Assessment of cardiovascular effects of non-insulin blood-glucose-lowering agents Nelly Herrera Comoglio

To cite this version:

Nelly Herrera Comoglio. Assessment of cardiovascular effects of non-insulin blood-glucose-lowering agents. Human health and pathology. Université de Bordeaux, 2019. English. ￿NNT : 2019BORD0355￿. ￿tel-02969556￿

HAL Id: tel-02969556 https://tel.archives-ouvertes.fr/tel-02969556 Submitted on 16 Oct 2020

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés.

THÈSE PRÉSENTÉE

POUR OBTENIR LE GRADE DE

DOCTEUR DE

L’UNIVERSITÉ DE BORDEAUX

ÉCOLE DOCTORALE

SPÉCIALITÉ PHARMACOÉPIDEMIOLOGIE

Option Pharmaco-épidémiologie, pharmaco-vigilance

Par Nelly Raquel HERRERA COMOGLIO

TITRE :

ÉVÉNEMENTS CARDIOVASCULAIRES MAJEURS ET MORTALITÉ EN PATIENTS TRAITÉS AVEC DES HYPOGLYCÉMIANTS NON INSULINIQUES Étude de cohortes basée sur une population de Catalogne, Espagne

Sous la direction de : Xavier VIDAL GUITART

Soutenue le 17 Décembre 2019

Membres du jury :

Mme. AGUSTI ESCASANY, Antonia Prof. Fundacio Institut Catala de Farmacologia Président, rapporteur Mme. BOSCH Montserrat Prof. Associée Fundacio Institut Catala de Farmacologia Examinateur M. SALVO, Francesco Prof. Université de Bordeaux Examinateur

Titre : ÉVÉNEMENTS CARDIOVASCULAIRES MAJEURS ET MORTALITÉ EN PATIENTS TRAITÉS AVEC DES HYPOGLYCÉMIANTS NON INSULINIQUES Étude de cohortes basée sur une population de Catalogne, Espagne

Résumé :

Le diabète mellitus Type 2 (DMT2) est une maladie chronique et progressive causée par multiples facteurs. Plus de 422 millions de personnes dans tout le monde ont diabète; la maladie a un profond impact social et économique. La maladie cardiovasculaire est la cause principale de la morbilité et la mortalité chez les patients diabetiques, qui ont des taux de mortalité plues élévées que la population non-diabétique.

La définition de la DMT2 est basée sur ses manifestations métaboliques – surtout celles de glucose sanguin – qui servent comme marqueurs du contrôle et de l’évolution de la maladie. Cependant, tandis qu’on reconnait l’effet du contrôle de la glucose sanguin sur les complications microvasculaires, son impact sur les complications macrovasculaires ne sont pas clairs.

Depuis 2008, les nouveaux agents hypoglycémiants doivent démontrer leur sécurité cardiovasculaire, soit à travers d’une meta-analyse ou d’essais cliniques évaluant les résultat cliniques cardiovasculaires; quelques nouveaux agents ont montré une réduction des effets cliniques (comme infarctus du myocarde et accident cérébro- vasculaire) et de la mortalité. Toutefois, les populations qui faisaient partie de ces essais cliniques a grande échelle ont différences avec la population générale; donc, les résults de ces essais ne sont pas completement généralisables.

Tandis que les essais cliniques randomisés sont toujours considérés le “gold- standard” pour la génération de l’évidence scientifique, les études observationelles qui sont fait à partir de grande bases de données utilisés pour d’autre propos, dites “secondaires”, sont de plus en plus utilisées pour la génération de l’évidence scientifique complémentaire ou confirmatoire de celle provenant des essais cliniques, surtout quand ces essais ne sont pas disponibles ou sont impracticables.

Ce travail montre les résultats d’une étude observationnelle de cohortes, basée sur la population enregistrée en SIDIAP, une large base de données des médecins généraux de Catalogne, qui reccueil les régistres de plus de 5,5 millions de personnes. Les évenements cliniques cardiovasculares et la mortalité ont été évalués dans la population générale, non-sélectionnée, traitée avec des agents hypoglycémiants non-insuliniques. On attend que les résultats de cette investigation soient útiles pour la prise de decisions, tant au niveau des cliniciens comme au niveau de la santé publique.

Mots clés : événements cardiovasculaires majeurs ; mortalité ; diabete mellitus type 2 ; hypoglycémiants non-insuliniques

Title: Assessment of cardiovascular effects of non-insulin blood-glucose-lowering agents

Cardiovascular outcomes and mortality in Type 2 diabetes mellitus patients treated with non-insulin blood glucose-lowering drugs in Catalonia: a six-year retrospective population-based cohort study

Abstract :

Type 2 diabetes mellitus (T2DM) is a multifactorial, chronic, progressive disease, affecting more than 422 million people over the world, and having a significant societal and economic impact. Cardiovascular disease is the leading cause of morbidity and mortality in T2DM patients, who have higher rates of mortality than the non-diabetic population.

T2DM is defined by its metabolic -mainly glucose-related- manifestations which serve as markers for controlling the evolution of disease. However, while the effect of control serum glucose levels on microvascular complications is acknowledged, its impact on macrovascular complications remains uncertain.

Since 2008, new blood glucose-lowering agents have to demonstrate cardiovascular safety, and some have shown to reduce cardiovascular outcomes and mortality. However, the populations included in these large cardiovascular outcome trials differ from the general population, making results no fully generalisable.

While randomised controlled trials are the gold standard for generating scientific evidence, observational studies conducted with secondary data of Electronic medical records (EMRs) are increasingly used as a source of complementary or confirmatory evidence, especially when RCTs are not feasible or unavailable.

This work report an observational, population-based cohort study conducted in

SIDIAP, a large Catalan general practitioners database that contains health data of 5,5 million people. We assessed cardiovascular outcomes and mortality in general, unselected T2DM population treated with non-insulin blood-glucose-lowering agents. The results are expected to be useful both for clinical and public health decision-making.

Keywords : cardiovascular outcomes ; mortality ; Type 2 diabetes mellitus ; non- insulin blood-glucose-lowering agents

Unité de recherche

Institut Catala de Farmacologia

European Programme de Pharmacovigilance and Pharmacoepidemiology

PhD Thesis

Assessment of cardiovascular effects of non-insulin blood glucose-lowering agents

Cardiovascular outcomes and mortality in Type 2 diabetes mellitus patients treated with non-insulin blood glucose-lowering drugs in Catalonia: a six-year retrospective population-based cohort study.

PhD candidate Nelly Raquel Herrera Comoglio Prof. Xavier Vidal Guitart Director Universitat Autonoma de Barcelona, Spain Supervisors Prof. Antonia Agusti Prof. Montserrat Bosch Plenary Doctoral Committee Prof. Antonia Agusti President of Jury Prof. Montserrat Bosch Prof. Francesco Salvo

Year: 2019 Bordeaux, France.

PhD Thesis

Assessment of cardiovascular effects of non- insulin blood glucose-lowering agents

Cardiovascular outcomes and mortality in Type 2 diabetes mellitus patients treated with non-insulin blood glucose-lowering drugs in Catalonia: a six-year retrospective population-based cohort study

PhD candidate Nelly Raquel Herrera Comoglio

Director Prof. Xavier Vidal Guitart Fundacio lnstitut Catala de Farmacología Universitat Autonoma de Barcelona, Spain

Supervisors Prof. Antonia Agusti Fundacio lnstitut Catala de Farmacología

Prof. Montserrat Bosch Fundacio lnstitut Catala de Farmacología

Plenary Doctoral Committee President of Jury: Prof. Antonia Agusti

Prof. Montserrat Bosch

Prof. Francesco Salvo Université de Bordeaux

Year: 2019

Bordeaux, France

Declaration of good academic conduct

“I Nelly Raquel HERRERA COMOGLIO, hereby certify that this dissertation, Which is

47,571 words in length, has been written by me’that it is a record ofwork carried out by me, and that it has not been submitted in any previous application for a higher degree.

All sentences or passages quoted in this dissertation from other people’s work (with or without trivial changes) have been placed within quotation marks, and specifically acknowledged by reference to the author, WOrk and page. I understand that plagiarism -

the unacknowledged use of such passages - Will be considered grounds for fail皿e in

this dissertation and in the degree prograITme aS a Whole. I also a触m that’Wi血the exception of the specific acknowledgements, the following dissertation is entirely my

own work.一一

Signature of the Acknowledgements

To the thesis director, Prof. Xavier Vidal, who was always present and supported all the instances of this project, since its early beginnings to the final result.

To the director of Catalan Institut of Pharmacology, Prof. Albert Figueras, to all the team of professionals and its founder, Prof. Joan-Ramon Laporte, for the permanent contribution to the education and investigation in Pharmacoepidemiology in Spain and Latin-American countries.

To the members of Institut Jordi Gol, for the participation in this research

iv Table of contents

Abstract x

Abbreviations xii

Part I Background 1

I Introduction 2

II Cardiovascular Outcomes trials assessing the effect of non-insulin blood-glucose-lowering agents on major cardiovascular adverse events (MACE) and mortality 14

III Generalisability of Cardiovascular Outcomes Trials to the Real World: Implications for 32 Clinical Practice

Part II Cardiovascular outcomes and mortality among type 2 diabetes mellitus patients prescribed first and second-line blood glucose-lowering drugs: a population-based cohort study in the Catalan electronic medical record database, SIDIAP, 2010- 43 2015

IV Protocol Rationale and Design 44

v V Cardiovascular outcomes and mortality in type 2 diabetes mellitus patients prescribed first-line non-insulin blood-glucose-lowering agents as monotherapy 78

VI Cardiovascular outcomes and mortality in type 2 diabetes mellitus patients prescribed second-line, metformin-based non-insulin blood-glucose-lowering agents dual therapies 107

VII Discussion and Conclusion 138

References 144

VIII Appendix A 180

IX Annexes 187

IX.1 Cardiovascular outcomes, heart failure and mortality in type 2 diabetic patients treated with glucagon-like peptide 1 receptor agonists (GLP-1 RAs): A systematic review and meta-analysis of 189 observational cohort studies

IX.2 Linagliptin and Cardiac Failure 208

IX.3 Glibenclamide/glyburide and palpitations in 221 Asian population

vi vii Major cardiovascular outcomes (MACE), mortality and heart failure in Type 2 diabetes mellitus patients treated with non-insulin blood glucose-lowering drugs in Catalonia: a six-year retrospective population-based cohort study Abstract

Diabetes mellitus is a chronic, progressive disease, that affects an increasing number of people worldwide and present with microvascular and macrovascular complications. People with Type 2 diabetes mellitus have 2-4 fold of cardiovascular disease, the leading cause of morbidity and mortality for diabetic patients. Management of T2DM is based on control of blood- glucose and CV risk factors. Therapies for Type 2 diabetes mellitus encompass insulins, sulfonylureas, metformin, meglitinides, thiazolidinediones, dipeptidyl-peptidase inhibitors, glucagon-like peptide 1 receptor agonists, sodium-glucose 2 cotransporter inhibitors and other agents. Since 2008, all new blood glucose-lowering agents have to show CV safety to comply with regulatory recommendations; usually accomplished through large cardiovascular outcomes randomised trials (CVOTs). As the clinical outcomes assessed are relatively rare, the populations of these trials are mostly high CV risk patients. Agents of two classes, GLP-1 RAs and SGLT-2, have shown 13-14% of MACE risk reduction in T2DM patients, the results are driven by all-cause mortality for liraglutide and empagliflozin for CV death. The question that arises is to what extent the results of these CVOTs are generalisable to unselected populations.

