu, and Y. Huang. 2016. 'Efficacy and safety of empagliflozin as add-on to metformin for type 2 diabetes: a systematic review and meta- analysis', Eur J Clin Pharmacol , 72: 655-63.

Zhou, Fang Liz, Fen Ye, Vineet Gupta, Rishab Gupta, Jennifer Sung, Paulos Berhanu, and Lawrence Blonde. 2017. "Older adults with type 2 diabetes (T2D) experience less hypoglycemia when switching to insulin glargine 300 U/mL (Gla-300) vs. other basal insulins (DELIVER 3 Study)." In Diabetes , A256-A56. AMER DIABETES ASSOC 1701 N BEAUREGARD ST, ALEXANDRIA, VA 22311-1717 USA.

Zick, R., B. Petersen, M. Richter, and C. Haug. 2007. 'Comparison of continuous blood glucose measurement with conventional documentation of hypoglycemia in patients with Type 2 diabetes on multiple daily insulin injection therapy', Diabetes Technol Ther , 9: 483-92.

Zinman, B., S. P. Marso, E. Christiansen, S. Calanna, S. Rasmussen, and J. B. Buse. 2018. 'Hypoglycemia, Cardiovascular Outcomes, and Death: The LEADER Experience', Diabetes Care , 41: 1783-91.

Zinman, B., C. Wanner, J. M. Lachin, D. Fitchett, E. Bluhmki, S. Hantel, M. Mattheus, T. Devins, O. E. Johansen, H. J. Woerle, U. C. Broedl, S. E. Inzucchi, and Empa-Reg Outcome Investigators. 2015. 'Empagliflozin, Cardiovascular Outcomes, and Mortality in Type 2 Diabetes', N Engl J Med , 373: 2117-28.

Zobel, E. H., T. W. Hansen, P. Rossing, and B. J. von Scholten. 2016. 'Global Changes in Food Supply and the Obesity Epidemic', Curr Obes Rep , 5: 449-55.

Zorzela, L., Y. K. Loke, J. P. Ioannidis, S. Golder, P. Santaguida, D. G. Altman, D. Moher, S. Vohra, and P. RISMAHarms Group. 2016. 'PRISMA harms checklist: improving harms reporting in systematic reviews', BMJ , 352: i157.

Zoungas, S., H. Arima, H. C. Gerstein, R. R. Holman, M. Woodward, P. Reaven, R. A. Hayward, T. Craven, R. L. Coleman, and J. Chalmers. 2017. 'Effects of intensive glucose control on microvascular outcomes in patients with type 2 diabetes: a meta-analysis of individual participant data from randomised controlled trials', Lancet Diabetes Endocrinol , 5: 431-37.

Zoungas, S., A. Patel, J. Chalmers, B. E. de Galan, Q. Li, L. Billot, M. Woodward, T. Ninomiya, B. Neal, S. MacMahon, D. E. Grobbee, A. P. Kengne, M. Marre, and S. Heller. 2010. 'Severe hypoglycemia and risks of vascular events and death', N Engl J Med , 363: 1410- 8.

78

Appendices

Appendix I: Scoping Review Flow Diagram

PRISMA FLOW SCOPING REVIEW

Systematic Reviews Records identified from PUBMED (=99)

Reasons for exclusion (n=40): Study Design: - Network Meta-Analysis (n=22)

Identification - Systematic Review (no pooling) (n=10) Background AHA’s allowed: - CKD population (n=7) - CVOT trials (n=1)

Records screened (n=59)

Reasons for exclusion (n=27): Comparator: - SU (n=9) - Insulin (n=2)

- Any Active (n=9) - Other injectables (n=1) - non-SGLT2i (n=1) - non-insulin (n=1) Screening - metformin (n=2) - DPP4i (n=1) - GLP1RA (n=1)

Reasons for exclusion (n=30): Background: - SU (n=1) - Insulin (n=4) - Any (n=25) - Hypo not meta-analyzed (n=1)

Records screened (n=2) Eligibility

DPP4i (n=1) DPP4i or SGLT2i (n=1) Included Liu et al. 2014 Kawalec et al. 2014

79

Appendix II: PRISMA 2015 Checklist

PRISMA-P 2015 Checklist

This checklist has been adapted for use with protocol submissions to Systematic Reviews from Table 3 in Moher D et al: Preferred reporting items for systematic review and meta- analysis protocols (PRISMA-P) 2015 statement. Systematic Reviews 2015 4:1

Information Line Section/topic # Checklist item reported number(s) Yes No ADMINISTRATIVE INFORMATION Title Identify the report as a protocol of a 1-3 Identification 1a systematic review If the protocol is for an update of a previous Update 1b systematic review, identify as such If registered, provide the name of the 36 Registration 2 registry (e.g., PROSPERO) and registration number in the Abstract Authors Provide name, institutional affiliation, and e- 4, 248-263 mail address of all protocol authors; provide Contact 3a physical mailing address of corresponding author Describe contributions of protocol authors 265-267 Contributions 3b and identify the guarantor of the review If the protocol represents an amendment of a previously completed or published Amendments 4 protocol, identify as such and list changes; otherwise, state plan for documenting important protocol amendments Support Indicate sources of financial or other 269-270 Sources 5a support for the review Provide name for the review funder and/or Sponsor 5b sponsor Describe roles of funder(s), sponsor(s), Role of 5c and/or institution(s), if any, in developing sponsor/funder the protocol INTRODUCTION Describe the rationale for the review in the 39-89 Rationale 6 context of what is already known Provide an explicit statement of the 90-93 Objectives 7 question(s) the review will address with 80

Information Line Section/topic # Checklist item reported number(s) Yes No reference to participants, interventions, comparators, and outcomes (PICO)

METHODS Specify the study characteristics (e.g., 97-140 PICO, study design, setting, time frame) and report characteristics (e.g., years Eligibility criteria 8 considered, language, publication status) to be used as criteria for eligibility for the review Describe all intended information sources 142-152, (e.g., electronic databases, contact with Information 9 study authors, trial registers, or other grey sources literature sources) with planned dates of coverage Present draft of search strategy to be used 246-247 for at least one electronic database, Search strategy 10 including planned limits, such that it could be repeated STUDY RECORDS Describe the mechanism(s) that will be 153-209 Data 11a used to manage records and data management throughout the review State the process that will be used for 154-160 selecting studies (e.g., two independent Selection 11b reviewers) through each phase of the process review (i.e., screening, eligibility, and inclusion in meta-analysis) Describe planned method of extracting data 161-166, from reports (e.g., piloting forms, done 190-192 Data collection 11c independently, in duplicate), any processes process for obtaining and confirming data from investigators List and define all variables for which data 161-166 will be sought (e.g., PICO items, funding Data items 12 sources), any pre-planned data assumptions and simplifications List and define all outcomes for which data 134-140 Outcomes and will be sought, including prioritization of 13 prioritization main and additional outcomes, with rationale Describe anticipated methods for assessing 170-178 Risk of bias in 14 risk of bias of individual studies, including individual studies whether this will be done at the outcome or

81

Information Line Section/topic # Checklist item reported number(s) Yes No study level, or both; state how this information will be used in data synthesis DATA Describe criteria under which study data 179-192 15a will be quantitatively synthesized If data are appropriate for quantitative 193-196, synthesis, describe planned summary 201-209 measures, methods of handling data, and 15b methods of combining data from studies, Synthesis including any planned exploration of consistency (e.g., I 2, Kendall’s tau) Describe any proposed additional analyses 220-232 15c (e.g., sensitivity or subgroup analyses, meta-regression) If quantitative synthesis is not appropriate, 15d describe the type of summary planned Specify any planned assessment of meta- 197-200 Meta-bias(es) 16 bias(es) (e.g., publication bias across studies, selective reporting within studies) Confidence in 210-216 Describe how the strength of the body of cumulative 17 evidence will be assessed (e.g., GRADE) evidence

82

Appendix III: Medline OVID Search Strategy

MEDLINE OVID HSSS 6.4.d sensitivity and precision max

1. randomized controlled 2. controlled clinical trial.pt. 3. randomized.ab. trial.pt.

4. placebo.ab. 5. clinical trials as topic.sh. 6. randomly.ab.

7. trial.ti. 8. 1 or 2 or 3 or 4 or 5 or 6 or 9. exp animals/ not 7 humans.sh.

10. 8 not 9 11. exp Diabetes Mellitus, 12. (type adj2 ("2" or "II" or Type 2/ "two") adj2 (diabet* or DM)).mp.

13. ((noninsulin or (non adj1 14. NIDDM*.mp. 15. T2D*.mp. insulin)) adj2 depend* adj2 diabet*).mp.

16. or/11-15 17. Biguanides/ 18. exp metformin/

19. biguanide*.mp. 20. metformin*.mp. 21. metaformin*.mp.

22. dimethylbiguanid*.mp. 23. dimethylguanid*.mp. 24. Glifor*.mp.

25. metfogamma*.mp. 26. Siofor*.mp. 27. Diaformin*.mp.

28. Diabex*.mp. 29. Dianben*.mp. 30. Gluformin*.mp.

31. Obimet*.mp. 32. Glumetza*.mp. 33. Fortamet*.mp.

34. Riomet*.mp. 35. Carbophage*.mp. 36. Glucophage*.mp.

83

37. or/17-36 38. exp Dipeptidyl Peptidase 39. exp Dipeptidyl-Peptidase 4/ IV Inhibitors/

40. ((dipeptidylpeptidase* or 41. gliptin*.mp. 42. exp sitagliptin (dipeptidyl adj1 peptidase*) phosphate/ or dpp*) adj2 ("4" or "iv" or "four") adj2 inhibit*).mp.

43. sitagliptin*.mp. 44. januvia*.mp. 45. janumet*.mp.

46. tesavel*.mp. 47. xelevia*.mp. 48. MK0431.mp.

49. (MK adj2 "0431").mp. 50. istavel*.mp. 51. istamet*.mp.

52. saxagliptin*.mp. 53. onglyza*.mp. 54. BMS477118.mp.

55. (BMS adj2 "477118").mp. 56. kombiglyze*.mp. 57. exp linagliptin/

58. linagliptin*.mp. 59. trajenta*.mp. 60. tradjenta*.mp.

61. jentadueto*.mp. 62. jenttadueto*.mp. 63. Ondero*.mp.

64. BI1356.mp. 65. (BI adj2 "1356").mp. 66. alogliptin*.mp.

67. allogliptin*.mp. 68. SYR322.mp. 69. (SYR adj2 "322").mp.

70. kazano*.mp. 71. nesina*.mp. 72. vipidia*.mp.

73. vildagliptin*.mp. 74. galvus*.mp. 75. jalra*.mp.

76. xiliarx*.mp. 77. LAF237.mp. 78. (LAF adj2 "237").mp.

79. eucreas*.mp. 80. galvumet*.mp. 81. teneligliptin*.mp.

84

82. tenelia*.mp. 83. MP513.mp. 84. (MP adj2 "513").mp.

85. dutogliptin*.mp. 86. PHX1149T.mp. 87. (PHX adj2 1149T).mp.

88. omarigliptin*.mp. 89. MK3102.mp. 90. (MK adj2 "3102").mp.

91. gemigliptin*.mp. 92. LC150444.mp. 93. (LC adj2 "150444").mp.

94. (LC15 adj2 "0444").mp. 95. anagliptin*.mp. 96. suiny*.mp.

97. K726J96838.mp. 98. evogliptin*.mp. 99. suganon*.mp.

100. DA1229.mp. 101. (DA adj2 "1229").mp. 102. retagliptin*.mp.

103. SP2086.mp. 104. (SP adj2 "2086").mp. 105. carmegliptin*.mp.

106. camegliptin*.mp. 107. R1579.mp. 108. RG1579.mp.

109. (RG adj2 "1579").mp. 110. RO4876904.mp. 111. (RO adj2 "4876904").mp.

112. trelagliptin*.mp. 113. zafatek*.mp. 114. SYR472.mp.

115. (SYR adj2 "472").mp. 116. bisegliptin*.mp. 117. denagliptin*.mp.

118. GW823093.mp. 119. (GW adj2 "823093").mp. 120. (psn adj2 "9301").mp.

121. psn9301.mp. 122. *.mp. 123. PF00734200.mp.

124. (PF adj2 125. (PF adj2 "734200").mp. 126. melogliptin*.mp. "00734200").mp.

127. GRC8200.mp. 128. (GRC adj2 "8200").mp. 129. septagliptin*.mp.

85

130. or/38-129 131. exp Glucagon-Like 132. exp Glucagon like Peptide 1/ peptide 1 receptor/

133. (((glucagon adj1 like 134. exp liraglutide/ 135. liraglutide*.mp. adj1 peptide) or GLP) adj2 ("1" or "one") adj4 (agonist* or analog* or RA*)).mp.

136. victoza*.mp. 137. saxenda*.mp. 138. NN2211.mp.

139. (NN adj2 "2211").mp. 140. NNC901170.mp. 141. (NNC adj1 "90" adj1 "1170").mp.

142. (NNC90 adj1 143. exenatide*.mp. 144. byetta*.mp. "1170").mp.

145. bydureon*.mp. 146. AC2993.mp. 147. (AC adj2 "2993").mp.

148. AC3174.mp. 149. (AC adj2 "3174").mp. 150. ITCA650.mp.

151. (ITCA adj2 "650").mp. 152. lixisenatide*.mp. 153. lyxumia*.mp.

154. adlyxin*.mp. 155. AVE0010.mp. 156. (AVE adj2 "0010").mp.

157. ZP10A.mp. 158. (ZP adj2 "10A").mp. 159. dulaglutide*.mp.

160. trulicity*.mp. 161. LY2189265.mp. 162. (LY adj2 "2189265").mp.

163. albiglutide*.mp. 164. tanzeum*.mp. 165. GSK716155.mp.

166. (GSK adj2 167. eperzan*.mp. 168. taspoglutide*.mp. "716155").mp.

86

169. (BIM adj2 "51077").mp. 170. BIM51077.mp. 171. (ITM077 or (ITM adj2 "077") or R1583 or (R adj2 "1583") or RO5073031 or (RO adj2 "5073031")).mp.

172. (RO adj2 173. RO5073031.mp. 174. (R adj2 "1583").mp. "5073031").mp.

175. R1583.mp. 176. (ITM adj2 "077").mp. 177. ITM077.mp.

178. semaglutide*.mp. 179. ozempic*.mp. 180. NN9535.mp.

181. (NN adj2 "9535").mp. 182. elsiglutide*.mp. 183. albenatide*.mp.

184. cjc1131.mp. 185. "cjc ad2 1131".mp. 186. efpeglenatide*.mp.

187. HM11260C.mp. 188. (HM adj2 "11260C").mp. 189. LY307161.mp.

190. (LY adj2 "307161").mp. 191. pegapamodutide*.mp. 192. or/131-191

193. exp Sodium-Glucose 194. exp sodium-glucose 195. gliflozin*.mp. Transport Proteins/ transporter 2/

196. (sodium* adj4 glucose* 197. (sglt* adj4 inhibit*).mp. 198. exp Canagliflozin/ adj4 (cotransport* or (co adj2 transport*) or transport*) adj4 inhibit*).mp.

199. canagliflozin*.mp. 200. invokana*.mp. 201. JNJ24831754.mp.

87

202. (JNJ adj2 203. TA7284.mp. 204. (TA adj2 "7284").mp. "24831754").mp.

205. invokamet*.mp. 206. dapagliflozin*.mp. 207. farxiga*.mp.

208. forxiga*.mp. 209. xigduo*.mp. 210. BMS512148.mp.

211. (BMS adj2 212. qtern*.mp. 213. empagliflozin*.mp. "512148").mp.

214. jardiance*.mp. 215. synjardy*.mp. 216. (BI adj2 "10773").mp.

217. BI10773.mp. 218. BI4487.mp. 219. (BI adj2 "4487").mp.

220. glyxambi*.mp. 221. ertugliflozin*.mp. 222. PF049717129.mp.

223. (PF adj2 224. MK8835.mp. 225. (MK adj2 "8835").mp. "049717129").mp.

226. steglujan*.mp. 227. steglatro*.mp. 228. segluromet*.mp.

229. steglaro*.mp. 230. ipragliflozin*.mp. 231. suglat*.mp.

232. ASP1941.mp. 233. (ASP adj2 "1941").mp. 234. remogliflozin*.mp.

235. GSK189075.mp. 236. (GSK adj2 237. tofogliflozin*.mp. "189075").mp.

238. apleway*.mp. 239. deberza*.mp. 240. CSG452.mp.

241. (CSG adj2 "452").mp. 242. R7201.mp. 243. (R adj2 "7201").mp.

244. atigliflozin*.mp. 245. bexagliflozin*.mp. 246. EGT0001474.mp.

88

247. (EGT adj2 248. EGT0001442.mp. 249. (EGT adj2 "0001474").mp. "0001442").mp.

250. EGT1442.mp. 251. (EGT adj2 "1442").mp. 252. luseogliflozin*.mp.

253. lusefi*.mp. 254. (TS adj2 "071").mp. 255. TS071.mp.

256. sergliflozin*.mp. 257. GW869682X.mp. 258. (GW adj2 869682X).mp.

259. *.mp. 260. lx4211.mp. 261. (LX adj2 "4211").mp.

262. velagliflozin*.mp. 263. (ISIS adj2 "388626").mp. 264. ISIS388626.mp.

265. (ISIS adj2 SGLT2Rx).mp. 266. henagliflozin*.mp. 267. SHR3824.mp.

268. (SHR adj2 "3824").mp. 269. mizagliflozin*.mp. 270. GSK1614235.mp.

271. (GSK adj2 272. BI44847.mp. 273. (BI adj2 "44847").mp. "1614235").mp.

274. or/193-273 275. 37 or 130 or 192 or 274 276. 10 and 16 and 275

89

Appendix IV: PRISMA Flow Diagram

Records identified from OVID Records identified from Records identified from Cochrane MEDLINE (n=4,742) EMBASE (n=5,484) Library (n=5,915)

Records identified through database searching (n=16,141) Identification

Duplicate records removed (n=7,167)

Records screened

(n=8,974) Reasons for exclusion (n=8710) : - Not AHA of Interest (n=2936) - Background Therapy (n=133) - Not Pbo Comparator (n=204) Screening - Not in English (n=264) - Not T2DM (n=894) - Not an RCT (n=3968) Full -text articles - Study Design (n=311)

assessed for eligibility (n=264)

Eligibility Additional records identified through manual search (n=3 9)

Reasons for exclusion (n=162): - Background Therapy (n=27)

- Duplicate (n=40) - Hypo not Evaluated (n=29) - Study Design (n=66) Included Studies included in qualitative synthesis (n= 141 )

Met DPP4i DPP4i GLP1RA GLP1RA SGLT2i SGLT2i + Dual Triple Non-Met MONO MONO + MET MONO + MET MONO MET Initiation Therapy Dual (n=8) (n=23) (n=45) (n=9) (n=8) (n=17) (n=15) (n=6) (n=7) (n=3)

90

Appendix V: Excluded Studies

Excluded Studies (which at first appear eligible)

Study Reason for exclusion

Ahren et al. 2004 4-weeks only

Inhibition of dipeptidyl peptidase-4 reduces glycemia, sustains insulin levels, and reduces glucagon levels in type 2 diabetes

Barnett et al. 2012 Hypoglycemia reported for extension Linagliptin monotherapy in type 2 diabetes patients for whom phase metformin is inappropriate: an 18-week randomized, double-blind, placebo-controlled phase III trial with a 34-week active-controlled extension

Bode et al. 2013 Background therapy allowed Efficacy and safety of canagliflozin treatment in older subjects with type 2 diabetes mellitus: a randomized trial

Chacra et al. 2017 Background therapies

A randomized, double-blind, trial of the safety and efficacy of

omarigliptin (a once-weekly DPP-4 inhibitor) in subjects with

type 2 diabetes and renal impairment

Dou et al. 2017 Not vs placebo

Efficacy and safety of saxagliptin in combination with

metformin as initial therapy in Chinese patients with type

91

2 diabetes: Results from the START study, a multicentre, randomized, double-blind, active-controlled, phase 3 trial

Frederich et al. 2012 Hypo data included extension phase The efficacy and safety of the dipeptidyl peptidase-4 inhibitor saxagliptin in treatment-naive patients with type 2 diabetes mellitus: a randomized controlled trial

Garber et al. 1997 Metformin open label

Efficacy of Metformin in Type II Diabetes: Results of a Double- Blind Placebo-controlled Dose-Response Trial

Hadjadj et al. 2016 Not vs placebo

Initial Combination of Empagliflozin and Metformin in Patients With Type 2 Diabetes

Hoffman et al. 1997 Open-label

Efficacy of 24-week Monotherapy with Acarbose, Metformin or Placebo in Dietary-treated NIDDM Patients: The Essen II Study

Jadzinsky et al. 2009 Not vs placebo

Saxagliptin given in combination with metformin as initial therapy improves glycemic control in patients with type 2 diabetes compared with either monotherapy: a randomized controlled trial

Ji et al. 2015 Not vs placebo

Efficacy and safety of linagliptin co-administered with low-dose metformin once daily versus high-dose metformin twice daily in

92

treatment-naive patients with type 2 diabetes: a double-blind randomized trial

Kadowaki et al. 2009 Background Therapies

Exenatide Exhibits Dose-Dependent Effects on Glycemic Control over 12 weeks in Japanese Patients with Suboptimally controlled T2DM

Kaku et al. 2016 Not vs placebo

Randomized, double-blind, phase III study to evaluate the efficacy and safety of once-daily treatment with alogliptin and metformin hydrochloride in Japanese patients with type

2 diabetes

Kutoh et al. 2012 Not vs placebo

Alogliptin as an initial therapy in patients with newly diagnosed, drug naı¨ve type 2 diabetes: a randomized, control trial

Lambers Heerspink et al. 2013 Background Therapies

Dapagliflozin a glucose-regulating drug with diuretic properties in subjects with type 2 diabetes

Lavalle-Gonzalez et al. 2013 Hypoglycemia reported only for the Efficacy and safety of canagliflozin compared with placebo and extension phase sitagliptin in patients with type 2 diabetes on background metformin monotherapy: a randomised trial

Lewin et al. 2015 Not vs placebo

93

Initial Combination of Empagliflozin and Linagliptin in Subjects With Type 2 Diabetes

Lim et al. 2017 Not vs placebo

Efficacy and safety of initial combination therapy with gemigliptin and metformin compared with monotherapy with either drug in patients with type 2 diabetes: A double-blind randomized controlled trial (INICOM study)

Lu et al. 2013 Background Therapies

Safety and efficacy of twice-daily exenatide in Taiwanese patients with inadequately controlled type 2 diabetes mellitus

Lukashevich et al. 2011 Background Therapies

Safety and efficacy of vildagliptin versus placebo in patients with type 2 diabetes and moderate or severe renal impairment: a prospective 24-week randomized placebo-controlled trial

Meneilly et al. 2003 Background Therapies

Effects of 3 Months of Continuous Subcutaneous Administration of Glucagon-Like Peptide 1 in Elderly Patients With Type 2 Diabetes

Mu et al. 2017 Not vs placebo

Efficacy and safety of linagliptin/metformin single-pill combination as initial therapy in drug-naı¨ve Asian patients with type 2 diabetes

Park et al. Not vs placebo

94

Effect of gemigliptin on glycemic variability in patients with type 2 diabetes

Strain et al. 2013 Background

Individualised treatment targets for elderly patients with type 2 Therapies diabetes using vildagliptin add-on or lone therapy (INTERVAL): a 24 week, randomised, double-blind, placebo-controlled study

Su et al. 2014 Background

A randomized controlled clinical trial of vildagliptin plus metformin Therapies combination therapy in patients with type II diabetes mellitus

TinaJones et al. 2016 Open-label empa

Linagliptin as add-on to empagliflozin and metformin in In both arms patients with type 2 diabetes: Two 24-week randomized, double-blind, double-dummy, parallel-group trials

Umpierrez et al. 2011 Background

The effects of LY2189265, a long-acting glucagon-like peptide-1 Therapies analogue, in a randomized, placebo-controlled, double-blind study of overweight/obese patients with type 2 diabetes: the EGO study

Yale et al. 2014 Background

Efficacy and safety of canagliflozin over 52 weeks in patients with Therapies type 2 diabetes mellitus and chronic kidney disease

Yamout et al. 2014 Background

Therapies

95

Efficacy and Safety of Canagliflozin in Patients with Type 2 Diabetes and Stage

3 Nephropathy

Yoon et al. 2017 Background

Efficacy, safety and albuminuria-reducing effect of gemigliptin in Therapies Korean type 2 diabetes patients with moderate to severe renal impairment: A 12-week, double-blind randomized study (the GUARD Study)

96

Appendix VI: Characteristics of Included Studies

Characteristics of Included Studies – Metformin Monotherapy

Study ID Dose Study n= Mean Age (SD) Gender (% male) Countries Ethnicity (%) Duration of Diabetes in Duration Studied Years (SD)

Chiasson met 500mg 36 weeks met 500mg met 500mg TID: 57.9 met 500mg TID: 73.5; pbo: Canada met 500mg TID: Caucasian 88, Black met 500mg TID: 7.5 (7.4); JL 2001 TID TID: 83; (8.6); pbo: 57.7 (9.9) 67.5 1.2, Asian 7.2, Other 3.6; pbo: pbo: 5.1 (4.9) pbo: 83 Caucasian 91.6, Black 1.2, Asian 4.8, Other 2.4

DeFronzo met 850mg 29 weeks met 850mg met 850mg TID: 53 met 850mg TID: 43.4; pbo: USA White, Black, Hispanic ( n not given) met 850mg TID: 6.0 (0.5); RA 1995 TID TID: 143; (1); pbo: 53 (1) 42.5 pbo: 6.0 (0.6) pbo: 146

Fonseca met 1500mg 12 weeks met met 1500mg: 53.1 met 1500mg 58; pbo: 46.4 USA, met 1500mg: White 46.4, non-White met 1500mg: 4.13 (4.71); VA 2013 QD 1500mg: (11.7); pbo: 53.4 (9.7) Mexico, 53.6; pbo: White 47.8, non-White 52.2 pbo: 4.64 (5.93) 69; pbo: Columbia, 69 India, Philippines

Goldstein met 500mg 24 weeks met 500mg met 500mg bid: 53.4 met 500mg bid: 48.9; met multinational; met 500mg bid: White 47.8, Black 6.6, met 500mg bid: 4.5 (3.9); BJ 2007 BID; met bid: 182; (10.2); met 1000mg 1000mg bid: 45.1; placebo: not specified Hispanic 30.2, Asian 7.7, Other 7.7; met met 1000mg bid: 4.4 (4.4); 1000mg BID met bid: 53.2 (9.6); 52.8 1000mg bid: White 58.2, Black 4.9, placebo: 4.6 (4.9) 1000mg placebo: 53.6 (10.0) Hispanic 21.4, Asian 5.5, Other 9.9; bid: 182; placebo: White 46.0, Black 9.7, placebo: Hispanic 26.7, Asian 6.8, Other 10.8 176

97

Haak T met 500mg 24 weeks met 500mg met 500mg bid: 52.9 met 500mg bid: 56.9; met 14 countries; met 500mg bid: White 64.6, Asian 35.4, met 500mg BID: ≤1: 41.1, 2012 BID; met bid: 144; (10.4); met 1000mg 1000mg bid: 53.1; pbo: not specified Black 0, Hawaiian/Pacific Islander 0.0; > 1 to 5: 35.5, > 5: 23.4; 1000mg BID met bid: 55.2 (10.6); pbo: 50.0 met 1000mg bid; White 64.6, Asian met 1000mg BID: ≤1: 1000mg 55.7 (11.0) 34.0, Black 1.4, Hawaiian/Pacific 34.8, > 1 to 5: 45.7, > 5: bid: 147; Islander 0.0; pbo: white 63.9, Asian 19.6; pbo: ≤1: 30.8, > 1 to pbo: 72 36.1, Black 0, Hawaiian/Pacific Islander 5: 35.4, > 5: 33.8 0.0

Ji L 2016 met 500mg 24 weeks met 500mg met 500mg BID: 52.6 met 500mg BID: 54.8; met China met 500mg BID: Asian: 100; met met 500mg BID: 1.0 (0.2); BID; met BID: 126; (9.5); met 850mg 850mg BID: 60.5; pbo: 850mg BID Asian: 100; pbo Asian 100 met 850mg BID: 1.1 (0.2); 850mg BID met 850mg BID: 53.0 (10.3); 68.5 pbo 1.1 (0.2) BID: 124; pbo: 53.6 (9.7) pbo: 126

Ji L 2017 met 500mg 26 weeks met met 500mg BID: met 500mg BID: 50.6; China met 500mg BID: Asian 99.4, American not reported BID 500mg 53.6 (9.91); pbo: 52.2 pbo: 58.3 Indian or Alaskan Native 0, multiracial BID: 161; (10.17) 0.6; pbo: Asian 98.8, American Indian pbo: 161 or Alaskan Native 1.2, Multiracial 0

Pratley RE met 500mg 26 weeks met 500mg met 500mg BID: 54.6 met 500mg BID: 41.2; met worldwide; [RACE] met 500mg BID: Asian 16.7, met 500mg BID: 3.8 (3.9); 2014 BID; met BID: 114; (10.20); met 1000mg 1000mg BID: 45.9; pbo: not specified Black or African American 5.3, White met 1000mg BID: 4.1 1000mg BID met BID: 52.6 (11.30); 50.5 74.6, Other 3.5; met 1000mg BID: (4.59); pbo: 4.3 (4.78) 1000mg pbo: 53.1 (9.60) Asian 18.0, Black or African American BID: 111; 5.4, White 71.2, Other 5.4; pbo: Asian pbo: 109 18.3, Black or African American 7.3, White 69.7, Other 4.6

98

Characteristics of Included Studies – Metformin Monotherapy continued

Study ID Baseline Mean Change Mean Difference HYPO def Ascertainment of Hypo Rescue Medication Excl of Pts w events? A1C % in HbA1C % in HbA1c % vs At screening or (SD) vs. baseline pbo during study)

Chiasson met 500mg met 500mg TID: met 500mg TID: - not reported Patient-reported not reported no JL 2001 TID: 8.2 -0.85 (0.12); 1.25 (n/a) (0.9); pbo: pbo: 0.38 (0.12) 8.1 (0.7)

DeFronzo met 850mg met 850mg TID: not reported not reported Report of symptoms compatible with not reported no RA 1995 TID: 8.4 -1.4 (0.1); pbo: hypoglycemia; no biochemical (0.1); pbo: 0.4 (0.1) (not documentation of hypoglycemia 8.2 (0.2) specified as necessary. LSM)

Fonseca met 1500mg: [LSM] met [LSM] met Serious defined by development Patient-reported (blood glucose not reported no VA 2013 8.03 (0.90); 1500mg: pbo: 1500mg: -0.72 (- of hypoglycemic coma requiring unconfirmed) Patients were provided pbo: 7.84 0.26 0.97 to -0.48) hospitalization or requiring with a glucometer and asked to (0.78) discontinuation of study drug. monitor

capillary blood glucose twice daily (fasted and 2 h after a meal) and also if symptoms of hypoglycemia occurred.

Goldstein met 500mg [LSM] met [LSM] metformin not reported not reported safety and efficacy endpoints no BJ 2007 BID: 8.9 500mg BID: - 500mg BID: -0.99 excluded data after rescue (1.0); met 0.82 (-0.98 to - (-1.22 to -0.75); with glyburide 1000mg 0.66); met metformin BID: 8.7 1000mg BID: - 1000mg BID: -1.3 (0.9); pbo: 1.13 (-1.29 to - (-1.53 to -1.06)

8.7 (1.0)

99

0.97); pbo: 0.17 (0.0 to 0.33)

Haak T met 500mg adjusted mean adjusted mean Severe hypoglycemia defined as Reported by the investigator. The use of rescue therapy unlikely, but 2012 bid: 8.7 (SE) met 500mg (SE) met 500mg requiring the assistance of Hypoglycemic episodes were (with sulphonylureas, hypoglycemic events (0.9); met bid: -0.6 (0.1); bid: -0.8 (0.1) (- another person to actively recorded and analyzed thiazolidinediones or insulin) graded at investigators 1000mg bid: met 1000mg bid 1.0 to -0.5); met administer carbohydrate, was permitted. Values discretion; criteria not 8.5 (0.9); -1.1 (0.1); pbo: 1000mg bid -1.2 glucagon or other. separately from other AEs. obtained after rescue specified and could pbo: 8.7 0.1 (0.1) (0.1) (-1.5 to -0.9) Hypoglycemic event intensity medication was initiated lead to patient (1.0) were not used in the LOCF. exclusion was graded according to the investigator’s discretion.

Ji L met 500mg [LSM] met [LSM] met 500mg Any episode with symptoms not reported Rescue therapy with open- no BID: 8.7 500mg BID: - BID: -0.70 (-1.01 consistent with hypoglycemia label glipizide permitted. 2016 (1.0); met 1.29 (-1.54 to - to -0.39); met (e.g., weakness, dizziness, The primary approach to 850mg BID: 1.04); met 850mg BID: -0.97 shakiness, increased sweating, analyzing safety data treated 8.7 (1.1); 850mg BID: - (-1.28 to -0.66) palpitations or confusion) was data obtained after the pbo: 9.0 1.56 (-1.8 to - reported as an episode of initiation of rescue therapy as (1.1) 1.32); pbo: -0.59 symptomatic hypoglycemia missing; a secondary (-0.84 to -0.34) without a requirement for approach included all data, confirmatory blood glucose regardless of rescue therapy. values. Asymptomatic hypoglycemia was defined as an

episode without symptoms of hypoglycemia, but with fingerstick glucose level ≤3.9 mmol/L (≤70 mg/dL). Severe hypoglycemia

100

was defined as any episode requiring assistance, either medical or non-medical. Episodes with a markedly depressed level of consciousness, loss of consciousness or seizure were to be classified as having required medical assistance, whether or not medical assistance was obtained.