The evidence from pharmacoepidemiologic safety studies conducted in large electronic healthcare databases has increasing importance as complementary or confirmatory evidence in regulatory or payers’ decision-making. Observational research also has a unique significance to assess the effect of drug or drug classes in a particular setting and real-world conditions. However, observational research can be flawed by bias in design and analyses and should be rigorously conducted to provide compelling insights and to minimise the inherent confounding by indication of non-randomised studies.

viii The present work hypothesises that, in the study period, the treatment with new classes of blood glucose-lowering drugs in an adult, T2DM population in Catalonia, is not associated with a clinically relevant benefit, defined as a 10% reduction in cardiovascular morbidity and mortality compared with the use of reference non-insulin glucose-lowering agents, metformin and sulphonylureas (SU).

This work presents a longitudinal population-based cohort study to assess CV outcomes and mortality among adults Type 2 diabetes mellitus patients treated with non-insulin blood-glucose-lowering agents in Catalonia. Patients should have been registered in the Catalan nationwide healthcare system and their data recorded in the general practitioners’ Information System for the Development of Research in Primary Care (SIDIAP) database. We used a new-user design to avoid prevalent-user bias and assessed exposures through an as-treated approach, following patients from the first prescription of a given agent to its discontinuation, switching or the addition of another antidiabetic drug. To minimise bias, cohorts of patients were compared at the same line of treatment. Crude incident rates of CV outcomes and mortality were adjusted by demographic, clinical and socio-economic variables through a Cox multivariate analyses. Although we minimised selection bias, other biases such as information bias are likely to be significant in health medical records databases, and residual confounding can not be ruled out.

ix ABBREVIATIONS

3-p MACE 3-point major adverse cardiovascular event 4-p MACE 4-point major adverse cardiovascular event AGE Advanced glycation end products AHT Arterial hypertension AMI Acute myocardial infarction BL Baseline BMI Body mass index BNP Brain natriuretic peptide CABG Coronary arterial by-pass graft CHD Coronary heart disease CHF Congestive heart failure CI Confidence interval CKD Chronic kidney disease CV Cardiovascular CVD Cardiovascular disease DBP Diastolic blood pressure DCCT Diabetes Control and Complications Trial DM Diabetes mellitus DPP-4 Dipeptidyl peptidase – 4 DPP-4i Dipeptidyl peptidase – 4 inhibitor eGFR Estimated glomerular filtration rate EMA European Medicines Agency EU European Union FDA US Food & Drug Administration FPG Fasting plasma glucose GIP Glucose-dependent insulinotropic peptide GLP-1 Glucagon-like peptide 1 GLP-1 RA Glucagon-like peptide 1 receptor agonist HbA1c Glycated haemoglobin HDL-C High-density lipoprotein colesterol HF Heart failure HHF Hospitalisation for heart failure HOPE Heart Outcomes Prevention Evaluation HR Hazard ratio

x HUA Hospitalisation for unstable angina LDL-C Low-density lipoprotein colesterol LV Left ventriculum, left ventricular MACE Major adverse cardiovascular events MET metformin MI Myocardial infarction NIAD Non-insulin blood-glucose-lowering “antidiabetic” drug N-BNP N-terminal pro-Brain natriuretic peptide PAD Peripheral arterial disease PCO Primary composite outcome PTCA Percutaneous transluminal coronary angioplasty RCT Randomised controlled trial RF Renal failure RR Relative risk SBP Systolic blood pressure SCO Secondary composite outcome SGLT-2 Sodium-glucose cotransporter-2 SGLT-2 i Sodium-glucose cotransporter-2 inhibitors SIDIAP Information System for the Development of Research in Primary Care SU Sulfonylurea T1DM Type 1 Diabetes Mellitus T2DM Type 2 Diabetes Mellitus TC Total cholesterol TG Triglycerides TIA Transient ischemic attack TZD Thiazolidinediones UK United Kingdom UKPDS United Kingdom Prospective Diabetes Study UA Unstable angina US United States VADT Veterans Affairs Diabetes Trial VLDL-C Very low-density lipoprotein cholesterol WHO World Health Organization

xi

Part I: Background I. Introduction I. Introduction

The prevalence and trends of diabetes mellitus

Diabetes mellitus (DM) affects more than 422 million people; by 2035, its prevalence is foreseen to rise to 592 million. The number of people with diabetes increased almost 4-fold from 1980 to 2014. [1] The global prevalence of diabetes among adults over 18 years of age has risen from 4.7% in 1980 to 8.5% in 2014 (1 every 12 people). [1, 2] Diabetes mellitus Type 2 (T2DM) accounts or around 90% of all diabetes cases worldwide.[2]

The substantial increase in diabetes prevalence observed both in developed and developing countries might be due to either an increased incidence or longer survival.[3] Diagnosed type 2 diabetes mellitus’ prevalence has been estimated to increase more than twice between 2000 and 2013 in the UK, up to 5.32%. [4] In Catalonia diagnosed T2DM prevalence was 7.6% in 2009, being 3-fold higher in patients aged 75 yr. or older, [5] which is consistent with data reporting a 25% of US population aged ≥65 years having diabetes. [6] Some more recent studies report a stabilisation or fall in diabetes incidence in some countries, to which preventive strategies could have contributed. A recently published review reported an increase of diagnosed diabetes in most populations from the 1960s to the early 2000s, after which a pattern emerged of stable trends in 30% and declining trends in 36% of the reported populations. [3] However, data are limited in low and middle-income countries, where trends in diabetes incidence could be different.

Introduction 3

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 Diabetes mellitus vascular complications and mortality

Diabetes is a significant cause of blindness, kidney failure, heart attacks, stroke and lower limb amputation, [1] WHO projects that diabetes will be the 7th leading cause of death in 2030, and it has been estimated that diabetes caused 4.9 million deaths in 2014. [1,2] The highest number of people with diabetes is between 40 and 59 years of age. Patients with Type 2 diabetes mellitus (T2DM) are more likely to die from any cause and cardiovascular (CV) causes; risks vary and are higher with younger age, worse glycemic control, and greater severity of renal complications; for younger people, the risks of dying persists even for those with acceptable glycaemic control. [7, 8]

Diabetes-related microvascular complications can lead to significant morbidity and premature mortality; however, the most important cause of death in people with diabetes is for cardiovascular disease (CVD). [9] It has long been recognised that diabetes is an independent risk factor for CVD, affecting all components of the cardiovascular system: microvasculature, larger arteries, the heart, as well as the kidneys; and imparting a 2- to 4-fold risk of CVD. Also, many diabetic patients often have other risk factors for CVD, such as obesity, hypertension and dyslipidemia. [10] Patients with diabetes have twice the risk of incident myocardial infarction and stroke as that of the general population, many do not survive their first event, or their mortality rate is generally higher than that of the general population. As many as 80% of patients with type 2 diabetes mellitus will develop and possibly die of macrovascular disease. [11, 12] Older adults with diabetes are at substantial risk for both acute and chronic microvascular and cardiovascular complications of the disease. However, cardiovascular disease prevalence is not affected by older-age onset diabetes. [6]

T2DM people often present with other risk factors for cardiovascular disease (CVD). A third of people with T2DM have CVD: 29.1% had atherosclerosis, 21.2% had coronary heart disease (CHD), 14.9% had heart failure (HF), 14.6% had angina, 10.0% had had a myocardial infarction (MI), and 7.6% had experienced a stroke. CVD causes death in 50% of T2DM patients. [13] In Catalonia, in 2009, the prevalence of CVD prevalence was 22.0%, being coronary heart disease (18.9%) and peripheral ischemia (4.5%) the more frequent manifestations. [14]

Introduction 4

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

A study published in 2009 report that adults with diabetes have experienced a 50% reduction in the rate of incident CVD, although remaining at a consistent, approximate 2- fold excess for CVD events compared with those without diabetes.[15] Marked reductions in cardiovascular disease mortality were seen in the last decades as a result of new therapies and proactive diagnosis. [16, 17] In diabetic patients, CVD mortality rates have decreased in a greater extent than in non-diabetic, thus reducing the difference. Regional differences in mortality in T2DM populations have been reported in Spain and the UK. [17, 18] In US adult diabetic population, 10-year relative changes in mortality were significant for major CVD (by -33%), ischemic heart disease (by -40 %), and stroke (by -30%), but not heart failure (by -0.5%, non-significant) or arrhythmia (-12.0%) [16]

The pathogenesis of heart failure includes not only coronary artery disease but also hypertension and diabetic cardiomyopathy, not fitting clearly into the traditional, binary classification of diabetes complications as either microvascular or macrovascular.[19] In the Framingham study, which has found that in non-diabetic patients the incidence rate of heart failure was higher for men than for women, it has been estimated that in diabetic patients treated with insulin, diabetes confers more than a two-fold increase in the risk of heart failure in men and five-fold higher risk in women. [20, 21] As with stroke and myocardial infarction, in a heart-failure setting in patients with diabetes, mortality rates are about twice that of the non-diabetic population; individuals with diabetes aged 45–54 years are almost 9-fold more likely to develop heart failure, and the relative risk falls to 1.8 for those aged 75–84 years. [19] Results of 4-yrs follow-up of an international registry found that diabetes mellitus was associated with a 33% greater risk of hospitalisation for heart failure. In patients with diabetes mellitus, heart failure at baseline was independently associated with cardiovascular death, increasing fatal outcome 2.5-fold.[21]

Heart failure is the second most frequent cardiovascular presentation in people with diabetes, (14,1%), being peripheral artery disease the first one, with 16,2%. A study conducted in England during 1998-2010 and using data of four linked databases (primary care, hospital admission, disease registry, and death certificate records) found that 17.9% people with type 2 diabetes had a first cardiovascular presentation. Patients with Type 2 diabetes were at about three times higher risk of peripheral arterial disease (HR 2.98), and

Introduction 5

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 at increased risk of ischaemic stroke, stable angina, heart failure, and non-fatal myocardial infarction. [22]

In a decade, in the UK, the proportion of diabetic patients increased from 18% in 2002- 2004 to 26% in 2012-2014.[23] However, the absolute number of newly diagnosed heart failure individuals increased by 12%, and the estimated absolute number of prevalent heart failure cases increased even more, by 23%, this mainly due to an increase in population size and age. Patient age increased 0.79 years and patients had more multi- morbidity at first presentation of heart failure, from 3.4 to 5.4. In the same period, diabetes mellitus was the fifth most prevalent comorbidity for incident CVD (11.2%), but the frequency was higher between 60-69 and 70-79 years (16.3% and17.9% respectively). [24]

The increased mortality of people with diabetes is due not only to CV death but also to cancer-related deaths and other causes. [25]

Both the increased prevalence of DM and diabetes-related comorbidities impact on healthcare costs. Average annual healthcare costs associated with patients with type 2 diabetes are substantially more expensive (72.4%) compared with non-diabetic subjects. They are higher among diabetic patients with poor glycemic control and macrovascular complications. [26]

Glycemic markers and DM complications

Despite the extensive clinical research devoted to, diabetes is still defined by its biochemical manifestations (elevated fasting plasma glucose, glycated haemoglobin, hyperglycaemia and glucosuria) and complications, the pathogenesis of type 2 diabetes and its complications remains unknown. [27] Several mechanisms have been proposed to explain hyperglycemia to increased cardiovascular morbidity and mortality. It has been suggested that hyperglycemia may produce advanced glycation end products in diabetic patients and even in those who are prone to developing diabetes before diabetes onset, contributing to endothelial dysfunction, atherosclerosis and microangiopathy, relevant factors to CVD and heart failure. [28, 29] Blood glucose binds irreversibly with proteins, the rate and extent of nonenzymatic glycation of proteins depend mainly on the prevailing

Introduction 6

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 glucose concentration and the protein life span. Covalent, nonreversible glycation of proteins - the formation of advanced glycation end products (AGE)-, is the final stage of a sequential process that starts with reversible, non-covalent glycation. The presence of various AGE is thought to be linked to the normal ageing process and the chronic complications of diabetes mellitus.[30] Glycated haemoglobin (HbA1c), which indicates the glycemic level during the previous 3-months – the lifespan time of red blood cells-, is the surrogate marker that has been the gold standard outcome in diabetic trials for more than 40 yrs. [31] In healthy subjects, levels of “stable” HbA1C are 5–6% of total HbA, these values can increase up to 15% or more in diabetic individuals.∼ However, there can also be a “labile” HbA1C formed during the early, reversible stages of the glycation process and which reflects ambient vs longer-term glucose levels; this reversible HbA1c may overestimate HbA1C by up 2–3% in healthy subjects and by 10% in subjects with diabetes. It should be kept in mind that hyperglycemia does not provide the complete answer to the aetiology of increased early glycated products, given that glycated haemoglobin is also present in some non-diabetic conditions, including chronic renal failure. [32]

The beneficial effect of intensive therapy on microvascular outcomes have been established for insulin-dependent diabetes mellitus in 1993, showing a direct relationship between increased glycemic levels and microvascular complications. The observational study UKPD 35 found that in type 2 diabetes mellitus patients, previous hyperglycemia was strongly associated with microvascular and macrovascular complications, being any reduction in HbA1c likely to reduce the risk of complications, with the lowest risk being in those with HbA1c values in the normal range (<6.0%). [33] Each 1% reduction in updated mean HbA1c was associated with reductions in risk of macrovascular and microvascular complications: non-significant 14% for myocardial infarction and a significant 37% for microvascular complications. [34]

A substantial amount of increased cardiovascular risk and all-cause mortality caused by T2D cannot be explained by traditional vascular risk factors. Only 35% of the excess cardiovascular risk and 42% of the excess mortality risk caused by T2D have been found to be mediated by the classical cardiovascular risk factors. For CVD, the most considerable mediated effects were by insulin resistance, elevated triglycerides and

Introduction 7

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 micro‐albuminuria. For mortality, the largest mediated effects were by micro‐albuminuria and insulin resistance. [35]

The UKPDS 33, published in 1998, compared the effects of pharmacologic blood-glucose control (“intensive group”, either sulfonylureas or insulin) with diet in patients with type 2 diabetes. The “intensive treatment” decreased the risk of microvascular complications, but not the macrovascular disease. In this study, neither sulfonylureas or insulin showed an adverse effect on cardiovascular outcomes but increased the risk of hypoglycaemia.[36] Since then, the beneficial effect of blood glucose-lowering agents on microvascular complications of diabetes mellitus has been almost unanimously acknowledged by most published statements (77%–100%) and guidelines (95%). [37] However, their effect on macrovascular complications, such as coronary, cerebral and peripheral macroangiopathy, remains uncertain.[11, 38, 39] A meta-analysis of 16 guidelines and 328 statements found that this evidence reported no significant impact of tight glycemic control on the risk of dialysis/transplantation/renal death, blindness, or neuropathy, and a consistent 15% relative risk reduction of non-fatal myocardial infarction, with no significant effect on all-cause mortality, cardiovascular mortality, or stroke. [37] These results are consistent with a previous meta-analysis of more-intensive vs less intensive glucose control found the same risk reduction of 15% for MI, favouring the more intensive control. Exploratory analysis in this MA also suggested that participants with no history of macrovascular disease obtained the benefit, whereas those with a prior macrovascular disease did not. [40]

Epidemiological studies and meta-analyses of RCTs have clearly shown a direct relationship between HbA1c and CVD, but the potential of intensive glycemic control to reduce CVD events has been less clearly defined. [9] A meta-analysis of 102 clinical trials showed that DM confers about a two-fold excess risk for a wide range of vascular diseases. Independently from other conventional risk factors, after adjustment for other risk factors, an increase of 1% in the glycated haemoglobin level is associated with an increase of 18% in the risk of cardiovascular events.[41] The prospective observational study UKPDS 35, published in 2000, found that the incidence of clinical complications was significantly associated with glycaemia reduction, being each 1% mean HbA1c reduction associated with reductions in risk of 21% for diabetes-related deaths, 14% for

Introduction 8

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

MI, 21% for any endpoint related to diabetes and 37% for microvascular complications, retinopathy or renal failure. Interestingly, no threshold of risk was observed for these effects. [34] The association between higher levels of HbA1c and increased CV risks have been confirmed with more or less consistent results in studies using secondary data from healthcare databases. [42, 43] It also has been suggested that in no diabetic patients, the relation between glycated haemoglobin and cardiovascular events would have a linear association in non-extreme values. [44] The Heart Outcomes Prevention Evaluation (HOPE) found that in diabetic participants, a 1% absolute rise in the updated HbA1c predicted future CV events after adjusting for confounders and treatment, and the analysis of diabetic and non-diabetic patients showed that a 1 mmol/l rise in fasting plasma glucose was related to an increased risk of CV outcomes, after adjusting for presence or absence of diabetes, thus indicating an independent progressive relationship between indices of glycaemia and incident CV events, renal disease and death. [45]

It also has been suggested that the current target of HBA1c level does not predict a better coronary microcirculatory function in T2DM patients and that there is a possible link between coronary microvascular disease and LV diastolic function in Type2 diabetic patients. [46, 47]

Ideally, glycemic control should be attained with no hypoglycaemic events. Hypoglycaemia produces significant metabolic stress that could trigger major vascular events such as myocardial infarction and stroke. [48] A decade ago, the potential CV dangers of intensive treatment regimens and strict glycemic control in T2DM people who have CV disease (CVD) arose in three trials in which excess mortality was observed. [49 -51]

Intensive blood-glucose control and clinical outcomes

The Diabetes Control and Complications Trial (DCCT, 1993) randomly assigned 1441 patients with insulin-dependent diabetes mellitus to receive intensive therapy or standard therapy with insulin. In this study, tight glycemic control in type 1 diabetes patients significantly reduced the development and progression of chronic diabetic complications,

Introduction 9

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 such as retinopathy, nephropathy, and neuropathy. [33] Long-term follow-up of these patients demonstrated beneficial effects on macrovascular outcomes in the Epidemiology of Diabetes Interventions and Complications study. The risk of the primary composite CVD outcome was reduced by 42% in the original and that of fatal or non-fatal MI or stroke (MACE) by 57% in the intensive vs the control group, but the limited number of patients with events (only 12) was inadequate to draw conclusions. [52]

The United Kingdom Prospective Diabetes Study (UKPDS 33, 1998) was designed in order to assess micro and macrovascular complications of diabetes in 3867 newly diagnosed patients with type 2 diabetes, median age 54 years. After three months of diet, patients were randomly assigned to standard dietary therapy or pharmacological therapy based either on sulfonylureas (chlorpropamide, glibenclamide and glipizide) or with insulin. Patients assigned to diet received pharmacological treatment only if they had hyperglycemic symptoms or a FPG higher than 15 mm/L. The goal of pharmacological therapy was to maintain FPG < 6.0 mm/L, with stepwise addition of other hypoglycaemic agents (metformin or insulin) when the glycaemic goals were not met (i.e., patients assigned to any of the three sulfonylureas could be given metformin; oral agents could later be replaced by insulin). Follow-up was up to ten years. HA1c was 7.0% in the intensive group compared with 7.9% in the conventional group - an 11% reduction, with no difference in HbA1c among agents in the intensive group. Compared with the conventional group, the risk in the intensive group was 12% lower for the composite of any diabetes-related endpoint (sudden death, death from hyperglycaemia or hypoglycaemia, fatal or non-fatal myocardial infarction, angina, heart failure, stroke, renal failure, amputation, vitreous haemorrhage, retinopathy requiring photocoagulation, blindness in one eye, or cataract extraction); 10% lower for any diabetes-related death (death from myocardial infarction, stroke, peripheral vascular disease, renal disease, hyperglycaemia or hypoglycaemia, and sudden death); and 6% lower for all-cause mortality. Most of the risk reduction in any diabetes-related aggregate endpoint was due to a 25% risk reduction in microvascular endpoints. [36]

In the United Kingdom Prospective Diabetes Study (UKPDS 34, 1998), 753 overweight patients were included in a randomised controlled trial and were followed for 10.7 years. Four hundred eleven patients were allocated in standard treatment, primarily with diet