Ji L 2017 met 500mg [LSM] (SE) met [LSM] not not reported self-monitor blood glucose levels, Rescue confounding. no BID: 8.4 500mg BID: - reported keep a hypoglycemic diary Hyperglycemic rescue (0.78); pbo: 1.04 (0.11) (95% permitted. 8.21 (0.77) CI -0.70 to - 0.278); pbo - 0.19 (n/a)

Pratley not reported; [LSM] (SE) met [LSM] (SE) Mild to moderate hypoglycaemia Use of a home glucose Hyperglycemic rescue (SU or no RE 2014 The majority 500mg BID: - manually (blood glucose <70 mg/dl (3.89 other) was permitted. of patients 0.65 (0.094); calculated: met mmol/l), symptomatic or monitor and diary to record (60% met 1000mg 500mg BID: - asymptomatic). All hypoglycemic episodes. overall) BID: -1.11 0.80; met 1000mg hypoglycemic episodes were entered with (0.092); pbo: BID: -1.26 associated with a blood glucose a baseline 0.15 (n/a) <70 mg/dl (3.89 mmol/l). Severe episodes required assistance. HbA1c of 8.5% or lower.

101

Characteristics of Included Studies – DPP4i Monotherapy

Study ID Dose Study n= Mean Age Gender (% Countries Studied Ethnicity (%) Duration of Diabetes in Years (SD) Duration (SD) male)

Agarwal P teneli 20mg 16 weeks teneli: 158; teneli: 49.6 teneli: 63.9; India Indian not reported 2018 QD; 40mg pbo: 79 (8.79); pbo: pbo: 54.4 QD 48.9 (7.84)

Aschner P sita 100mg 24 weeks sita 100mg: sita 100mg: sita 100mg: USA sita 100mg: Asian 13.4, Black 4.2, Hispanic sita 100mg: 4.3 (4.9); 200mg: 4.3 (4.7); 2006 QD; 200mg 238; 200mg: 53.4 (9.5); 57.1; 24.4, Caucasian 51.3, other 6.7; 200mg: pbo: 4.6 (4.7) QD 250; pbo: 200mg: 54.9 200mg: Asian 14.8, Black 4.8, Hispanic 21.2, 253 (10.1); pbo: 46.8; pbo: Caucasian 52.8, other 6.4; pbo: Asian 13.4, 54.3 (10.1) 51.4 Black 6.3, Hispanic 25.3, Caucasian 50.2, other 4.7

Barnett lina 5mg 18 weeks lina 5mg: lina 5mg: lina 5mg: Canada, Mexico, lina 5mg: White 70.2, Asian 27.8, Other 2.0; lina 5mg: ≤1 year 21.8, >1 to 5 years AH 2012 QD (extension 151; pbo: 76 56.4 (10.6); 36.4; pbo: Philippines, pbo: White 67.1, Asian 27.6, Other 5.3 51.0, >5years 27.2; pbo: ≤1 year 24.7, >1 data not pbo: 56.7 43.4 Romania, Russia, to 5 years 54.8, >5years 20.5 captured) (9.7) Ukraine, and USA. (7 countries)

Barzilai N sita 50mg 24 weeks sita: 102; sita: 71.6 sita: 47; pbo: USA sita: White 75, Black 10, Hispanic 9, Asian sita: 7.8 (0.8); pbo: 7.8 (0.7) 2011 QD; sita pbo: 104 (6.1); pbo: 47 3, Other 4; pbo: White 83, Black 9, 100mg QD 72.1 (6.0) Hispanic 6, Asian 3, Other 0 depending on renal function as study done in elderly population

102

Chen Y lina 5mg 24 weeks lina 5mg: lina 5mg: lina 5mg: China, Malaysia, [COUNTRY] lina 5mg: Chinese 86.0; lina 5mg: ≤ 1 year 50.0, > 1-5 years 31.6, 2015 QD 200; pbo: 99 54.6 (101); 58; pbo: Philippines Malaysian 7.5; Philippine 6.5; pbo; Chinese > 5 years 18.4; pbo: ≤ 1 year 52.1, > 1-5 pbo: 54.1 59.6 88.9; Malaysian 7.1; Philippine 4.0 years 31.9, > 5 years 16.0[Time since (9.3) diagnosis, %]

DeFronzo alo 12.5mg 26 weeks alo 12.5: 53.4 (11.1) 53.2 Argentina, Brazil, White: 66.9 not reported 2008 QD or alo 133; alo 25: (no Chile, Dominican 25mg QD 131; pbo: 64 breakdown) Republic, Guatemala, Hungary, India, Mexico, Netherlands, New Zealand, Poland, South Africa, United Kingdom, United States

Dejager S vilda 50mg 24 weeks vilda 50mg vilda 50mg vilda 50mg USA, Russia, vilda 50mg QD: Caucasian 73.1, vilda 50mg QD: 2.1 (3.6); 50mg BID: 2.1 2007 QD; vilda QD: 163; QD: 55.3 QD: 41.3; Tunisia Hispanic/Latino 13.5, Black 9.6, all other (3.3); 100mg QD: 2.4 (4.2); pbo: 1.6 (2.5) 50mg BID; 50mg BID: (11.4); 50mg 50mg BID: 3.8; 50mg BID: Caucasian 3.3, vilda 152; 100mg BID: 52.8 46.7; 100mg Hispanic/Latino 13.3, Black 10.0, all other 100mg QD QD: 157; (9.6); 100mg QD: 53.3; 33.4; 100mg QD: Caucasian 76.1, pbo: 160 QD: 53.6 pbo: 47.9 Hispanic/Latino 15.2, Black 4.3, all other (10.8); pbo: 4.4; pbo: Caucasian 69.1, Hispanic/Latino 52.2 (11.2) 11.7, Black 12.8, all other 6.4

Gantz I omari 25mg 24 weeks sita: 165; sita: 60 (9); sita: 69.7; Japan not reported sita: 7.4 (5.3); pbo: 8.6 (5.1) 2017 OW; sita pbo: 83 pbo: 61 (9) pbo: 68.7 50mg QD

Gantz I omari 25mg 24 weeks omari: 166; omari: 60 omari: 62.7; Japan not reported omari: 7.4 (5.5); pbo: 8.6 (5.1) 2017 OW; sita pbo: 83 (11); pbo: 61 pbo: 68.7 50mg QD (9)

103

Goldstein sita 100mg 24 weeks sita 100mg: sita 100mg: sita 100mg: multinational; not [RACE] sita 100mg: White 52.0, Black 6.1, sita 100mg: 4.4 (4.6); pbo: 4.6 (4.9) BJ 2007 QD 179; pbo: 53.3 (10.2); 52.0; pbo: specified Hispanic 29.1, Asian 3.4, Other 9.5; pbo: 176 pbo: 53.6 52.8 White 46.0, Black 9.7, Hispanic 26.7, Asian (10.0) 6.8, Other 10.8

Haak T lina 5mg 24 weeks lina 5mg: lina 5mg: lina 5mg: 14 countries; not lina 5mg: White 68.3, Asian 31.7, Black 0.0, lina 5mg: ≤1: 40.0, > 1 to 5: 34.8, > 5: 2012 QD 142; pbo: 72 56.2 (10.8); 56.3; pbo: specified Hawaiian/Pacific Islander 0.0; pbo: White 25.2; pbo: ≤1: 30.8, > 1 to 5: 35.4, > 5: pbo: 55.7 50.0 63.9, Asian 36.1, Black 0.0, 33.8 (11.0) Hawaiian/Pacific Islander 0.0

Hanefeld sita 25mg 12 weeks sita 25mg sita 25mg sita 25mg Germany, [RACE] sita 25mg QD: Asian 0.9, Black 3.6, sita 25mg QD: 3.6 (3.4); 50mg QD: 3.3 M 2007 QD; sita QD: n = QD: 55.1 QD: 51.4; Hungary, Iceland, White 88.3, other 7.2; sita 50mg QD: Asian (3.9); 100mg QD: 3.6 (3.9); 50mg BID: 50mg QD; 111; 50mg (9.6); 50mg 50mg QD: Lithuania, Poland, 0, Black 8.0, White 85.7, other 6.3; sita 4.5 (5.9); pbo: 3.3 (3.4) sita 100mg QD: n = QD: 55.3 45.5; 100mg United Kingdom, 100mg QD: Asian 0, Black 5.5, White 88.2, QD; sita 112; 100mg (10.3); QD: 55.5; United States other 6.4; sita 50mg BID: Asian 0.9, Black 50mg BID QD: n = 100mg QD: 50mg BID: 6.3, White 81.1, other 11.7; pbo: Asian 0.9, 110; 50mg 56.0 (7.9); 44.1; pbo: Black 7.2, White 78.4, other 13.5 BID: n = 50mg BID: 63.1 111; pbo: 55.2 (9.5); 111 pbo: 55.9 (9.3)

Home P omari 25mg 24 weeks omari 25mg omari 25mg omari 25mg Bulgaria, omari 25mg QW: White 68.5, Asian 28.5, omari 25mg QW: 5.4 (3.8); pbo: 5.7 (4.7) 2018 QW QW: 165; QW: 57.4 QW: 57.6; Germany, Black 2.4, multi-racial 0, American pbo: 164 (9.2); pbo: pbo: 59.1 Hungary, Italy, Indian/Alaska Native 0.6; pbo: White 67.7, 57.0 (9.7) Netherlands, Asian 26.2, Black 5.5, multi-racial 0.6, Philippines, American Indian/Alaska Native 0 Romania, South Korea, Taiwan, United States

104

Hong S teneli 20mg 24 weeks teneli 20mg: teneli 20mg: teneli 20mg: South Korea Korean alo: 4.59; pbo: 4.59 2016 QD 99; pbo: 43 56.64 52.53; pbo: (10.07); pbo: 65.12 57.93 (11.90)

Inagaki N trela 12 weeks trela 12.5mg trela 12.5mg trela 12.5mg Japan Japanese trela 12.5mg QW: 7.82; 25mg QW: 5.96; 2014 12.5mg QW: 54; QW: 60.6 QW: 61; 50mg QW: 5.78; 100mg QW: 6.09; QW; trela 25mg QW: (10.24); 25mg QW: 200mg QW: 6.99; pbo: 6 25mg QW; 52; 50mg 25mg QW: 65; 50mg trela 50mg QW: 51; 58.5 (10.49); QW: 53; QW; trela 100mg QW: 50mg QW: 100mg QW: 100mg QW; 55; 200mg 61.0 (10.18); 51; 200mg trela 200mg QW: 54; 100mg QW: QW: 67; QW pbo: 55 57.8 (10.38); pbo: 66 200mg QW: 60.5 (11.26); pbo: 61.6 (9.79)

Inagaki et alo 25mg 24 weeks alo 25mg: [MEDIAN] alo 25mg: Japan Korean alo: 7.07; pbo: 7.54 al. 2015 QD 92; pbo: 50 alo: 60 (53- 75; pbo: 86 65); pbo: 62 (54-67)

Inagaki et trela 100mg 24 weeks trela 100mg [MEDIAN] trela 100mg Japan Korean trela: 6.25; pbo: 7.54 al. 2015 QW QW: 101; trela 100mg QW: 72; pbo: 50 QW: 58 (52- pbo: 86 65); pbo: 62 (54-67)

105

Iwamoto Y sita 25mg 12 weeks sita 25mg sita 25mg sita 25mg Japan Japanese sita 25mg QD: 4.7 (4.3); 50mg QD: 5.6 2010 QD; 50mg QD: n=80; QD: 59.9 QD: 63.8; (6.4); 100mg QD: 5.4 (5.4); 200mg QD: QD; 100mg 50mg QD: (7.9); 50mg 50mg QD: 5.1 (4.9); pbo: 6.4 (5.5) QD; 200mg n=72; QD: 60.2 65.3; 100mg QD 100mg QD: (9.4); 100mg QD: 51.4; n=70; QD: 58.3 200mg QD: 200mg QD: (9.5); 200mg 58.8; pbo: n=68; pbo: QD: 60.6 68.5 73 (7.7); pbo: 60.2 (8.0)

Ji L 2016 sita 100mg 24 weeks sita 100mg: sita 100mg: sita 100mg: China Chinese sita 100mg: 1.1 (0.2); pbo: 1.1 (0.2) QD 120; pbo: 51.7 (10.2); 61.7; pbo: 127 pbo: 53.6 68.5 (9.7)

Ji L 2017 alo 12.5mg 26 weeks alo 12.5mg alo 12.5mg alo 12.5mg China, Malaysia, alo 12.5mg BID: Asian 99.4, American not reported BID BID: 163; BID: 55.4 BID: 60.1; South Korea and Indian or Alaskan Native 0.6, Multiracial 0; pbo: 163 (9.62); pbo: pbo: 58.3 Taiwan pbo: Asian 98.8, American Indian or 52.2 (10.17) Alaskan Native 1.2, Multiracial 0

Jung CH evo 2.5mg 12 weeks evo 2.5mg evo 2.5mg evo 2.5mg South Korea Korean evo 2.5mg QD: 4.00 (3.57); 5mg QD: 3.98 2015 QD; 5mg QD: 40; QD: 52.10 QD: 55.00; (3.88); 10mg QD: 3.95 (3.88); pbo: 3.55 QD; 10mg 5mg QD: (11.48); 5mg 5mg QD: (2.85) QD 43; 10mg QD: 54.21 58.14; 10mg QD: 37; (9.74); 10mg QD: 54.05; pbo: 38 QD: 53.16 pbo: 68.42 (9.60); pbo: 54.42 (9.85)

Kadowaki teneligliptin 12 weeks teneli 10: teneli 10: teneli 10: Japan Korean teneli 10: 6.2 (5.2); teneli 20: 6.3 (6.4); T 2013 10mg QD; 84; teneli 57.7 (9.1); 59.5; teneli teneli 40: 6.5 (6.1); pbo: 5.8 (5.0) teneligliptin 20: 79; teneli 20: 20: 74.7;

106

20mg QD + teneli 40: 59.2 (9.5); teneli 40: teneligliptin 81; pbo: 80 teneli 40: 65.4; pbo: 40mg QD 57.5 (10.4); 63.8 pbo: 58.5 (9.6)

Kawamori linagliptin 26 weeks lina 5mg: lina 5mg: lina 5mg: Japan Japanese lina 5mg: ≤1 year 11.9, > 1-5 years 38.4, > R 2012 5mg QD; 159; lina 60.3 (9.4); 69.8; lina 5 years 49.7; lina 10mg: ≤1 year 11.9, > 1- linagliptin 10mg: 160; lina 10mg: 10mg: 70.0; 5 years 36.9, > 5 years 51.3; pbo: ≤1 year 10mg QD pbo: 80 61.3 (10.0); pbo: 71.3 8.8, > 1-5 years 45.0, > 5 years 46.3 pbo: 59.7 (8.9)

Kikuchi M vilda 10mg 12 weeks vilda 10mg vilda 10mg vilda 10mg Japan Japanese vilda 10mg BID: 4.5 (4.2); 25mg BID: 4.7 2009 BID; 25mg BID: n=71; BID: 58.9 BID: 73.2; (4.5); 50mg BID: 4.7 (4.3); pbo: 7.1 (5.5) BID; 50mg 25mg BID: (8.6); 25mg 25mg BID: BID n=72; 50mg BID: 57.8 63.9; 50mg BID: n=76; (8.5); 50mg BID: 67.1; pbo: 72 BID: 58.8 pbo: 63.9 (8.6); pbo: 60.4 (8.1)

Kumar saxa 5mg 24 weeks saxa 5mg: not reported not reported India Indian not reported PKM 2014 QD 107; pbo: 106

Mohan V sita100mg 18 weeks sita 100mg: sita 100mg: sita 100mg: China, India, South sita 100mg: Chinese 46, Indian 36, Korean sita 100mg: 2.1 (1.7); pbo: 1.9 (1.6) 2009 QD 352; pbo: 50.9 (9.3); 57; pbo: 60 Korea 18; pbo: Chinese 46, Indian 35, Korean 19 178 pbo: 50.9 (9.3)

Nonaka K sita 100mg 12 weeks sita 100mg: sita 100mg: sita 100mg: Japan Japanese sita 100mg: 4 (4.1); pbo: 4.1 (4.6) 2008 QD 75; pbo: 76 55.6 (8.6); 60; pbo: 66

107

pbo: 55.0 (8.0)

Pan CY saxa 5mg 24 weeks saxa 5mg: saxa 5mg: saxa 5mg: China, India, [REGION] saxa 5mg: China 59.5, India saxa 5mg: 0.8 (1.4); pbo: 1.2 (2.6) 2012 QD 284; pbo: 51.2 (10.0); 56.3; pbo: Philippines, South 21.8, South Korea 6.3, Philippines 12.3; pbo: 284 pbo: 51.6 54.6 Korea China 58.5, India 21.1, South Korea 7.7, (10.3) Philippines 12.7

Pan CY alo 25mg 16 weeks alo 25mg: alo 25mg: alo 25mg: China, Taiwan, [COUNTRY OR REGION] alo 25mg: China alo 25mg: 1.9 (2.4); pbo: 2.1 (2.8) 2017 QD 92; pbo: 92 51.6 (10.4); 59.8; pbo: Hong Kong 97.8, Hong Kong 2.2, Taiwan 0; pbo: China pbo: 53.1 58.1 97.8, Hong Kong 1.1, Taiwan 1.1 (8.9)

Park J evo 5mg 24 weeks evo 5mg: evo 5mg: evo 5mg: South Korea Korean evo 5mg: 4.74 (3.81); pbo: 4.25 (4.10) 2017 QD 80; pbo: 80 57.6 (11.0); 48.8; pbo: pbo: 56.8 57.5 (9.8)

Pi-Sunyer vilda 50mg 24 weeks vilda 50mg vilda 50mg vilda 50mg USA, India, [RACE] vilda 50mg QD: Caucasian 54.5, vilda 50mg QD: 1.8 (2.7); 50mg BID: 2.4 FX 2007 QD; vilda QD: 88; QD: 50.6 QD: 55.7; Slovakia Hispanic or Latino 18.2, Asian (Indian (3.2); 100mg QD: 2.1 (2.9); pbo: 2.5 (3.7) 50mg BID; 50mg BID: (10.4); 50mg 50mg BID: subcontinent) 15.9, Asian (non-Indian vilda 83; 100mg BID: 50.2 56.6; 100mg subcontinent) 3.4, Black 8.0; 50mg BID: 100mg QD QD: 91; (12.7); QD: 53.8; Caucasian 53.0, Hispanic or Latino 21.7, pbo: 92 100mg QD: pbo: 54.3 Asian (Indian subcontinent) 18.0, Asian 52.0 (11.7); (non-Indian subcontinent) 1.2, Black 6.0; pbo: 52.0 100mg QD: Caucasian 58.2, Hispanic or (12.0) Latino 12.1, Asian (Indian subcontinent) 16.5, Asian (non-Indian subcontinent) 1.1, Black 12.1; pbo: Caucasian 51.1, Hispanic or Latino 18.5, Asian (Indian subcontinent) 16.3, Asian (non-Indian subcontinent) 1.1, Black 13.0

108

Pratley RE vilda 25mg 12 weeks vilda 25mg vilda 25mg vilda 25mg South America, vilda 25mg BID: Black 2.9, Caucasian 47.1, vilda 25mg BID: 4.6 (5.6); pbo: 3.5 (5.7) 2006 BID BID: 70; BID: 56.9 BID: 40; Mexico Oriental 1.4, Other 48.6; pbo: Black 0, pbo: 28 (9.4); pbo: pbo: 60 Caucasian 46.4, Oriental 0, Other 53.6 52.8 (10.0)

Pratley RE alo 25mg 26 weeks alo 25mg alo 25mg alo 25mg worldwide; not alo 25mg QD: Asian 15.2, Black or African alo 25mg QD: 3.6 (4.12); alo 12.5mg BID: 2014 QD; alo QD: 112; alo QD: 52.6 QD: 42.9; specified American 2.7, White 75.0, Other 7.1; alo 4.0 (4.80); pbo: 4.3 (4.78) 12.5mg 12.5mg (9.38); alo alo 12.5mg 12.5mg BID: Asian 18.6, Black or African BID BID: 113; 12.5mg BID: 55.8; American 2.7, White 73.5, Other 5.3; pbo: pbo: 109 BID: pbo: 50.5 Asian 18.3, Black or African American 7.3, 53.7(9.7); White 69.7, Other 4.6 pbo: 53.1 (9.60)

Raz I 2006 sita 100mg 18 weeks sita 100mg sita 100mg sita 100mg Multinational, sita 100mg QD: White 69.3, Black 7.8, sita 100mg QD: 4.5 (4.3); 200mg QD: 4.5 QD; sita QD: 205; QD: 54.5 QD: 53.7; nations not Hispanic 18.0, Asian 3.9, Other 1.0; 200mg (3.9); pbo: 4.7 (5.0) 200mg QD 200mg QD: (10.0); 200mg QD: specified QD: White 70.9, Black 5.3, Hispanic 18.9, 206; pbo: 200mg QD: 50.5; pbo: Asian 3.4, Other 1.5; pbo: White 61.8, Black 110 55.4 (9.2); 62.7 10.9, Hispanic 20.0, Asian 4.5, Other 2.7 pbo: 55.5 (10.1)

Rhee EJ gemi 50mg 12 weeks gemi 50mg: gemi 50mg: gemi 50mg: Korea Korea gemi 50mg: 4.53 (3.83); 100mg: 4.51 2010 QD; gemi 35; 100mg: 52.43 (9.63); 71.43; (5.22); 200mg: 3.18 (3.25); pbo: 4.31 100mg QD; 37; 200mg: 100mg: 100mg: (4.90) gemi 35; pbo: 34 53.22 62.16; 200mg QD (11.92); 200mg: 200mg: 51.43; pbo: 54.29 67.65 (10.33); pbo: 51.26 (8.59)

109

Ristic S vilda 25mg 12 weeks vilda 25mg vilda 25mg vilda 25mg USA, Russia vilda 25mg BID: Caucasian 80.4; 25mg QD: vilda 25mg BID: 3.28 (3.81); 25mg QD: 2005 BID; vilda BID: 51; BID: 55.6 BID: 47.1; Caucasian 79.6; 50mg QD: Caucasian 77.4; 3.10 (5.16); 50mg QD: 2.71 (3.24); 100mg 25mg QD; 25mg QD: (10.9); 25mg 25mg QD: 100mg QD: Caucasian 74.6, pbo: Caucasian QD: 3.03 (4.22); pbo: 2.28 (2.99) vilda 50mg 54; 50mg QD: 57.4 63.0; 50mg 87.9 QD; vilda QD: 53; (10.2); 50mg QD: 49.1; 100mg QD 100mg QD: QD: 57.0 100mg QD: 63; pbo: 58 (10.2); 55.6; pbo: 100mg QD: 56.9 56.2 (10.1); pbo: 54.6 (10.6)

Roden empa 10mg 12 weeks empa 10mg: empa 10mg: empa 10mg: Belgium, Canada, empa 10mg: Asian 64, White 34, empa 10mg: ≤1 year: 39, >1-5 years: 41, 2013 QD; empa 224; 25mg: 56.2 (11.6); 63; 25mg: China, Germany, Black/African-American 1, American- >5-10 years: 13, >10 years: 7; 25mg: ≤1 25mg QD 224; pbo: 25mg: 53.8 65; pbo: 54 India, Ireland, Indian/Alaska Native 0, Hawaiian/Pacific year: 41, >1-5 years: 37, >5-10 years: 17, 228 (11.6); pbo: Japan, Switzerland Islander 1, 25mg: Asian 64, White 33, >10 years 6; pbo: ≤1 year: 32, >1-5 years: 54.9 (10.9) and USA Black/African-American 3, American- 46, >5-10 years: 15, >10 years: 8 Indian/Alaska Native 0, Hawaiian/Pacific Islander 0; pbo: Asian 64, White 33, Black/African-American 3, American- Indian/Alaska Native 0, Hawaiian/Pacific Islander 0

Rosenstock saxa 2.5mg 12 weeks saxa 2.5mg saxa 2.5mg saxa 2.5mg USA saxa 2.5mg QD: Caucasian 85, African- saxa 2.5mg QD: 1.0 (0.0-14.0); 5mg QD: J 2008 QD; saxa QD: 55; QD: 52.5 QD: 40; American 11, Other 4; 5mg QD: Caucasian 0.8 (0.0 to 8.2); 10mg QD: 0.7 (0.0 to 5mg QD; 5mg QD: (10.53); 5mg 5mg QD: 87, African-American 13, Other 0; 10mg 12.9); 20mg QD: 1.7 (0.0 to 13.0); 40mg saxa 10mg 47; 10mg QD: 53.7 53; 10mg QD: Caucasian 84, African-American 8, QD: 1.3 (0.0 to 19.0); 100mg QD: 0.5 (0.0 QD; saxa QD: 63; (10.14); QD: 63; Other 8; 20mg QD: Caucasian 87, African- to 26.0); pbo: 1.8 (0.0 to 23.0) 20mg QD; 20mg QD: 10mg QD: 20mg QD: American 7, Other 6; 40mg QD: Caucasian saxa 40mg 54; 40mg 54.5 (8.61); 70; 40mg 92, African-American 4, Other 4; 100mg QD QD: 52; 20mg QD: QD: 58; QD: Caucasian 77, African-American 16, 100mg QD: 53.6 (8.59); 100mg QD: Other 7; pbo: Caucasian 87, African- 44; pbo: 67 40mg QD: 61; pbo: 63 American 10, Other 3 110

54.1 (11.12); (100mg QD 100mg QD: for 6 51.4 (9.91); weeks) pbo: 52.8 (10.18)

Rosenstock saxa 2.5mg 24 weeks saxa 2.5mg saxa 2.5mg saxa 2.5mg USA saxa 2.5mg QD: White 87.3, Black/African saxa 2.5mg QD: 3.1 (3.5); 5mg QD: 2.5 J 2009 A QD; saxa QD: 102; QD: 53.27 QD: 56.9; American 4.9, Asian 4.9, Other 2.9; 5mg (3.3); 10mg QD: 2.3 (3.1); pbo: 2.3 (2.7) 5mg QD; 5mg QD: (10.06); 5mg 5mg QD: QD: White 87.7, Black/African American saxa 10mg 106; 10mg QD: 53.91 50.9; 10mg 4.7, Asian 3.8, Other 3.8; 10mg QD: White QD QD: 98; (11.57); QD: 45.9; 81.6, Black/African American 6.1, Asian pbo: 95 (did 10mg QD: pbo: 49.5 6.1, Other 6.1; pbo: White 83.2, not include 52.72 Black/African American 6.3, Asian 3.2, open label (11.27); pbo: Other 7.4 cohort) 53.91 (12.32)

Scherbaum vilda 50mg 52 weeks vilda 50mg vilda 50mg vilda 50mg Finland, France, vilda 50mg QD: Caucasian 99.4, Other 0.6; vilda 50 mg QD: 2.5 (2.9); pbo: 2.7 (3.2) WA 2008 QD QD: 156; QD: 63.3 QD: 59.6; Germany, pbo: Caucasian 99.3, Other 0.7 pbo: 150 (10.2); pbo: pbo: 59.3 Romania, Spain, 62.8 (11.0) Sweden

Scott R sita 5mg 12 weeks sita 5mg sita 5mg sita 5mg Multinational; sita 5mg BID: Asian 5.6, Black 6.4, Multi- sita 5mg BID: 4.3 (4.1); 12.5mg BID: 4.9 2007 BID; sita BID: 125; BID: 55.1 BID: 49.6; nations not racial 6.4, White 68.8, Other 12.8; 12.5mg (5.0); 25mg BID: 5.0 (5.2); 50mg BID: 4.2 12.5mg 12.5mg (9.5); 12.5mg specified BID: Asian 4.9, Black 4.9, Multi-racial 5.7, (4.0); pbo: 4.8 (4.7) BID; sita BID: 123; 12.5mg BID: 48.0; White 63.4, Other 21.1; 25mg BID: Asian 25mg BID; 25mg BID: BID: 56.2 25mg BID: 4.9, Black 8.9, Multi-racial 6.5, White 61.0, sita 50mg 123; 50mg (9.0); 25mg 57.7; 50mg Other 18.7; 50mg BID: Asian 2.4, Black 4.8, BID BID: 124; BID: 55.6 BID: 52.4; Multi-racial 7.3, White 69.4, Other 16.1; pbo: 125 (9.0); 50mg pbo: 62.4 pbo: Asian 2.4, Black 8.0, Multi-racial 7.2, BID: 55.1 White 66.4, Other 16.0

111

(9.8); pbo: 55.3 (9.7)

Sheu omari 12 weeks omari omari omari 21 countries; not omari 0.25mg once weekly: White 53.1, omari 0.25mg once weekly: 4.8 (4.2); 1mg WHH 0.25mg 0.25mg once 0.25mg once 0.25mg once reported Asian 29.2, Multiracial 5.3, American Indian once weekly: 5.3 (4.3); 3mg once weekly: 2015 QW; omari weekly: 113; weekly: 54.3 weekly: or Alaska Native 7.1, Black or African 5.4 (3.9); 10mg once weekly: 5.1 (4.6); 1mg 1mg once (8.9); 1mg 57.5; 1mg American 4.4, Native Hawaiian or other 25mg once weekly: 5.9 (5.2); pbo: 5.8 QW; omari weekly: 115; once once Pacific Islander 0.9; 1mg once weekly: (4.6) 3mg QW; 3mg once weekly: 55.7 weekly: White 60.9, Asian 28.7, Multiracial 3.5, omari 10mg weekly: 114; (8.5); 3mg 58.3; 3mg American Indian or Alaska Native 5.2, Black QW; omari 10mg once once once or African American 1.7, Native Hawaiian 25mg QW weekly: 115; weekly: 55.3 weekly: or other Pacific Islander 0; 3mg once 25mg once (8.5); 10mg 57.0; 10mg weekly: White 61.4, Asian 23.7, Multiracial weekly: 114; once once 4.4, American Indian or Alaska Native 1.8, pbo: 114 weekly: 54.4 weekly: Black or African American 7.0, Native (10.0); 25mg 48.7; 25mg Hawaiian or other Pacific Islander 1.8; 10mg once once once weekly: White 55.7, Asian 26.1, weekly: 55.1 weekly: Multiracial 7.0, American Indian or Alaska (8.8); pbo: 60.5; pbo: Native 7.8, Black or African American 2.6, 55.9 (8.4) 57.0 Native Hawaiian or other Pacific Islander 0.9; 25mg once weekly: White 54.4, Asian 26.3, Multiracial 10.5, American Indian or Alaska Native 2.6, Black or African American 6.1, Native Hawaiian or other Pacific Islander 0; pbo: White 56.1, Asian 28.1, Multiracial 8.8, American Indian or Alaska Native 3.5, Black or African American 2.6, Native Hawaiian or other Pacific Islander 0.9

Wu W lina 5mg 24 weeks lina 5mg: lina 5mg: lina 5mg: China Chinese not reported 2015 QD 33; pbo: 22 52.5 (11.0); 65.7; pbo: pbo: 51.2 50 (7.5) 112

Yang SJ gemi 50mg 24 weeks gemi 50 mg: gemi 50 mg: gemi 50 mg: Korea and India gemi 50 mg: Indian 56 Korean 44; pbo: gemi 50 mg: 3.24 (3.84); pbo: 2.86 (4.36) 2013 QD 87; pbo: 87 54 (49-60); 57; pbo: 44 Indian 60, Korean 40 pbo: 52 (45 - 60)

Yang HK ana 100mg 24 weeks ana 100mg ana 100mg ana 100mg Republic of Korea Korean ana 100mg BID: 3.17 (5.53); 200mg BID: 2015 BID; ana BID: 37; BID: 54.43 BID: 40.54; (South Korea) 3.43 (3.40); pbo: 4.14 (4.10) 200mg BID 200mg BID: (9.86); 200mg BID: 30; pbo: 38 200mg BID: 60.00; pbo: 57.70 (9.71); 63.16 pbo: 56.74 (9.72)

Characteristics of Included Studies – DPP4i Monotherapy continued

Study ID Baseline Mean Change Mean HYPO def Ascertainment of Hypo Rescue Medication Excl of Pts w A1C % (SD) in HbA1C % Difference in events? At vs. baseline HbA1c % vs screening or pbo during study)

Agarwal P teneli: 7.75 [LSM] (SE) [LSM] (95% not reported not reported Metformin used as rescue medication. no 2018 (n/a); pbo: teneli: -0.304 CI) teneli: - Data collected after initiation of rescue 7.74 (0.118); pbo: 0.555 (0.176 therapy were treated as missing. 0.251 (0.1565) to 0.934)

Aschner P sita 100mg: [LSM] [LSM] not reported not reported Rescue therapy (metformin) permitted. no 2006 8.0 (0.9); sita100mg: - sita100mg: - Sita 100mg included one episode after 200mg: 8.1 0.61 (-0.74 to - 0.79 (-0.96 to initiation of rescue. Data still included

113

(0.9); pbo: 8.0 0.49); 200mg: - -0.62); since same patient could have had other (0.8) 0.76 (-0.88 to - 200mg: -0.94 episodes not attributed to rescue therapy. 0.64); pbo: 0.18 (-1.11 to - (0.06 to 0.30) 0.77)

Barnett AH lina 5mg: 8.1 lina 5mg: -0.39 lina 5mg: - Hypoglycaemia was not reported Rescue therapy was initiated with no 2012 (1.0); pbo: 8.1 (0.14); pbo: 0.60 (-0.88 to defined according to initially and, if blood glucose (0.9) 0.21 (0.16) -0.32) American Diabetes remained higher than the levels described [FAS, LOCF, Association guidelines as above, insulin was introduced with or Model includes either asymptomatic with without continuation of pioglitazone. continuous plasma glucose ≤3.9 baseline mmol/l (≤70 mg/dl), HbA1c, number symptomatic with plasma of prior diabetes glucose ≤3.9 mmol/l (≤70 drugs, reason of mg/dl) or severe (requiring metformin third-party assistance to intolerance and administer resuscitative treatment. action).