Introduction 10

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 alone, and 342 patients were allocated in pharmacological treatment with metformin, aiming for FPG < 6 mmol/L. A secondary analysis compared the 342 patients allocated metformin with 951 overweight patients allocated intensive blood-glucose control with chlorpropamide (n=265), glibenclamide (n=277), or insulin (n=409). Metformin has found to have a 34% reduction on cardiovascular outcomes in overweight patients; sulfonylureas showed a non-significant reduction in risk of myocardial infarction (MI).[53] It has been noted that these results were obtained in a randomised subgroup of obese patients (342 patients in the metformin group and 411 in the conventional group) and have never been reproduced, suggesting design and methodological drawbacks. [54] In a supplementary trial, patients on maximal doses of sulfonylureas who attained an HbA1c ≤ 6.1 mmol/L were allocated to be added metformin or to continue on sulfonylurea alone. Patients who were added metformin had a significant 60% higher all- cause death compared with those given sulfonylurea alone. [53]

Post-trial monitoring aimed to determine whether this improved glucose control persisted and whether such therapy had a long-term effect on macrovascular outcomes: 3277 patients were followed through clinical visits or annual questionnaires for five years, with no intervention to maintain their previously assigned therapies all patients in years 6 to 10 were assessed through questionnaires. Although differences in glycated haemoglobin levels were lost after the first year, the relative reduction in risk of microvascular outcomes persisted at ten years and reduction in risk on some CV outcomes emerged. In the sulfonylurea-insulin group, relative reductions in risk persisted at ten years for any diabetes-related endpoint (9%, P=0.04) and microvascular disease (24%, P=0.001), risk reductions for myocardial infarction (15%, P=0.01) and death from any cause (13%, P=0.007) emerged over time. In the metformin group, significant risk reductions persisted for any diabetes-related endpoint (21%, P=0.01), myocardial infarction (33%, P=0.005), and death from any cause (27%, P=0.002). [55]

In the Veterans Affairs Diabetes Trial, (VADT, 2009) no significant effect on the rates of major cardiovascular events, death, or microvascular complications - except progression of albuminuria- was obtained through an intensive glucose control in patients with poorly controlled type 2 diabetes. [56] In this study, 1791 military veterans (mean age, 60.4 years, mean time from diagnosis of diabetes 11.5, yrs., 40% with a history of a previous

Introduction 11

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 cardiovascular event) were randomly assigned to receive intensive vs the standard pharmacological therapy. Intensive therapy started at maximal doses and standard therapy at half of the maximal doses. [56] The primary outcome was the time from randomisation to the first occurrence of a major cardiovascular event, a composite of myocardial infarction, stroke, death from cardiovascular causes, congestive heart failure, surgery for vascular disease, inoperable coronary disease, and amputation for ischemic gangrene. The median follow-up was 5.6 yrs. Patients with a BMI ≥ 27 were given metformin plus rosiglitazone 27, and those who had a BMI ≤ 27 were started on glimepiride plus rosiglitazone.[56] In the follow-up extension of VADT trial, after 9.8 years of follow-up, patients with type 2 diabetes who had been randomly assigned to intensive glucose control for 5.6 years had fewer major cardiovascular events than those assigned to standard therapy, but no improvement was seen in the rate of overall survival (VADT follow-up, 2015).[57]

In the ADVANCE trial (2008), with glucose intensive control there were no significant effects on major macrovascular events ( HR 0.94; 95% CI, 0.84 to 1.06; P=0.32), death from cardiovascular causes (HR 0.88; 95% CI, 0.74 to 1.04; P=0.12), or death from any cause (HR 0.93; 95% CI, 0.83 to 1.06; P=0.28). In this study,11,140 patients with type 2 diabetes were allocated to receive either standard glucose control or intensive glucose control, the latter defined as the use of gliclazide (modified release) plus other drugs as required to achieve a glycated haemoglobin value of 6.5% or less. [58] After a median of 5 years of follow-up, the haemoglobin target was achieved in the intensive-control group (6.5%), while in the standard-control group was 7.3%. Intensive control reduced the incidence of combined major macrovascular and microvascular events, primarily because of a reduction in the incidence of nephropathy. Severe hypoglycemia was more frequent common in the intensive-control group (2.7%, vs 1.5% in the standard-control group; hazard ratio, 1.86; 95% CI, 1.42 to 2.40; [58] However, with intensive control

In the ACCORD trial (2008), the intensive group therapy was discontinued after a follow- up of 3.5 yrs. because of higher mortality (HR ratio, 1.22; 95% CI, 1.01 to 1.46; P = 0.04).This study assessed the effect of intensive therapy vs glucose-lowering standard on 10,251 patients (mean age, 62.2 years, 38% women, 35% with a history of a cardiovascular event) with a median glycated haemoglobin level of 8.1%. The target of

Introduction 12

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 the intensive therapy group was an HbA1c level below 6.0%, and the standard therapy target was from 7.0 to 7.9%. The primary outcome - a composite of non-fatal myocardial infarction, non-fatal stroke, or death from cardiovascular causes- was no significant reduced in the intensive therapy group (HR, 0.90; [CI], 0.78 to 1.04; P = 0.16). This result was due to a lower rate of nonfatal MI in the intensive group than in the standard therapy group (3.6% vs. 4.6%; HR, 0.76; 95% CI, 0.62 to 0.92; P = 0.004), and a higher rate of death from cardiovascular causes in the intensive group (2.6% vs. 1.8%; hazard ratio, 1.35; 95% CI, 1.04 to 1.76; P = 0.02); with no significant difference in the rate of nonfatal stroke (1.3% vs. 1.2%; HR, 1.06; 95% CI, 0.75 to 1.50; P = 0.74). Of note, rates of the primary outcome began to separate in the two study groups after three years.[59] After the intensive therapy was discontinued, the target for glycated haemoglobin level was set from 7 to 7.9% for all participants, and the median HbA1c in this group rose from 6.4% to 7.2%, and the use of glucose-lowering medications and rates of severe hypoglycemia were similar in the two groups. The follow-up continued until the planned end of the trial (5 yrs). The trends in CV mortality and MI persisted during the entire follow-up period (HR for death, 1.19; 95% CI, 1.03 to 1.38; and HR for non-fatal myocardial infarction, 0.82; 95% CI, 0.70 to 0.96). [60]

Before the ACCORD trial’s results were published, in 2008, a majority of statements declared valuable to achieve tight glycemic control to prevent macrovascular complications (47%–59%). In 2009, only 21% of statements favoured strict glycemic control. [37] The concentrations of glycated haemoglobin (HbA1c), which are used as a surrogate marker for outcomes that are important to patients, such as blindness or amputation, do not have a linear relationship with CV outcomes.[61] An intensive glucose control – aiming to maintain HbA1c levels close to those of healthy patients, has failed to demonstrate benefits for CV mortality, though showing a trend towards lower MI risks. Being the control of CV risk factors one of the goals of the diabetes care and CVD, other surrogates, like blood pressure, lipids, albumin excretion rates, and C reactive protein have been used to predict CVD outcomes and mortality.

In 2008, as a result of the findings of an increased number of MI in a trial with rosiglitazone, the CV safety of blood glucose-lowering drugs was required to be assessed through major adverse cardiovascular events (MACE) endpoints.

Introduction 13

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

II. Cardiovascular Outcomes trials assessing the effect of non-insulin blood-glucose-lowering agents on major cardiovascular adverse events (MACE) and mortality

II. Cardiovascular Outcomes trials assessing the effect of non-insulin blood-glucose-lowering agents on major cardiovascular adverse events (MACE) and mortality

II.1. Randomised controlled trials assessing cardiovascular outcomes before the FDA guidance

Sulfonylureas and biguanides

The first RCT for the assessment of cardiovascular effects and mortality in diabetic patients began in 1961: the University Group Diabetes Program (UGDP) was initiated as a result of a congressional request about the impact of the treatment with the first- generation tolbutamide on the cardiovascular complications of diabetes. “The UGDP was a randomised, controlled, multicenter clinical trial designed to evaluate the effectiveness of long-term hypoglycaemic drug therapy in preventing or delaying the vascular complications of diabetes (newly diagnosed, non-insulin dependent, adult-onset diabetes). The tolbutamide and phenformin treatments were terminated in 1969 and 1971, respectively, because of lack of efficacy.” [62] It was one of the first large‐scale cooperative clinical trials designed and implemented in the United States. Patients were allocated to placebo, tolbutamide, phenformin, or insulin. The study investigators concluded in 1969 that the combination of diet and tolbutamide therapy was no more effective than diet alone in prolonging life. [62, 63] Interestingly, the initiative was impulsed by a congressman who had a daughter in treatment with tolbutamide. The study was stopped eight years later because of an increase in cardiovascular deaths in those receiving tolbutamide.

Cardiovascular Outcomes trials 15

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Sulfonylureas, metformin and insulin

In the late 1970s, the UKPDS was set up in Oxford. It included more than 5102 out of 7600 subjects considered for inclusion at 23 centres across the UK. It was the most extensive study, and the median follow-up was ten years. The primary aim was to determine the effect of intensive glycaemic control on the incidence of complications; the secondary objective was to assess whether there were differences between treatments. Subjects were randomised to “conventional” (diet) or “intensive” treatment; when diet failed to achieve glycaemic targets, subjects were randomised to sulfonylureas, insulin or metformin if they were obese. [64, 36, 53] The primary outcome measures were aggregates of any diabetes-related clinical endpoint, diabetes-related death, and all-cause mortality. The results of the UKPDS 33 (3867 patients) showed that over ten years, patients in the intensive group had a reduction of HbA1c of 0.9% compared with conventional therapy (7.0% vs 7.9%) with no difference among agents in the intensive group. [36] The UKPDS 34 included 1704 overweight patients who were randomized to diet alone versus intensive blood-glucose control policy with metformin, or chlorpropamide, glibenclamide or insulin. The reduction of HbA1c was 0.6% in metformin-treated patients (7.4% vs 8.0%), and they had risk reductions of 32% for any diabetes-related endpoint, 42% for diabetes-related death, and 36% for all-cause mortality. The early addition of metformin in sulfonylurea-treated patients increased the risk of diabetes-related death compared with a continued sulfonylurea alone. [53] The 10- years post-trial monitoring showed that the benefit for glycaemic control was evident over time risk for MI (15%) and death from any cause (13%) in the sulfonylurea insulin group; in the metformin group, reductions for MI and mortality were 33% and 27%, respectively. The benefit remained even when between-group differences in glycated haemoglobin levels were lost after the first year. [55] As mentioned in the “Introduction” section, the ACCORD, the ADVANCE and the VADT trials, aiming to reach a stricter glycemic control failed to demonstrate a beneficial effect of intensive glucose lowering on CV risk. A meta-analysis indicated a modestly reduced risk of non-fatal myocardial infarction (0.85, 0.74 to 0.96), similar results concerning MI were obtained in the observation UKPDS 35 [65, 34]

Cardiovascular Outcomes trials 16

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Dual agonists of peroxisome proliferator-activated receptor: muraglitazar