Barzilai N sita: 7.8 (0.8); [LSM] sita: -0.5 [LSM] sita: - not reported All randomized patients were asked to not reported no 2011 pbo: 7.8 (0.7) (-0.7 to -0.2); 0.7 (-0.9 to - measure and record self-monitored pbo: 0.2 (0.0 to 0.5) blood glucose (SMBG) data at four 0.5) specific times on each of the following 3 days: ‘day 2’ ‘day 3’ and ‘day 7’. Times for SMBG were (1) immediately before breakfast (after fasting for at least 8 hours), (2) 2 hours after breakfast, (3) immediately before the evening meal, and (4) 2 hours after the evening meal.

Chen Y lina 5mg: lina 5mg: -0.68 lina 5mg: - Hypoglycemia was not reported; investigator-defined The use of metformin as rescue no 2015 7.95 (0.89); (0.07); pbo: - 0.50 (0.11) classified by investigators medication was permitted. 0.18 (0.10) (adjusted as asymptomatic

114

pbo: 8.09 (adjusted mean mean (0.91) (SE)) difference with glucose concentration (SE)) ≤70 mg/dL, documented symptomatic

with glucose concentration of 54–70 mg/dL, documented

symptomatic with glucose concentration <54 mg/dL but no

need for external assistance, and severe hypoglycemia requiring the assistance of another person.

DeFronzo 7.9 (0.08) [LSM] alo 12.5: [LSM] not reported not reported Rescue medication permitted but type not no 2008 -0.56 (p<0.001); manually known. alo 25: -0.59 calculated alo (p<0.001); pbo: 12.5: -0.54; -0.02 alo 25: -0.57

Dejager S vilda 50mg vilda 50mg vilda 50mg Hypoglycemia was defined Confirmed by SMBG measurement < not reported no; but a patient 2007 QD: 8.2 (0.8); QD: -0.8 (0.1); QD: -0.5 as symptoms suggestive of 3.1 mmol / l plasma glucose could also be 50mg BID: 50mg BID: -0.8 (0.2); 50mg low blood glucose equivalent discontinued due 8.6 (0.8); (0.1); 100mg BID: -0.5 confirmed by self- to UTE 100mg QD: QD: -0.9 (0.1); (0.2); 100mg monitored blood glucose [unsatisfactory 8.4 (0.8); pbo: pbo: -0.3 (0.1) QD: -0.6 (0.2) (SMBG) measurement < therapetuci effect] 8.4 (0.8) 3.1 mmol/l plasma glucose equivalent.

115

Gantz I sita: 8.0 (0.8); [LSM] sita: - [LSM] sita: - Symptomatic not reported not reported no 2017 pbo: 8.1 (0.7) 0.65 (-0.74 to - 0.78 (-0.94 to hypoglycaemia: episode 0.55); pbo: 0.13 -0.61) with clinical symptoms (-0.0 to 0.27) attributed to based on an hypoglycaemia, without LDA model regard to glucose level with terms for treatment, prior AHA therapy status (yes/no), time, and interaction of time by treatment, time by prior AHA therapy status, and time by treatment by prior AHA therapy status, with the constraint that the mean baseline was the same for all treatment groups). s.e., standard error

Gantz I omarigliptin: [LSM] omari: - [LSM] omari: Symptomatic not reported not reported no 2017 7.9 (0.7); pbo: 0.66 (-0.76 to - -0.80 (-0.96 hypoglycaemia: episode 8.1 (0.7) 0.57); pbo: 0.13 to -0.63) with clinical symptoms (-0.0 to 0.27) attributed to 116

based on an hypoglycaemia, without LDA model regard to glucose level with terms for treatment, prior AHA therapy status (yes/no), time, and interaction of time by treatment, time by prior AHA therapy status, and time by treatment by prior AHA therapy status, with the constraint that the mean baseline was the same for all treatment groups). s.e., standard error

Goldstein sita 100mg: [LSM] sita [LSM] sita not reported not reported Glycemic rescue therapy (glyburide no BJ 2007 8.9 (1.0); pbo: 100mg: -0.66 (- 100mg: -0.83 [glibenclamide]) permitted. Safety and 8.7 (1.0) 0.83 to -0.50); (-1.06 to - efficacy endpoints excluded data after pbo: 0.17 (0.0 0.60) rescue with glyburide. to -0.33)

Haak T lina 5mg: 8.7 adjusted mean adjusted Severe hypoglycemia Hypoglycemic episodes were Rescue therapy (with sulphonylureas, unlikely, but 2012 (1.0); pbo: 8.7 (SE) lina 5mg: - mean (SE) defined as requiring the recorded and analyzed thiazolidinediones or insulin) was hypoglycemic (1.0) assistance of another permitted. Values obtained after rescue events graded at 117

0.5 (0.1); pbo: lina 5mg: -0.6 person to actively medication was initiated were not used in investigators 0.1 (0.1) (0.1) administer carbohydrate, separately from other AEs. the LOCF. discretion; criteria glucagon or other. Hypoglycemic event intensity not specified and could lead to was graded according to the patient exclusion investigator’s discretion

Hanefeld M sita 25mg sita 25mg QD: - sita 25mg The determination of Confirmed by investigator via. patient not reported no 2007 QD: 7.7 (0.9); 0.28 (-0.42 to - QD: -0.39 (- hypoglycemia was made logbook (subjective hypoglycemia + 50mg QD: 0.14); 50mg 0.59 to -0.20); by the study site fingerstick glucose levels if measure + 7.6 (1.0); QD: –0.44 (– 50mg QD: - investigator based upon debrief with patient) 100mg QD: 0.58 to –0.30); 0.55 (-0.75 to information provided by 7.8 (0.9); 100mg QD: – -0.36); 100mg the patient in a logbook. 50mg BID: 0.44 (–0.58 to – QD: -0.56 (- 7.8 (0.9); pbo: 0.30); 50mg 0.75 to -0.36); A fingerstick blood 7.6 (0.9) BID: –0.43 (– 50mg BID: - glucose determination 0.56 to –0.29); 0.54 (-0.74 to concurrent with the pbo: 0.12 (-0.02 -0.35) episode was not required to to 0.26) assess an episode as hypoglycemia, although investigators could include the fingerstick glucose measurement, if it was available, in their assessment of the episode. Pt regularly measured blood glucose via. fingerstick 7x daily.

Home P omari 25mg [LSM] omari [LSM] omari Symptomatic A questionnaire Rescued by adding open-label metformin no 2018 QW: 8.0 25mg QW: - 25mg QW: - hypoglycemia: episode and after by adding open-label (0.9); pbo: 8.1 0.49 (-0.73 to - 0.39 (-0.59 to with clinical symptoms was provided to participants to collect glimepiride. (1.0) 0.24); pbo: - -0.19) attributed to hypoglycemia, data on

118

0.10 (-0.34 to without regard to glucose 0.14) level. hypoglycemia.

Severe hypoglycemia: episode that required assistance, either medical or non-medical.

Asymptomatic hypoglycemia: self- measured glucose values ≤3.9 mmol/L without symptoms

Hong S teneli 20mg: [LSM] (SE) [LSM] teneli Review of patient data for The incidence of hypoglycemia was Rescue medication not specified. no 2016 7.74 (0.61); teneli 20mg: - 20mg: -0.94 signs and symptoms of assessed by reviewing patient data for pbo: 7.74 0.90 (0.09); (-1.22 to - hypoglycemia in addition signs and symptoms of hypoglycemia (0.53) pbo: 0.03 (0.12) 0.65) to reviewing self- in addition to reviewing self- monitoring of blood monitoring of blood glucose (SMBG) glucose data data.

Inagaki N trela 12.5mg [LSM] (SE) [LSM] not reported. Investigator- not reported Rescue not permitted. Patients requiring no 2014 QW: 8.18 trela 12.5mg manually defined, criteria not rescue medication were withdrawn from (0.89); 25mg QW: -0.37 calculated specified study. QW: 7.99 (0.068); 25mg trela 12.5mg (0.77); 50mg QW: -0.32 QW: -0.72; QW: 8.07 (0.070); 50mg 25mg QW: - (0.86); 100mg QW: -0.42 0.67; 50mg QW: 8.41 (0.070); 100mg QW: -0.77; (0.97); 200mg QW: -0.54 100mg QW: - QW: 7.84 (0.068) 200mg 0.89; 200mg (0.76); pbo: QW: -0.55 QW: -0.90 8.15 (0.95) (0.069); pbo: 0.35 (0.068)

119

Inagaki et alo 25mg: [LSM] alo [LSM] alo not reported. Investigator- A data not reported. no al. 2015 7.87 (0.86); 25mg: -0.46 25mg: -0.70 defined, criteria not pbo: 7.72 (0.63); pbo: (-0.905 to - specified safety monitoring board was not set (0.77) 0.24 (0.52) 0.493) up in this study.

Inagaki et trela 100mg [LSM] trela [LSM] trela not reported. Investigator- A data not reported. no al. 2015 QW: 7.73 100mg QW: - 100mg QW: - defined, criteria not (0.85); pbo: 0.32 (0.59); 0.56 (-0.754 specified safety monitoring board was not set 7.72 (0.77) pbo: 0.24 (0.52) to -0.368) up in this study.

Iwamoto Y sita 25mg [LSM] sita [LSM] sita Patients were counseled to Hypoglycemia was assessed by the not reported no 2010 QD: 7.49 25mg QD: -0.41 25mg QD: - record results of self- study site investigators through (0.82); 50mg (-0.52 to -0.29); 0.69 (-0.85 to monitored blood glucose reviewing patient self-reports of signs QD: 7.57 50mg QD: -0.71 -0.52); 50mg levels and symptoms of and symptoms of hypoglycemia. A (0.84); 100mg (-0.83 to -0.59); QD: -0.99 (- hypoglycemia (e.g., fingerstick blood glucose QD: 7.56 100mg QD: - 1.16 to -0.82); sweating, anxiety, determination concurrent with the (0.80); 200mg 0.69 (-0.81 to - 100mg QD: - palpitations, headache, episode was not required to assess an QD: 7.65 0.56); 200mg 0.96 (-1.14 to blurred vision, clouding of episode as hypoglycemia, although (0.82); pbo: QD: -0.76 (- -0.79); 200mg consciousness in diaries) investigators could include the 7.74 (0.93) 0.89 to -0.64); QD: -1.04 (- for proper assessment of fingerstick glucose measurement, if it pbo: 0.28 (0.16 1.21 to -0.86) hypoglycemic events was available, in their assessment of to 0.40) during the study. the episode.

120

Ji L 2016 sita 100mg: [LSM] sita [LSM] sita Any episode with not reported Rescue therapy with open-label glipizide. no 8.7 (1.1); pbo: 100mg: -0.99 (- 100mg: -0.40 symptoms consistent with The primary approach to analyzing safety 9.0 (1.1) 1.24 to -0.75); (-0.71 to - hypoglycemia (e.g., data treated data obtained after the pbo: -0.59 (- 0.09) weakness, dizziness, initiation of rescue therapy as missing; a 0.84 to -0.34) (p=0.011) shakiness, increased secondary approach included all data, sweating, palpitations or regardless of rescue therapy. confusion) was reported as an episode of symptomatic hypoglycemia without a requirement for confirmatory blood glucose values. Asymptomatic hypoglycemia was defined as an episode without symptoms of hypoglycemia, but with fingerstick glucose level ≤3.9 mmol/L (≤70 mg/dL). Severe hypoglycemia was defined as any episode requiring assistance, either medical or non-medical. Episodes with a markedly depressed level of consciousness, loss of

121

consciousness or seizure were to be classified as having required medical assistance, whether or not medical assistance was obtained.

Ji L 2017 alo 12.5mg [LSM] (SE) alo [LSM] alo not reported not reported Rescue confounding. Hyperglycemic no BID: 8.48 12.5mg BID: - 12.5mg BID: rescue permitted. (0.71); pbo: 0.86 (0.11); -0.68 (-0.889 8.21 (0.77) pbo: -0.19 to -0.467)

Jung CH evo 2.5mg [LSM] evo [LSM] evo not reported not reported If rescue therapy required, patients were no 2015 QD: 7.67 2.5mg QD: - 2.5mg QD: - dropped from the study. (0.83); 5mg 0.59 (0.10); 0.47 (-0.80 to QD: 7.58 5mg QD: -0.70 -0.15); 5mg (0.71); 10mg (0.10); 10mg QD: -0.57 (- QD: 7.71 QD: -0.64 0.86 to -0.29); (0.88); pbo: (0.11) 10mg QD: - 7.57 (0.87) 0.53 (-0.83 to -0.23)

Kadowaki teneli 10: 7.9 [LSM] teneli [LSM] teneli not reported Patients were provided a glucose not reported no T 2013 (0.7); teneli 10: -0.8 (0.1); 10: -0.9 (-1.0 20: 7.8 (0.7); teneli 20: -0.8 to -0.7); teneli meter and were given diaries in which teneli 40: 7.7 (0.1); teneli 40: 20: -0.9 (-1.1 to record hypoglycemic (0.7); pbo: 8.0 -0.9 (0.1); pbo: to -0.7); teneli (0.7) 0.1 (0.1) 40: -1.0 (-1.2 symptoms and self-monitored blood to -0.9) glucose (SMBG) concentrations.

Hypoglycaemia was assessed by the investigators who reviewed the patient’s self-reported signs/symptoms of

122

hypoglycaemia and SMBG data.

Kawamori lina 5mg: lina 5mg: -0.24 lina 5mg: - Hypoglycaemia was not reported not reported no R 2012 8.07 (0.66); (0.06); lina 0.87 (-1.04, - defined according to lina 10mg: 10mg: -0.25 0.70); lina American Diabetes 7.98 (0.68); (0.06); pbo: 10mg: -0.88 Association guidelines pbo: 7.95 0.63 (0.08) (-1.05, -0.71) (0.67)

Kikuchi M vilda 10mg [LSM] vilda [LSM] (SE) Symptoms suggestive of none not reported no 2009 BID: 7.40 10mg BID: - vilda 10mg hypoglycemia were (0.8); 25mg 0.53; 25mg BID: -0.8 classified as a BID: 7.40 BID: -0.67; (0.10); 25mg hypoglycemic event if (0.9); 50mg 50mg BID: - BID: -1.0 reversed by therapeutic BID: 7.40 0.92; pbo: 0.28 (0.10); 50mg actions (intake of sucrose, (0.8); pbo: 7.4 (SE n/a) BID: -1.2 etc.) or as an AE, if not (0.8) (0.10) reversed. The severity of hypoglycemic events was graded on a scale of 2. Grade 1 was defined as signs or symptoms suggestive of hypoglycemia that could be managed by the patient either with sucrose or in any other appropriate way. Grade 2 was defined as any hypoglycemic episode wherein the patient required assistance of others or hospitalization. Plasma glucose was not measured in any patient 123

during the occurrence of hypoglycemic events.

Kumar not reported, saxa 5mg: - saxa 5mg: - not reported not reported not reported no PKM 2014 inclusion 0.51; pbo: -0.05 0.46 (0.14) (- criteria were (SE n/a) 0.73 to -0.18) HbA1c (p=0.0011) between 7- adjusted 10% mean absolute reductions (SE)

Mohan V sita 100mg: [LSM] sita [LSM] sita not reported not reported not reported no 2009 8.7 (1.0); pbo: 100mg: -0.7 (- 100mg: -1.0 8.8 (1.1) 0.8 to -0.6); (-1.2 to -0.8) pbo: 0.3 (0.1 to 0.5)

Nonaka K sita 100mg: [LSM] sita [LSM] sita Symptoms of Patients were given diaries on which not reported no 2008 7.5 (0.9); pbo: 100mg: -0.65 (- 100mg: -1.05 hypoglycemia include to record study drug intake, results of 7.7 (0.9) 0.80 to -0.50); (-1.27 to - sweating, anxiety, twice a- pbo: 0.41 (0.26 0.84) palpitations, headache, to 0.56) blurred vision, clouding of week home self-monitoring of blood consciousness. Severe glucose levels (SMBG), hypoglycemia not specified. symptoms of hypoglycemia, and other symptoms. In the event

of symptoms of hypoglycemia (e.g., sweating, anxiety,

palpitations, headache, blurred vision, clouding of consciousness), patients

124

were to obtain a SMBG, take countermeasures (e.g., ingest glucose), and notify the investigator. Patients were

to notify the investigator if SMBG showed glucose levels <60 or >270 mg/dL.

Pan CY saxa 5mg: 8.1 saxa 5mg: - saxa 5mg: - Reported hypoglycaemia none; no episodes of confirmed Rescue therapy with open-label metformin no 2012 (0.8); pbo: 8.2 0.84; pbo: -0.34 0.50 (-0.65 to was defined as signs and hypoglycaemia (symptoms and was permitted. Data on or after rescue (0.8) -0.34) symptoms consistent with plasma glucose level ≤2.8 mmol/ therapy excluded from study. hypoglycaemia with or without a documented L) glucose measurement. Confirmed hypoglycemic events were defined as those associated with symptoms of hypoglycaemia and a documented plasma glucose level of ≤2.8 mmol/L.

Pan CY alo 25mg: [LSM] alo [LSM] alo Mild-to-moderate not reported Rescue therapy permitted. no 2017 8.04 (0.92); 25mg: -0.99; 25mg: -0.58 hypoglycemia, whether pbo: 7.86 pbo: -0.42 (SE (-0.78 to - symptomatic or (0.78) n/a) 0.37) asymptomatic, was defined as plasma glucose levels <3.9mmol/L. Severe hypoglycemia was defined as any hypoglycemic episode that required the assistance of another

125

person to actively administer carbohydrate, glucagon, or other resuscitative actions, and was associated with a documented plasma glucose level <3.9 mmol/L.

Park J 2017 evo 5mg: evo 5mg: -0.23; evo 5mg: - Hypoglycaemia was not reported not reported no 7.21 (0.56); pbo: 0.05 0.28 defined as signs and/or pbo: 7.20 (p<0.001) symptoms consistent with (0.63) hypoglycaemia with or without a documented glucose measurement, or as plasma glucose level ≤3.9 mmol/L without signs or symptoms.

Pi-Sunyer vilda 50mg average mean vilda 50mg Patients were provided Patient reported; with or without not reported no FX 2007 QD: 8.4 (0.9); change (SD) QD: 0.011; with glucose monitoring confirmed blood glucose 50mg BID: vilda 50mg QD: 50mg BID: devices and supplies and measurement 8.4 (0.9); -0.5 (0.1); 50mg <0.001; instructed on their use. 100mg QD: BID: -0.7 (0.1); 100mg QD: Confirmed hypoglycemia 8.3 (0.8); pbo: 100mg QD: -0.8 <0.001 was defined as symptoms 8.5 (0.8) (0.1); pbo: 0.0 suggestive of low blood (0.1) glucose confirmed by self- monitored blood glucose (SMBG) measurement <3.1 mmol/L plasma glucose equivalent. Instances of SMBG measurement <3.1 mmol/L plasma glucose equivalent without accompanying 126

symptoms were recorded as asymptomatic low blood glucose. Severe hypoglycemia was defined as any episode requiring the assistance of another party.

Pratley RE vilda 25mg vilda 25mg vilda 25mg Plasma glucose less than or Subjects were provided with glucose not reported no 2006 BID: 8.0 BID: -0.6 (0.1); BID: -0.6 equal to 3.1 mmol/l monitoring devices and supplies and (0.9); pbo: 8.1 pbo: 0.0 (0.2) (0.2) instructed on their use. An episode of (1.2) (p=0.0012) hypoglycemia was defined as symptoms consistent with hypoglycemia accompanied by a glucose measurement less than or equal to 3.1 mmol/l.

Pratley RE not reported; [LSM] (SE) alo [LSM] (SE) Mild to moderate Use of a home glucose Hyperglycemic rescue (SU or other) was no 2014 The majority 25mg QD: -0.56 manually hypoglycaemia (blood permitted. If a sulfonylurea was of patients (0.093); alo calculated: glucose <70 mg/dl (3.89 monitor and diary to record contraindicated or inappropriate, other (60% overall) 12.5mg BID: - alo 25mg QD: mmol/L), symptomatic or hypoglycemic episodes. rescue medications were prescribed at the entered with a 0.65 (0.094); -0.72; alo asymptomatic). All investigator’s discretion; these patients baseline pbo 0.15 (nss) 12.5mg BID: hypoglycemic episodes discontinued study drug but continued -0.80 were associated with a with visits and procedures. HbA1c of blood glucose <70 mg/dl 8.5% or (3.89 mmol/l). Severe lower. episodes required assistance.

Raz I 2006 sita 100mg [LSM] sita [LSM] sita not reported not reported Rescue therapy (metformin) permitted. no QD: 8.0 (0.8); 100mg QD: - 100mg QD: - Data collected prior to rescue were 200mg QD: 0.48 (-0.61 to - 0.60 (-0.82 to included in the efficacy analyses. 8.1 (0.9); pbo: 0.35); 200mg -0.39); 200mg 8.0 (0.9) QD: -0.36 (-

127

0.48 to -0.23); QD: -0.48 (- pbo: 0.12 (-0.05 0.70 to -0.26) to 0.30)

Rhee EJ gemi 50mg: gemi 50mg: - gemi 50mg: - not reported no not reported no 2010 8.24 (1); 0.98 (0.76); 0.92 (0.76); 100mg: 8.18 100mg: -0.74 100mg: -0.68 (1.09); (0.79); 200mg: - (0.78); 200mg: 8.16 0.78 (0.78); 200mg; -0.72 (1.13); pbo; pbo: -0.06 (0.77) 8.2 (1.21) (0.76)

Ristic S vilda 25mg [LSM] (SE): [LSM] (SE): All events (symptomatic or Determined by investigator not reported no 2005 BID: 7.64 25mg BID: - 25mg BID: - asymptomatic) determined (0.69); 25mb 0.31 (0.11); 0.18 (0.15); by the investigator as QD: 7.73 25mg QD: -0.27 25mb QD: - hypoglycaemia were (0.80); 50mg (0.10); 50mg 0.14 (0.14); included in the adverse QD: 7.70 QD: -0.56 50mg QD: - event summary. There (0.82); 100mg (0.10); 100mg 0.43 (0.14); were only two cases of QD: 7.64 QD: -0.53 100mg QD: - symptomatic confirmed (0.75); pbo: (0.10); pbo: - 0.40 (0.14) hypoglycaemia (plasma 7.76 (0.83) 0.13 (0.10) glucose 3.7 mmol/l).

Roden empa 10mg: empa 10mg: - empa 10mg: - Confirmed hypoglycemic Confirmed hypoglycemia either Rescue medication permitted. The no 2013 7.87 (0.88); 0.66 (-0.76 to - 0.74 (-0.88 to adverse events plasma glucose <3·9 mmol/L or initiation, choice, and dose of rescue 25mg: 7.86 0.56); 25mg: - -0.59); 25mg: (hypoglycemic episodes requiring assistance, or both medication used were at the discretion of (0.85); pbo: 0.78 (-0.88 to - -0.85 (-0.99 reported as adverse events the investigator. 7.91 (0.78) 0.67); pbo: 0.08 to -0.71) when plasma glucose was (-0.03 to 0.18) <3·9 mmol/L, required assistance, or both)

128

Rosenstock saxa 2.5mg (SE) [95% CI]: (SE) [95% Confirmed hypoglycaemia Confirmed hypoglycaemia, defined as Rescue medication permitted. no J 2008 QD: 7.7 saxa 2.5mg QD: CI]: saxa was defined as a finger- a finger-stick blood glucose value 50 (0.97); 5mg -0.72 (0.12) (- 2.5mg QD: - stick blood glucose value mg/dl. QD: 7.9 0.97 to -0.48); 0.45 (0.17) (- of ≤50 mg/dl (2.78 (1.09); 10mg 5mg QD: -0.90 0.78 to -0.13); mmol/L) associated with QD: 8.0 (0.14) (-1.17 to 5mg QD: - classical symptoms. Severe (1.14); 20mg -0.63); 10mg 0.63 (0.18) [- hypoglycemia defined as QD: 7.9 QD: -0.81 0.97 to -0.29] requiring the assistance of (0.99); 40mg (0.11) (-1.03 to ; 10mg QD: - another person to actively QD: 7.8 -0.58); 20mg 0.54 (0.16) [- administer carbohydrate, (1.00); 100mg QD: -0.74 0.85 to -0.23]; glucagon or other. QD: 7.8 (0.12) (-0.98 to 20mg QD: - (1.01); pbo: -0.50); 40mg 0.47 (0.17) (- 8.0 (0.98) QD: -0.80 0.80 to -0.14); (0.12) (-1.04 to 40mg QD: - -0.56); 100mg 0.53 (0.17) (- QD: n/a; pbo: 0.86 to -0.20) 0.27 (0.11) (- 0.49 to -0.05)

Rosenstock saxa 2.5mg (SE) saxa manually Symptoms of Reported by pt; no cases confirmed Open-label metformin as rescue therapy. no J 2009 A QD: 7.9 (0.9); 2.5mg QD: - calculated hypoglycemia and with fingerstick Efficacy and safety measurements 5mg QD: 8.0 0.43; 5mg QD: - saxa 2.5mg confirmed hypoglycemia, obtained after rescue were not included (1.1); 10mg 0.46; 10mg QD: QD: -0.24; defined as symptoms of QD: 7.8 (0.9); -0.54; pbo: 0.19 5mg QD: - hypoglycemia with a in analyses. pbo: 7.9 (0.9) 0.27; 10mg fingerstick glucose ≤50 QD: -0.35 mg/dL (2.8 mmol/L), were also recorded.

129

Scherbaum vilda 50mg (adjusted mean vilda 50mg Hypoglycaemia was Self-confirmed with blood glucose not reported No, not WA 2008 QD: 6.7 (0.4); change) vilda QD: -0.3 (0.1) defined as symptoms measurement specifically. But pbo: 6.8 (0.4) 50mg QD: -0.2 (p<0.001) suggestive of low blood patients with (0.1) (p<0.001); adjusted glucose confirmed by self- diabetic pbo: 0.1 (0.1) mean change monitored blood glucose complications or measurement <3.1 mmol/l lab abnormalities plasma glucose equivalent. within the last 6 Severe hypoglycaemia was months were defined as any episode excluded. requiring the assistance of another party.

Scott R sita 5mg BID: [LSM] sita 5mg [LSM] sita not reported Hypoglycaemia was assessed by study not reported No, but they did 2007 7.9 (1.0); BID: -0.15 (- 5mg BID: - site investigators through reviewing exclude patients 12.5mg BID: 0.29 to -0.01); 0.38 (-0.58 to with 7.9 (0.9); 12.5mg BID: - -0.19); daily glucose logs and patient self- hypoglycemia 25mg BID: 0.41 (-0.55 to - 12.5mg BID: report of signs and symptoms of from glipizide arm 7.9 (0.9); 0.27); 25mg -0.64 (-0.84 hypoglycaemia. of study (data not 50mg BID: BID: -0.43 (- to -0.45); recorded here). 7.8 (1.0); pbo: 0.56 to -0.29); 25mg BID: - 7.9 (1.0) 50mg BID: - 0.66 (-0.85 to 0.54 (-0.68 to - -0.47); 50mg 0.40); pbo: 0.23 BID: -0.77 (- (0.10 to 0.37) 0.96 to -0.58)

Sheu WHH omari 0.25mg [LSM] omari [LSM] omari not reported not reported Rescued with open-label metformin. Data no 2015 once weekly: 0.25mg once 0.25mg once acquired after the initiation of rescue 8.1 (0.9); 1mg weekly: -0.14 (- weekly: -0.28 therapy were treated as missing. once weekly: 0.30 to 0.01); (-0.50 to - 8.0 (0.9); 3mg 1mg once 0.06); 1mg once weekly: weekly: -0.36 (- once weekly:

130

7.9 (0.9); 0.51 to -0.20); -0.50 (-0.71 10mg once 3mg once to -0.28); weekly: 8.0 weekly: -0.35 (- 3mg once (0.9); 25mg 0.50 to -0.19); weekly: -0.49 once weekly: 10mg once (-0.70 to - 8.1 (1.0); pbo: weekly: -0.53 (- 0.27); 10mg 8.1 (0.9) 0.68 to -0.37); once weekly: 25mg once -0.67 (-0.88 weekly: -0.57 (- to -0.45); 0.73 to -0.42); 25mg once pbo: 0.14 (-0.01 weekly: -0.72 to 0.30) (-0.93 to - 0.50)

Wu W lina 5mg: [sd not manually Hypoglycemia was defined Self-reported capillary blood glucose not reported no 2015 7.97 (0.68); reported] lina calculated: - according to American levels of 3.5 mmol/L, 3.7 mmol/L, pbo: 8.00 5mg; -1.2 (0.7); 0.8 (n/a) Diabetes Association and 3.7 mmol/L. (0.69) pbo: -0.4 (0.4) guidelines (Seaquist et al. 2013)

Yang SJ gemi 50 mg: [LSM] not [LSM] gemi not reported no Rescue therapy with metformin permitted. no 2013 8.2 (1.0); pbo: reported 50 mg: -0.71 8.3 (1.1) (-1.04 to - 0.37)

Yang HK ana 100mg ana 100mg ana 100mg not reported not reported not reported no 2015 BID: 7.13 BID: -0.50 BID: -0.73; (0.72); 200mg (0.45); 200mg 200mg BID: - BID: 7.19 BID: -0.51 0.74 (CI (n/a) (0.73); pbo: (0.55); pbo: 7.11 (0.63) 0.23 (0.62)

131

Characteristics of Included Studies – GLP1RA Monotherapy

Study ID Dose Study n= Mean Age Gender Countries Ethnicity (%) Duration of Diabetes in Duration (SD) (% Studied Years (SD) male)

Fonseca VA lixi 20μg 12 weeks lixi 2-step lixi 2-step lixi 2- Belgium, India, lixi 2-step dose increase: Caucasian 73.3, Asian 22.5, Black [MEDIAN DURATION w 2012 QD dose dose step Israel, Japan, 0, Other 4.2; 1-step dose increase: Caucasian 71.4, Asian range] 2-step dose increase: 1.4 increase: increase: dose South Korea, 24.4, Black 2.5, Other 1.7; pbo: Caucasian 73.8, Asian 19.7, (0.2-21.5); 1-step dose 120; 1-step 53.3 (9.7); increase: Mexico, Poland, Black 2.5, Other 4.1 increase: 1.1 (0.2-23.9); pbo: dose 1-step dose 52.5; 1- Romania, 1.4 (0.2-12.5) increase: increase: step Russian 119; pbo: 53.8 (10.9); dose Federation, 122 pbo: 54.1 increase: Tunisia, (11.0) 52.9; Ukraine, USA pbo: 49.2

Grunberger G dula 0.1mg 12 weeks dula dula 0.1mg dula 7 countries; not dula 0.1mg QW: Caucasian 83, Asian 11, Black or African- dula 0.1mg QW: 3.9 (3.2); 2012 QW; dula 0.1mg: 35; QW: 56.3 0.1mg specified American 3, Others 3; 0.5mg: Caucasian 82, Asian 15, Black 0.5mg: 3.7 (3.8); 1.0mg: 3.3 0.5mg 0.5mg: 34; (9.2); QW: or African-American 3, Others 0; 1.0mg: Caucasian 77, Asian (2.5); 1.5mg: 4.6 (4.1); pbo: QW; dula 1.0mg: 34; 0.5mg: 56.9 31.4; 15, Black or African-American 0, Others 9; 1.5mg: Caucasian 3.9 (4.7) 1.0mg 1.5mg: 29; (9.1); 0.5mg: 83, Asian 14, Black or African-American 3, Others 0; pbo: QW; dula pbo: 32 1.0mg: 57.2 47.1; Caucasian 78, Asian 16, Black or African-American 3, Others 1.5mg QW (8.8); 1.0mg: 3 1.5mg: 57.5 47.1; (7.9); pbo: 1.5mg: 55.0 (9.3) 44.8; pbo:56.3

Madsbad S lira 12 weeks lira lira 0.045mg lira Scandinavia, not reported lira 0.045mg QD: 4.1 (3.7); 2004 0.045mg 0.045mg QD: 53 0.045mg United 0.225mg QD: 4.4 (4.0); QD; lira QD: 26; (9.0); QD: Kingdom 0.45mg QD: 4.5 (4.6); 0.60mg 0.225mg 0.225mg 0.225mg 84.61; 132

QD; lira QD: 24; QD: 58 0.225mg QD: 4.6 (4.6); 0.75mg QD: 6.1 0.45mg 0.45mg (7.5); QD: (7.9); pbo: 3.4 (2.9) QD; lira QD: 27; 0.45mg QD: 62.5; 0.60mg 0.60mg 57 (11.3); 0.45mg QD; lira QD: 30; 0.60mg QD: QD: 0.75mg 0.75mg 57 (7.7); 66.67; QD QD: 28; 0.75mg QD: 0.60mg pbo: 29 58 (9.7); QD: pbo: 57 66.67; (9.4) 0.75mg QD: 57.14; pbo: 68.96

Moretto TJ exen 5μg 24 weeks exen 5μg: exen 5μg: exen USA, Puerto exen 5μg: White 65, Asian 29, Hispanic 6, Black 0; exen exen 5μg: 2 (3); exen 10μg: 2 2008 SC BID; 77; exen 54 (10); 5μg: 52; Rico, Romania, 10μg: White 72, Asian 23, Hispanic 1, Black 4; pbo: White (3); pbo: 1 (2) exen10μg 10μg: 78; exen 10μg: exen Russia, India 66, Asian 27, Hispanic 3, Black 4 SC BID pbo: 77 55 (10); 10μg: pbo: 53 (9) 62; pbo 55

Nauck MA albi 30mg 52 weeks albi 30mg: albi 30mg: albi USA, Mexico albi 30mg: White 84.2, African American/African 9.9, Asian albi 30mg: 3.4 (3.7); 50mg: 4.2 2016 QW; albi 101; 50mg: 53.6 (10.9); 30mg: 1.0; 50mg: White 78.8, African American/African 14.1, (4.6); pbo 4.3 (4.0) 50mg QW 99; pbo 50mg: 52.0 57.4; Asian 1.0; pbo: White 78.2, African American/African 13.9, 101 (11.8); pbo: 50mg: Asian 5.0. Ethnicity Hispanic/Latino: albi 30mg: 29.7; 50mg: 53.1 (11.7) 50.5; 26.3; pbo: 28.7 pbo: 57.4

Seino Y 2008 lira 0.1mg 14 weeks lira 0.1mg lira 0.1mg lira Japan Japanese lira 0.1mg SC QD: 7.15 (5.14); SC QD; SC QD: SC QD: 0.1mg 0.3mg SC QD: 6.78 (4.69); lira 0.3mg 45; 0.3mg 56.5 (8.4); SC QD: 0.6mg SC QD: 8.87 (6.77); SC QD; SC QD: 0.3mg SC 68.89;

133

lira 0.6mg 46; 0.6mg QD: 56.8 0.3mg 0.6mg SC QD: 7.62 (4.92); SC QD; SC QD: (8.8); 0.6mg SC QD: pbo: 7.48 (5.65) lira 0.6mg 45; 0.6mg SC QD: 69.57; SC QD SC QD: 60.0 (7.0); 0.6mg 44; pbo: 46 0.6mg SC SC QD: QD: 55.5 62.22; (7.6); pbo: 0.6mg

57.5 (8.7) SC QD: 70.45;

pbo: 63.04

Seino Y 2014 albi 15mg 16 weeks albi 15mg albi 15mg albi Japan Japanese albi 15mg QW: 6.3 (4.6); 30mg QW; albi QW: 52; QW: 53.3 15mg QW: 7.8 (5.4); 30mg BIW: 7.2 30mg QW; 30mg QW: (10.3); QW: (5.6); pbo: 6.7 (7.8) albi 30mg 54; 30mg 30mg QW: 61.5; BIW BIW: 53; 58.0 (9.3); 30mg pbo: 53 30mg BIW: QW: 59.1 (8.5); 70.4; pbo: 57.5 30mg (11.1) BIW: 77.4; pbo 69.8

Sorli C 2017 sema 30 weeks sema sema 0.5mg sema Canada, Italy, ETHNIC Origin: sema 0.5mg SC QD: Hispanic/Latino 27; sema 0.5mg SC QD: 4.81 0.5mg SC 0.5mg SC SC QD: 0.5mg Japan, Mexico, sema 1.0mg SC QD hispanic/latino: 35; pbo hispanic/latino: (6.1); 1.0mg SC QD: 3.62 QD; sema QD: 128; 54.6 (11.1); SC QD: Russia, South 28; RACE: sema 0.5mg SC QD White 65, Black or African (4.88); pbo: 4.06 (5.48) 1.0mg SC 1.0mg SC 1.0mg SC 47; Africa, UK, American 9, Asian 20; sema 1.0 mg SC QD White 68, Black QD QD: 130; QD: 52.7 1.0mg USA or African American 8, Asian 19; pbo: White 64, Black or pbo: 129 (11.9); pbo: SC QD: African American 8, Asian 21 53.7 (11.3)

134

62; pbo: 54

Vilsbøll T lira 14 weeks lira lira 1.90mg: lira Denmark, not reported lira 1.90mg: 4.0 (1-29); 2007 0.65mg 1.90mg: 55.4 (11.4); 1.90mg: France, 1.25mg: 7.0 (0-21); 0.65mg: QD; lira 41; 1.25mg: 73.17; Netherlands, 6.0 (1-25); pbo: 5.0 (1-23) 1.25mg 1.25mg: 53.8 (10.7); 1.25mg: Slovakia QD; lira 42; 0.65mg: 54.76; 1.90 mg 0.65mg: 56.5 (9.3); 0.65mg: QD 40; pbo: 40 pbo: 57.7 67.50; (8.2) pbo: 47.50

135

Characteristics of Included Studies – GLP1RA Monotherapy continued

Study ID Baseline A1C Mean Mean HYPO def Ascertainment of Hypo Rescue Excl of Pts w % (SD) Change in Difference in Medication events? At HbA1C % HbA1c % vs screening or vs. baseline pbo during study)

Fonseca 2-step dose [LSM] 2-step [LSM] [no sd] Symptomatic hypoglycemia was defined as Safety and tolerability were assessed Rescue no VA 2012 increase: 7.98 dose increase: 2-step dose symptoms consistent with hypoglycemia, (metformin) (0.9); 1-step -0.77; 1-step increase: -0.54; with accompanying blood glucose <3.3 by physical examination, blood pressure, permitted. dose increase: dose increase: 1-step dose mmol/L (60 mg/dL) and/or prompt recovery 8.07 (0.9); pbo: -0.94; pbo: - increase: -0.66 with carbohydrate. Severe symptomatic heart rate, 12-lead electrocardiogram, 8.07 (0.9) 0.27 hypoglycemia was defined as symptomatic standard laboratory measurements, anti hypoglycemia in which the patient required lixisenatide antibodies, and adverse the assistance of another person and which was associated either with a plasma glucose events reporting (including, in particular, level <36 mg/dL (2.0 mmol/L) or, if no plasma glucose measurement was available, symptomatic and severe symptomatic with prompt recovery with carbohydrate.

hypoglycemia, local intolerability at injection

site, allergic or allergic-like reactions,

suspected pancreatitis, and major

cardiovascular events).

Grunberger dula 0.1mg [LSM] dula [LSM] Referencing the American Diabetes not reported Use of no G 2012 QW: 7.1 (0.6); 0.1mg QW: - manually Association definition, hypoglycaemia was additional 0.5mg: 7.2 0.37 (-0.69 to calculated dula defined as plasma glucose ≤ 3.9 mmol (≤ 70 oral (0.6); 1.0mg: -0.06); 0.1mg QW: - mg⁄dl) and ⁄ or symptoms and ⁄ or signs antidiabetes 7.3 (0.7); 0.5mg: -0.89 0.38; 0.5mg: - attributable to hypoglycaemia. Severe drugs was 1.5mg: 7.3 (-1.21 to - 0.90; 1.0mg: - hypoglycaemia was defined as an episode permitted 0.57); 1.0mg: only when 136

(0.4); pbo: 7.4 -1.04 (-1.36 1.05; 1.5mg: - requiring the assistance of another person to needed for (0.6) to -0.72); 1.05 actively administer therapy rescue. 1.5mg: -1.04 (-1.39 to - 0.70); pbo: 0.01 (SE 0.13)

Madsbad S lira 0.045mg lira 0.045mg not reported Symptoms and Minor reported. Minor Safety parameters (adverse events, not reported no 2004 QD: 7.4 (0.8); QD: 0.25; hypoglycemia defined in the study as a blood hypoglycemic episodes, weight, standard 0.225mg QD: 0.225mg QD: glucose < 2.8 mmol/l, thus included as hematology and biochemistry profile, vital 7.9 (0.8); -0.34; 0.45mg severe. signs, and electrocardiogram) were assessed at 0.45mg QD: 7.7 QD: -0.30; each visit; Each patient was seen on seven (1.0); 0.60mg 0.60mg QD: - occasions: screening; baseline; after 1, 4, 8, QD: 7.4 (1.2); 0.70; 0.75mg and 12 weeks of treatment; and follow-up. 0.75mg QD: 7.4 QD: -0.75; (0.9); pbo: 7.4 pbo: n/a Patients were supplied with a blood glucose (1.2) meter (One Touch Profile Glucometer) and instructed in its use. Fasting blood glucose was measured every morning. HbA1c,

fasting serum glucose, insulin, C-peptide, and glucagon were measured every 4 weeks.

Moretto TJ exen 5μg: 7.9 [LSM] exen [LSM] Hypoglycemia was defined as signs or Patient-reported not reported no 2008 (1.0); exen 5μg: -0.7 manually symptoms associated with hypoglycemia, or 10μg: 7.8 (1.0); (0.1); exen calculated exen an SMBG value <64 mg/dL (3.5 mmol/L), pbo: 7.8 (0.9) 10μg: -0.9 5μg: -0.5 exen regardless of whether this concentration was (0.1); pbo: - 10μg: -0.7 0.2 (0.1) considered to be associated with signs, symptoms, or treatment. Severe hypoglycemia was defined as an episode

137

with signs or symptoms consistent with hypoglycemia during which the patient required the assistance of

another person and that was associated with an SMBG value <54 mg/dL (3.0 mmol/L) or prompt recovery after administration

of oral carbohydrate, glucagon injection, or IV glucose.

Nauck MA albi 30mg: 8.0 [LSM] not [LSM] albi American Diabetes Association criteria; not reported Rescue no 2016 (0.8); 50mg: 8.2 reported 30mg: -0.84 (- Severe event requiring another person to medication (0.9); pbo 8.0 1.11, -0.58); administer a resuscitative action; (metformin, (0.9) 50mg: -1.04 (- sulfonylureas 1.31, -0.77) Documented symptomatic—plasma glucose insulin and concentration ≤3.9 mmol/l (70 mg/dl) and sitagliptin) presence of hypoglycemic symptoms. permitted. HYPOGLYC EMIA COUNT IS PRE- RESCUE.

Seino Y lira 0.1mg SC not reported lira 0.1mg SC Minor was defined as symptomatic events Seven-point plasma glucose profile which not reported no 2008 QD: 8.50 QD: -0.79 (- confirmed by plasma glucose <3.1 mmol/L. (0.84); 0.3mg 1.08, -0.50); Major not defined. was measured before and approximately 2 h SC QD: 8.24 0.3mg SC QD: - after each meal, and at bedtime by self- (0.92); 0.6mg 1.22 (-1.50, - monitoring at home using glucose SC QD: 8.21 0.93); 0.6mg SC (0.83); 0.6mg QD: -1.64 (- meters before start of treatment and end of SC QD: 8.12 1.93, -1.35); study, 0.6mg SC QD: -

138

(0.98); pbo: 1.85 (-2.14, - 8.43 (1.02) 1.56) ANOVA model with dose group and pre- treatment as fixed effects and baseline value as covariate

Seino Y albi 15mg QW: albi 15mg albi 15mg QW: not reported. not reported not reported no 2014 8.54 (0.82); QW: -0.63 ( - -0.89 (-1.22, - 30mg QW: 8.54 0.61); 30mg 0.57); 30mg (0.86); 30mg QW: -1.29 (- QW: -1.55 (- BIW: 8.54 1.27); 30mg 1.88, -1.23); (0.81); pbo: BIW: -0.84 (- 30mg BIW: - 8.51 (0.73) 0.82); pbo: 1.10 (-1.43, - 0.27 (0.28) 0.78) [model- adjusted difference from pbo]

Sorli C sema 0.5mg SC sema 0.5mg sema 0.5mg SC Severe hypoglycaemia or blood glucose- not reported Rescue no 2017 QD: 8.09 SC QD: -1.45 QD: -1.43 (- confirmed hypoglycaemia (<3·1 mmol/L permitted. (0.89); 1.0mg (-1.65, -1.26); 1.71, -1.15); Data before SC QD: 8.12 1.0mg SC 1.0mg SC QD: - initiation of (0.81); pbo 7.95 QD: -1.55 (- 1.53 (-1.81, - any rescue (0.85) 1.74, -1.36); 1.25) medication pbo: -0.02 (- used. 0.23 to 0.18)

Vilsbøll T lira 1.90mg: 8.5 lira 1.90mg: - lira 1.90mg: - Major or minor not reported not reported no 2007 (0.9); 1.25mg: 1.45; 1.25mg: 1.74 (-2.18 to - 8.3 (0.8); -1.4; 0.65mg: 1.29); 1.25mg: -

139

0.65mg: 8.1 -0.98; pbo: 1.69 (-2.13 to - (0.6); pbo: 8.2 0.29 1.24); 0.65mg: - (0.7) 1.27 (-1.72 to - 0.82)

Characteristics of Included Studies – SGLT2i Monotherapy

Study ID Dose Study n= Mean Age (SD) Gender (% male) Countries Studied Ethnicity (%) Duration of Diabetes in Years Duration (SD)

Bailey CJ dapa 1mg 24 weeks dapa 1mg dapa 1mg QD: dapa 1mg QD: 52.8; USA, Canada, Mexico, not reported dapa 1mg QD: 1.6 (2.55); 2.5mg 2012 QD; dapa QD: 72; 53.7 (9.04); 2.5mg 2.5mg QD: 45.9; 5mg Russia, India, South QD: 1.5 (2.19); 5mg QD: 1.4 2.5mg QD; 2.5mg QD: QD: 53.5 (10.61); QD: 47.1; pbo: 54.4 Africa, Puerto Rico (3.24); pbo: 1.1 (1.95) dapa 5mg 74; 5mg 5mg QD: 51.3 QD QD: 68; (11.51); pbo: 53.5 pbo: 68 (11.08)

Ferrannini E dapa 2.5mg 24 weeks dapa dapa 2.5mg QD: dapa 2.5mg QD: 55.4; USA, Canada, Mexico, not reported [MEDIAN] dapa 2.5mg QD: 0.50 2010 QD; dapa 2.5mg QD: 53.0 (11.7); 5mg 5mg QD: 48.4; 10mg Russia (0.1 to 2.90); 5mg QD: 0.25 (0.10 5mg QD; 65; 5mg QD: 52.6 (10.9); QD: 48.6; pbo: 41.3 to 1.40); 10mg QD: 0.45 (0.10 to dapa 10mg QD: 64; 10mg QD: 50.6 3.40); pbo: 0.50 (0.10 to 3.40) QD in the 10mg QD: (9.97); pbo 52.7 MORNING 70; pbo: 75 (10.3)

(evening doses were exploratory)

140

Ferrannini E empa 5mg 12 weeks empa 5mg empa 5mg QD: empa 5mg QD: 56.8; Argentina, Croatia, empa 5mg QD: White 63.0, not reported 2013 QD; empa QD: 81; 59.0 (37-78); 10mg QD: 49.4; 25mg Estonia, Germany, Italy, Asian 35.8, Other 1.2; 10mg 10mg QD; 10mg QD: 10mg QD: 58.0 QD: 50.0; pbo: 54.9 korea, Lithuania, QD: White 64.2, Asian 34.6, empa 25mg 81; 25mg (30-76); 25mg Romania, Russia, Other 1.2; 25mg QD: White QD QD: 82; QD: 57.0 (30-79); Sweden, Slovakia, 65.9, Asian 32.9, Other 1.2; pbo: 82 pbo: 58.0 (28-80) Taiwan, Ukraine pbo: White 65.9, Asian 34.1, (median range) Other 0

Fonseca VA ipra 12.5mg 12 weeks ipra ipra 12.5mg QD: ipra 12.5mg QD: 55.7; not reported ipra 12.5mg QD: White 58.6, ipra 12.5mg QD: 4.08 (3.24); 2013 QD; ipra 12.5mg 53.9 (9.6); 50mg 50mg QD: 50.7; 150mg Non-white 41.4; 50mg QD: 50mg QD: 4.61 (4.65); 150mg 50mg QD; QD: 70; QD: 52.6 (10.7); QD: 42.6; 300mg QD: White 52.2, Non-white 47.8; QD: 5.11 (6.46); 300mg QD: 4.48 ipra 150mg 50mg QD: 150mg QD: 54.2 54.4; pbo: 46.4 150mg QD: White 54.4, (4.91); pbo: 4.64 (5.93) QD; ipra 67; 150mg (10.3); 300mg Non-white 45.6; 300mg QD: 300mg QD QD: 68; QD: 54.2 (10.7); White 52.9, Non-white 47.1; 300mg pbo: 53.4 (9.7) pbo: White 47.8, Non-white QD: 68; 52.2 pbo: 69

Inagaki N cana 50mg 12 weeks cana 50mg cana 50mg QD: cana 50mg QD: 61; Japan Japanese not reported 2013 QD; cana QD: 82; 57.4 (10.8); 100mg QD: 70.3; 200mg 100mg QD; 100mg 100mg QD: 57.7 QD: 64.5; 300mg QD: cana 200mg QD: 74; (10.5); 200mg 73.3; pbo 72.0 QD; cana 200mg QD: 57.0 (1.07); 300mg QD QD: 76; 300mg QD: 57.1 300mg (10.1); pbo: 57.7 QD: 75; (11.0) pbo: 75

141

Inagaki N cana 100mg 24 weeks cana cana 100mg QD: cana 100mg QD: 65.6; Japan Japanese cana 100mg QD: 4.72 (4.59); 2014 QD; cana 100mg 58.4 (10.4); 200mg QD: 81.8; pbo: 200mg QD: 5.88 (5.93); pbo: 200mg QD QD: 90; 200mg QD: 57.4 64.5 5.63 (5.76) 200mg (11.1); pbo: 58.2 QD: 88; (11.0) pbo: 93

Ji L 2014 dapa 5mg 24 weeks dapa 5mg: dapa 5mg: 53.0 dapa 5mg: 65.6; 10mg: China, South Korea, dapa 5mg: Chinese 89.1, dapa 5mg: 1.15 (2.0); 10mg: 1.67 QD; dapa 128; (11.07); 10mg: 64.7; pbo: 65.9 Taiwan, India Asian Indian 6.3, Korean 4.7, (2.8); pbo: 1.30 (2.0) 10mg QD 10mg: 51.2 (9.89); pbo: Japanese 0, Other Asian 0; 133; pbo: 49.9 (10.87) 10mg: Chinese 88.0, Asian 132 Indian 6.8, Korean 3.8, Japanese 0.8, Other Asian 0.8; pbo: Chinese 88.6, Asian Indian 6.1, Korean 3.8, Japanese 0.8, Other Asian 0.8

Kadowaki T empa 5mg 12 weeks empa 5mg empa 5mg QD: empa 5mg QD: 76.4; Japan Japanese (438) not reported 2014 QD; empa QD: 110; 57.3 (11.2); 10mg 10mg QD: 70.6; 25mg 10mg QD; 10mg QD: QD: 57.9 (9.4); QD: 77.1; 50mg QD: empa 25mg 109; 25mg 25mg QD: 57.2 77.3; pbo: 73.4 QD; empa QD: 109; (9.7); 50mg QD: 50mg QD 50mg QD: 56.6 (10.3); pbo: 110; pbo: 58.7 (8.7) 109

Kaku K dapa 1mg 12 weeks dapa 1mg dapa 1mg QD: dapa 1mg QD: 79.7; Japan Japanese dapa 1mg QD: 4.89 (4.37); 2.5mg 2013 QD; dapa QD: 59; 55.9 (9.7); 2.5mg 2.5mg QD: 69.6; 5mg QD: 4.41 (3.97); 5mg QD: 5.34 2.5mg QD; 2.5mg QD: QD: 57.7 (9.3); QD: 81.0; 10mg QD: (4.51); 10mg QD: 4.73 (4.73); dapa 5mg 56; 5mg 5mg QD: 58.0 75.0; pbo: 79.6 pbo: 4.74 (3.82) QD; dapa QD: 58; (9.5); 10mg QD: 10mg QD 10mg QD: 56.5 (11.5); pbo: 52; pbo: 54 58.4 (10.0)

142

Kaku K dapa 5mg 24 weeks dapa 5mg: dapa 5mg: 58.6 dapa 5mg: 58.1; 10mg: Japan Japanese dapa 5mg: 4.59 (5.56); 10mg: 2014 A QD; dapa 86; 10mg: (10.4); 10mg: 57.5 60.2; pbo: 59.8 4.93 (4.52); pbo 5.29 (6.17) 10mg QD 88; pbo: 87 (9.3); pbo: 60.4 (9.7)

Kaku K tofo 10mg 24 weeks tofo 10mg tofo 10mg QD: tofo 10mg QD: 66.7; Japan Japanese tofo 10mg QD: 6.3(7.1); 20mg 2014 B QD; tofo QD: 57; 58.6 (9.8); 20mg 20mg QD: 67.2; 40mg QD: 6.4 (5.1); 40mg QD: 6.7 20mg QD; 20mg QD: QD: 56.6 (10.2); QD: 67.2; pbo: 66.1 (5.5) tofo 40mg 58; 40mg 40mg QD: 57.0 QD QD: 58; (9.1); pbo: 56.8 pbo: 56 (9.9)

Kashiwagi ipra 12.5mg 12 weeks ipra ipra 12.5mg QD: ipra 12.5mg QD: 58.90; Japan not reported manually calculated: ipra 12.5mg A 2014 QD; ipra 12.5mg 55.3 (10.2); 25mg 25mg QD: 66.22; 50mg QD: 6.31; 25mg QD: 6.40; 50mg 25mg QD; QD: 73; QD: 57.0 (10.4); QD: 59.72; 100mg QD: QD: 6.64; 100mg QD: 7.8; pbo: ipra 50mg 25mg QD: 50mg QD: 55.9 68.06; pbo: 71.01 6.31 QD; ipra 74; 50mg (11.4); 100mg 100mg QD QD: 72; QD: 56.0 (10.4); 100mg pbo: 55.2 (9.7) QD: 72; pbo: 69

Roden 2013 empa 10mg 12 weeks empa empa 10mg: 56.2 empa 10mg: 63; 25mg: Belgium, Canada, China, empa 10mg: Asian 64, White empa 10mg: ≤1 year: 39, >1-5 QD; empa 10mg: (11.6); 25mg: 53.8 65; pbo: 54 Germany, India, Ireland, 34, Black/African-American years: 41, >5-10 years: 13, >10 25mg QD 224; (11.6); pbo: 54.9 Japan, Switzerland and 1, American-Indian/Alaska years: 7; 25mg: ≤1 year: 41, >1-5 25mg: (10.9) USA Native 0, Hawaiian/Pacific years: 37, >5-10 years: 17, >10 224; pbo: Islander 1, 25mg: Asian 64, years 6; pbo: ≤1 year: 32, >1-5 228 White 33, Black/African- years: 46, >5-10 years: 15, >10 American 3, American- years: 8 Indian/Alaska Native 0, Hawaiian/Pacific Islander 0; pbo: Asian 64, White 33, Black/African-American 3,

143

American-Indian/Alaska Native 0, Hawaiian/Pacific Islander 0

Seino Y luseo 1mg 12 weeks luseo 1mg luseo 1mg QD: luseo 1mg QD: 72.7; Japan not reported luseo 1mg QD: 4.7 (4.1); 2.5mg 2014 A QD; luseo QD: 55; 58.5 (9.1); 2.5mg 2.5mg QD: 67.9; 5mg QD: 4.6 (4.4); 5mg QD: 4.5 (4.2); 2.5mg QD; 2.5mg QD: QD: 57.4 (9.3); QD: 75.9; 10mg QD: 10mg QD: 6.2 (5.4); pbo 5.1 (4.6) luseo 5mg 56; 5mg 5mg QD: 57.3 63.8; pbo: 71.9 QD; luseo QD: 54; (11.4); 10mg QD: 10mg QD 10mg QD: 59.6 (7.8); pbo: 58; pbo: 57 57.1 (10.0)

Seino Y luseo 0.5mg 12 weeks luseo luseo 0.5mg QD: luseo 0.5mg QD: 68.3; Japan not reported luseo 0.5mg QD: 4.90 (4.49); 2014 B QD; luseo 0.5mg QD: 55.2 (10.1); 2.5mg 2.5mg QD: 57.4; 5mg 2.5mg QD: 6.15 (6.50); 5mg QD: 2.5mg QD; 60; 2.5mg QD: 58.3 (9.4); QD: 72.1; pbo: 74.1 5.77 (5.55); pbo: 7.30 (6.43) luseo 5mg QD: 61; 5mg QD: 56.8 QD 5mg QD: (9.3); pbo: 57.6 61; pbo: 54 (11.0)

Stenlöf K cana 100mg 26 weeks cana cana 100mg QD: cana 100mg QD: 41.5; 17 countries; not cana 100mg QD: White 63.6, cana 100mg QD: 4.5 (4.4); 2013 QD; cana 100mg 55.1 (10.8); 300mg QD: 45.2; pbo: specified Black or African American 300mg QD: 4.3 (4.7); pbo: 4.2 300mg QD QD: 195; 300mg QD: 55.3 45.8 9.2, Asian 13.8, Other 13.3; (4.1) 300mg (10.2); pbo: 55.7 300mg QD: White 69.5, QD: 197; (10.9) Black or African American pbo: 192 7.1, Asian 14.7, Other 8.6; pbo: White 69.8, Black or African American 4.7, Asian 15.1, Other 10.4

Terra SG ertu 5mg 52 weeks ertu 5mg ertu 5mg QD: ertu 5mg QD: 57.1; USA, Canada, Israel, ertu 5mg QD: American ertu 5mg QD: 5.11 (5.09); 15mg 2017 QD; ertu QD: 156; 56.8 (11.4); 15mg 15mg QD: 59.2; pbo: Italy, Mexico, South Indian or Alaska Native 0, QD: 5.22 (5.55); pbo: 4.63 (4.52) 15mg QD 15mg QD: QD: 56.2 (10.8); 53.6 Africa, UK Asian 6.4, Black or African- 152; pbo: pbo: 56.1 (10.9) American 6.4, Multiple 1.3, 153 White 85.9; 15mg QD:

144

American Indian or Alaska Native 0, Asian 9.2, Black or African-American 6.6, Multiple 1.3, White 82.9; pbo: American Indian or Alaska Native 0.7, Asian 9.8, Black or African-American 5.9, Multiple 1.3, White 82.4

Characteristics of Included Studies – SGLT2i Monotherapy continued

Study ID Baseline Mean Mean HYPO def Ascertainment of Hypo Rescue Medication Excl of Pts w A1C % Change in Difference in events? At (SD) HbA1C % vs. HbA1c % vs screening or during baseline pbo study)

Bailey CJ dapa 1mg dapa 1mg QD: dapa 1mg QD: - Major hypoglycaemia was defined as a not reported Rescue confounding. no 2012 QD: 7.8 -0.68 (-0.91 to 0.69 (-1.02, - symptomatic episode that required third-party Measurements obtained after rescue (0.98); -0.45); 2.5mg 0.37); 2.5mg assistance and was associated with plasma were not included in the efficacy 2.5mg QD: -0.72 (- QD: -0.74 (- glucose <3.00 mmol/l (<54 mg/dl) and prompt analyses but were included in the QD: 8.1 0.95 to -0.49); 1.07, -0.41); recovery after glucose or glucagon safety analysis. (1.07); 5mg QD: - 5mg QD: -0.84 administration. Minor hypoglycaemia was 5mg QD: 0.82 (-1.06 to (-1.17, -0.50) either a symptomatic or asymptomatic episode 7.9 (1.03); -0.58); pbo: associated with plasma glucose <3.50 mmol/l pbo: 7.8 0.02 (-0.22 to (<63 mg/dl), that does not qualify as a major (1.12) 0.25) episode. Additional accounts of hypoglycemic symptoms that did not meet these criteria were categorized as ‘other’.

145

Ferrannini E dapa dapa 2.5mg manually Severe: major episode, defined as a Patient-reported. Rescue confounding. Rescue no 2010 2.5mg QD: -0.58 calculated dapa symptomatic episode requiring third-party Patients were instructed (metformin) permitted. Data QD: 7.92 (0.11); 5mg 2.5mg QD: - assistance due to severe impairment in to self-monitor their obtained after rescue were excluded (0.90); QD: -0.77 0.35; 5mg QD: - consciousness or behavior, with a capillary or blood glucose daily and from efficacy analyses but not 5mg QD: (0.11) 0.54; 10mg QD: plasma glucose value 54 mg/dl (3.0 mmol/L), to report any unusually safety analysis. 7.86 (p=0.0005); -0.66 and prompt recovery after glucose or glucagon high or low blood (0.94); 10mg QD: - administration. glucose event or any 10mg QD: 0.89 (0.11) Assessed in symptoms suggestive of 8.01 (p<0.0001); patients without hypoglycemia. (0.96); pbo: -0.23 missing pbo: 7.84 (0.10) baseline and (0.87) week 24 values with last observation carried forward; Mean value after adjustment for baseline

value.

Ferrannini E empa 5mg empa 5mg manually Symptomatic [not specified] or laboratory- not reported not reported no 2013 QD: 7.9 QD: -0.4 (- calculated empa documented [values not given] (0.8); 0.61 to -0.25); 5mg QD: -0.5; 10mg QD: 10mg QD: - 10mg QD: -0.6; 8.0 (0.8); 0.5 (-0.66 to - 25mg QD: -0.7 25mg QD: 0.30); 25mg 7.8 (0.8); QD: -0.6 (- pbo: 7.8 0.81 to -0.45); (0.8) pbo: 0.1 (- 0.09 to 0.27)

146

Fonseca VA ipra [LSM] ipra [LSM] ipra Serious, as defined by development of Patient-reported (blood not reported no 2013 12.5mg 12.5mg QD: - 12.5mg QD: - hypoglycemic coma requiring glucose unconfirmed). QD: 7.95 0.22; 50mg 0.49 (-0.73 to - Patients were provided (0.78); QD: -0.39; 0.24); 50mg hospitalization, or requiring discontinuation of with a glucometer 50mg QD: 150mg QD: - QD: -0.65 (- study drug. (Roche Accu-Chek®; 8.05 0.47; 300mg 0.90 to -0.40); (0.81); QD: -0.55; 150mg QD: - Hoffmann-La Roche 150mg pbo: 0.26 0.73 (-0.98 to - Ltd, Basel, Switzerland) QD: 7.83 (LOCF) 0.48); 300mg and asked to monitor (0.65); QD: -0.81 (- 300mg 1.06 to -0.56) capillary blood glucose QD: 7.90 twice daily (fasted and 2 (0.67); h after a meal) and pbo: 7.84 also, if symptoms of (0.78) hypoglycemia occurred.

Inagaki N cana [LSM] cana [LSM] not reported Hypoglycemic not reported no 2013 50mg QD: 50mg QD: - manually symptoms, were 8.13 0.61 (p<0.01); calculated cana recorded throughout the (0.78); 100mg QD: - 50mg QD: - study. 100mg 0.80 (p<0.01); 0.72; 100mg QD: 8.05 200mg QD: - QD: -0.91; (0.86); 0.79 (p<0.01); 200mg QD: - 200mg 300mg QD: - 0.90; 300mg QD: 8.11 0.88 (p<0.01); QD: -0.99 (0.88); pbo: 0.11 300mg data represents QD: 8.17 last observation (0.81); carried forward pbo: 7.99 (0.77)

147

Inagaki N cana [LSM] cana [LSM] cana Asymptomatic hypoglycemia -typical Patients were instructed not reported no 2014 100mg 100mg QD: - 100mg QD: - hypoglycemic symptoms were absent but the to measure blood QD: 7.98 0.74 (0.07); 1.03 (0.10); blood glucose level was low (≤ 70 mg/dl or glucose levels (0.73); 200mg QD: - 200mg QD: - 3.89 mmol/L). 200mg 0.76 (0.07); 1.05 (0.10) on ‡ 3 days/week using a QD: 8.04 pbo: 0.29 Symptomatic hypoglycemia -typical glucose meter as early in (0.77); (0.07) hypoglycemic symptoms were present the morning pbo: 8.04 irrespective of the blood glucose level. (0.70) (Severe not defined) as possible in the fasting state throughout the study. Data

were to be recorded in a patient diary. The patients were also

instructed to measure glucose levels, if possible, in the event of

symptoms suggestive of hypoglycemia. The investigator

classified hypoglycemia as either asymptomatic hypoglycemia

(typical hypoglycemic symptoms were absent but the blood

148

glucose level was £ 70 mg/dl) or symptomatic hypoglycemia

(typical hypoglycemic symptoms were present irrespective of

the blood glucose level).