Peroxisome proliferator-activated receptors (PPARs) are nuclear transcription factors that modulate gene expression, regulating glucose and fatty acid metabolism, apoptosis, angiogenesis, cell proliferation and differentiation, and immune response. Peroxisome proliferator-activated receptors gamma agonists increase insulin sensitivity (“glitazones” rosiglitazone and pioglitazone). The first dual alpha -gamma agonist was muraglitazar. In 2005, a meta-analysis of documents about phase 2 and 3 clinical trials released under public disclosure laws for the FDA advisory committee meeting evaluated the incidence of death, myocardial infarction (MI), stroke, congestive heart failure (CHF), and transient ischemic attack (TIA) in diabetic patients treated with muraglitazar compared with controls. The primary outcome was a composite of incidence of death, non-fatal MI, or non-fatal stroke; an extended composite outcome included these events plus the incidence of CHF and TIA. In the muraglitazar-treated patients, the primary outcome occurred in 1.47% patients compared with 0.67% patients in the combined placebo and pioglitazone treatment groups (controls) (relative risk 2.23; 95% CI 1.07-4.66). For the expanded MACE the RR was 2.62; 95% CI, 1.36-5.05. Components of the composite endpoint exceeded 2.1 but were not statistically significant. [66, 67]

Thiazolidinediones

FDA issued the marketing authorisation for rosiglitazone in late May 1999, and European authorities did so in July 2000 but required a post-approval clinical outcome trial, known as the RECORD (Rosiglitazone Evaluated for Cardiovascular Outcomes and Regulation of glycemia in Diabetes) trial which was published in 2009. Concerns about the safety of another thiazolidinedione, the pioglitazone, based on preclinical data, prompted that a cardiovascular safety trial was conducted, the PROActive trial.

The PROActive trial (2005) assessed the effect of pioglitazone on secondary prevention of macrovascular events in 5238 patients. Patients were followed for a mean of 2.85 years. The primary endpoint was the composite of all-cause mortality, non-fatal myocardial infarction (including silent myocardial infarction), stroke, acute coronary syndrome, endovascular or surgical intervention in the coronary or leg arteries, and amputation

Cardiovascular Outcomes trials 17

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 above the ankle; no significant results were achieved for the primary endpoint (HR, 0.90). The secondary endpoint (composite of all-cause mortality, MI and stroke) was significantly less frequent in the pioglitazone group (HR, 0.84); meanwhile the incidence of heart failure hospitalisations was higher in the pioglitazone group. In a subgroup of 2,445 patients with previous MI, pioglitazone achieved a statistically significant beneficial effect on the prespecified end point of fatal and non-fatal MI (28%) and acute coronary syndrome (ACS) (37%), but not in the primary endpoint; the incidence of heart failure and fatal heart failure were higher in the pioglitazone group. [68, 69]

The weaknesses of the design of the RECORD study (the composite of death and cardiovascular hospitalisations) and conduction (the low rate of events) have been criticised. [70] The results of an interim analysis were published in 2007 as a response to the meta-analysis of Nissen. In this meta-analysis had suggested increased CV risk for patients treated with rosiglitazone, with a significant odds ratio for myocardial infarction of 1.43 (95% confidence interval: 1.03 to 1.98, p = 0.03) and a border-line significant increase of the risk of CV mortality. [71, 72] Instead, the interim results from the RECORD study reported that rosiglitazone was associated with a small, non-significant increase in the risk of the primary outcome of all hospitalizations and deaths from CV cause (HR, 1.08; 95% CI 0.89 to 1.31), and for the fatal or non-fatal myocardial infarction outcome, the HR ratio was 1.16 (95% CI 0.75 to 1.81). [71] The sponsor did a meta- analysis with data similar to that by Nissen and Wolski had been provided to the FDA and the European Medicines Agency in August 2006, and prompted the information was included in product labels in Europe two months later. [70] Observational research using health care database found that the treatment with TZD monotherapy was associated with a significantly increased risk of congestive heart failure (adjusted rate ratio [RR], 1.60; 95% CI, 1.21-2.10), acute myocardial infarction (RR, 1.40; 95% CI 1.05-1.86), and death (RR, 1.29 95% CI 1.02-1.62) compared with other oral hypoglycemic agent combination therapies. The increased risk of congestive heart failure, acute myocardial infarction, and mortality associated with TZD use appeared limited to rosiglitazone. [73]

Cardiovascular Outcomes trials 18

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

II.2. The “post-rosiglitazone era”: regulatory guidances for new non-insulin glucose-lowering agents

Three new classes have been introduced since 2005, the glucagon-like peptide-1 (GLP- 1) receptor agonists, the dipeptidyl peptidase-4 (DPP- 4) inhibitors, and the sodium- glucose cotransporter-2 (SGLT-2) inhibitors. Exenatide b.i.d., the first GLP-1 RA, was approved in the US in 2005 and sitagliptin, the first DPP-4 i, in 2006, and one year later in the UE.[74]

In September 2010 US FDA significantly restricted the use of rosiglitazone to patients who cannot control their Type 2 diabetes on other medications, and required that GSK develop a restricted access program for Avandia (rosiglitazone) under a risk evaluation and mitigation strategy - or REMS - available to new patients only if they are unable to achieve glucose control on other medications and are unable to take pioglitazone, the only other drug in the class of thiazolidinediones. [75] FDA performed a re-evaluation of the Rosiglitazone Evaluated for Cardiovascular Outcomes and Regulation of Glycemia in Diabetes (RECORD) trial and decided to modify the rosiglitazone REMS program requirements in November 2013. [76] Rosiglitazone was withdrawn from the EU market in September 2010; the marketing authorisation for Avandia (Rosiglitazone) expired on 11 July 2015 following the decision of the marketing authorisation holder, SmithKline Beecham Ltd., not to apply for a renewal of the marketing authorisation. [77, 78]

In December 2008, the US Food and Drug Administration (FDA) issued a Guidance for Industry recommending that “to establish the safety of a new antidiabetic therapy to treat type 2 diabetes, sponsors should demonstrate that the therapy will not result in an unacceptable increase in cardiovascular risk”. At the time of NDA submission, all applicants have to compare the incidence of important CV events occurring with their investigational agent to the incidence of the same types of events in the control group. At least three major cardiovascular events (MACE) should be prospectively adjudicated: CV death, non-fatal myocardial infarction and non-fatal stroke, and can include other endpoints. This assessment can be accomplished through a meta-analysis of phase 2 and phase 3 clinical trials and/or throughout a single, large safety trial. [38]

Cardiovascular Outcomes trials 19

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

In 2012, the EMA issued guidance stating that a new glucose-lowering agent should preferably show a neutral or beneficial effect on parameters associated with cardiovascular risk(e.g. body weight, blood pressure, lipid levels), recommending that “the emphasis will be on major cardiovascular events (MACE) (CV death, non-fatal myocardial infarction and stroke) but hospitalization for unstable angina could also be included in a composite endpoint if the main objective is to exclude a safety signal. Other events, such as revascularisation and/or worsening of heart failure, will also be evaluated. [39, 79]

As a result of these regulatory recommendations, an increasing number of large randomised controlled trials have been designed and conducted to assess the impact of non-insulin glucose-lowering agents on major cardiovascular outcomes. Due to randomised allocation and double-blind design, well designed and conducted RCTs are considered the “gold standard” for scientific evidence: every patient in a study has a known (usually equal) chance of receiving each of the treatments, the selection bias is minimised, and both known (and unknown) confounding factors are likely to be distributed in an unbiased manner between the groups. Random assignment of a large number of subjects into treatment groups usually leads to a good balance of observed and unobserved risk factors in all groups. Nevertheless, randomised controlled trials have major limitations when they are used to assess the role of medications in the aetiology and management of chronic diseases. The primary limitations arise from selected populations, the long-time required from trial design to completion, the relatively short duration of exposure, and under representativeness of frail elderly patients. Results obtained from trials can be misleading if generalised to the general population because effect sizes, baseline risks, and comorbidity have been shown to differ between trial populations and the broader population not represented in trials. [80, 81] Although longer, with larger sample sizes, and including older patients, CV outcomes large trials for hypoglycemic agents are not completely free of these limitations. In particular, RCTs include selected populations (i.e., patients at high cardiovascular risk, exclusion of patients at the end stage of chronic renal disease, good treatment compliance); and patients are followed in conditions different from the clinical practice.

Cardiovascular Outcomes trials 20

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

FDA’s and EMA guidances recommend that outcomes in RCTs evaluating glucose- lowering agents for T2DM should include a 3 p. MACE (cardiovascular death, non-fatal infarction and non-fatal stroke), and possibly another expanded MACE, including unstable angina, revascularization procedures; EMA included heart failure [79] Being death the most critical clinical event, it has a very low expected rate in T2DM trials; the event rate of the rest of CV outcomes are foreseen to be low, even in high CV risk populations. Then, to reduce the sample size and the length of the study, these RCTs have a primary composite outcome (PCO) of three or four individual components: cardiovascular death, and non-fatal events of similar clinical importance. However, analysis, interpretation and reporting of COs are complex and can be even misleading. [81]

Up to date, fifteen cardiovascular outcome trials comparing drugs vs placebo have been published; an additional one, the CAROLINA trial, that assessed the safety of linagliptin vs placebo. Out of them, all those belonging to the class of dipeptidyl-peptidase -4 inhibitors showed non-inferiority vs placebo but failed to show superiority.

Apart from other studies terminated because of safety concerns (fasiglifam) and some others finished (ACE [acarbose]) or terminated (omarigliptin, taspoglutide). The trial assessing omarigliptin in patients with T2DM and CVD, OMNEON (A Study to Assess Cardiovascular Outcomes Following Treatment With Omarigliptin) was terminated because of commercial reasons; interim results showed no effect on MACE.

Peroxisome proliferator-activated receptors (PPARs)

. AleCardio: Aleglitazar is a dual agonist of PPARs with insulin-sensitising and glucose-lowering actions. The AleCardio trial enrolled 7226 patients hospitalised for acute coronary syndrome. The planned follow-up – at least 2.5 years- was terminated after a median of 104 weeks, upon recommendation of the data and safety monitoring board due to futility for efficacy and increased rates of safety endpoints (hospitalisation due to heart failure and changes in renal function). In July 2013, the sponsor announced that following the results of a regular safety review, the independent Data and Safety Monitoring Board (DSMB) has

Cardiovascular Outcomes trials 21

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

recommended halting the trial due to safety signals and lack of efficacy. The 3- point MACE was non-significant (HR 0.96). There were increased rates of heart failure and gastrointestinal bleeding and renal impairment. Heart failure is an established risk of PPAR-gamma activators and thought to be due to fluid retention. The increased risk for heart failure associated with aleglitazar in the AleCardio trial (HR, 1.22) was similar to that attributed to pioglitazone in a meta- analysis (HR, 1.41) Increased serum creatinine is also a known effect of PPAR- alpha activators and was associated with aleglitazar in this trial. [82-84]

Dipeptidyl peptidase-4 inhibitors

The incretin-based therapies include the oral dipeptidyl peptidase 4 inhibitors (DPP-4 i) and glucagon-like peptide-1 receptor (GLP-1R) agonists. While GLP-1RAs exert glucoregulatory actions by binding to GLP-1 receptors, DPP-4 i prevent inactivation of GLP-1.

Four CVOTs assessed DPP-4 inhibitors vs placebo: TECOS (sitagliptin)], EXAMINE, (alogliptin), SAVOR-TIMI 53 (saxagliptin) and CARMELINA (linagliptin). None of these trials has shown to reduce the risk of MACE in the treatment group. Saxagliptin has shown a significant increased frequency of heart failure and alogliptin a non-significant increased risk of HF. Table II.1 shows the characteristics and results of the CVOTs assessing DPP-4 inhibitors vs placebo.

. TECOS (sitagliptin): Sitagliptin was the first marketed dipeptidyl-peptidase inhibitor, agent approved by the US FDA in October 2006; the European Commission granted a marketing authorisation valid throughout the European Union in March 2007. The TECOS evaluated long-term effects on cardiovascular outcomes of sitagliptin, or placebo added to existing therapy in 14,671 patients aged ≥50 years with glycated haemoglobin level, 6.5 to 8.0%, established CVD and no severe renal insufficiency, the median follow-up was 3 yrs. In the TECOS Study, sitagliptin was non-inferior to placebo for the primary 4-points MACE (+ and hospitalisation for unstable angina), HR, 0.98; 95% CI, 0.88 to 1.09, or hospitalization for heart failure.[85]

Cardiovascular Outcomes trials 22

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table II.1: Characteristics of cardiovascular outcomes trials (CVOTs) assessing the effects of dipeptidyl peptidase-4 inhibitors (DPP-4 i) vs placebo

CVOTs N Follow- MACE All- CV AMI Stroke HHF up Cause Mortality DPP-4 i Mortality Median

TECOS 14,671 3.0 yrs 0.98 (0.89– 1.01 1.03 0.95 0.97 1.00 [85] 1.08) (4 p) (0.90– (0.89– (0.81– (0.79– (0.83– Sitagliptin 1.14) 1.19) 1.11) 1.19) 1.20) 0.99 (0.89– 1.10) (3-p)

EXAMINE 5,380 1.5 yrs 0.96 (≤1.16) 0.88 0.85 1.08 0.91 (non- [86, 87] (0.71– (0.66– (0.88– (0.55– significant Alogliptin 1.09) 1.10) 1.33) 1.50) increase)

SAVOR- 16,492 2.1 yrs 1.00 (0.89– 1.11 1.03 0.95 1.11 1.27 TIMI 53 1.12) (3-p) (0.96– (0.87– (0.80– (0.88– (1.07– [88] 1.27) 1.22) 1.12) 1.39) 1.51) Saxagliptin 1.02 (0.94– 1.11)

CARMELINA 6,991 2.2 yrs 1.02 (0.89- 0.98 0.96 1.12 0.91 0.90 [93] 1.17) (0.84- (0.81- (0.90- (0.67- (0.74- Linagliptin 1.13) 1.14) 1.40) 1.23) 1.08)

CVOTs: cardiovascular outcomes trials; DPP-4 i: dipeptidyl peptidase-4 inhibitors; MACE: major adverse cardiovascular events (composite outcome); 5-p: 5 points MACE; 4-p: 4 points MACE; 3-p: 3-points MACE. CV: cardiovascular; AMI: acute myocardial infarction; HHF: hospitalisation for heart failure. All subjets were randomized 1:1 to investigational product and placebo. Significant results are highlighted in bold.

. EXAMINE (alogliptin): Alogliptin is a selective DPP-4 i, approved for the treatment of type 2 diabetes in January 2013 in the US and in September 2013 in EU. Examination of Cardiovascular Outcomes with Alogliptin versus Standard of Care (EXAMINE study) assessed the primary 3-point MACE in 5380 T2DM patients with an acute coronary syndrome (ACS) within the previous 15 to 90 days, and showed no difference between groups, although the glycated haemoglobin levels were significantly lower with alogliptin than with placebo.[86]

Cardiovascular Outcomes trials 23

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

The exploratory extended MACE endpoint (plus urgent revascularisation due to unstable angina, and hospital admission for heart failure) did not show differences, HR 0.98, 95% CI 0.86–1.12, either the hospital admission for heart failure HR 1.07, 95% CI 0.79–1.46).65 Alogliptin had no effect on composite events of cardiovascular death and hospital admission for heart failure in the post hoc analysis (HR 1.00, 95% CI 0.82–1·21) and results did not differ by baseline BNP concentration. Patients with a history of heart failure at baseline were older, more frequently women, and had higher baseline BNP concentrations and lower eGFR values, than patients with no history of heart failure. [87]

. SAVOR-TIMI 53 (saxagliptin): Saxagliptin is DPP-4 inhibitor approved in July 2009 in the US and in October 2009 in EU. The SAVOR-TIMI 53 trial [88] included 16,492 patients with T2DM, HbA1c 6.5% to 12.0%, and either a history of established cardiovascular disease (78%) or multiple risk factors for vascular disease; the follow-up had a median of 2.1 years. Results showed neutral effects of saxagliptin on primary composite of 3-point MACE, HR, 1.00; 95% CI 0.89 to 1.12, as well as on the major secondary 5-point MACE (plus hospitalization for unstable angina, coronary revascularization, or heart failure) HR 1.02; 95% CI 0.94 to 1.11; P = 0.66.

However, hospitalization for heart failure was more frequent in the saxagliptin group than in the placebo group (3.5% vs 2.8%; hazard ratio, 1.27; 95% CI, 1.07 to 1.51). [89] These results were consistent irrespective of the renal function. Overall, the risk of hospitalisation for heart failure among the three eGFR severity groups of patients was 2.2% (reference), 7.4% (adjusted HR 2.38), and 13.0% (adjusted HR 4.59), respectively. The relative risk of hospitalisation for heart failure with saxagliptin was similar in patients with different levels of eGFR. Saxagliptin and placebo groups showed similar results in the change in eGFR and safety renal outcomes, including doubling of serum creatinine, initiation of chronic dialysis, renal transplantation, or serum creatinine >6.0 mg/dL. However, patients with renal impairment who were treated with saxagliptin achieved similar reductions in microalbuminuria than those of the overall trial population. [90] The change in albumin/creatinine ratio (ACR) did not correlate with that in HbA1c.

Cardiovascular Outcomes trials 24

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[91] In the SAVOR TIMI 53 trial, baseline HbA1c ≥7% was associated with increased risk of cardiovascular death, myocardial infarction, or ischemic stroke (adjusted hazard ratio 1.35; 95% CI 1.17-1.58) but not with hospitalisation for heart failure (adjusted HR 1.09; 95% CI, 0.88-1.36). [92] . CARMELINA (linagliptin): The trial CARMELINA trial included 6979 patients (mean age, 65.9 years; eGFR, 54.6 mL/min/1.73 m; 80.1% with renal impairment), the median follow-up was 2.2 years. The HR for the 3-points MACE was 1.02; 95% CI, 0.89-1.17. No differences were observed for the kidney outcome (time to first occurrence of adjudicated death due to renal failure, end- stage renal disease, ESRD, or sustained 40% or higher decrease in eGFR from baseline) HR, 1.04; 95% CI, 0.89-1.22. No difference was found in hypoglycemia, but there were more cases of confirmed acute pancreatitis in the linagliptin group.[93]

Glucagon-like peptide 1 receptor agonists (GLP-1 Ras)

Glucagon-like peptide-1 (GLP-1 potentiates the insulin secretion from pancreatic beta cells and lowers inappropriate high glucagon secretion in a glucose-dependent manner; it also has effects in extrapancreatic tissues (gastrointestinal tract, heart, vasculature, and central and peripheral nervous system). Seven studies of the class of the glucagon-like peptide-1 receptor agonists: LEADER (liraglutide), SUSTAIN-6 (semaglutide), HARMONY (albiglutide) and the REWIND (dulaglutide) trials showed benefits on MACE; the PIONEER trial (oral semaglutide), the EXSCEL trial (exenatide) and the ELIXA (lixisenatide) showed non-inferiority but no beneficial effects on CV outcomes. Table II.2 shows the characteristic and results of CVOTs assessing GLP-1 RAs vs placebo.

. ELIXA: Lixisenatide is a GLP-1 RA, with a short half-life (i.v. 30 min and 2−3 h after s.c. administration. The ELIXA study included 6068 patients with acute coronary syndrome within the previous 180 days, mean follow-up was 2.1 years. The intervention showed neutral results on the 4-points MACE (+ unstable angina) and in its components or heart failure. [94]

Cardiovascular Outcomes trials 25

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table II.2: Characteristics of cardiovascular outcomes trials (CVOTs) assessing effects of glucagon-like peptide 1 receptor agonists (GLP-1 RAs) vs placebo

CVOTs n Follow- MACE All-Cause CV AMI Stroke HHF GLP-1 RAs up Mortality Mortality Median

ELIXA 6,068 2.1 yr. 1.02 (0.89- 0.94 (0.78- 0.98 (0.78– 1.03 (0.87– 1.12 (0.79– 0.96 (0.75 - [94] 1.17) 1.13) 1.22) 1.22) 1.58) 1.23) Lixisenatide

LEADER 9,340 3.8 yr. 0.87 * 0.85* 0.78* 0.86 (0.73– 0.86 (0.71– 0.87 (0.73– [95] (0.78- 0.97) (0.74- 0.97) (0.66- 0.93) 1.00 1.06) 1.05) Liraglutide

SUSTAIN-6 3,297 2.1 yr. 0.74 * 1.05 (0.74– 0.98; (0.65 0.74 (0.51- 0.61 (0.38 1.11 (0.77– [96] (0.58- 0.95) 1.50) 1.48) 1.08) -0.99) 1.61) Semaglutide

EXSCEL 14,752 3.2 yr. 0.91 (0.83- 0.86* 0.88 0.97 0.85 0.94 [97] 1.00) (0.77−0.97) (0.76−1.02) (0.85−1.10) (0.70−1.03) (0.78−1.13) Exenatide LAR

HARMONY 9,463 1·6 yr. 0.78 * 0.95 (0.79– 0.93 (0·73– 0.75 * 0.86 (0.66– prior HF [98] (0·68–0·90) 1.16) 1·19) (0.61–0.90) 1.14) [0.70 (0.54- 0.90)]* Albiglutide