Ji L 2014 dapa 5mg: dapa 5mg: - dapa 5mg: - Major hypoglycemia was defined as not reported Rescue confounding. Rescue no 8.14 1.04 (-1.18 to 0.75; 10mg: - symptomatic episodes requiring external (metformin) permitted. For the (0.74); -0.90); 10mg: 0.82 (SE or p assistance due to severely impaired efficacy analysis, observations after 10mg: -1.11 (-1.24 to value (n/a)) consciousness or behavior, with capillary or the 8.28 -0.98); pbo: - plasma glucose values <54 mg/dL (0.95); 0.29 (-0.43 to (<3.0mmol/L) and prompt recovery after initiation of rescue therapy was pbo: 8.35 -0.16) glucose or glucagon administration. Minor excluded. Safety data after rescue (0.95) hypoglycemia was defined as any therapy included.

episode (symptomatic or asymptomatic) with a capillary or plasma glucose measure <63 mg/dL (<3.5mmol/L) that did not qualify as a major episode. Other episodes of hypoglycemia were defined as episodes reported by the investigator that were suggestive of hypoglycemia but did not meet the aforementioned criteria.

Kadowaki T empa 5mg empa 5mg empa 5mg QD: Confirmed events, plasma glucose <70 mg/dL patient-reported Rescue therapy permitted and data no 2014 QD: 7.92 QD: -0.42 (SE -0.72 (SE ± (<3.9 mmol/L) and/or requiring assistance. obtained after treated as missing. (0.70); ± 0.089); 0.08) (-0.87 to - Severe not explicitly defined but results 10mg QD: 10mg QD: - 0.57); 10mg suggest no assistance required. 7.93 0.40 (SE ± QD: -0.70 (SE ±

149

(0.71); 0.09); 25mg 0.08) (-0.85 to - 25mg QD: QD: -0.65 (SE 0.55); 25mg 7.93 ± 0.09); 50mg QD: -0.95 (SE ± (0.78); QD: -0.61 (SE 0.08) (-1.10 to - 50mg QD: ± 0.09); pbo: 0.80); 50mg 8.02 0.30 (SE ± QD: -0.91 (SE ± (0.65); 0.09) 0.08) (-1.06 to - pbo: 7.94 0.76) (0.74)

Kaku K dapa 1mg dapa 1mg QD: dapa 1mg QD: - not reported not reported not reported no 2013 QD: 8.10 -0.12 (0.07); 0.49 (0.10) (- (0.79); 2.5mg QD: - 0.68 to -0.29); 2.5mg 0.11 (0.07); 2.5mg QD: - QD: 7.92 5mg QD: - 0.48 (0.10) (- (0.74); 0.37 (0.07); 0.67 to -0.28); 5mg QD: 10mg QD: - 5mg QD: -0.74 8.05 0.44 (0.07); (0.10) (-0.93 to (0.66); pbo: 0.37 -0.54); 10mg 10mg QD: (0.07) QD: -0.80 8.18 (0.10) (-1.00 to (0.69); -0.61) pbo: 8.12 (0.71)

Kaku K dapa 5mg: [LSM] dapa [LSM] dapa not reported not reported Rescue permitted. Efficacy data no 2014 A 7.50 5mg: -0.41 (- 5mg: -0.35 (- after rescue not included but did (0.72); 0.53 to -0.29); 0.52 to -0.18); include body weight. 10mg: 10mg: -0.45 (- 10mg: -0.39 (- 7.46 0.57 to -0.33); 0.56 to -0.23) (0.61); pbo: -0.06 (- pbo: 7.50 0.18 to 0.06) (0.63)

150

Kaku K tofo 10mg [LSM] tofo [LSM] tofo Hypoglycemia was defined as either (1) Self-monitored blood not reported Yes. severe or 2014 B QD: 8.45 10mg QD: - 10mg QD: - signs/symptoms consistent with glucose frequent (0.75); 0.797 (-0.960 0.769; 20mg 20mg QD: to -0.634); QD: -0.990; hypoglycemia that resolved after ingesting episodes of 8.34 20mg QD: - 40mg QD: - food or administration of glucagon or glucose, hypoglycemia (0.81); 1.017 (-1.178 0.842 (CI n/a) or (2) blood glucose ≤ 50 mg/dL (2.78 within 4 weeks 40mg QD: to -0.856); mmol/L) with or without signs/symptoms of before the 8.37 40mg QD: - hypoglycemia. (Severe or major not defined, (0.77); 0.870 (-1.031 however events that occurred were described screening for pbo: 8.41 to -0.709) as mild or moderate). provisional (0.78) registration. Patients with a history of infection or

hypoglycemia was excluded because of the difficulty in

evaluating the safety of tofogliflozin.

Kashiwagi ipra ipra 12.5mg ipra 12.5mg not reported not reported not reported no A 2014 12.5mg QD: -0.11 QD: -0.61 (- QD: 8.39 (0.09); 25mg 0.85 to -0.36); (0.90); QD: -0.47 25mg QD: -0.97 25mg QD: (0.09); 50mg (-1.21 to -0.72); 8.32 QD: -0.79 50mg QD: -1.29 (0.83); (0.09); 100mg (-1.54 to -1.04); 50mg QD: QD: - 100mg QD: - 8.33 0.81(0.09); 1.31 (-1.55 to - (0.80); pbo: 0.50 1.06) ipra 100mg (0.09) adjusted mean,

151

QD: 8.25 difference vs (0.76); placebo (95% pbo: 8.36 CI) (0.79)

Roden 2013 empa empa 10mg: - empa 10mg: - Confirmed hypoglycemic adverse events Confirmed Rescue medication permitted. no 10mg: 0.66 (-0.76 to 0.74 (-0.88 to - (hypoglycemic episodes reported as adverse hypoglycemia either 7.87 -0.56); 25mg: 0.59); 25mg: - events when plasma glucose was <3·9 plasma glucose (0.88); -0.78 (-0.88 to 0.85 (-0.99 to - mmol/L, required assistance, or both). 25mg: -0.67); pbo: 0.71) <3·9 mmol/L or 7.86 0.08 (-0.03 to requiring assistance, or (0.85); 0.18) both. pbo: 7.91 (0.78)

Seino Y luseo 1mg [LSM] luseo [LSM] luseo not reported not reported not reported no 2014 A QD: 7.77 1mg QD: - 1mg QD: -0.51 (0.79); 0.29 (-0.41 to (-0.68 to -0.34); 2.5mg -0.17); 2.5mg 2.5mg QD: - QD: 8.05 QD: -0.39 (- 0.61 (-0.78 to - (0.75); 0.51 to -0.27); 0.44); 5mg QD: 5mg QD: 5mg QD: - -0.68 (-0.85 to - 7.86 0.46 (-0.58 to 0.51); 10mg (0.69); -0.34); 10mg QD: -0.64 (- 10mg QD: QD: -0.43 (- 0.81 to -0.48) 7.95 0.54 to -0.31); luseo difference (0.67); pbo: 0.22 vs. placebo; not pbo: 7.92 (0.10 to 0.34) adjusted (0.84)

Seino Y luseo [LSM] luseo [LSM] luseo not reported patient-reported not reported no 2014 B 0.5mg 0.5mg QD: - 0.5mg QD: - QD: 8.16 0.36 (-0.5 to - 0.42 (-0.6 to - (0.93); 0.2); 2.5mg 0.2); 2.5mg QD:

152

2.5mg QD: -0.62 (- -0.68 (-0.9 to - QD: 8.07 0.8, -0.5); 0.5); 5mg QD: - (0.90); 5mg QD: - 0.82 (-1.0 to - 5mg QD: 0.75 (-0.9, - 0.6) 8.16 0.6); pbo: 0.06 (0.96); (-0.1 to 0.2) pbo: 7.88 (0.72)

Stenlöf K cana [LSM] cana [LSM] cana Documented hypoglycaemia episodes included patient-documented Rescue therapy with metformin no 2013 100mg 100mg QD: - 100mg QD: - biochemically confirmed episodes (concurrent fingerstick glucose test permitted. QD: 8.1 0.77; 300mg 0.91 (-1.1 to - fingerstick or plasma glucose ≤3.9 mmol/l, (1.0); QD: -1.03; 0.7); 300mg irrespective of symptoms) and severe Efficacy data post rescue not 300mg pbo: 0.14 QD: -1.16 (-1.3 hypoglycaemia included. QD: 8.0 to -1.0) (1.0); pbo: episodes (i.e., requiring the assistance of 8.0 (1.0) another individual or resulting in seizure or loss of consciousness).

Terra SG ertu 5mg [LSM] ertu ertu 5mg QD: - The endpoint of symptomatic hypoglycaemia investigator-confirmed Rescue confounding only if severe. no 2017 QD: 8.16 5mg QD: - 0.99 (-1.22 to - consisted of episodes with clinical symptoms Safety analyses included all (0.88); 0.79 (-0.95 to 0.76); 15mg of hypoglycaemia reported by the investigator randomized participants who 15mg QD: -0.63); 15mg QD: -1.16 (- (i.e., biochemical documentation [FPG or received at least one dose of study 8.35 QD: -0.96 (- 1.39 to -0.93) finger-stick glucose] was not required]. medication and excluded data after (1.12); 1.12 to -0.80) Documented hypoglycaemia, defined as initiation of rescue medication with pbo: 8.11 episodes with a glucose level ≤3.9 mmol/L, the exception of summaries of (0.92) with or without symptoms, was also analyzed. serious AEs, deaths, and Severe hypoglycaemia was defined as an discontinuations because of AEs, episode that required assistance, either medical postural blood pressure and pulse or non-medical. rate which included data after rescue.

153

Characteristics of Included Studies – DPP4i Added to Metformin Background

Study ID Dose Study n= Mean Age (SD) Gender (% male) Countries Studied Ethnicity (%) Duration of Diabetes in Duration Years (SD)

Amin NB sita 12 weeks sita 100mg: sita 100mg: 53.3 sita 100mg: 72.7; Canada, India, South not reported sita 100mg: 6.3 (0.3 to 20.0); 2015 100mg 55; pbo: 54 (10.7); pbo 54 pbo: 55.6 Korea, Mexico, USA pbo: 6.4 (0.3 to 20.5) QD (8.1)

Bosi E 2007 vilda 24 weeks vilda 50mg vilda 50mg QD: vilda 50mg QD: USA, France, Italy, vilda 50mg QD: Caucasian 74.1, vilda 50mg QD: 6.8 (5.5); 50mg QD: 177; 54.3 (9.7); 100mg 57.3; 100mg QD: Sweden Hispanic or Latino 16.8, Black 6.3, 100mg QD: 5.8 (4.7); pbo: QD; vilda 100mg QD: QD: 53.9 (9.5); 61.5; pbo: 53.1 All other 2.8; 100mg QD: Caucasian 6.2 (5.3) 100mg 185; pbo 182 pbo 54.5 (10.3) 74.1, Hispanic or Latino 13.3, Black QD 9.1, All other 3.5; pbo: Caucasian 73.1, Hispanic or Latino 18.5, Black 6.9, All other 1.5

Charbonnel B sita 24 weeks sita 100mg: sita 100mg: 54.4 sita 100mg: 55.8; not reported sita 100mg: Asian 10.6, Black 6.7, sita 100mg: 6.0 (5.0); pbo: 2006 100mg 464; pbo: (10.4); pbo: 54.7 pbo: 59.5 Hispanic 15.5, White 63.1, Other 4.1; 6.6 (5.5) QD 237 (9.7) pbo: Asian 11.0, Black 5.9, Hispanic 11.8, White 67.1, Other 4.2

DeFronzo RA saxa 24 weeks saxa 2.5mg saxa 2.5mg QD: saxa 2.5mg QD: not reported saxa 2.5mg QD: Caucasian 79.7, saxa 2.5mg QD: 6.7 (5.6); 2009 2.5mg QD: 192; 54.7 (10.1); 5mg 43.2; 5mg QD: African American 4.2, Asian 4.2, 5mg QD: 6.4 (4.7); 10mg QD; saxa 5mg QD: QD: 54.7 (9.6); 53.9; 10mg QD: Other 12.0; 5mg QD: Caucasian 83.2, QD: 6.3 (4.4); pbo: 6.7 (5.6) 5mg QD; 191; 10mg 10mg QD: 54.2 52.5; pbo: 53.6 African American 5.8, Asian 1.6, saxa QD: 181; (10.1); pbo: 54.8 Other 9.4; 10mg QD: Caucasian 79.6, 10mg QD pbo: 179 (10.2) African American 7.7, Asian 2.8, Other 9.9; pbo: Caucasian 83.8, African American 3.9, Asian 2.2, Other 10.1

154

Forst T 2010 lina 1mg 12 weeks lina 1mg lina 1mg QD: lina 1mg QD: UK, Germany, France, lina 1mg QD: White 98, Black 0, lina 1mg QD: 6.9 (5.9); 5mg QD; lina QD: 65; 59.2 (8.4); 5mg 55.4; 5mg QD: Slovakia, Ukraine and Asian 2; 5mg QD: White 100, Black QD: 7.3 (7.5); 10mg QD: 8.2 5mg QD; 5mg QD: QD: 59.6 56.1; 10mg QD: Sweden 0, Asian 0; 10mg QD: White 98, (6.8); pbo: 6.2 (5.1) lina 10mg 66; 10mg (9.8);10mg QD: 53.0; pbo: 62.0 Black 2, Asian 0; White 98, Black 1, QD QD: 66; 61.8 (8.8); pbo: Asian 1 (manually calculated to add to pbo: 71 60.1 (8.1) 100)

Goodman M vilda 24 weeks vilda 100mg vilda 100mg QD vilda 100mg QD USA, Europe (not a vilda 100mg QD AM: Caucasian 63.2, not reported 2009 100mg QD AM: AM: 54.7 (10.3); AM: 52.8; 100mg country) Black 8.8, Asian 0.8, Hispanic or QD AM; 125; 100mg 100mg QD PM: QD PM: 52.8; Latino 24.8, Pacific Islander 0.8, 100mg QD PM: 55.2 (11.4); pbo: pbo: 52.8 Other 1.6; 100mg QD PM: Caucasian QD PM 123; pbo: 54.5 (9.7) 68.3, Black 7.3, Asian 1.6, Hispanic 122 or Latino 22.8, Pacific Islander 0, Other 0; pbo: Caucasian 68.0, Black 5.7, Asian 0, Hispanic or Latino 26.2, Pacific Islander 0, Other 0

Kadowaki T sita 50mg 52 weeks sita 50mg: not reported not reported Japan Japanese overall 7.5 2013 QD 76; pbo: 71

Nauck MA alo 26 weeks alo 12.5mg alo 12.5mg QD: alo 12.5mg QD: 15 countries; not [RACE]: alo 12.5mg QD: White 80, alo 12.5mg QD: 6 (5); 25mg 2009 12.5mg QD: 213; 55 (11); 25mg 47.4; 25mg QD: reported African American 2, Asian 8, Other QD: 6 (4); pbo: 6 (5) QD; alo 25mg QD: QD: 54 (11); pbo 54.3; pbo: 48 race 10; 25mg QD: White 76, African 25mg QD 210; pbo: 56 (11) American 6, Asian 9, Other race 9; 104 pbo: White 76, African American 7, Asian 6, Other race 11; [ETHNICITY]: alo 12.5: Hispanic or Latino 31, Not Hispanic or Latino 69; alo 25: Hispanic or Latino 32, Not Hispanic or Latino 68; pbo: Hispanic or Latino 24, Not Hispanic or Latino 76

155

Odawara M vilda 12 weeks vilda 50mg vilda 50mg BID: vilda 50mg BID: Japan Japanese vilda 50mg BID: 7.2 (6.18); 2014 50mg BID: 69; 58.7 (9.81); pbo: 63.8; pbo: 68.6 pbo: 7.0 (5.92) BID pbo: 70 57.5 (9.15)

Pan C 2012 vilda 24 weeks vilda 50mg vilda 50mg BID: vilda 50mg BID: China Chinese vilda 50mg BID: 4.92 (4.8); 50mg BID: 146; 54.2 (9.62); 50mg 50; 50mg QD: 50mg QD: 5.02 (4.42); pbo: BID; 50mg QD: QD: 53.7 (10.0); 44.6; pbo: 45.8 5.15 (4.58) vilda 148; pbo: pbo: 54.5 (9.68) 50mg QD 144

Pan CY 2017 alo 25mg 16 weeks alo 25mg alo 25mg QD: alo 25mg QD: China, Taiwan, Hong [country or region] alo 25mg: China alo 25mg QD: 1.9 (2.4); pbo: QD QD: 92; pbo: 51.6 (10.4); pbo: 59.8; pbo: 58.1 Kong 97.8, Hong Kong 2.2, Taiwan 0; pbo: 2.1 (2.8) 92 53.1 (8.9) China 97.8, Hong Kong 1.1, Taiwan 1.1

Raz I 2008 sita 18 weeks sita 100mg: sita 100mg: 53.6 sita 100mg: 51; multinational; not [RACE] sita 100mg: White 42, sita 100mg: 8.4 (6.5); pbo: 100mg 96; pbo: 94 (9.5); pbo 56.1 pbo: 41.5 reported Hispanic 32, Black 3, Multiracial 22, 7.3 (5.3) QD (9.5) Other 1; pbo: White 47, Hispanic 25, Black 1, Multiracial 25, Other 2

Rosenstock J goso 12 weeks goso 20mg overall: 56.5 not reported not reported not reported Mean (Range) goso 20mg 2011 20mg QD: 116; QD: 7.1 (0.1 to 25.2); 30mg QD; goso 30mg QD: QD: 8.3 (0.1 to 44.2); pbo: 30mg QD 116; pbo: 57 9.2 (0.2 to 30.2)

Rosenstock J sita 12 weeks sita 100mg: sita 100mg: 51.7 sita 100mg: 58; not reported not reported sita 100mg: 5.6 (4.7); pbo: 2012 100mg 65; pbo: 65 (8.1); pbo: 53.3 pbo: 48 6.4 (5.0) QD (7.8)

Ross SA 2012 lina 12 weeks lina 2.5mg lina 2.5mg BID: lina 2.5mg BID: India, Republic of lina 2.5mg BID: White 67.3, Asian lina 2.5mg BID: ≤1 year 7.9, 2.5mg BID: 223; 58.7 (9.9); 5mg 61.9; 5mg QD: Korea, Malaysia, 32.3, Other 0.4; 5mg QD: White 62.1, > 1-5 years 39.7, > 5 years BID; lina 5mg QD: QD: 58.4 (10.6); 54.0; pbo: 47.7 Belgium, France, Italy, Asian 36.6, Other 1.3; pbo: White 52.3; 5mg QD: ≤1 year 9.0, > 5mg QD 224; pbo 44 pbo: 59.9 (10.7) 72.7, Asian 27.3, Other 0 1-5 years 38.9, > 5 years

156

the Netherlands, Spain, 52.0; pbo: ≤1 year 2.3, > 1-5 Canada years 41.9, > 5 years 55.8

Scott R 2008 sita 18 weeks sita 100mg: sita 100mg: 55.2 sita 100mg: 55; multinational; not sita 100mg: Caucasian 61, Asian 38, sita 100mg: 4.9 (3.5); pbo: 100mg 94; pbo: 92 (9.8); pbo: 55.3 pbo: 59 reported Others 1; pbo: Caucasian 61, Asian 5.4 (3.7) QD (9.3) 39, Others 0

Shankar RR ertu 24 weeks omari 25mg omari 25mg QW: omari 25mg QW: not reported omari 25mg QW: White 86.1, Asian omari 25mg QW: 8.2 (5.2); 2017 25mg QW: 201; 57.5 (8.1); pbo: 50.2; pbo: 50.7 2.5, Black 9.5, Multi-racial 1.0, pbo: 7.4 (5.6) QW pbo: 201 56.8 (9.1) Native American 1.0; pbo: White 86.1, Asian 2.5, Black 8.0, Multi- racial 0, Native American 3.5

Taskinen MR lina 5mg 24 weeks lina 5mg lina 5mg QD: 56.5 lina 5mg QD: 53; Czech Republic, Finland, lina 5mg QD: White 75, Asian 22, lina 5mg QD: ≤ 1 year: 11, > 2011 QD QD: 523; (10.1); pbo: 56.6 pbo: 57 Greece, India, Israel, Other 3; pbo: White 79, Asian 18, 1-5 years: 34, > 5 years: 56; pbo: 177 (10.9) Mexico, New Zealand, Other 3 pbo: ≤ 1 year: 13, > 1-5 Russia, Sweden, USA years: 34, > 5 years: 53

Terra SG 2011 goso 2mg 12 weeks goso 2mg goso 2mg QD: goso 2mg QD: Colombia, Germany, goso 2mg QD: White 64.9, Black 8.1, goso 2mg QD: 9.4; 5mg QD: QD; goso QD: 37; 5mg 55.9 (9.0); 5mg 67.6; 5mg QD: Italy, Spain, Sweden, Asian 0, Other 27.0, Unspecified 0; 6.8; 10mg QD: 6.7; 20mg 5mg QD; QD: 38; QD: 56.4 (6.5); 73.7; 10mg QD: USA 5mg QD: White 57.9, Black 13.2, QD: 6.5; pbo: 7.2 (SD n/a) goso 10mg QD: 10mg QD: 55.7 59.7; 20mg QD: Asian 0, Other 28.9, Unspecified 0; 10mg 77; 20mg (8.4); 20mg QD: 65.8; pbo: 69.7 10mg QD: White 59.7, Black 6.5, QD; goso QD: 73; pbo: 55.9 (8.1); pbo: Asian 3.9, Other 29.9, Unspecified 0; 20mg QD 76 57.2 (7.8) 20mg QD: White 65.8, Black 11.0, Asian 0, Other 23.3, Unspecified 0; pbo: White 59.2, Black 11.8, Asian 0, Other 27.6, Unspecified 1.3

Wang W 2016 lina 5mg 24 weeks lina 5mg: lina 5mg: 55.1 lina 5mg: 49.8; China, Philippines, [Country] lina 5mg: China 89.8, lina 5mg: ≤1 year 16.3, > 1-5 QD 205; pbo: (10.7); pbo: 56.5 pbo: 50.0 Malaysia Malaysia 6.3, Philippines 3.9; pbo: 40.9, > 5 42.9; pbo: ≤1 year 100 (8.7) China 80.0, Malaysia 11.0, 19.6, > 1-5 33.0, > 5 47.4 Philippines 9.0

157

White JL 2014 saxa 12 weeks saxa 2.5mg saxa 2.5mg BID: saxa 2.5mg BID: USA, Germany, [RACE] saxa 2.5mg BID: White 86.5, saxa 2.5mg BID: 5.8 (6.37); 2.5mg BID: 74; 53.9 (10.35); pbo: 54.1; pbo: 52.3 Hungary, Puerto Rico Black/African American 10.8, pbo: 6.2 (4.21) BID pbo: 86 56.6 (9.97) American Indian or Alaska native 0, Asian 2.7. [ETHNICITY] hispanic or latino 39.2, non-hispanic or latino 45.9, not reported 14.9; pbo: White 93.0, Black/African American 3.5, American Indian or Alaska native 1.2, Asian 2.3. [ETHNICITY] hispanic or latino 40.7, non-hispanic or latino 41.9, not reported 17.4

Yang W 2011 saxa 5mg 24 weeks saxa 5mg: saxa 5mg: 53.8 saxa 5mg: 48.1; China, India, South [REGION] saxa 5mg: China 58.3, saxa 5mg: 5.1 (5.0); pbo: 5.1 QD 283; pbo: (10.4); pbo: 54.4 pbo: 48.4 Korea India 25.8, South Korea 15.9; pbo: (4.0) 287 (10.1) China 565.1, India 25.8, South Korea 18.1

Yang W 2012 sita 24 weeks sita 100mg: sita 100mg: 54.1 sita 100mg: 47; China Chinese sita 100mg: 6.4 (4.4); pbo: 100mg 197; pbo: (9.0); pbo: 55.1 pbo: 55 7.3 (4.6) QD 198 (9.8)

158

Characteristics of Included Studies – DPP4i Added to Metformin Background continued

Study ID Baseline Mean Mean HYPO def Ascertainment of Hypo Rescue Medication Excl of Pts w events? At A1C % Change in Difference screening or during study) (SD) HbA1C % in HbA1c vs. baseline % vs pbo

Amin NB sita [LSM pbo [LSM] Hypoglycaemia was captured as an AE based Home blood glucose monitoring Rescue therapy not no 2015 100mg: corrected, manually on signs/symptoms, home blood glucose using an ACCU-CHEK permitted. Patients 8.24 80% CI] sita calculated monitoring [defined as blood glucose requiring rescue (0.15); 100mg; -0.87 sita 100mg; ≤3.9mmol/l (70mg/dl) using an ACCU-CHEK home glucometer, or ≤4.1mmol/l were withdrawn pbo: 8.08 (-1.01 to - -0.76 home glucometer, or ≤4.1mmol/l (74mg/dl). (74mg/dl) using International from study. (0.14) 0.73); pbo: - Severe not defined. 0.11 (-0.25 Federation of Clinical Chemistry- to 0.04) referenced ACCU-CHEK glucometers; The recommended home glucose monitoring frequency was determined for each subject by the investigator; however, a frequency of at least once daily (and in event of hypoglycaemia)

was emphasized.

Bosi E 2007 vilda vilda 50mg vilda Hypoglycemia was defined as symptoms Patients were provided with glucose not reported no 50mg QD: -0.5 adjusted suggestive of low blood glucose confirmed by monitoring devices and supplies QD: 8.4 (0.1); 100mg mean self-monitored blood glucose measurement and instructed on their use. (0.9); QD: -0.9 change; <3.1 mmol/l plasma glucose equivalent. Severe 100mg (0.1); pbo: between- hypoglycemia was defined as any episode QD: 8.4 0.2 (0.1) treatment requiring the assistance of another party. (1.0); pbo difference 8.3 (0.9) (vildagliptin - placebo)

159

vilda 50mg QD: -0.7 (0.1) (p<0.001); 100mg QD: -1.1 (0.1) (p<0.001)

Charbonnel sita [LSM pbo [LSM pbo not reported Investigators evaluated each clinical Rescue therapy No. A B 2006 100mg: subtracted] subtracted] adverse experience for intensity permitted. Data 8.0 (0.8); (95% CI): (manually exclude after rescue. history of hypoglycemia was pbo: 8.0 sita -0.67 (- calculated): (mild, moderate, or severe), not an exclusion (0.8) 0.77 to - sita -0.65 duration, outcome, and relationship 0.57); pbo: - to study drug. criterion. 0.02 (-0.15 to 0.10)

DeFronzo saxa [LSM] [LSM] Reported provided in Table 2 but not defined. Patient reported hypoglycemia Rescue therapy no RA 2009 2.5mg adjusted (SE) adjusted Confirmed hypoglycemia defined by symptoms permitted. Data QD: 8.1 saxa 2.5mg (SE) saxa of hypoglycemia in the setting of a fingerstick adverse events, AND confirmed exclude after rescue. (1.0); QD: -0.59 2.5mg QD: - blood glucose value ≤50 mg/dl (2.8 mmol/l). 5mg QD: (0.07); 5mg 0.73 (0.10) hypoglycemia (fingerstick glucose 8.1 (0.8); QD: -0.69 (-0.92 to - value of 50 mg/dl associated with 10mg (0.07); 10mg 0.53); 5mg symptoms), QD: 8.0 QD: -0.58 QD: -0.83

(1.0); (0.07); pbo: (0.10) (-1.02 were recorded. pbo: 8.1 0.13 (0.07) to -0.63);

(0.9) 10mg QD: - 0.72 (0.10)

(-0.91 to -

0.52)

160

Forst T 2010 lina 1mg mean change manually not reported During the study, patients made not reported Yes, if severe. QD: 8.2 from calculated visits to the study center at weeks 2, (0.7); baseline lina 1mg 4, 8 and 12, with a follow-up visit at 5mg QD: (SD) (FAS QD: -0.38; week 14. Patients underwent safety 8.5 (0.8); LOCF) lina 5mg QD: - and efficacy assessments and were 10mg 1mg QD: - 0.74; 10mg withdrawn if they had a fasting QD: 8.4 0.14 (0.92); QD: -0.66 plasma glucose > 13.3 mmol ⁄ l (0.7); 5mg QD: - (measured on two separate days) at pbo: 8.4 0.50 (0.81); any visit; if they showed clinical (0.7) 10mg QD: - signs of severe hypoglycaemia or a 0.42 (0.87); blood glucose level < 2.5 mmol⁄ l; if pbo: 0.24 their dose of metformin changed. (0.74)

Goodman M vilda (SE) vilda vilda 100mg Hypoglycaemia defined as a plasma glucose Patients were educated about not reported no 2009 100mg 100mg QD QD AM: - <3.1 mmol/l. Severe hypoglycaemia is any hypoglycemic symptoms and QD AM: AM: -0.66 0.83 event where the patient required assistance treatment of hypoglycemic events, 8.5 (1); (0.11); (p<0.001); from a third party or hospitalization. and were instructed to take a blood 100mg 100mg QD 100mg QD glucose measurement if a QD PM: PM: -0.53 PM: -0.70 hypoglycemic event was suspected 8.5 (0.9); (0.11); pbo: (p<0.001) pbo: 8.7 0.17 (0.11) (no sd) (1.1)

Kadowaki T overall [LSM] sita [LSM] sita not reported no not reported no 2013 8.3 50 mg: -0.4 50mg: -0.70 (-0.6 to -0.2); (-0.9 to -0.5)

161

pbo: 0.3 (0.1 to 0.5)

Nauck MA alo [LSM] [SE] [LSM] Hypoglycaemia [blood glucose <60 mg ⁄dl Patient recorded; Visits included Patients requiring no 2009 12.5mg alo 12.5mg manually (<3.3 mmol⁄l) in presence of symptoms; blood assessment of vital signs, physical rescue therapy were QD: 7.9 QD: -0.6 calculated: glucose <50 mg⁄dl (<2.8 mmol⁄l) regardless of examination, concomitant terminated from (0.7); (0.1); 25mg alo 12.5mg: symptoms] and severe hypoglycaemia [defined medication review and adverse study. 25mg QD: -0.6 -0.5; alo as any episode requiring the assistance of event (AE) monitoring, review of QD: 7.9 (0.1); pbo: - 25mg: -0.5 another person to administer actively, diaries and glucometer readings, (0.8); 0.1 (0.1) carbohydrate, glucagon or other resuscitative laboratory assessments pbo: 8.0 actions, associated with blood glucose <60 (hematology, serum chemistry (0.9) mg⁄dl (< 3.3 mmol⁄l)]. and urinalysis) and documentation of drug dosing compliance, as determined by via pill count.

Odawara M vilda adjusted difference in Hypoglycemia was defined as symptoms Patient-reported; confirmed by Rescue medication no 2014 50mg mean change adjusted suggestive of hypoglycemia that was further blood glucose measurement (additional OADs or BID: 8 (SE) vilda mean confirmed by a self-monitored blood glucose (0.83); 50mg BID: change vilda measurement of <3.1 mmol/L. insulin) not pbo: 8.0 =-1.1 (0.06); 50 mg BID: permitted. (0.96) pbo: -0.1 -1.0 (0.09) The event was considered grade 1 if the patient (0.06) (p<0.001) was able to initiate self-treatment, and grade 2 if the patient required assistance of another person or hospitalization.

Pan C 2012 vilda adjusted vilda 50mg Hypoglycaemia was defined as the presence of All not reported no 50mg mean change BID: -0.51 symptoms suggestive of hypoglycaemia, BID: 8.09 (SE) vilda (0.11) confirmed by self-monitored glucose <56 patients were provided with one (0.85); 50mg BID: - (p<0.001); mg/dl (3.1 mmol/l). glucose meter each for self- 50mg 1.05 (0.08); not reported monitored blood glucose and were QD: 8.05 50mg QD: - for vilda 50 instructed about the use of

162

(0.84); 0.92 (0.08); qd because pbo: 8.01 pbo: -0.54 not Severe hypoglycaemia was defined as an the device at the entry of the study; (0.82) (0.08) significant episode requiring assistance of another party. Hypoglycaemia was defined between- treatment as the presence of symptoms difference suggestive of hypoglycaemia, (SE). confirmed by self-monitored glucose <56 mg/dl (3.1 mmol/l).

Severe hypoglycaemia was defined as an episode requiring

assistance of another party. Patients were required to record the event and associated information such as glucose value and time of occurrence in the study diary.

Pan CY 2017 alo 25mg [LSM] alo [LSM] alo Mild-to-moderate hypoglycemia, not reported Rescue permitted. no QD: 8.04 25mg QD: =- 25mg QD: - (0.92); 0.99; pbo: - 0.58 (-0.78 whether symptomatic or asymptomatic, was pbo: 7.86 0.42 to -0.37) (0.78) (p<0.001) defined as plasma glucose levels <3.9mmol/L. Severe hypoglycemia was defined as any hypoglycemic episode that required the assistance of another person to actively

administer carbohydrate, glucagon, or other resuscitative actions, and was associated with a documented plasma glucose level <3.9 mmol/L.