PIONEER-6 3,183 1.32 0.79; (0.57 0.51* 0.49 * 1.18 (0.73 - 0.74 (0.35- 0.86 (0.48– to 1.11) (0.31- 0.84) (0.27-0.92) 1.90) 1.57) 1.55) [99] Oral semaglutide

REWIND 9,901 5.4 yr. 0.88 * 0.90 (0.80– 0.91 (0.78– 0.96 (0.79– 0.76 * 0.93 (0.77– (0.79–0.99) 1.01) 1.06) 1.15) (0.62–0.94) 1.12) [100] [5.1– Dulaglutide 5.9]

CVOTs: cardiovascular outcomes trials; GLP-1 RAs: glucagon-like peptide 1 receptor agonists; MACE: major adverse cardiovascular events (composite outcome4-p: 4 points MACE; 3-p: 3-points MACE. CV: cardiovascular; AMI: acute myocardial infarction; HHF: hospitalisation for heart failure. Significant results are highlighted in bold. Values for follow-up are median. Participants in all trials were randomized 1 :1 to investigational drug or placebo. * no history of HF [0.82 (0.69-0.98)]

Cardiovascular Outcomes trials 26

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

. The LEADER (liraglutide) trial assessed the effects of liraglutide, a GLP-1 analogue administered once daily. The LEADER assessed the 3-point MACE in 9340 patients with high CV risk T2DM; the median follow-up was 3.8 years. The primary outcome for the liraglutide group was significant lower (HR, 0.87 95% CI 0.78-0.97), mainly due to a lower rate in CV death in the liraglutide group (hazard ratio, 0.78). Non-fatal MI and non-fatal stroke were non-statistically lower in the liraglutide group: HR, 0.88 95% CI 0.71-1.03 and HR, 0.89 95% CI 0.72-1.11 respectively. The rate of death from any cause was lower in the liraglutide group (HR, 0.85 95% CI 0.74-0.97). Hospitalisation for heart failure (HHF) was non-significantly reduced in the liraglutide arm (HR, 0.87 95% CI 0.73-1.05).The rate of composite outcome of renal or retinal microvascular events was lower in the liraglutide group than in the placebo group (hazard ratio, 0.84), driven by a lower rate of nephropathy events in the liraglutide group (hazard ratio, 0.78), retinopathy events were non significantly higher in the liraglutide group than in the placebo group (0.6 vs 0.5 events per 100 patient-years; HR, 1.15). [95]

. SUSTAIN-6: The SUSTAIN study was a non-inferiority large RCT assessing the effect of semaglutide, a long-acting glucagon-like peptide 1 (GLP-1) analogue with an extended half-life of approximately 1 week, on cardiovascular outcomes vs placebo 3297 Type 2 diabetic patients. The primary composite outcome was the 3 points MACE, HR 0.74; 95% CI 0.58 to 0.95. Nonfatal myocardial infarction was reduced, HR 0.74; 95% CI, 0.51 to 1.08; as well as nonfatal stroke (HR 0.61; 95% CI, 0.38 to 0.99. Rates of death from cardiovascular causes were similar in the two groups. [96]

. The EXSCEL (exenatide LAR) included 14,752 patients, 73.1% with a previous CVD; the median follow-up was 3.2 years. Patients treated with exenatide and placebo group did not show differences in the 3-point MACE, HR 0.91; 95% CI 0.83 to 1.00. The rates of death from cardiovascular causes, fatal or non-fatal myocardial infarction, fatal or non-fatal stroke, hospitalisation for heart failure, and hospitalization for acute coronary syndrome, and the incidence of acute

Cardiovascular Outcomes trials 27

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

pancreatitis, pancreatic cancer, medullary thyroid carcinoma, and serious adverse events did not differ significantly between the two groups. [97]

. The HARMONY (albiglutide) trial included 9463 participants with CVD; the median follow-up was at least 1.6 years. Albiglutide was superior to placebo in reducing the risk of 3 p. MACE (HR 0.78, 95% CI 0·68-0·90).[98]

. The PIONEER 6 (oral semaglutide) trial assessed the 3-point composite MACE in 3183 patients with high cardiovascular risk (mean age 66 years; 84.7% were 50 years of age or older and had cardiovascular or chronic kidney disease; mean follow-up was 15.9 months) The primary MACE for semaglutide vs placebo was non-significant, HR 0.79; 95% CI 0.57 to 1.11. For components of the MACE, death from cardiovascular causes, HR 0.49; 95% CI, 0.27 to 0.92; and death from any cause (HR 0.51; 95% CI, 0.31 to 0.84) were significantly reduced, while non- fatal MI and non-fatal stroke showed no differences. [99]

. The REWIND (dulaglutide) trial recruited 9901 patients who had either a previous cardiovascular event or cardiovascular risk factors, 31.5 % of patients had a history CVD, 22.2% had a baseline renal impairment, and 8.5% had a history of heart failure; mean HbA1c 7·2%; the mean follow-up was 5.4 years. Patients treated with dulaglutide had decreased risk of the primary outcome, the 3-point MACE, HR 0.88, 95% CI 0.79-0.99 and stroke HR 0.76 (0.62–0.94. All- cause and CV mortality, MI and HF did not differ between groups. [100] The exploratory analysis showed improved renal outcomes in patients treated with dulaglutide: HR 0.95, 95% CI 0.77-0.93, with a clear effect for microalbuminuria, HR 0.77, 95% CI 0.68–0.87. More participants assigned to dulaglutide reported a gastrointestinal adverse event during follow-up compared with participants assigned to placebo. [101]

Sodium-glucose co-transporter 2 inhibitors (SGLT-2)

The SGLT1 and SGLT2 are primarily responsible for intestinal glucose absorption and reabsorption of most of the filtered glucose in the kidney. The tree studies belonging to the class of sodium-glucose cotransporter 2 inhibitors: EMPA-REG (empagliflozin),

Cardiovascular Outcomes trials 28

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

CANVAS (canagliflozin) and DECLARE-TIMI 58 (dapagliflozin) showed reduction in heart failure outcomes. For the group of glucose co-transporter-2 inhibitors (SGLT-2 i), empagliflozin (EMPA-REG outcome) has shown a substantial relative risk reduction of death from cardiovascular causes, hospitalisation for heart failure and all-cause mortality. The hematocrit increases during treatment with SGLT2 inhibitors, which have a diuretic effect but do not cause sufficient hemoconcentration to increase the risk of cerebral infarction. Elevation of the hematocrit during SGLT2 inhibitor therapy is presumed to involve enhancement of erythropoiesis in addition to hemoconcentration.[102] The increase in the hematocrit and an HR of 1.33 for stroke have been found in a meta-analysis of RCTs with SGLT2 inhibitors. [103, 104] The characteristics and results of the CVOTs assessing sodium-glucose cotransporter-2 inhibitors vs placebo are shown in Table III.3.

Table III.3: Characteristics of cardiovascular outcomes trials (CVOTs) assessing effects of sodium-glucose cotransporter-2 inhibitors (SGLT-2 i) vs placebo

CVOTs N Follow- MACE All-Cause CV AMI Stroke HHF up Mortality Mortality SGLT-2 i Median

EMPAREG 7020 3.1 0.86 (0.74– 0.68 (0.57– 0.62 (0.49– 0.87 (0.70– 1.18 (0.89– 0.65 (0.50– OUTCOMES (1:1:1) 0.99) 0.82) 0.77) 1.09) 1.56) 0.85) [105, 106] Empagliflozin CANVAS 10,142 2.34 0.86 (0.75– 0.87 (0.74– 0.87 (0.72– 0.89 (0.73– 0.87 (0.69– 0.67 (0.52– 0.97) 1.01) 1.06) 1.09 1.09) 0.87) [107] (1:1) Canagliflozin DECLARE- 17,160 4.2 0.93 0.93 0.98 0.89 1.01 0.73 TIMI 58 (1:1) (0.84−1.03) (0.82−1.04) (0.82−1.17) (0.77−1.01) (0.84−1.21 (0.61−0.88) [108, 109] Dapagliflozin

CVOTs: cardiovascular outcomes trials; SGLT-2 i: sodium-glucose cotransporter-2 inhibitors; MACE: major adverse cardiovascular events (composite outcome4-p: 4 points MACE; 3-p: 3-points MACE. CV: cardiovascular; AMI: acute myocardial infarction; HHF: hospitalisation for heart failure. Significant results are highlighted in bold. Randomisation is shown in n column, as (1 :1), empagliflozin has three arms : placebo, empagliflozin 10 mg and empagliflozin 25 mg.

Cardiovascular Outcomes trials 29

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

. The EMPA-REG outcomes (empagliflozin) study assessed the effect of empagliflozin, in 7020 patients with a high risk for cardiovascular events; the median follow-up time was 3.1 yrs. The primary outcome was the 3-point MACE; the key secondary composite outcome was the primary outcome plus hospitalisation for unstable angina. Empagliflozin showed a reduction of CV death (HR 0.62, 95% CI 0.49-0.77), hospitalisation for HF (HR 0.65, 95% CI 0.50-0.85), and death from any cause (HR 0.68 95% CI 0.57-0.82), but not for the rates of fatal and non-fatal MI (HR 0.87 95%CI 0.87-1.09) or fatal and non-fatal stroke (HR 1.18, 95% CI 0.89-1.56). 10.1% had heart failure at baseline. The composite of heart failure hospitalisation or cardiovascular death (HR: 0.66, 95% CI 0.55–0.79), hospitalisation for heart failure (HR: 0.65, 95% CI 0.5–0.85), CV death (HR 0.62 95% CI 0.49–0.77), and all-cause mortality was lower in patients treated with empagliflozin than with placebo. Still, patients with HF at baseline had smaller non-significant reductions. [105, 106]

. CANVAS (canagliflozin) The CANVAS Program integrated data from two trials involving a total of 10,142 participants (mean age 63.3 years, 35.8% women, mean diabetes duration was 13.5 years, 65.6% had a history of cardiovascular disease. The primary outcome was a composite of 3-point MACE. The rate of the primary outcome was lower with canagliflozin than with placebo (HR 0.86; 95% CI 0.75 - 0.97). Canagliflozin showed a possible benefit of canagliflozin concerning the progression of albuminuria (HR 0.73 [95% CI 0.67-0.79]) and the renal composite (HR 0.60, 95% CI 0.47-0.77); an increased risk of amputations has also reported. [107]

. The DECLARE-TIMI 58 trial included 17, 160 patients; 60% didn’t have atherosclerotic CV disease, median follow-up was 4.2 years. Dapagliflozin did not result in a lower rate of MACE but showed a reduction in CV death, HR 0.83 95% CI 0.73 to 0.95, which reflected a lower rate of hospitalization for heart failure (HR 0.73; 95% CI, 0.61 to 0.88) By 10% of population included had heart failure. [108, 109]

Cardiovascular Outcomes trials 30

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Conclusion

This summary shows that most of the CVOTs assessing non-insulin blood glucose- lowering drugs include high CV risk patients, who, in some trials, are the 100%. The inclusion of high CV risk patients is aimed to achieve the goal of showing absence of negative CV outcomes in the shortest possible time. [110, 111] The EMA guidance states that “the study population included in the studies supporting the marketing authorisation application should resemble the target population for the medicinal product. However, a higher representation of subjects with a high baseline risk for cardiovascular diseases and complications (…) compared to the target population may be acceptable for assessing cardiovascular safety profile regardless of whether a meta-analytic or a dedicated cardiovascular outcome trial approach is used.” [79]

These large double-blind, randomised, event-driven RCTs have adopted a non-inferiority design – in some cases, superiority is tested.