163

Raz I 2008 sita [LSM] sita [LSM] sita not reported not reported Rescue permitted no 100mg: 100mg: -1.0 100mg: -1.0 but not included. 9.3 (0.9); (-1.2 to -0.8); (-1.4 to -0.7) pbo: 9.1 pbo: 0.0 (- (0.8) 0.2 to 0.3)

Rosenstock J goso [LSM] goso [LSM] goso not reported Laboratory testing for safety not reported no 2011 20mg 20mg QD: - 20mg QD: - parameters was carried out at QD: 8.24 0.92 (-1.09 0.79 (-1.10 (1.25); to -0.75); to -0.49); baseline and weeks 2, 4, 8 and 12, 30mg 30mg QD: - 30mg QD: - and adverse events were QD: 8.17 1.04 (-1.22 0.92 (-1.23 (1.39); to -0.87); to -0.61) recorded at all visits. pbo: 8.05 pbo: -0.13 (- (1.02) 0.38 to 0.13)

Rosenstock J sita sita 100mg: - sita 100mg: not reported no, patient-reported not reported no 2012 100mg: 0.74; pbo: - -0.52 7.64 0.22 (0.95); pbo: 7.75 (0.83)

Ross SA lina lina 2.5mg lina 2.5mg Hypoglycaemia episodes were classified by not reported Rescue permitted no 2012 2.5mg BID: -0.46 BID: -0.74 investigators but not included. BID: 7.96 (0.05); 5mg (-0.97 to - (0.78); QD: -0.52 0.52); 5mg according to guidelines from the American 5mg QD: (0.05); pbo: QD: -0.80 (- Diabetes 7.98 0.28 (0.11) 1.02 to - (0.72); (SE) 0.58) Association. pbo: 7.92 (0.74)

164

Scott R 2008 sita [LSM] sita [LSM] sita not reported not reported not reported no 100mg: 100mg: -0.73 100mg: - 7.8 (1.0); (-0.87 to - 0.51 (-0.70 pbo: 7.7 0.60); pbo: - to -0.32) (p (0.9) 0.22 (-0.36 ≤ 0.001) to -0.08)

Shankar RR omari [LSM] omari [LSM] Symptomatic hypoglycemia: episode with A standard questionnaire was Rescue permitted no 2017 25mg 25mg QW: - omari 25mg clinical symptoms attributed to hypoglycemia, provided to subjects to but data not QW: 8.1 0.54 (-0.69 QW: -0.55 without regard to glucose level. included. (0.9); to -0.40); (-0.75 to - collect hypoglycemia information. pbo: 8.0 pbo: 0.0 (- 0.34) Severe hypoglycemia: episode that required (0.9) 0.14 to 0.15) assistance, either medical or non-medical. Episodes with a markedly depressed level of consciousness, a loss of consciousness, or seizure were classified as having required medical assistance, whether or not medical assistance was obtained. Asymptomatic hypoglycemia: glucose values ≤3.9 mmol/L without symptoms.

Taskinen MR lina 5mg lina 5mg lina 5mg Plasma glucose concentration ≤3.9 mmol/l HbA1c, fasting plasma glucose Rescue confounding no 2011 QD: 8.09 QD: =-0.49 QD: -0.64 (FPG), adverse events (AEs), (0.86); (0.04); pbo: (0.07) (-0.78 pbo: 8.02 0.15 (0.06) to -0.50) serious adverse events (SAEs) and (0.88) (p<0.0001) vital signs were evaluated at adjusted mean every visit; Safety assessments were change from made at screening, baseline placebo run-in and weeks 0, 12, 24 and 25.

165

Terra SG goso 2mg [LSM] goso [LSM] goso not reported not reported not reported no 2011 QD: 8.00 2mg QD: - 2mg QD: - (0.88); 0.26 (-0.58 0.31 (-0.70 5mg QD: to 0.06); 5mg to 0.08); 8.52 QD: -0.69 (- 5mg QD: - (1.30); 1.00 to - 0.74 (-1.12 10mg 0.39); 10mg to -0.36); QD: 8.37 QD: -0.65 (- 10mg QD: - (1.08); 0.87 to - 0.70 (-1.02 20mg 0.42); 20mg to -0.38); QD: 8.33 QD: -0.70 (- 20mg QD: - (0.97); 0.93 to - 0.75 (-1.07 pbo: 8.38 0.48); pbo: to -0.43) (1.19) 0.05 (-0.17 (PPS) to 0.27) (PPS)

Wang W lina 5mg: lina 5mg: - lina 5mg: - Investigator-defined hypoglycemia was not reported Rescue medication no 2016 7.99 0.66 (0.05); 0.52 (0.09) categorized as asymptomatic with plasma (glimepiride) (0.83); pbo: -0.14 (-0.70 to - glucose ≤70 mg/dL (3.89 mmol/L), permitted. pbo: 8.00 (0.07) 0.34) (0.80) Adjusted Adjusted documented symptomatic with glucose 54–70 mean (SE) mean (SE) mg/dL (3.0-3.89 mmol/L), documented symptomatic with glucose <54 mg/dL (3.0 mmol/L) without need for external assistance, and severe (requiring another person’s assistance for active administration of resuscitative actions).

White JL saxa [LSM] (SE) saxa 2.5mg Confirmed hypoglycemia, defined by No; confirmed hypoglycemia, not reported no 2014 2.5mg saxa 2.5mg BID: -0.32 hypoglycemic symptoms plus a fingerstick defined by hypoglycemic symptoms BID: 7.92 BID: -0.56 (SE 0.12) glucose value ≤50 mg/dL (2.8 mmol/L). (0.961); p=0.006

166

pbo: 7.97 (0.09); pbo: - (0.819) 0.22 (0.08) plus, a fingerstick glucose value ≤50 mg/dL.

Yang W saxa saxa 5mg: - saxa 5mg: Reported presented but not defined. No, signs and symptoms are Rescue therapy was no 2011 5mg: 7.9 0.78; pbo: - adjusted Confirmed hypoglycemic events were those patient-reported not permitted during (0.8); 0.37 mean associated with symptoms of hypoglycemia the study. pbo: 7.9 change from and a documented plasma glucose level ≤2.8 (0.8) baseline - mmol/L, and reported hypoglycemic events 0.42 (-0.55 were defined as signs and symptoms consistent to -0.29) with hypoglycemia with or without documented glucose measurement

Yang W sita [LSM] sita [LSM] sita All hypoglycemic events reported in the Patients were counseled to self- Rescue therapy with no 2012 100mg: 100mg: -1.0 100mg: -0.9 present study were symptomatic (i.e. episode monitor their blood glipizide permitted. 8.5 (0.9); (1.2 to -0.9); (-1.1 to -0.7) with clinical symptoms attributed to Data obtained after pbo: 8.5 pbo: -0.1 (- hypoglycemia, without regard to glucose level glucose levels and immediately the initiation of (0.9) 0.3 to 0) - symptoms of hypoglycemia, such as notify investigators if rescue sweating, anxiety, and palpitations); marked severity (e.g. markedly depressed level of they experienced symptoms of therapy was treated consciousness loss of consciousness, or hypoglycemia, such as as missing in all seizures). Patients were instructed to also analyses. sweating, anxiety, and palpitations, report episodes of asymptomatic low blood for assessment of hypoglycemic glucose levels (i.e. ≤3.9 mmol⁄L). events during the study. Hypoglycemia was assessed by the study site investigators by reviewing patient self-reports of signs and symptoms of hypoglycemia. A fingerstick blood glucose determination was not required to support the documentation of a

167

symptomatic episode of hypoglycemia, but was

required to report an episode of asymptomatic hypoglycemia.

Patients were instructed to also report episodes of asymptomatic low blood glucose levels (i.e.

≤3.9 mmol⁄ L).

168

Characteristics of Included Studies – GLP1RA Added to Metformin Background

Study ID Dose Study n= Mean Age Gender (% Countries Studied Ethnicity (%) Duration of Diabetes in Duratio (SD) male) Years (SD) n

Ahrén B lixi 20μg QD 24 lixi 20μg lixi 20μg lixi 20μg AM: Australia, Canada (LMC), Chile, Czech lixi 20μg AM: Caucasian 86.7, lixi 20μg AM: 6.2 (5.3); 2013 SC AM; lixi weeks AM: 255; AM: 54.5; 38.4; 20μg PM: Republic, Germany, Croatia, Mexico, Black 2.7, Asian 8.6, Other 2.0; 20μg PM: 6.2 (5.4); pbo: 20μg QD SC 20μg 20μg PM: 44.7; pbo: 47.6 Morocoo, the Philippines, Romania, 20μg PM: Caucasian 89.4, Black 5.9 (4.7) PM PM: 255; 54.8; pbo: Russian Federation, South Africa, Spain, 2.4, Asian 7.8, Other 0.4; pbo: pbo: 170 55.0 (9.4) Ukraine, USA, Venezuela Caucasian 91.2, Black 2.4, Asian 6.5, Other 0

Bolli GB lixi one-step 24 lixi one- lixi one- lixi one-step: 15 countries; not specified lixi one-step: Caucasian 88, Black lixi one-step: 5.8 (3.9); 2014 dose weeks step: 161; step: 55.4 44; two-step: 1, Asian 8, Other 4; two-step: two-step: 6.0 (4.6); pbo: (main two-step: (8.9); two- 45; pbo:45 Caucasian 91, Black 1, Asian 7, 6.2 (4.7) increase (10 efficacy) 161; pbo: step: 54.6 Other 1; pbo: Caucasian 93, Black μg once daily 160 (8.9); pbo: 1, Asian 6, Other 1 for 2 weeks 76 58.2 (9.8) then 20 μg weeks QD; lixi two- (safety) step dose increase (10 μg once

daily for 1 week, 15 μg once daily for 1 week then 20 μg QD

Davies M sema 2.5mg 26 sema sema 2.5mg sema 2.5mg PO Austria, Bulgaria, Canada, Denmark, sema 2.5mg PO QD: White 81.4, sema 2.5mg PO QD: 6.1 2017 PO QD; sema weeks 2.5mg PO QD: QD: 64.3 Germany, Israel, Italy, Malaysia, Serbia, Black or African American 8.6, (6.0) 5mg PO QD: 5.3 5mg PO QD; 56.7 (9.9) Asian 10.0, American Indian or (4.7) 10mg PO QD: 5.8

169

sema 10mg PO QD: 5mg PO South Africa, Spain, Sweden, United Alaska Native 0, Other 0 5mg PO (4.8) 20mg PO QD: 7.0 PO QD; sema 70 QD: 55.7 5mg PO QD: Kingdom, USA QD: White 90.0, Black or African (5.3) 40mg PO QD: 7.7 20mg PO QD; (11.0) 67.1 American 2.9, Asian 5.7, American (5.9) 40mg PO QD slow 5mg PO sema 40mg 10mg PO Indian or Alaska Native 0, Other dose (8w) escalation: 6.6 QD: 70 10mg PO QD: PO QD; sema QD: 56.5 1.4 10mg PO QD: White 82.6, (4.9) 40mg PO QD fast 62.3 1mg SC QW (10.1) Black or African American 10.1, dose (2w) escalation: 5.6 10mg PO 20mg PO Asian 5.8, American Indian or (4.7) pbo: 6.7 (5.1) QD: 69 20mg PO QD: QD: 58.3 Alaska Native 0, Other 1.4 20mg 62.9 (10.4) PO QD: White 84.3, Black or 20mg PO African American 5.7, Asian 5.7, QD: 70 40mg PO QD: 40mg PO American Indian or Alaska Native 60.6 QD: 56.5 2.9, Other 1.4 40mg PO QD: White 40mg PO (10.2) 88.7, Black or African American QD: 71 40mg PO QD 40mg PO 5.6, Asian 4.2, American Indian or slow dose (8w) QD slow Alaska Native 0, Other 1.4 40mg 40mg PO escalation: 58.6 dose (8w) QD slow PO QD slow dose (8w) escalation: escalation: dose 40mg PO QD White 77.1, Black or African 57.1 (10.5) (8w) fast dose (2w) American 10.0, Asian 10.0, 40mg PO escalatio escalation: 62.9 American Indian or Alaska Native QD fast n: 70 0, Other 2.9 40mg PO QD fast dose dose (2w) pbo: 56.3 (2w) escalation: White 84.3, Black escalation: 40mg PO or African American 10.0, Asian 57.7 (10.8) QD fast 5.7, American Indian or Alaska pbo: 58.9 dose Native 0, Other 0 pbo: White 80.3, (10.3) (2w) Black or African American 8.5, escalatio Asian 9.9, American Indian or n: 70 Alaska Native 1.4, Other 0

pbo: 71

DeFronzo exen 5μg SC 30 exen 5μg exen 5μg exen 5μg BID: USA exen 5μg BID: Caucasian 77.3, exen 5μg BID: 6.2 (5.9); RA 2005 BID; exen weeks BID: BID: 53 51.8; 10μg BID: Black 10.9, Hispanic 7.3, Other 4.6; 10μg BID: 4.9 (4.7); pbo: 10μg SC BID 110; (11); 10μg 60.2; pbo: 59.3 10μg BID: Caucasian 79.6, Black 6.6 (6.1)

170

10μg BID: 52 8.8, Hispanic 8.0, Other 3.5; pbo: BID: (11); pbo: Caucasian 72.6, Black 13.3, 113; pbo: 54 (9) Hispanic 10.6, Other 3.5 113

Kim D exen LAR 13 exen exen 0.8mg exen 0.8mg not reported; exen 0.8mg LAR: Caucasian 56, exen 0.8mg LAR: 5 (3); 2007 0.8mg SC weeks 0.8mg LAR: 55 LAR: 75; exen Black 13, Hispanic 19, Other 13; exen 2.0mg LAR: 4 (5); QW; exen LAR: 16; (12); exen 2.0mgLAR: 67; exen 2.0mg LAR: Caucasian 60, pbo: 4 (4) 2.0mg SC QW exen 2.0mg pbo 36 Black 20, Hispanic 20, Other 0; LAR LAR: 51 pbo: Caucasian 64, Black 14, 2.0mg: (11); pbo: Hispanic 21, Other 0 15; pbo: 55 (9) 14

Nauck MA sema OW 12 sema sema QW sema QW Austria, Bulgaria, Finland, France, White 75% (breakdown n/a) sema QW 0.1mg: 3.6 (5.0); 2016 weeks QW 0.1mg 55.2 0.1mg: 66; SC Germany, Hungary, India, Italy, Serbia, SC QW 0.2mg: 2.3 (2.7); without dose 0.1mg: (10.1); SC QW 0.2mg: 70; South Africa, Spain, Switzerland, Turkey, SC QW 0.4mg: 2.0 (2.3); escalation 47; SC QW 0.2mg: SC QW 0.4mg: United Kingdom SC QW 0.8mg: 3.0 (3.0); (0.1–0.8 mg); QW 54.7 (10.0); 77; SC QW SC 0.8mg E: 2.6 (2.1); SC or with dose 0.2mg: SC QW 0.8mg: 52; SC QW 1.6mg E: 1.8 (2.0); escalation (E) 44; SC 0.4mg 53.8 0.8mg E 63; SC pbo: 2.4 (3.3) (0.4 mg steps QW (10.2); SC QW 1.6mg E: to 0.4mg: QW 0.8mg: 55; pbo: 61 49; SC 55.0 (9.7); 0.8 or 1.6mg E QW SC 0.8mg over 1–2 0.8mg: E: 55.9 weeks) 44; SC (7.9); SC 0.8mg E: QW 1.6mg 45; SC E: 56.4 QW (10.5); pbo: 1.6mg E: 55.3 (10.6) 45; pbo:46

171

Ratner RE [within 1h of 13 lixi 5μg lixi 5μg SC lixi 5μg SC QD: multinational; not specified lixi 5μg SC QD: Caucasian 69.1, lixi 5μg SC QD: 7.2 (4.9); 2010 breakfast] lixi weeks SC QD: QD: 56.8 47.3; 10μg SC Black 9.1, Other 21.8; 10μg SC 10μg SC QD: 6.2 (4.1); 5μg SC QD; 55; 10μg (7.8); 10μg QD: 59.6; 20μg QD: Caucasian 69.2, Black 11.5, 20μg SC QD: 6.4 (6.8); lixi 10μg SC SC QD: SC QD: SC QD: 50.9; Other 19.3; 20μg SC QD: 30μg SC QD: 6.0 (4.8) QD; lixi 20μg 52; 20μg 55.4 (9.2); 30μg SC QD: Caucasian 81.8, Black 1.8, Other 5μg SC BID: 6.2 (6.0); SC QD; lixi SC QD: 20μg SC 50 5μg SC BID: 16.4; 30μg SC QD: Caucasian 79.6, 10μg SC BID: 6.4 (5.0); 30μg SC QD 55; 30μg QD: 55.4 47.2; 10μg SC Black 11.1, Other 9.3; 5μg SC BID: 20μg SC BID: 6.6 (5.1); SC QD: (9.9); 30μg BID: 51.8; 20μg Caucasian 86.8, Black 3.8, Other 30μg SC BID: 7.0 (5.4); 54 SC QD: SC BID: 37.0; 9.4; 10μg SC BID: Caucasian 80.4, pbo: 7.1 (5.4) 56.5 (8.7) 30μg SC BID: Black 3.6, Other 16.0; 20μg SC lixi 5μg SC 42.6; pbo: 56.0 BID: Caucasian 77.8, Black 3.7, BID; lixi 10μg Other 18.5; 30μg SC BID: SC BID; lixi 5μg SC Caucasian 64.8, Black 16.7, Other 20μg SC BID; BID: 53; 5μg SC 18.5; pbo: Caucasian 77.1, Black lixi 30μg SC 10μg SC BID: 57.1 11.0, Other 11.9 BID BID: 56; (8.2); 10μg 20μg SC SC BID: BID: 54; 56.0 (7.9); 30μg SC 20μg SC BID: 54; BID: 56.7 pbo 109 (8.3); 30μg SC BID: 55.3 (9.1); pbo: 56.3 (9.2)

Rosenstock albi 4mg SC 16 albi 4mg albi 4mg albi 4mg SC USA, Mexico, Chile, Dominican Republic not reported albi 4mg SC QW: 4.4 J 2009 B QW; albi weeks SC QW: SC QW: QW: 42.9; (4.1); 15mg SC QW: 4.7 15mg SC QW; 34; 15mg 50.4 (10.3); 15mg SC QW: (4.6); 30mg SC QW: 5.2 albi 30mg SC SC QW: 15mg SC 51.4; 30mg SC (5.4) 15mg SC biweekly: QW 34; 30mg QW: 55.5 QW: 25.8 15mg 4.3 (4.3); 30mg SC SC QW: (10.5); SC biweekly: biweekly: 5.5 (4.5); 50mg 29 30mg SC 42.4; 30mg SC SC biweekly: 5.2 (5.5)

172

QW: 54.2 biweekly: 50.0; 50mg SC monthly: 5.3 albi 15mg SC 15mg SC (9.7) 50mg SC (3.7); 100mg SC monthly: biweekly; albi biweekly: biweekly: 54.3 4.3 (3.7); pbo: 3.9 (3.0) 30mg SC 30; 30mg 15mg SC 50mg SC biweekly; albi SC biweekly: monthly: 48.6; 50mg SC biweekly: 52.5 (9.6); 100mg SC biweekly 32; 50mg 30mg SC monthly: 55.9; SC biweekly: pbo: 54.9 biweekly: 55.5 (9.9); 34 50mg SC albi 50mg SC biweekly: monthly; albi 50mg SC 51.1 (10.3) 100mg SC monthly: monthly 35; 50mg SC 100mg monthly: SC 54.1 (11.3); monthly: 100mg SC 33; pbo: monthly: 50 54.4 (9.9); pbo: 54.0 (10.6)

173

Characteristics of Included Studies – GLP1RA Added to Metformin Background continued

Study ID Baseline Mean Change Mean HYPO def Ascertainment of Rescue Excl of Pts w events? At A1C % in HbA1C % Difference in Hypo Medication screening or during (SD) vs. baseline HbA1c % vs study) pbo

Ahrén B lixi 20μg [LSM] lixi 20μg [LSM] lixi 20μg Symptomatic hypoglycemia was defined as symptoms of not reported Rescue One patient (0.4%) 2013 AM: 8.0 AM: -0.9% (SE AM: -0.5 (SE ± hypoglycemia with an accompanying blood glucose <3.3 confounding. (0.9); 20μg ±0.07); pbo: - 0.09) (95% CI - mmol/L (60 mg/dL) and/or prompt recovery with oral Rescue in the lixisenatide morning PM: 8.1 0.4 (SE ±0.08); 0.66 to -0.31); carbohydrate, intravenous glucose, or glucagon injection. (sulfonylureas as injection arm and three (0.9); pbo: 20μg PM: -0.8 20μg PM: -0.4 Severe symptomatic hypoglycemia was defined as the first option) patients (1.2%) in the 8.1 (0.9) (SE ±0.07); (SE ± 0.09) symptomatic hypoglycemia that required the assistance of permitted. lixisenatide pbo: -0.4 (SE (95% CI -0.54 another person, and that was associated either with a plasma ±0.08) to -0.19) glucose level <2.0 mmol/L (36 mg/dL) or, if no plasma evening injection arm glucose measurement was obtainable, with prompt recovery discontinued with carbohydrate, intravenous glucose, or glucagon injection. study therapy owing to a

treatment-emergent AE of hypoglycemia

during the 24-week treatment period.

Bolli GB lixi one- [LSM] lixi one- [LSM] lixi one- Symptomatic hypoglycaemia defined as event with clinical Patient-reported Rescue no 2014 step: 8.0 step: -0.9 (SE step: -0.5 (-0.7 symptoms with either plasma glucose < 60 mg/dl (3.3 [medications not (0.9); two- ±0.10); two- to -0.3); two- mmol/L) or prompt recovery after oral carbohydrate specified] step: 8.1 step: -0.8 (SE step: -0.4 (-0.6 administration (if no plasma glucose measurement was permitted. (0.9); pbo: ±0.1); pbo: -0.4 to -0.2) available), intravenous glucose or glucagon administration. 8.0 (0.8) (SE ±0.10) Severe hypoglycaemia—event with clinical symptoms considered to result from hypoglycaemia in which the participant required the assistance of another person, because

174

the participant could not treat him/herself as a result of acute neurological impairment directly resulting from the hypoglycemic event, and one of the following: event was associated with plasma glucose < 2.0 mmol/l (36 mg/dl); if no plasma glucose measurement is available, then the event was associated with prompt recovery after oral carbohydrate, intravenous glucose or glucagon administration.

Davies M sema 2.5mg pooled sema sema 2.5mg PO Severe or BG-confirmed’ hypoglycemic episodes are defined Patient-reported; Rescue no 2017 PO QD: 8.0 40mg: -1.8; pbo QD: -0.4 (-0.7, - as severe by ADA classification or BG-confirmed with a Plasma glucose medication (0.7) 5mg 0.3 (CI 0.1) 5mg PO plasma glucose value of <56 mg/dL (<3.1 mmol/L) should always be permitted. PO QD: 7.8 provided for QD: -0.9 (-1.2, - with/without measured and (0.6) 10mg ETD) 0.6) 10mg PO recorded when a PO QD: 7.8 QD: -1.2 (-1.5, - symptoms consistent with hypoglycemia; ADA classified, ≤70 hypoglycemic (0.7) 20mg 0.9) 20mg PO mg/dL (≤3.9 mmol/L); c ADA-classified ‘severe’ episode is suspected. PO QD: 7.9 QD: -1.4 (-1.7, - hypoglycemia is an episode requiring the assistance of another The record should (0.7) 40mg 1.0) 40mg PO person to actively administer carbohydrate, glucagon, or take include the PO QD: 8.0 QD: -1.6 (-1.9, - other corrective actions. following (0.7) 40mg 1.3) 40mg PO information: PO QD QD slow dose The typical signs and symptoms of confirmed hypoglycaemia slow dose (8w) escalation: include hunger, slight headache, nausea, light-headedness, Date and time of (8w) -1.4 (-1.7, -1.1) palpitations and sweating. Major hypoglycaemia (severe) may hypoglycemic escalation: 40mg PO QD produce loss of consciousness. episode, The plasma 8.0 (0.7) fast dose (2w) glucose level before Symptoms of confirmed and not major hypoglycaemia should 40mg PO escalation: -1.3 treating the episode be treated by ingestion of carbohydrates. QD fast (-1.6, -1.0) (if available), dose (2w) Whether the episode

escalation: was symptomatic, 7.8 (0.8) Whether the subject pbo: 8.0 was able to treat (0.8) him/herself, Date and time of last trial product administration prior 175

to episode, Type of last trial product prior to episode, Date and time of last main meal prior to episode, Whether the episode occurred in relation to increased physical activity The answer to the question: “Was subject able to treat him/herself?” must be answered “No” if oral carbohydrates, glucagon or IV glucose had to be administered to the subject by another person.

DeFronz exen 5μg [LSM] (SE) [LSM] Treatment-emergent adverse events were defined as those Patient-reported not reported no o RA BID: 8.3 exen 5μg BID: - (manually occurring upon or after receiving the first randomized dose. 2005 (1.1); 10μg 0.4 (0.1); 10μg calculated, no The intensity of hypoglycemic episodes was defined as BID: 8.2 BID: -0.8 (0.1); SE or CI) exen mild/moderate or severe. For mild/moderate hypoglycemia, (1.0); pbo: pbo: 0.1 (0.1) 5μg BID: -0.5; subjects reported symptoms consistent with hypoglycemia that 8.2 (1.0) 10μg BID: -0.9 may have been documented by a plasma glucose concentration value <3.3 mmol/l. For severe hypoglycemia, subjects required the assistance of another person to obtain treatment for their hypoglycemia, including intravenous glucose or intramuscular glucagon.

176

Kim D exen 0.8mg exen 0.8mg exen 0.8mg not reported Patient-reported not reported No, but one patient 2007 LAR: 8.6 LAR: -1.4 (SE LAR: -1.8; exen withdrew due to loss of (1.1); ± 0.3); 2.0mg 2.0mg LAR: - glycemic control. 2.0mg exen exen LAR: -1.7 2.1 LAR: 8.3 (SE ± 0.3); pbo: For self-monitored (1.1); pbo: 0.4 (SE ± 0.3) manually blood glucose 8.6 (1.4) calculated profiles, subjects were given blood glucose

meters and instructed to perform

measurements by fingerstick at the fingertip.

Pre-prandial glucose was measured 15

min before each meal, postprandial glucose

was measured 1.5–2 h after each

meal, and an additional glucose measurement

was taken at 0300 h. Measurements 177

were recorded on 3 separate days for both

baseline and week 15.

Nauck sema QW [LSM] sema [LSM] sema Minor hypoglycemia defined as a plasma glucose level <3.1 patient-reported not reported no MA 2016 0.1mg: 8.2 QW 0.1mg: - QW 0.1mg: -0.1 mmol/L and therefore counted as any. Major hypoglycemia (0.9); SC 0.6; SC QW (-0.5 to 0.3); SC defined as requiring assistance and captured as severe. QW 0.2mg: 0.2mg: -0.9 QW 0.2mg: -0.4 8.2 (0.9); (p<0.05); SC (-0.8 to -0.0); SC QW QW 0.4mg: -1.1 SC QW 0.4mg: 0.4mg: 8.1 (p<0.001) ; SC -0.6 (-1.0 to - (0.9); SC QW 0.8mg:-1.5 0.2); SC QW QW 0.8mg: (p<0.0001); SC 0.8mg: -1.0 (- 8.2 (0.9); 0.8mg E:-1.4 1.4 to -0.6); SC SC 0.8mg (p<0.0001); SC 0.8mg E: -1.0 (- E: 8.0 (0.8); QW 1.6mg E:- 1.3 to -0.6); SC SC QW 1.7 (p<0.0001); QW 1.6mg E: - 1.6mg E: pbo: -0.5 1.2 (-1.6 to -0.8) 8.0 (0.7); pbo: 8.1 (0.8)

Ratner lixi 5μg SC [LSM] lixi 5μg [LSM] Symptomatic hypoglycaemia was defined as symptoms Patient-reported not reported no RE 2010 QD: 7.58 SC QD: -0.47 manually consistent with hypoglycaemia, with an accompanying blood (0.7); 10μg (p=0.0056); calculated lixi glucose < 3.3 mmol⁄ l or prompt recovery with carbohydrate. SC QD: 10μg SC QD: - 5μg SC QD: - Severe reported but not defined. 7.52 (0.6); 0.50 0.29; 10μg SC 20μg SC (p=0.0033); QD: -0.32; 20μg QD: 7.58 20μg SC QD: - SC QD: -0.51; (0.7); 30μg 0.69 30μg SC QD: -

178

SC QD: (p<0.0001)); 0.58; 5μg SC 7.52 (0.7) 30μg SC QD: - BID: -0.47; 5μg SC 0.76 (p<0.0001) 10μg SC BID: - BID: 7.60 5μg SC BID: - 0.60; 20μg SC (0.6); 10μg 0.65 BID: -0.57; SC BID: (p<0.0001); 30μg SC BID: - 7.54 (0.6); 10μg SC BID: - 0.69 20μg SC 0.78 BID: 7.61 (p<0.0001); (0.7); 30μg 20μg SC BID: - SC BID: 0.75 p<0.0001; 7.46 (0.5); 30μg SC BID: - pbo: 7.53 0.87 (p<0.001); (0.6) pbo: -0.18 (n/a)

Rosensto albi 4mg albi 4mg SC not reported. not reported not reported not reported no ck J 2009 SC QW: QW: -0.11 The primary B 8.1 (1.0); (1.16); 15mg efficacy end 15mg SC SC QW: -0.49 point was QW: 8.0 (0.74); 30mg change from (0.9); 30mg SC QW: -0.87 baseline A1C at SC QW: (0.65) week 16 versus 8.0 (0.9) placebo across 15mg SC 15mg SC different doses biweekly: biweekly: -0.56 within each 8.2 (1.0); (0.97); 30mg schedule 30mg SC SC biweekly: - (weekly, biweekly: 0.79 (0.98); biweekly, and 8.0 (1.0); 50mg SC monthly). 50mg SC biweekly: -0.79 biweekly: (1.04) 8.0 (0.7) 50mg SC 50mg SC monthly: -0.55

179

monthly: (1.01); 100mg 7.9 (0.8); SC monthly: - 100mg SC 0.87 (0.87) monthly: 8.0 (1.0); pbo: 7.9 (0.9)

Characteristics of Included Studies – SGLT2i Added to Metformin Background

Study ID Dose Study n= Mean Age (SD) Gender Countries Studied Ethnicity (%) Duration of Diabetes in Years Duration (% male) (SD)

Amin NB ertu 1mg 12 weeks ertu 1mg QD: ertu 1mg QD: ertu 1mg Canada, India, South Korea, Mexico, not reported ertu 1mg QD: 6.3 range (0.1 to 24); 2015 QD; ertu 54; 5mg QD: 55; 53.1 (9.1); 5mg QD: 63; USA 5mg QD: 6.7 (0.3 to 30.0); 10mg 5mg QD; 10mg QD: 55; QD: 54.7 (7.7); 5mg QD: QD: 6.1 (0.2 to 20.0); 25mg QD: ertu 10mg 25mg QD: 55; 10mg QD: 57.3 74.5; 6.0 (0.3 to 18.2); pbo: 6.4 (0.3 to QD; ertu pbo: 54 (6.5); 25mg 10mg QD: 20.5) 25mg QD QD: 54.2 (8.8); 56.4; pbo: 54 (8.1) 25mg QD: 67.3; pbo:56.6

Bailey CJ dapa 24 weeks dapa 2.5mg QD: dapa 2.5mg dapa USA, Canada, Argentina, Mexico, not reported dapa 2.5mg QD: 6.0 (6.2); 5mg 2010 2.5mg 137; 5mg QD: QD: 55.0 (9.3); 2.5mg Brazil QD: 6.4 (5.8); 10mg QD: 6.1 (5.4); QD; dapa 137; 10mg QD: 5mg QD: 54.3 QD: 51; pbo: 5.8 (5.1) 5mg QD; 135; pbo: 137 (9.4); 10mg 5mg QD: dapa 10mg QD: 52.7 (9.9); 50; 10mg QD pbo: 53.7 (10.3) QD: 57; pbo: 55

180

Bolinder J dapa 10mg 24 weeks dapa 10mg QD: dapa 10mg: dapa Bulgaria, Czech Republic, Hungary, dapa 10mg white: 100; pbo: dapa 10mg: 6.0 (4.5); pbo: 5.5 2012 QD 89; pbo: 91 60.6 (8.2); pbo: 10mg: Poland and Sweden 100 (5.3) 60.8 (6.9) 55.1; pbo: 56

Häring HU empa 24 weeks empa 10mg QD: empa 10mg empa Canada, China, France, Germany, empa 10mg QD: White 52, empa 10mg QD: ≤ 1 year: 9, >1-5 2014 10mg QD; 217; 25mg QD: QD: 55.5 (9.9); 10mg QD: India, South Korea, Mexico, Slovakia, Asian 46, Black/African years: 36, >5-10 years 31, >10 empa 213; pbo: 207 25mg QD: 55.6 58; 25mg Slovenia, Taiwan, Turkey, USA American 2, American years 24; 25mg QD: ≤ 1 year 9, >1- 25mg QD (10.2); pbo: QD: 56; Indian/Alaska Native 1; 5 years 32, >5-10 years 35, >10 56.0 (9.7) pbo: 56 25mg QD: White 53, Asian years 24; pbo: ≤ 1 year 9, >1-5 46, Black/African American years 40, >5-10 years 31, >10 0, American Indian/Alaska years 19 Native 1; pbo: White 55, Asian 44, Black/African American 1, American Indian/Alaska Native 0

Ikeda S tofo 2.5mg 12 weeks tofo 2.5mg QD: tofo 2.5mg QD: tofo multinational; not reported tofo [WHITE %] 2.5mg QD: tofo 2.5mg QD: 4.88 (3.915); 5mg 2015 QD; tofo 66; 5mg QD: 65; 53.3 (10.86); 2.5mg 69.7; 5mg QD: 63.1; 10mg QD: 5.01 (3.599); 10mg QD: 5.77 5mg QD; 10mg QD: 66; 5mg QD: 54.8 QD: 51.5; QD: 60.6; 20mg QD: 70.3; (4.385); 20mg QD: 5.21 (3.934); tofo 10mg 20mg QD: 64; (10.53); 10mg 5mg QD: 40mg QD: 58.2; pbo: 63.6 40mg QD: 6.44 (5.811); pbo: 5.98 QD; tofo 40mg QD: 67; QD: 54.5 47.7; (5.287) 20mg QD; pbo: 66 (10.70); 20mg 10mg QD: tofo 40mg QD: 56.3 51.5; QD (9.79); 40mg 20mg QD: QD: 57.5 67.2; (9.31); pbo: 40mg QD: 53.9 (11.12) 46.3; pbo: 54.5

Kashiwagi ipra 50mg 24 weeks ipra 50mg: 112; ipra 50mg: 56.2 ipra Japan Japanese (112) ipra 50mg: 7.49; pbo: 8.05 A 2015 QD pbo: 56 (10.67); pbo: 50mg: 57.7 (9.24)

181

58.9; pbo: 58.9

Lu CH ipra 50mg 24 weeks ipra 50mg: 87; ipra 50mg: 53.9 ipra South Korea, Taiwan ipra 50mg: Korean (43), ipra 50mg: 6.49; pbo: 5.82 2016 QD pbo: 83 (11.3); pbo: 50mg: Taiwanese (44) 53.4 (11.3) 50.6; pbo: 39.8

Qiu R 2014 cana 50mg 18 weeks cana 50mg BID: cana 50mg BID: cana 7 countries; not reported cana 50mg BID: White 80.6, cana 50mg BID: 6.7 (4.9); 150mg BID; cana 93; 150mg BID: 58.6 (8.9); 50mg Black or African American BID: 7.3 (6.0); pbo: 7.0 (6.4) 150mg 93; pbo: 93 150mg BID: BID: 5.4, Asian 3.2, Other 10.8; BID 56.7 (10.3); 43.0; 150mg BID: White 89.2, pbo: 57.0 (9.3) 150mg Black or African American BID: 1.1, Asian 6.5, Other 3.2; 47.3; pbo: pbo: White 78.5, Black or 49.5 African American 4.3, Asian 9.7, Other 7.5

Rosenstock cana 50mg 12 weeks cana 50mg QD: cana 50mg QD: cana not reported not reported cana 50mg QD: 5.6 (5.0); 100mg J 2012 QD; cana 64; 100mg QD: 53.3 (8.5); 50mg QD: QD: 6.1 (4.7); 200mg QD: 6.4 100mg 64; 200mg QD: 100mg QD: 53; (5.7); 300mg QD: 5.9 (5.2); 300mg QD; cana 65; 300mg QD: 51.7 (8.0); 100mg BID: 5.8 (4.6); pbo: 6.4 (5.0) 200mg 64; 300mg BID: 200mg QD: QD: 56; QD; cana 64; pbo: 65 52.9 (9.6); 200mg 300mg 300mg QD: QD: 51; QD; cana 52.3 (6.9); 300mg 300mg 300mg BID: QD: 56; BID 55.2 (7.1); pbo: 300mg 53.3 (7.8) BID: 44; pbo: 48

Rosenstock empa 1mg 12 weeks empa 1mg QD: empa 1mg QD: empa 1mg 16 countries; not specified empa 1mg QD: Non- not reported J 2013 QD; empa 71; 5mg QD: 71; 57 (8.8); 5mg QD: 58; Hispanic White 86, Hispanic 5mg QD; 10mg QD: 71; QD: 60 (7.3); 5mg QD: White 13, Non-Hispanic

182

empa 25mg QD: 70; 10mg QD: 59 41; 10mg Black 0, Other 1; 5mg QD: 10mg QD; 50mg QD: 70; (9.0); 25mg QD: 47; Non-Hispanic White 83, empa pbo: 71 QD: 59 (8.1); 25mg QD: Hispanic White 16, Non- 25mg QD; 50mg QD: 56 53; 50mg Hispanic Black 0, Other 1; empa (9.4); pbo: 60 QD: 56; 10mg QD: Non-Hispanic 50mg QD (8.5) pbo: 47 White 78, Hispanic White 20, Non-Hispanic Black 3, Other 0; 25mg QD: Non- Hispanic White 83, Hispanic White 16, Non-Hispanic Black 1, Other 0; 50mg QD: Non-Hispanic White 86, Hispanic White 11, Non- Hispanic Black 3, Other 0; pbo: Non-Hispanic White 90, Hispanic White 9, Non- Hispanic Black 1, Other 0

Rosenstock sota 75mg 12 weeks sota 75mg QD: sota 75mg QD: sota 75mg USA sota 75mg QD: not reported J 2015 QD; sota 59; 200mg QD: 56.1 (9.6); QD: 57.6; Black/African American 8.5, 200mg 60; 200mg BID: 200mg QD: 200mg White 81.4, Other [American QD; sota 60; 400mg QD: 55.6 (9.3); QD: 28.3; Indian/Alaska 200mg 60; pbo: 60 200mg BID: 200mg Native/Asian/Other/Multiple] BID; sota 56.4 (8.8); BID: 10.2; 200mg QD: 400mg 400mg QD: 48.3; Black/African American QD 56.1 (9.5); pbo: 400mg 15.0, White 85.0, Other 55.1 (9.8) QD: 48.3; [American Indian/Alaska pbo: 43.3 Native/Asian/Other/Multiple] 0; 200mg BID: Black/African American 8.3, White 88.3, Other [American Indian/Alaska Native/Asian/Other/Multiple]

183

3.3; 400mg QD: Black/African American 10.0, White 85.0, Other [American Indian/Alaska Native/Asian/Other/Multiple] 5.0; pbo: Black/African American 10.0, White 81.7, Other [American Indian/Alaska Native/Asian/Other/Multiple] 8.3

Ross S empa 16 weeks empa 12.5mg empa 12.5mg empa Countries not specified; Europe, empa 12.5mg BID: White (time since dx.) empa 12.5mg BID: 2015 12.5mg BID: 215; 25mg BID: 57.6 (9.9); 12.5mg North American, Latin America 81.9, Black/African- ≤ 1 year 12.6; >1 to ≤5 years 35.8; BID; empa QD: 214; 5mg 25mg QD: 58.2 BID: American 7.9, Asian 7.0, >5 years 51.6; 25mg QD: ≤ 1 year 25mg QD; BID: 215; 10mg (10.2); 5mg 57.2; Other 3.3; 25mg QD: White 7.0; >1 to ≤5 years 39.3; >5 years empa 5mg QD: 214; pbo: BID: 58.8 (9.8); 25mg QD: 89.3, Black/African- 53.7; 5mg BID: ≤ 1 year 9.3; >1 to BID; empa 107 10mg QD: 58.5 53.3; 5mg American 4.7, Asian 4.2, ≤5 years 35.8; >5 years 54.9; 10mg 10mg QD (10.8); pbo: BID: Other 1.9; 5mg BID: White QD: ≤ 1 year 6.5; >1 to ≤5 years 57.9 (11.2) 55.8; 87.9, Black/African- 31.8; >5 years 61.7; pbo: ≤ 1 year 10mg QD: American 7.9, Asian 2.8, 8.4; >1 to ≤5 years 29.0; >5 years 50.5; pbo: Other 1.4; 10mg QD: White 62.6 51.4 84.1, Black/African- American 6.5, Asian 4.7, Other 4.7; pbo: White 86.9, Black/African-American 7.5, Asian 1.9, Other 3.7

Schumm- dapa 16 weeks dapa 2.5mg BID: dapa 2.5mg dapa Countries not specified; Europe, dapa 2.5mg BID: White 79.0, dapa 2.5mg BID: 4.80 (3.87); 5mg Draeger 2.5mg 100; 5mg BID: BID: 58.3 (9.0); 2.5mg South Africa Black 10.0, Asian 8.0, Other BID: 5.12 (4.2); 10mg QD: 5.45 PM 2015 BID; dapa 99; 10mg QD: 5mg BID: 55.3 BID: 3.0; 5mg BID: White 84.8, (4.05); pbo: 5.53 (4.23) 5mg BID; 99; pbo: 101 (9.3); 10mg 37.0; 5mg Black 5.1, Asian 7.1, Other dapa 10mg QD: 58.5 (9.8); BID: 3.0; 10mg QD: White 81.8, QD pbo: 58.5 (9.4) 46.5; Black 6.1, Asian 3.0, Other 184

10mg QD: 9.1; pbo: White 81.2, Black 49.5; pbo: 5.0, Asian 9.9, Other 4.0 46.5

Wilding ipra 12 weeks ipra 12.5mg QD: ipra 12.5mg ipra not reported ipra 12.5mg QD: White 91.3, ipra 12.5mg QD: 6.8 (6.4); 50mg JPH 2013 12.5mg 69; 50mg QD: QD: 56.6 (8.5); 12.5mg Non-white 8.7; 50mg QD: QD: 6.0 (5.3); 150mg QD: 5.7 QD; ipra 68; 150mg QD: 50mg QD: 58.6 QD: 47.8; White 97.1, non-white 2.9; (4.8); 300mg QD: 5.5 (4.8); pbo 5.7 50mg QD; 67; 300mg QD: (7.6); 150mg 50mg QD: 150mg QD: white 97.0, non- (3.2) ipra 72; pbo: 66 QD: 58.1 (8.2); 47.1; white 3.0; 300mg QD: white 150mg 300mg QD: 150mg 90.3, non-white 9.7; pbo: QD; ipra 56.6 (8.9); pbo: QD: 56.7; white 95.5, non-white 4.5 300mg 57.3 (8.6) 300mg QD QD: 50.0; pbo: 54.5

Yang W dapa 5mg 24 weeks dapa 5mg QD: dapa 5mg QD: dapa 5mg China, India, South Korea dapa 5mg QD: Asian Indian dapa 5mg QD: 4.2 (3.8); 10mg QD: 2016 QD; dapa 147; 10mg QD: 53.1 (9.1); QD: 45.6; 7.5, Chinese 86.4, Korean 5.3 (4.6); pbo 5.3 (4.4) 10mg QD 152; pbo: 145 10mg QD: 54.6 10mg QD: 6.1; 10mg QD: Asian Indian (9.5); pbo: 53.5 57.9; pbo: 8.6, Chinese 84.9, Korean (9.2) 59.3 6.6; pbo: Asian Indian 6.9, Chinese 86.9, Korean 6.2

185

Characteristics of Included Studies – SGLT2i Added to Metformin Background continued

Study ID Baseline Mean Change Mean Difference HYPO def Ascertainment of Hypo Rescue Medication Excl of Pts w events? At screening A1C % (SD) in HbA1C % in HbA1c % vs or during study) vs. baseline pbo

Amin NB ertu 1mg QD: [LSM] (80% [LSM] manually Hypoglycaemia was captured Home blood glucose monitoring Rescue not permitted. no 2015 8.01 (0.17); CI) ertu 1mg calculated ertu 1mg as an AE based on using an ACCU-CHEK 5mg QD: QD: -0.56 (- QD: -0.45; 5mg signs/symptoms, home blood 7.88 (0.13); 0.69 to -0.42); QD: -0.69; 10mg glucose monitoring [defined home glucometer, or ≤4.1mmol/l 10mg QD: 5mg QD: - QD: -0.62; 25mg as blood glucose ≤3.9mmol/l (74mg/dl) using International 8.13 (0.17); 0.80 (-0.84 to QD: -0.72 (70mg/dl) using an ACCU- 25mg QD: -0.66); 10mg CHEK home glucometer, or Federation of Clinical Chemistry- 8.30 (0.16); QD: -0.73 (- ≤4.1mmol/l (74mg/dl) using referenced pbo: 8.08 0.87 to -0.58); International Federation of ACCU-CHEK glucometers; The (0.14) 25mg QD: - Clinical Chemistry- recommended home glucose 0.83 (-0.98 to referenced ACCU-CHEK monitoring frequency -0.69); pbo: - glucometers]. Severe not 0.11 (-0.25 to defined. was determined for each subject 0.04) by the investigator; however,

a frequency of at least once daily (and in event of hypoglycaemia)

was emphasized.

Bailey CJ dapa 2.5mg dapa 2.5mg manually calculated Major event, defined as a not reported Rescue confounding. no 2010 QD: 7.99 QD: -0.67 (- dapa 2.5mg QD: - symptomatic episode Rescue (pioglitazone (0.90); 5mg 0.81 to -0.53); 0.37; 5mg QD: - requiring third party or acarbose) permitted. QD: 8.17 5mg QD: - 0.40; 10mg QD: - assistance because of severe For rescued patients, (0.96); 10mg 0.70 (-0.85 to 0.54 impairment in consciousness measurements QD: 7.92 -0.56); 10mg or behavior, with a capillary obtained after QD: -0.84 (- or plasma glucose initiation of rescue

186

(0.82); pbo: 0.98 to -0.70); concentration <3 mmol/L, medication were not 8.11 (0.96) pbo: -0.30 (- and prompt recovery after included in the 0.44 to -0.16) glucose or glucagon efficacy analysis but administration. were included in the safety analysis.

Bolinder J dapa 10mg: dapa 10mg: - dapa 10mg: -0.28 Major hypoglycemia was not reported Rescue confounding. Patients could be discontinued due 2012 7.19 (0.44); 0.39; pbo: - (-0.42 to -0.15) defined as a symptomatic Patients could receive to inadequate glycemic control at pbo: 7.16 0.10 (n/a) episode requiring external rescue therapy the discretion of the study (0.53) assistance due to severely exclusively with investigator. impaired consciousness or sitagliptin. Safety, not behavior, with capillary or efficacy analysis plasma glucose levels below includes patients after 3.0 mmol/liter and recovery rescue therapy. after glucose or glucagon administration. Minor hypoglycemia was defined as either symptomatic episode with capillary or plasma glucose levels below 3.5 mmol/liter, irrespective of the need for external assistance, or an asymptomatic episode with capillary or plasma glucose levels below 3.5 mmol/liter that does not qualify as a major episode. Other hypoglycemia was defined as an episode with symptoms suggestive of hypoglycemia but without confirmative measurement.

187

Häring HU empa 10mg empa 10mg empa 10mg QD: - Events consistent with Patient-reported. Rescue permitted. Yes. Where hyperglycemia or 2014 QD: 7.94 QD: -0.70 0.57 (-0.70 to - hypoglycemia and with Data after rescue set to hypoglycemia (0.79); 25mg (0.05); 25mg 0.43); 25mg QD: - plasma glucose levels of ≤3.9 missing and LOCF for QD: 7.86 QD: -0.77 0.64 (-0.77 to -0.50) mmol/L and/or requiring efficacy values. could not be controlled, the patient (0.87); pbo: (0.05); pbo: - differences in assistance. Severe not 7.9 (0.88) 0.13 (0.05) adjusted mean defined. was discontinued from the trial. values versus placebo

Ikeda S tofo 2.5mg [LSM] tofo [LSM] manually Plasma glucose ≤2.8mmol/l. not reported not reported no 2015 QD: 7.99 2.5mg QD: - calculated tofo Severe not separately (0.759); 5mg 0.440 (-0.588 2.5mg QD: -0.171; defined. QD: 8.01 to -0.293); 5mg QD: -0.348; (0.657); 10mg 5mg QD: - 10mg QD: -0.425; QD: 8.00 0.617 (-0.763 20mg QD: -0.499; (0.713); 20mg to -0.471); 40mg QD: -0.563 QD: 7.92 10mg QD: - (0.790); 40mg 0.694 (-0.840 QD: 7.92 to -0.549); (0.780); pbo: 20mg QD: - 7.88 (0.694) 0.768 (-0.916 to -0.620); 40mg QD: - 0.832 (-0.976 to -0.688); pbo: -0.269 (- 0.415 to - 0.123) LOCF

Kashiwagi ipra 50mg: ipra 50mg: - ipra 50mg: -1.30 (- not reported Patients were instructed how to not reported no A 2015 8.25 (0.719); 0.87 (0.655); 1.501 to -1.095) recognize and treat the symptoms pbo: 8.38 pbo: 0.38 Adj mean of hypoglycemia. (0.738) (0.703) difference

188

Lu CH ipra 50mg: ipra 50mg: - ipra 50mg: -0.46 (- not reported not reported not reported no 2016 7.74 (0.78); 0.94 (0.747); 0.66 to -0.27) pbo: 7.75 pbo: -0.47 (0.71) NGSP (0.806) (National Glycohemogl obin Standardizati on Program)

Qiu R 2014 cana 50mg [LSM] cana [LSM] cana 50mg Documented hypoglycemia biochemically documented only not reported no BID: 7.6 50mg BID: - BID: -0.44; 150mg episodes included (0.9); 150mg 0.45; 150mg BID: -0.60 biochemically documented BID: 7.6 BID: -0.61; (p<0.001 for both) episodes (concurrent (0.9); pbo: 7.7 pbo: -0.01 fingerstick or plasma glucose (0.9) ≤3.9 mmol/L with or without symptoms) and severe episodes (i.e., those requiring the assistance of another individual or resulting in seizure or loss of consciousness.

Rosenstock cana 50mg [LSM] cana [LSM] manually not reported patient-reported not reported no J 2012 QD: 8.00 50mg QD: - calculated cana (0.99); 100mg 0.79; 100mg 50mg QD: -0.57; QD: 7.83 QD: -0.76; 100mg QD: -0.54; (0.96); 200mg 200mg QD: - 200mg QD: -0.48; QD: 7.61 0.70; 300mg 300mg QD: -0.70; (0.80); 300mg QD: -0.92; 300mg BID: -0.73 QD: 7.69 300mg BID: - (1.02); 300mg 0.95; pbo: - BID: 7.73 0.22

189

(0.89); pbo: 7.75 (0.83)

Rosenstock empa 1mg empa 1mg manually calculated Defined by preferred Events were assessed at screening not reported no J 2013 QD: 7.8 (0.7); QD: -0.09 (- empa 1mg QD: - MedDRA terms.; severe and 5mg QD: 8.0 0.24 to 0.07); 0.24; 5mg QD: - plasma glucose levels (0.7); 10mg 5mg QD: - 0.38; 10mg QD: - visits 2−7 QD: 7.9 (0.7); 0.23 (-0.39 to 0.71; 25mg QD: - <3.0 mmol/l. 25mg QD: -0.08); 10mg 0.70; 50mg QD: - 8.1 (0.8); QD: -0.56 (- 0.64 50mg QD: 0.71 to -0.41); 7.9 (0.7); pbo: 25mg QD: - 8.0 (0.7) 0.55 (-0.70 to -0.40); 50mg QD: -0.49 (- 0.64 to -0.33); pbo: 0.15 (- 0.00 to 0.30)

Rosenstock sota 75mg [LSM] sota [LSM] manually not reported not reported Rescue permitted. Yes, exclusion criteria: Has had 2 or J 2015 QD: 8.0 (0.9); 400 mg: -0.92 calculated sota Efficacy data collected more emergency room visits, 200mg QD: (p<0.001); 75mg QD: -0.33; after initiation of doctors’ visits, or hospitalizations 8.3 (1.0); pbo: -0.09 200mg QD: -0.43; rescue therapy were due to hypoglycemia within the 6 200mg BID: (p=0.403) 200mg BID: -0.71; excluded from analysis months -> but not excluded for 8.4 (0.9); 400mg QD: -0.83 and hypoglycemic events occurring in 400mg QD: current study. 8.1 (1.0); pbo: subjected to the LOCF 7.9 (0.9) algorithm as needed.

Ross S empa 12.5mg empa 12.5mg empa 12.5mg BID: Confirmed hypoglycemic not reported Rescue medication no 2015 BID: 7.78 BID: -0.84 -0.63 (0.10); 25mg AEs were defined as AEs permitted. Values (0.79); 25mg (0.05); 25mg QD: -0.52 (0.10); observed after a QD: 7.73 QD: -0.72 5mg BID: -0.47 patient started rescue (0.79); 5mg (0.05); 5mg (0.10); 10mg QD: -

190

BID: 7.79 BID: -0.68 0.45 (0.10) (all therapy were set to (0.88); 10mg (0.05); 10mg p<0.001) with plasma glucose missing and imputed QD: 7.84 QD: -0.66 ≤3.9mmol/l and/or requiring (0.75); pbo: (0.05); pbo: - assistance. using a LOCF. 7.69 (0.72) 0.21 (0.08)

Schumm- dapa 2.5mg dapa 2.5mg Difference in Hypoglycemia was defined not reported Rescue not allowed not reported Draeger BID: 7.77 BID: -0.52 adjusted mean as a low blood glucose and patients were PM 2015 (0.75); 5mg (0.0594); 5mg change from reading [<3.5mmol/l (<63 discontinued from BID: 7.78mg BID: -0.65 baseline versus mg/dl)], with or without study. An increase in (0.76); (0.0600); placebo+MET-IR symptoms; a symptomatic metformin dosing. QD10mg QD: 10mg QD: - (SE) (CI). dapa event without a blood 7.71 (0.71); 0.59 (0.0598); 2.5mg BID: -0.22 glucose reading was pbo: 7.94 pbo: -0.30 (0.0840) [-0.38 to - considered ‘suggestive of (0.85) (0.0593) 0.05]; 5mg BID: - hypoglycemia’. Major Difference in 0.35 (0.0843) [-0.52 defined as a symptomatic adjusted mean to -0.18]; 10mg event requiring third-party change from QD: -0.29 (0.0844) assistance with a capillary or baseline (SE) [-0.45 to -0.12] plasma glucose value (<3.0mmol/l (<54 mg/dl)).

Wilding ipra 12.5mg [LSM] (95% [LSM] manually not reported Patients were provided with a not reported no JPH 2013 QD: 7.78 CI of LS calculated ipra glucometer (Accu-Chek (0.64); 50mg mean) ipra 12.5mg QD: -0.22; Performa; Hoffmann-La Roche QD: 7.76 12.5mg QD: - 50mg QD: -0.34; Ltd, Basel, Switzerland) and (0.66); 150mg 0.53 (-0.71 to 150mg QD: -0.41; asked to monitor capillary blood QD: 7.73 -0.36); 50mg 300mg QD: -0.48 glucose twice daily (fasted and 2 (0.69); 300mg QD: -0.65 (- h after a meal) and also when QD: 7.87 0.83 to -0.47); symptoms of hypoglycemia (0.82); pbo 150mg QD: - occurred. As classified by the 7.68 (0.60) 0.72 (-0.90 to investigator (not reported). -0.53); 300mg QD: -0.79 (-

191

0.97 to -0.62); pbo -0.31 (- 0.50 to -0.13)

Yang W dapa 5mg dapa 5mg QD: dapa 5mg QD: - Reported hypoglycemic Patient-reported; not Rescue confounding. no 2016 QD: 8.09 -0.82 (-0.94 to 0.59 (-0.76 to - episodes were classified as biochemically confirmed Rescue (pioglitazone) (0.72); 10mg -0.70); 10mg 0.42); 10mg QD: - major, minor, or ‘other’ permitted. Safety was QD: 8.17 QD: -0.85 (- 0.62 (-0.79 to -0.45) according to symptoms, summarized (0.84); pbo 0.96 to -0.73); capillary or plasma glucose descriptively and 8.13 (0.85) pbo -0.23 (- levels, and the requirement included data after 0.35 to -0.11) for external assistance. A major event was defined as a rescue. symptomatic episode requiring external (third party) assistance due to severe impairment in consciousness or

behavior with a capillary or plasma glucose value <3 mmol/L (<54 mg/dL) and prompt recovery after glucose or glucagon administration.

192

Characteristics of Included Studies – Second AHA Added to non-metformin Background

Study ID Dose Study n= Mean Age (SD) Gender (% male) Countries Ethnicity (%) Duration of Diabetes in Years (SD) Duration Studied

Kadowaki cana 100mg QD 24 weeks cana cana 100mg: 58.4 cana 100mg: 77.1; Japan Japanese cana 100mg: 8.34 (7.74); pbo: 6.50 (3.89) T 2017 100mg: 70; (8.9); pbo: 56.0 (9.5) pbo: 77.9 (as add-on to teneli) pbo: 68

Kadowaki teneli 20mg QD 24 weeks teneli teneli 20mg: 55.9 teneli 20mg: 83.1; Japan Japanese teneli 20mg: 8.15 (5.86); pbo: 7.34 (5.34) T 2018 20mg: 77; (8.3); pbo: 54.1 (10.2) pbo: 75.3 (as add-on to cana) pbo: 77

Odawara M FDC of vilda 14 weeks vilda met: vilda met: 57.5 (10.9); vilda met: 71.3; Japan Japanese vilda met: 7.0 (6.5); vilda pbo: 7.1 (6.9) 2015 50mg/met 250mg bid 115; vilda vilda pbo: 56.2 (9.8) vilda pbo: 71.4 or vilda 50mg/ met pbo: 56 500 mg bid

(patients previously on vilda monotherapy)

193

Characteristics of Included Studies – Second AHA Added to non-metformin Background continued

Study ID Baseline Mean Change Mean HYPO def Ascertainment of Hypo Rescue Medication Excl of Pts w events? A1C % in HbA1C % Difference in At screening or (SD) vs. baseline HbA1c % vs during study) pbo

Kadowaki T cana [LSM] (SE) [LSM] (SE) Severe hypoglycemia: An event Subjects were instructed to fill out the not reported no 2017 100mg: cana 100mg: - cana 100mg: - requiring assistance of another person required information in their patient 8.18 (0.90); 0.97 (0.10); 0.88 (-1.15 to - to administer carbohydrate, glucagon, diaries. If patients observed symptoms pbo: 7.87 pbo: -0.10 0.60) or other resuscitative actions of hypoglycemia, they performed self- (0.83) (0.10) Symptomatic hypoglycemia: Typical monitoring of blood glucose (SMBG) hypoglycemia symptoms with blood when possible and visited their doctor glucose levels of ≤70 mg/dL (3.89 immediately if the administration of mmol/L) at the time of onset sucrose (sugar) did not alleviate the Asymptomatic hypoglycemia: No symptoms. Subjects recorded typical hypoglycemia symptoms, but symptoms of hypoglycemia and data blood glucose levels were ≤70 mg/dL of SMBG (if possible) in their patient (3.89 mmol/L) Suspected diary. Investigators instructed every symptomatic hypoglycemia: Typical subject to bring their patient diaries to hypoglycemia symptoms, but blood each study center visit for viewing. If glucose levels were not measured. the investigator considered any events Relative hypoglycemia: Typical in the diary to be hypoglycemia that hypoglycemia symptoms with blood occurred after initiation of the blinded glucose levels >70 mg/dL at the time study drugs, they noted this in the of onset. hypoglycemia field in the clinical report form. Any such events that were considered adverse events but not hypoglycemia was noted in the adverse event field of the clinical report form.

194

Kadowaki T teneli [LSM] (SE) [LSM] (SE) not reported not reported not reported no 2018 20mg: 7.98 teneli 20mg: - teneli 20mg: - (0.80); pbo: 0.94 (0.08); 0.94 (-1.16 to - 8.09 (0.85) pbo: 0.00 0.72) (0.08)

Odawara M ≤ 8% vilda vilda met vilda met Hypoglycemia was defined as Patients were asked to record Rescue not permitted no 2015 met: 67%; 50/250mg bid: 50/250mg bid: - symptoms suggestive of vilda pbo: -0.6 (p<0.001); 0.7; vilda met hypoglycemia, further confirmed by hypoglycemic events in a study diary. 66.1%; >8 vilda 50/500mg bid: - self-monitored blood glucose to ≤ 9% met50/500mg 1.1 measurement of <3.1 mmol/L. The vilda met: bid: -1.0 event was considered severe if the 20.0%; (p<0.001); (manually patient required assistance of another vilda pbo: vilda pbo: 0.1 calculated) person or hospitalization. 19.6%; (FAS) >9% vilda met: 13.0%; vilda pbo: 14.3%

195

Characteristics of Included Studies – Dual Therapy Initiation

Study ID Dose Study n= Mean Age (SD) Gender Countries Ethnicity (%) Duration of Diabetes in Years (SD) Duration (% male) Studied

Goldstein sita 50mg + 24 weeks sitagliptin sitagliptin 50mg sitagliptin multinational; sitagliptin 50mg + metformin 500mg BID: White sitagliptin 50mg + metformin 500mg BJ 2007 met 500mg 50mg + + metformin 50mg + not specified 53.7, Black 6.8, Hispanic 28.9, Asian 4.7, Other BID: 4.5 (4.7); sitagliptin 50mg + BID; sita metformin 500mg BID: metformin 5.8; sitagliptin 50mg + metformin 1000mg BID: metformin 1000mg BID: 4.4 (4.2); pbo: 50mg + met 500mg BID: 54.1 (10.0); 500mg White 52.2, Black 7.7, Hispanic 26.9, Asian 6.0, 4.6 (4.9) 1000mg BID 190; sitagliptin 50mg BID: Other 7.1; pbo: White 46.0, Black 9.7, Hispanic sitagliptin + metformin 55.3; 26.7, Asian 6.8, Other 10.8 50mg + 1000mg BID: sitagliptin metformin 53.3 (9.6); 50mg + 1000mg placebo 53.6 metformin BID: 182; (10.0) 1000mg placebo: BID: 176 42.3; pbo: 52.8

Haak T lina 2.5mg 24 weeks lina 2.5mg lina 2.5mg BID lina 14 countries; lina 2.5mg BID + 500mg met BID: white 72.0, lina 2.5mg BID + 500mg met BID: ≤1: 2012 BID + BID + + 500mg met 2.5mg not specified Asian 25.9, Black 1.4, Hawaiian Pacific Islander 38.0, > 1 to 5: 33.6, > 5: 28.5; lina 2.5mg 500mg met 500mg met BID: 55.6 BID + 0.7; lina 2.5mg BID + 1000mg met BID: white BID + 1000mg met BID: ≤1: 36.4, > 1 to BID; lina BID: 143; (11.2); lina 500mg 65.7, Asian 33.6, Black 0.7, Hawaiian/Pacific 5: 35.7, > 5: 27.9; pbo: ≤1: 30.8, > 1 to 5: 2.5mg BID lina 2.5mg 2.5mg BID + met BID: Islander 0.0; pbo: White 63.9, Asian 36.1, Black 35.4, > 5: 33.8 + 1000mg BID + 1000mg met 51.0; lina 0.0, Hawaiian/Pacific Islander 0.0 met BID 1000mg met BID: 56.4 2.5mg BID: 143; (10.7); pbo: 55.7 BID + pbo: 72 (11.0) 1000mg met BID: 53.8; pbo: 50.0

196

Ji L 2016 sita 50mg 24 weeks sita 50mg sita 50mg BID + sita 50mg China sita 50mg BID + met 500mg BID: Asian 100; sita sita 50mg BID + met 500mg BID: 1.1 BID + met BID + met met 500mg BID: BID + 50mg BID + met 850mg BID: Asian 100; pbo (0.3); sita 50mg BID + met 850mg BID: 500mg BID; 500mg BID: 52.6 (11.3); sita met 100 1.1 (0.3); pbo: 1.1 (0.2) sita 50mg 122; sita 50mg BID + met 500mg BID + met 50mg BID 850mg BID: BID: 850mg BID + met 52.4 (9.3); pbo 69.7; sita 850mg BID: 53.6 (9.7) 50mg 125; pbo: BID + 126 met 850mg BID: 53.6; pbo: 68.5

Ji L 2017 alo 12.5mg 26 weeks alo + met: alo+met: 53.4 alo+met: China alo+met: Asian 100; pbo: Asian 98.8, American not reported + met 158; pbo: (10.46); pbo: 57.2; pbo: Indian or Alaskan Native 1.2, Multiracial 0 500mg FDC 161 52.2 (10.17) 58.3 BID

Miller S ertu 5mg 26 weeks ertu 5mg ertu 5mg QD + ertu 5mg USA ertu 5mg QD + sita 100mg QD: White 93.9, ertu 5mg QD + sita 100mg QD: 5.7 (5.0); 2018 QD + sita QD + sita sita 100mg QD: QD + sita Black or African American 2.0, American Indian ertu 15mg QD + sita 100mg QD: 6.5 100mg QD; 100mg QD: 56.4 (9.3); ertu 100mg or Alaska Native 4.1, Multiple 0.0, Native (6.5); pbo: 6.8 (6.5) ertu 15mg 98; ertu 15mg QD + sita QD: 58.2; Hawaiian or other Pacific Islander (0), Hispanic QD + sita 15mg QD + 100mg QD: 56.1 ertu 15mg or Latino 34.7; ertu 15mg QD + Sita 100mg QD: 100mg QD sita 100mg (10.1); pbo: 54.3 QD + sita White 84.4, Black or African American 7.3, QD: 96; (10.3) 100mg American Indian or Alaska Native 6.3, Multiple pbo: 97 QD: 55.2; 1.0, Native Hawaiian or other Pacific Islander pbo: 58.8 1.0, Hispanic or Latino 35.4; pbo: White 92.8, Black or African American 4.1, American Indian or Alaska Native 2.1, Multiple 1.0, Native Hawaiian or other Pacific Islander (0), Hispanic or Latino 38.1

197

Pratley alo 12.5mg 26 weeks alo 12.5mg alo 12.5mg + alo worldwide; not alo 12.5mg + met 500mg BID: White 68.5, Asian alo 12.5mg + met 500mg BID: 4.1 RE 2014 + met + met met 500mg BID: 12.5mg + specified 18.0, Black or African American 5.4, Other (4.78); alo 12.5mg + met 1000mg BID: 500mg BID; 500mg BID: 53.7 (11.59); alo met [American Indian or Alaska Native, Native 4.2 (4.97); pbo: 4.3 (4.78) alo 12.5mg 111; alo 12.5mg + met 500mg Hawaiian or Other Pacific Islander and + met 12.5mg + 1000mg BID: BID: Multiracial 8.1; alo 12.5mg + met 1000mg BID: 1000mg BID met 1000mg 54.6 (10.42); 43.2; alo white 68.4, Asian 22.8, Black or African BID: 114; pbo: 53.1 (9.60) 12.5mg + American 4.4, Other [American Indian or Alaska pbo: 109 met Native, Native Hawaiian or Other Pacific 1000mg Islander and Multiracial 4.4; pbo: white 69.7, BID: Asian 18.3, Black or African American 7.3, 54.4; pbo: Other [American Indian or Alaska Native, Native 50.5 Hawaiian or Other Pacific Islander and Multiracial 4.6

Characteristics of Included Studies – Dual Therapy Initiation continued

Study ID Baseline A1C % Mean Change in Mean Difference HYPO def Ascertainment of Hypo Rescue Medication Excl of Pts w events? (SD) HbA1C % vs. baseline in HbA1c % vs At screening or pbo during study)

Goldstein sita 50mg + met [LSM] sita 50mg + met [LSM] sita 50mg not reported not reported Rescue (glyburide) no BJ 2007 500mg BID: 8.8 500mg BID: -1.40 (-1.56 + met 500mg permitted. Safety and (1.0); sita 50mg + to -1.24); sita 50mg + BID: -1.57 (-1.8 efficacy endpoints met 1000mg BID: met 1000mg BID: -1.90 to -1.34); sita excluded data after 8.7 (0.9); pbo: 8.7 (-2.06 to -1.74); pbo: 50mg + met rescue. (1.0) 0.17 (0.0 to 0.33) 1000mg BID: - 2.07 (-2.30 to - 1.84)

198

Haak T lina 2.5mg BID + adjusted mean (SE) lina adjusted mean Severe hypoglycemia defined as Hypoglycemic episodes Rescue (with unlikely, but 2012 500mg met BID: 8.7 2.5mg BID + 500mg met (SE) lina 2.5mg requiring the assistance of were recorded and sulphonylureas, hypoglycemic events (1.0); lina 2.5mg BID BID: -1.2 (0.1); lina BID + 500mg met another person to actively analyzed separately from thiazolidinediones or graded at investigators + 1000mg met BID: 2.5mg BID + 1000mg BID: -1.3 (0.1); administer carbohydrate, other AEs. Hypoglycemic insulin) permitted. discretion; criteria not 8.7 (1.0); pbo: 8.7 met BID: -1.6 (0.1); pbo: lina 2.5mg BID + glucagon or other. event intensity was graded Values obtained after specified and could (1.0) 0.1 (0.1) 1000mg met BID: according to the rescue medication was lead to patient -1.7 (0.1) investigator’s discretion. initiated were not used exclusion in the LOCF.

Ji L 2016 sita 50mg BID + met [LSM] sita 50mg BID + [LSM] sita 50mg Any episode with symptoms not reported Rescue therapy no 500mg BID: 8.5 met 500mg BID: -1.67 (- BID + met 500mg consistent with hypoglycemia (glipizide) permitted. (1.0); sita 50mg BID 1.92 to -1.43); sita 50mg BID: -1.08 (-1.39 (e.g., weakness, dizziness, The primary approach + met 850mg BID: BID + met 850mg BID: - to -0.78); sita shakiness, increased sweating, to analyzing safety 8.6 (0.9); pbo: 9.0 1.83 (-2.07 to -1.58); 50mg BID + met palpitations or confusion) was data treated data (1.1) pbo: -0.59 (-0.84 to - 850mg BID: -1.24 reported as an episode of obtained after the 0.34) (-1.55 to -0.93) symptomatic hypoglycemia initiation of rescue without a requirement for therapy as missing; a confirmatory blood glucose secondary approach values. Asymptomatic included all data, hypoglycemia was defined as an regardless of rescue episode without symptoms of therapy. hypoglycemia, but with fingerstick glucose level ≤3.9 mmol/L (≤70 mg/dL). Severe hypoglycemia was defined as any episode requiring assistance, either medical or non-medical. Episodes with a markedly depressed level of consciousness, loss of consciousness or seizure were to be classified as having

199

required medical assistance, whether or not medical assistance was obtained.

Ji L 2017 alo+met: 8.39 (0.81); [LSM] alo+met: -1.53; [LSM] manually not reported Self-monitor blood glucose Rescue confounding. no pbo: 8.21 (0.77) pbo: -0.19 calculated levels, keep a Hyperglycemic rescue alo+met: -1.34 hypoglycemic diary permitted.

Miller S ertu 5mg QD + sita [LSM] ertu 5mg QD + [LSM] ertu 5mg Symptomatic not reported Rescue confounding. no 2018 100mg QD: 8.9 (0.9); sita 100mg QD: -1.6 (- QD + sita 100mg Rescue therapy erti 15mg QD + sita 1.8 to -1.4); ertu 15mg QD: -1.2 (-1.5 to - hypoglycemia (defined as (glimepiride) 100mg QD: 9.0 (0.9); QD + sita 100mg QD: - 0.8); ertu 15mg episodes with permitted. pbo: 9.0 (0.9) 1.7 (-1.9 to -1.5); pbo: - QD + sita 100mg 0.4 (-0.7 to -0.2) QD: -1.2 (-1.6 to - clinical symptoms reported by the Data following 0.9) investigator initiation of glycemic rescue were included as hypoglycemia; biochemical for the analysis of documentation serious AEs (SAEs), deaths, and not required). discontinuations due to AEs, and excluded for Documented hypoglycemia the other endpoints. (symptomatic and asymptomatic), defined as episodes with a glucose level <70 mg/dL (3.89 mmol/L), were recorded. Severe hypoglycemia defined as requiring non-medical assistance.

Pratley RE not reported; The [LSM] (SE) alo 12.5mg [LSM] manually Mild to moderate hypoglycemia Use of a home glucose Rescue (SU or other) no 2014 majority of patients + met 500mg BID: -1.22 calculated alo (blood glucose <70 mg/dl (3.89 permitted. (0.094); alo 12.5mg + 12.5mg + met mmol/L), symptomatic or monitor and diary to record hypoglycemic episodes. 200

(60%overall) entered met 1000mg BID -1.55 500mg BID: - asymptomatic). All with a baseline (0.09); pbo 0.15 (n/a) 1.37; alo 12.5mg hypoglycemic episodes were + met 1000mg associated with a blood glucose HbA1c of 8.5% or BID -1.70 <70 mg/dl (3.89 mmol/l). Severe lower. required assistance from another person.

Characteristics of Included Studies – Third AHA added to dual therapy background

Study ID Dose Study n= Mean Age Gender Countries Studied Ethnicity (%) Duration of Diabetes in Duration (SD) (% Years (SD) male)

Dagogo- ertu 5mg QD; ertu 24 wks ertu 5mg QD: ertu 5mg ertu USA, Argentina, ertu 5mg QD: White 73.1, Asian 21.2, Black or ertu 5mg QD: 9.9 (6.1); Jack S 15mg QD with 156; 15mg QD: 59.2 5mg Colombia, Czech African American 1.3, American Indian or Alaska 15mg QD: 9.2 (5.3); pbo 2018 extension QD: 153; (9.3); 15mg QD: Republic, Hungary, Native 0.6, Multiple 3.8, Hispanic or Latino 14.7; 9.4 (5.6) until 52 pbo: 153 QD: 59.7 51.9; Israel, Romania, 15mg QD: White 75.2, Asian 18.3, Black or weeks (8.6); pbo 15mg Slovakia, Republic of African American 2.6, American Indian or Alaska (added to met plus 58.3 (9.3) QD: Korea (South Korea), Native 3.3, Multiple 0.7, Hispanic or Latino 16.3; sita) 53.6; Malaysia, Bulgaria, pbo: White 70.6, Asian 21.6, Black or African pbo Finland American 2.0, American Indian or Alaska Native 65.4 3.3, Multiple 2.6, Hispanic or Latino 15.7

Jabbour dapa 10mg QD 24 weeks dapa 10mg: dapa 10mg: dapa Argentina, Germany, dapa 10mg: White 72.2, Black 4.9, Asian 0.9, dapa 10mg: 5.7 (4.87); SA 2014 223; pbo: 224 54.8 (10.4); 10mg: Mexico, Poland, UK, Other 22.0; pbo: White: 76.3, Black 2.7, Asian 0.9, pbo: 5.64 (5.40) pbo: 55.0 57.0; USA Other 20.1 (10.2) pbo: (added to met with or 52.7 without sita)

201

Ludvik B dula 1.5mg QW; dula 24 weeks dula 1.5mg dula 1.5mg dula Spain, Isreal, US, dula 1.5mg QW: White 89, American Indian or dula 1.5mg QW: 9.21 2018 0.75mg QW QW: 142; QW: 56.17 1.5mg Austria, Alaska Native: 1, Asian 0, Black of African (5.74); dula 0.75mg QW: dula 0.75mg (9.26); dula QW: American 2, Multiple 8; dula 0.75mg QW: White 10.05 (6.56); pbo 8.87 QW: 141; 0.75mg QW: 54.0; 90, American Indian or Alaska Native: 1, Asian 1, (6.13) pbo: 140 58.55 (9.14); dula Black of African American 2, Multiple 6; pbo: (SGLT2i with or pbo: 57.1 0.75mg White 89, American Indian or Alaska Native 3, without metformin) (9.59) QW: Asian 0, Black of African American 4, Multiple 4 49; pbo: 47

Mathieu dapa 10mg QD 24 weeks dapa 10mg: dapa 10mg: dapa USA, Puerto Rico, dapa 10mg: White 93.8, African American 5.0, dapa 10mg: 7.2 (5.7); pbo: C 2015 160; pbo: 160 55.2 (8.6); 10mg: Romania, Russian Asian 0.6, Other 0.6; pbo: White 91.9, African 8.0 (6.6) pbo 55.0 43.7; Federation, Poland, American 6.3, Asian 0.6, Other 1.3 (9.6) pbo: Mexico, UK (added to saxa plus 47.5 met)

Matthaei saxa 5mg/day QD 24 weeks saxa 5mg: saxa 5mg: saxa USA, Puerto Rico, saxa 5mg: White 88.9, Black 7.2, Asian 3.3, Other saxa 5mg: 8.1 (7.0); pbo: S 2015 153, pbo: 162 54.7 (9.8); 5mg: Canada, Romania, 0.7; pbo: White 87.0, Black 5.6, Asian 4.9, Other 7.4 (5.8) pbo: 54.5 47.7; Russian Federation, 2.5 (9.3) pbo: Poland, Mexico, Czech (added to dapa plus 46.9 Republic, Hungary met)

Rodbard cana 100mg; cana 26 weeks cana 108; pbo cana: 57.4 cana: 5 countries; not cana: White 74.8, Black African American 5.6, cana: 9.8 (5.4); pbo: 10.1 HW 2016 300mg 108 (9.3); pbo: 61.7; specified Asian 18.7, Other [Native Hawaiian or other (5.9) 57.5 (10.1) pbo: Pacific Islander and other] 0.9; pbo: White 72.6, 51.9 Black African American 15.1, Asian 011.3, Other [Native Hawaiian or other Pacific Islander and (added to sita plus other] 0.9 met)

202

Søfteland empa 24 weeks empa 10mg: empa 10mg: empa Australia, Brazil, empa 10mg: White 61.5, Asian 23.9, Other 14.7; empa 10mg: ≤1 year 5.5, E 2017 112; empa 54.3 (9.6); 10mg: Canada, France, Korea, empa 25mg: White 59.1, Asian 27.3, Other 13.6; >1-5 years 27.5, >5-10 (added to lina plus 25mg: 110; empa 25mg: 60.6; New Zealand, Norway, pbo: White 54.6, Asian 29.6, Other 15.7 years 38.5, >10 years 28.4; metformin) pbo: 110 55.4 (9.9); empa Spain, Taiwan, USA empa 25mg: ≤1 year 6.4, pbo 55.9 25mg: >1-5 years 37.3, >5-10

(9.7) 64.5; years 31.8, >10 years 24.5; pbo: pbo: ≤1 year 8.3, >1-5 single-pill 55.6 years 28.7, >5-10 years combination 35.2, >10 years 27.8

of empa 10mg/lina 5mg QD

or empa 25mg/lina 5mg QD

PLUS

unchanged background metformin

203

Characteristics of Included Studies – Third AHA added to dual therapy background continued

Study ID Baseline Mean Mean Difference HYPO def Ascertainment of Hypo Rescue Medication Excl of Pts w events? A1C % Change in in HbA1c % vs At screening or (SD) HbA1C % pbo during study) vs. baseline

Dagogo- ertu 5mg [LSM] ertu [LSM] ertu 5mg: Documented hypoglycaemia (symptomatic and not reported Rescue (glimepiride or no Jack S QD: 8.1 5mg: -0.8 (- -0.7 (-0.9 to - asymptomatic), defined as episodes with a glucose level insulin) permitted. Data 2018 (0.9); 0.9 to -0.6); 0.5); 15mg: -0.8 ≤3.9 mmol/L (70 mg/dL), with or without symptoms, post rescue excluded for 24 15mg 15mg: -0.9 (- (-0.9 to -0.6) was recorded. Severe hypoglycemia not defined. weeks but not 52-week QD: 8.0 1.0 to -0.7); analysis, with the (0.8); Pbo: -0.1 (- exception pbo: 8.0 0.2 to 0.0) (0.9) of those related to hypoglycaemia.

Jabbour dapa [LSM] dapa [LSM] adjusted not reported not reported Rescue therapy no SA 2014 10mg: 10mg: -0.5 (- mean change (glimepiride) permitted. 7.9 (0.8); 0.6 to -0.4); from baseline Change in hba1c as pbo: 8.0 pbo: 0.0 (- (LOCF), primary endpoint excluded (0.8) 0.1 to 0.1) placebo- data after rescue therapy; corrected change: unclear whether safety data dapa 10mg: -0.5 excluded data after rescue (-0.6 to -0.3)

Ludvik B dula [LSM] dula [LSM] dula Total hypoglycaemia (plasma glucose ≤70 mg/dL [3∙9 documented a criterion Rescue permitted. Oral Patients who had 2018 1.5mg 1.5mg QW: - 1.5mg QW vs mmol/L]). Severe Hypoglycaemia: An episode AHA or insulin, results severe persistent QW: 8.04 1.34 (SE pbo: -0.79 (-0.97 requiring the assistance of another person to actively provided with and without hyperglycemia or (0.65); 0.06); dula to -0.61); dula administer carbohydrate, glucagon, or other rescue. hypoglycaemia dula 0.75mg QW: 0.75mg QW vs resuscitative actions. These episodes may have been between study visits 0.75mg -1.21 (SE pbo: -0.66 (-0.84 associated with sufficient neuroglycopenia to induce were advised to QW: 8.04 0.06); pbo: - to -0.49) seizure or coma. Plasma glucose (PG) measurements contact the (0.61); may not have been available during such an event, but investigative site. For 204

pbo: 8.05 0.54 (SE neurological recovery attributable to the restoration of repeated episodes of (0.66) 0.06) PG to normal was considered sufficient evidence that hypoglycaemia, the the event was induced by a low PG concentration. metformin dose could Documented Symptomatic Hypoglycaemia: Any time a be reduced, patient felt that he/she was experiencing symptoms proceeding to and/or signs associated with hypoglycemia and had a complete withdrawal, PG level of ≤70 mg/dL (3∙9 mmol/L), or PG <54 mg/dL as deemed necessary (3∙0 mmol/L). Asymptomatic Hypoglycaemia: Any by the investigator. event not accompanied by typical symptoms of Continued risk of hypoglycaemia but with a measured PG of ≤70 mg/dL hypoglycaemia, (3∙9 mmol/L), or PG <54 mg/dL (3∙0 mmol/L). despite discontinuation Nocturnal Hypoglycaemia: Any hypoglycemic event of metformin, that occurred between bedtime and waking. Probable warranted dose Symptomatic Hypoglycaemia: An event during which reduction or complete symptoms of hypoglycaemia were not accompanied by withdrawal of the a PG determination (but that was presumably caused by SGLT2 inhibitor. a PG concentration of ≤70 mg/dL [3∙9 mmol/L]), or PG Patients were to <54 mg/dL (3∙0 mmol/L). Total hypoglycaemia continue injectable included any event that met criteria for severe, study drug in either documented symptomatic, asymptomatic, nocturnal, or case. probable symptomatic hypoglycaemia as defined above.

Mathieu C dapa dapa 10mg: dapa 10mg: -0.72 Hypoglycemia episodes were classified as minor not reported Rescue permitted. Data no 2015 10mg: -0.82 (-0.96 (-0.91 to -0.53) (symptomatic or asymptomatic with plasma glucose after the receipt of rescue 8.24 to -0.69); concentration of <63 mg/dL (3.5 mmol/L), regardless of medication were excluded (0.96); pbo -0.10 (- in safety analysis (one pbo: 8.17 0.24 to 0.04) the need for external assistance), major (symptomatic page after rescue). (0.98) requiring third-party assistance because of severe impairment in consciousness or behavior with or without a plasma glucose concentration of <54 mg/dL (3.0 mmol/L) and prompt recovery after glucose or glucagon administration), and other (suggestive episode not meeting the criteria for major or minor).

205

Matthaei S saxa saxa 5mg: - saxa 5mg: -0.35 Hypoglycemic episodes were classified as minor not reported Rescue therapy permitted. no 2015 5mg: 0.51 (-0.63 (-0.52 to -0.18) (symptomatic or asymptomatic, with plasma glucose Primary efficacy (change 7.97 to -0.39); concentration <63 mg/dL (3.5 mmol/L), regardless of in baseline hba1c) (0.83); pbo: -0.16 (- need for external assistance); major (symptomatic, excluded data after rescue. pbo: 7.86 0.28 to - requiring third-party assistance because of severe Whether safety analysis (0.93) 0.04) impairment in consciousness or behavior, with or excluded data after rescue without plasma glucose concentration <54 mg/dL (3.0 was not addressed. mmol/L), and prompt recovery after glucose or glucagon administration); and other (a suggestive episode not meeting the criteria for major or minor).

Rodbard cana: 8.5 [LSM] cana: [LSM] cana: - Documented hypoglycaemia [i.e. concurrent fingerstick Patients were instructed Rescue not permitted and no HW 2016 (0.9); -0.91; pbo: - 0.89 (-1.19 to - or plasma glucose ≤3.9mmol/l (≤70mg/dl) with or to record information on patients were discontinued pbo: 8.4 0.01 (CI n/a) 0.59) without symptoms or severe episodes (i.e. requiring the signs and symptoms from the study if requiring (0.8) assistance from another person or resulting in seizure or of hypoglycaemia, rescue. loss of consciousness)]. as well as associated SMBG measurements, if available.

Søfteland empa empa 10mg: empa 10mg: - Confirmed hypoglycemic AEs (plasma glucose values not reported Rescue therapy permitted no E 2017 10mg: -0.65 (n/a); 0.79 (-1.02 to - ≤3.9 mmol/L and/or requiring assistance) (but not DPP4i, GLP1RA 7.97 empa 25mg: 0.55); empa or SGLT2i). Data post (0.84); -0.56 (n/a); 25mg: -0.70 (- rescue set to missing for all empa pbo: 0.14 0.93 to -0.46) primary analysis. 25mg: 7.97 (0.82); pbo: 7.97 (0.85)

206

Appendix VII: Forest Plots of Any Hypoglycemia

207

208

209

210

211

212

Appendix VIII: Forest Plots of Severe Hypoglycemia

213

Appendix IX: Forest Plots of Including Studies with Zero Any Hypoglycemia

214

215

216

217

218

219

Appendix X: Forest Plots Including Studies with Zero Severe Hypoglycemia

220

221

222

223

224

225

Appendix XI: Post hoc Analyses if Severe Hypoglycemia was defined as ≤3.1 mmol/L

1 Monotherapy: Metformin vs PBO

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate

1.1 ANY HYPO 7 1954 Risk Ratio (M-H, Random, 1.36 [0.71, 2.63] 95% CI)

1.2 ANY HYPO (w BA0E) 8 2058 Risk Difference (M-H, 0.01 [-0.00, 0.03] Random, 95% CI)

1.3 SH 1 315 Risk Ratio (M-H, Random, 0.26 [0.01, 6.14] 95% CI)

1.4 SH (w BA0E) 8 2058 Risk Difference (M-H, 0.00 [-0.01, 0.01] Random, 95% CI)

2 Monotherapy: DPP4i vs PBO

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate

2.1 ANY HYPO 33 10545 Risk Ratio (M-H, Random, 1.07 [0.76, 1.51] 95% CI)

2.2 ANY HYPO (w BAOE) 43 13485 Risk Difference (M-H, 0.00 [-0.00, 0.01] Random, 95% CI)

2.3 SH 3 1036 Risk Ratio (M-H, Random, 1.03 [0.17, 6.15] 95% CI)

2.4 SH (w BA0E) 42 13218 Risk Difference (M-H, 0.00 [-0.00, 0.00] Random, 95% CI)

3 Metformin Background: DPP4i vs PBO

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate

226

3.1 ANY HYPO 18 7279 Risk Ratio (M-H, Random, 0.88 [0.61, 1.27] 95% CI)

3.2 ANY HYPO (w BA0E) 23 8218 Risk Difference (M-H, -0.00 [-0.01, 0.00] Random, 95% CI)

3.3 SH 6 2657 Risk Ratio (M-H, Random, 0.90 [0.31, 2.67] 95% CI)

3.4 SH (w BA0E) 23 8218 Risk Difference (M-H, 0.00 [-0.00, 0.00] Random, 95% CI)

4 Monotherapy: GLP1RA vs PBO

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate

4.1 ANY HYPO 6 1434 Risk Ratio (M-H, Random, 1.77 [0.85, 3.67] 95% CI)

4.2 ANY HYPO (w BA0E) 9 2210 Risk Difference (M-H, 0.01 [-0.00, 0.02] Random, 95% CI)

4.3 SH 1 164 Risk Ratio (M-H, Random, 0.66 [0.03, 15.85] 95% CI)

4.4 SH (w BA0E) 9 2210 Risk Difference (M-H, 0.00 [-0.01, 0.01] Random, 95% CI)

5 Metformin Background: GLP1RA vs PBO

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate

5.1 ANY HYPO 8 3352 Risk Ratio (M-H, Random, 1.30 [0.79, 2.14] 95% CI)

5.2 ANY HYPO (w BA0E) 8 3352 Risk Difference (M-H, 0.02 [0.01, 0.03] Random, 95% CI)

5.3 SH 2 946 Risk Ratio (M-H, Random, 0.94 [0.25, 3.50] 95% CI)

227

5.4 SH (w BA0E) 8 3352 Risk Difference (M-H, 0.00 [-0.00, 0.01] Random, 95% CI)

6 Monotherapy: SGLT2i vs PBO

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate

6.1 ANY HYPO 17 6036 Risk Ratio (M-H, Random, 1.29 [0.79, 2.09] 95% CI)

6.2 ANY HYPO (w BA0E) 17 6036 Risk Difference (M-H, 0.01 [0.00, 0.01] Random, 95% CI)

6.3 SH 1 461 Risk Ratio (M-H, Random, 2.49 [0.12, 51.59] 95% CI)

6.4 SH (w BA0E) 17 6036 Risk Difference (M-H, 0.00 [-0.00, 0.00] Random, 95% CI)

7 Metformin Background: SGLT2i vs PBO

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate

7.1 ANY HYPO 12 5213 Risk Ratio (M-H, Random, 0.92 [0.56, 1.53] 95% CI)

7.2 ANY HYPO (w BA0E) 15 5850 Risk Difference (M-H, 0.00 [-0.00, 0.01] Random, 95% CI)

7.3 SH 1 394 Risk Ratio (M-H, Random, 0.30 [0.05, 1.77] 95% CI)

228

7.4 SH (w BA0E) 15 5850 Risk Difference (M-H, -0.00 [-0.00, 0.00] Random, 95% CI)

8 Dual Initiation vs PBO

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate

8.1 ANY HYPO 6 1788 Risk Ratio (M-H, Random, 2.12 [0.89, 5.05] 95% CI)

8.2 ANY HYPO (w BA0E) 6 1788 Risk Difference (M-H, 0.02 [0.01, 0.04] Random, 95% CI)

8.3 SH 1 291 Risk Ratio (M-H, Random, 2.51 [0.12, 51.83] 95% CI)

8.4 SH (w BA0E) 6 1788 Risk Difference (M-H, 0.00 [-0.01, 0.01] Random, 95% CI)

9 Triple Therapy vs Dual Therapy + PBO

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate

9.1 ANY HYPO 7 2516 Risk Ratio (M-H, Random, 1.12 [0.69, 1.83] 95% CI)

9.2 ANY HYPO (w BA0E) 7 2516 Risk Difference (M-H, 0.01 [-0.01, 0.02] Random, 95% CI)

229

9.3 SH 3 1217 Risk Ratio (M-H, Random, 0.72 [0.11, 4.53] 95% CI)

9.4 SH (w BA0E) 6 2065 Risk Difference (M-H, 0.00 [-0.01, 0.01] Random, 95% CI)

10 Non-Metformin Background Dual Therapy vs Monotherapy + PBO

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate

10.1 ANY HYPO 0 0 Risk Ratio (M-H, Random, Not estimable 95% CI)

10.2 ANY HYPO (w 3 463 Risk Difference (M-H, 0.00 [-0.02, 0.02] BA0E) Random, 95% CI)

10.3 SH 0 0 Risk Ratio (M-H, Random, Not estimable 95% CI)

10.4 SH (w BA0E) 3 463 Risk Difference (M-H, 0.00 [-0.02, 0.02] Random, 95% CI)

230

Appendix XII: Summary of Findings (SoF) Table

Question : Monotherapy: Met compared to PBO in patients with T2D

Certainty assessment № of patients Effect

Certainty № of Study Risk of Other Monotherapy: Relative Absolute Inconsistency Indirectness Imprecision PBO studies design bias considerations Met (95% CI) (95% CI)

ANY HYPO

7 randomised very not serious not serious not serious none 53/1512 (3.5%) 12/442 RR 1.36 10 more ⨁⨁ ◯◯ trials serious (2.7%) (0.71 to per 1,000 LOW a,b 2.63) (from 8 fewer to 44 more)

CI: Confidence interval; RR: Risk ratio

Explanations a. attrition bias b. Harms outcome in a placebo-controlled trial

Question : Monotherapy: DPP4i compared to PBO in patients with T2D

Certainty assessment № of patients Effect

Certainty № of Study Risk of Other Monotherapy: Relative Absolute Inconsistency Indirectness Imprecision PBO studies design bias considerations DPP4i (95% CI) (95% CI)

ANY HYPO

33 randomised serious a not serious not serious not serious none 151/7567 38/2978 RR 1.07 1 more ⨁⨁⨁ ◯ trials (2.0%) (1.3%) (0.76 to per 1,000 MODERATE 1.51) (from 3 fewer to 7 more)

CI: Confidence interval; RR: Risk ratio

Explanations a. Harms outcome in a placebo-controlled trial

231

Question : Monotherapy: GLP1RA compared to PBO in patients with T2D

Certainty assessment № of patients Effect

Certainty № of Study Risk of Other Monotherapy: Relative Absolute Inconsistency Indirectness Imprecision PBO studies design bias considerations GLP1RA (95% CI) (95% CI)

ANY HYPO

6 randomised very not serious not serious not serious none 43/1020 (4.2%) 8/414 RR 1.77 15 more ⨁⨁ ◯◯ trials serious (1.9%) (0.85 to per 1,000 LOW a,b 3.67) (from 3 fewer to 52 more)

CI: Confidence interval; RR: Risk ratio

Explanations a. attrition bias b. Harms outcome in a placebo-controlled trial

Question : Monotherapy: SGLT2i compared to PBO in patients with T2D

Certainty assessment № of patients Effect

Certainty № of Study Risk of Other Monotherapy: Relative Absolute Inconsistency Indirectness Imprecision PBO studies design bias considerations SGLT2i (95% CI) (95% CI)

ANY HYPO

17 randomised serious a not serious not serious not serious none 78/4531 (1.7%) 17/1505 RR 1.29 3 more ⨁⨁⨁ ◯ trials (1.1%) (0.79 to per 1,000 MODERATE 2.09) (from 2 fewer to 12 more)

CI: Confidence interval; RR: Risk ratio

Explanations a. Harms outcome in a placebo-controlled trial

232

Question : Metformin Background: DPP4i compared to PBO in patients with T2D

Certainty assessment № of patients Effect

Certainty Metformin № of Study Risk of Other Relative Absolute Inconsistency Indirectness Imprecision Background: PBO studies design bias considerations (95% CI) (95% CI) DPP4i

ANY HYPO

18 randomised serious a not serious not serious not serious none 87/4853 (1.8%) 45/2426 RR 0.88 2 fewer ⨁⨁⨁ ◯ trials (1.9%) (0.61 to per 1,000 MODERATE 1.27) (from 7 fewer to 5 more)

SH

3 randomised not not serious not serious not serious none 4/839 (0.5%) 4/466 RR 0.62 3 fewer ⨁⨁⨁⨁ trials serious (0.9%) (0.14 to per 1,000 HIGH 2.67) (from 7 fewer to 14 more)

CI: Confidence interval; RR: Risk ratio

Explanations a. Harms outcome in a placebo-controlled trial

Question : Metformin Background: GLP1RA compared to PBO in patients with T2D

Certainty assessment № of patients Effect

Certainty Metformin № of Study Risk of Other Relative Absolute Inconsistency Indirectness Imprecision Background: PBO studies design bias considerations (95% CI) (95% CI) GLP1RA

ANY HYPO

8 randomised very not serious not serious not serious none 163/2618 23/734 RR 1.30 9 more ⨁⨁ ◯◯ trials serious a,b (6.2%) (3.1%) (0.79 to per 1,000 LOW 2.14) (from 7 fewer to 36 more)

CI: Confidence interval; RR: Risk ratio

233

Explanations a. attrition bias b. Harms outcome in a placebo-controlled trial

Question : Metformin Background: SGLT2i compared to PBO in patients with T2D

Certainty assessment № of patients Effect

Certainty Metformin № of Study Risk of Other Relative Absolute Inconsistency Indirectness Imprecision Background: PBO studies design bias considerations (95% CI) (95% CI) SGLT2i

ANY HYPO

12 randomised serious a not serious not serious not serious none 70/4069 (1.7%) 20/1144 RR 0.92 1 fewer ⨁⨁⨁ ◯ trials (1.7%) (0.56 to per 1,000 MODERATE 1.53) (from 8 fewer to 9 more)

CI: Confidence interval; RR: Risk ratio

Explanations a. Harms outcome in a placebo-controlled trial

Question : Dual Initiation compared to PBO in patients with T2D

Certainty assessment № of patients Effect

Certainty № of Study Risk of Other Dual Relative Absolute Inconsistency Indirectness Imprecision PBO studies design bias considerations Initiation (95% CI) (95% CI)

ANY HYPO

6 randomised very not serious not serious not serious none 60/1477 4/311 RR 2.12 14 more ⨁⨁ ◯◯ trials serious a,b (4.1%) (1.3%) (0.89 to per 1,000 LOW 5.05) (from 1 fewer to 52 more)

CI: Confidence interval; RR: Risk ratio

Explanations a. Harms outcome in a placebo-controlled trial 234

b. attrition bias

Question : Triple Therapy compared to Dual Therapy + PBO in patients with T2D

Certainty assessment № of patients Effect

Certainty Dual № of Study Risk of Other Triple Relative Absolute Inconsistency Indirectness Imprecision Therapy + studies design bias considerations Therapy (95% CI) (95% CI) PBO

ANY HYPO

7 randomised very not serious not serious not serious none 45/1459 25/1057 RR 1.12 3 more ⨁⨁ ◯◯ trials serious a,b (3.1%) (2.4%) (0.69 to per 1,000 LOW 1.83) (from 7 fewer to 20 more)

SH

3 randomised not not serious not serious not serious none 2/814 1/403 RR 0.72 1 fewer ⨁⨁⨁⨁ trials serious (0.2%) (0.2%) (0.11 to per 1,000 HIGH 4.53) (from 2 fewer to 9 more)

CI: Confidence interval; RR: Risk ratio

Explanations a. Harms outcome in a placebo-controlled trial b. attrition bias

235

Copyright Acknowledgements

Copyrights for figures and tables referenced in this thesis are available upon request.

236