Excepted the CAROLINA trial (that assessed CV outcomes in patients treated with linagliptin or glimepiride), all the investigational products are compared to placebo in addition to standard care.

All the compounds of the class of DPP-4 i showed no effects on MACE, but some agents (saxagliptin, alogliptin) were associated with an increase in the risk of heart failure. Conversely, because of the mechanism of action of SGLT-2, all these agents show benefits in terms of heart failure risk; however, the trials had different results. The differences in results might be caused by characteristics in populations (lower percentage of high CV risk patients in the dapagliflozin CVOT), but also by drugs’ characteristics. [112] In the class of GLP-1 RA, three agents have achieved reductions in risk of MACE: liraglutide, albiglutide and semaglutide; all are long-acting GLP-1 RA.

Liraglutide and empagliflozin received a US FDA approval for the reduction of the risk of cardiovascular death in adult patients with type 2 diabetes mellitus and cardiovascular disease prevention. [113, 114] The question arises if the beneficial effect achieved in high CV risk patients can be obtained when treating unselected populations.

Cardiovascular Outcomes trials 31

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

III. Generalizability of Cardiovascular Outcome Trials to the Real World: Implications for Clinical Practice III. Generalizability of Cardiovascular Outcome Trials to the Real World: Implications for Clinical Practice

Randomised clinical trials (RCTs) are considered the gold standard for assessing the efficacy of treatments because randomisation equally distributes known and unknown factors among control and intervention groups, reducing the potential for confounding: the different groups are therefore comparable and, if the study is sufficiently powered, the effect of the intervention can be identified.[80] RCTs generate evidence on the benefits and harms of therapeutic interventions and are the primary basis for many regulatory decisions – the first and the more important one is the marketing authorisation - and clinical guidelines.

Since 2008, large trials evaluating clinically significant CV outcomes (cardiovascular outcomes trials, CVOTs) have provided insights on the CV safety of new non-insulin blood glucose-lowering agents (NIADs, non-insulin “antidiabetic” drugs). [38, 115] In CVOTs, even if they have large sample size and a long follow-up period, mortality and events such as MI and stroke occur at a relatively low incidence rate. Because of this, CVOTs need to gather significantly clinical events in one primary composite endpoint; this allows sufficient power for evaluation. Most CVOTs assessing the effect of NIADs on Type2 diabetes mellitus population (T2DM) have a primary composite outcome of three-points (i.e. myocardial infarction, stroke and CV death) or four-points major adverse cardiovascular events (usually including unstable angina). [116 -118]

III. Generalizability of Cardiovascular Safety Trials 33

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 As the CV events included in the primary composite outcome are relatively rare, CVOTs generally include patients with a high risk of CV and with an expected high rate of CV events (e.g. patients with prior ischemic events occurred in a defined time) to ensure that the trial will have sufficient statistical power and to achieve the required number of events in the shortest time – usually more than two years. In CVOTs, many patients had established CVD, which ranged from 41% in DECLARE-TIMI (dapagliflozin) to 100% in EXAMINE (alogliptin), ELIXA (lixisenatide), EMPA‐REG OUTCOME (empagliflozin), and HARMONY (albiglutide). [108, 86, 94, 105, 98] These selected populations limit the generalizability of the results and prevent the assessment of primary prevention. [119] For instance, the rate of events in patients with a recent MI or ACS is much higher than in the general population, [120-122] As this high CV-risk population represents only a small part of the general T2DM population, the external validity of study results should be considered restricted to those who could have been enrolled in the study.

Based on the beneficial results of CVOTs, some agents have received the approval to reduce the risk of MACE, or death in patients with type 2 diabetes who have established CVD or CV risk factors, consistently to the populations included in the respective CVOTs: empagliflozin to reduce the risk of CV death (2016), liraglutide to reduce MACE (2017), canagliflozin to reduce CV events (2018) and dapagliflozin to reduce the risk of hospitalisation for heart failure (2018) and to reduce the risk of end-stage kidney disease, worsening of kidney function, heart-related death, and being hospitalised for heart failure in certain patients with type 2 diabetes and diabetic kidney disease (2019). To December 2019, other applications are still pending (injectable semaglutide to reduce the risk of major adverse cardiovascular events (MACE) and dulaglutide for the reduction of major adverse cardiovascular events (MACE) in adults with type 2 diabetes who have established cardiovascular (CV) disease or multiple cardiovascular risk factors.

The generalisability of RCTs’ results to patients in daily clinical practice is a significant issue both for clinical decision making and the incorporation of a given agent, or a class, in updated clinical guidelines. [123] Despite their similar characteristics in study design and definition of primary endpoints, CVOTs assessing agents of the same class of NIADs often differ in their inclusion/exclusion criteria. These different criteria impact on generalizability.

III. Generalizability of Cardiovascular Safety Trials 34

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

A meta-analysis of CVOTs that analysed results separately in patients with or without preexisting cardiovascular disease found that, compared with placebo, the use of both glucagon-like peptide-1 receptor agonists and sodium-glucose cotransporter-2 inhibitors was associated with a significant 14% lower MACE risk only in patients with preexisting cardiovascular disease at baseline; instead, the results were neutral for those patients without cardiovascular disease at baseline. [119]

The comparability of populations included in CVOTs and the “real-world” populations that could benefit from the good results of these trials is not only a clinical but also an economic issue. The cost of newer therapies as SGLT-2 i and GLP-1 RA are roughly 5 to 10-fold more expensive than older treatments. [124] For general populations, the benefit in terms of hospitalisations, adverse events (cardiovascular or not) and mortality remain to be assessed. When analysing hospital admissions costs, the differences in the event rates for MI, stroke or death for high CV risk populations and unselected populations should be taken into account, to not to oversize cost estimations. An economic-model analysis for “real-world” healthcare costs, but based on RCTs’ populations might be misleading. [125]

Follow-up duration should also be considered in the analysis of benefit because in primary care, patients usually maintain their treatments for much more extended periods than in CVOTs, and even observational studies.

III.1. Observational studies assessing the comparability of real- world populations with CVOTs’ populations

Glucagon-like peptide 1 receptor agonists

Hinton et al. conducted with data of the UK Royal College of General Practitioners (RCGP) Research and Surveillance Centre (RSC) network database. In this study, only 16.6% out of 84 394 T2DM patients met the inclusion criteria for established or high-risk CV disease of the LEADER trial (liraglutide). In the general population studied, patients were older (73.2 vs 64.2 years), had lower mean glycated haemoglobin (67.1 vs 71.6 mmol/mol) and mean body mass index (30.9 vs 32.5). [126]

III. Generalizability of Cardiovascular Safety Trials 35

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Boye et al. estimated the proportions of individuals in the reference population (US IQVIA Real-World Data Adjudicated Claims database) weighted by data from the National Health and Nutrition Examination Survey (NHANES), represented by subjets in the CVOTs for age, sex, body mass index (BMI), HbA1c, eGFR category, and prior MI. The cohort included 113,079 T2DM patients. Based on inclusion/exclusion criteria, 42.6% of the US reference population were eligible for enrolment in REWIND (dulaglutide), versus 15.9% in EXSCEL (exenatide LAR), 13.0% in SUSTAIN-6 (semaglutide), and 12.9% in LEADER (liraglutide). [127]

Wittbrodt et al. compared data from the US population National Health and Nutrition Examination Survey (NHANES) with published eligibility of CVOTs evaluating GLP-1 RAs, and estimated the percentage of US patients who would have met eligibility criteria for enrollment in the GLP-1 RAs CVOTs: 47.2% of US population fulfill criteria for EXSCEL, 22.4% for REWIND, 15.5% for FREEDOM-CVO, 12.8% for LEADER, 11.8% for SUSTAIN-6, 8% for HARMONY (albiglutide) and 6.4% for the ELIXA trial. [128]

Sodium-glucose co-transporter inhibitors

In European countries, there were consistent patterns of representativeness of CVOTs for enrolment criteria. The DECLARE-TMI 58 (dapagliflozin) trial had the highest representativeness, indicating that it included and examined patients who are most representative of the general T2D patients in the studied countries. [129]

Wittbrodt et al. conducted a cross-sectional retrospective study to evaluate the proportions of US adults with T2D meeting the eligibility criteria for each of the 4 sodium-glucose cotransporter-2 (SGLT2) inhibitor CVOTs. Data were extracted from the National Health and Nutrition Examination Survey, (NHANES) in the periods 2009- 2010 and 2011-2012. Weighted analysis of T2D diagnosis and other relevant clinical and demographic characteristics was used to estimate the percentage of US adults with T2D who met the eligibility criteria for the CANVAS program (CANagliflozin cardioVascular Assessment Study) (canagliflozin); and the DECLARE-TIMI 58 (dapagliflozin), EMPA- REG OUTCOME (empagliflozin), and VERTIS-CV (ertugliflozin) trials. 23,941,512 US

III. Generalizability of Cardiovascular Safety Trials 36

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 adults from data on key inclusion criteria and information indicating a diagnosis of T2D. Of these, 4.1% met the criteria for EMPA-REG OUTCOME, 4.8% for VERTIS-CV, 8.8% for the CANVAS program, and 39.8% for the DECLARE-TIMI 58 trial. [130]

Birkeland et al. conducted a cross-sectional study in four European countries. In 2015, a total T2D population of 803 836 patients was identified in Germany (29.79%), The Netherlands (4.51%), Norway (18.63%) and Sweden (47.07%). These cohorts showed that the general population had 25% to 44% cardiovascular (CV) disease baseline prevalence and high CV-preventive drug use (>80%). In brief, Type2 DM unselected population had less prevalent CV disease, and patients were slightly older than those included in the CVOTs. As in other studies assessing generasibility, the DECLARE-TIMI 58 trial had the highest representativeness, 59% compared to the general T2D population, 2-, 3- and 4-fold higher compared to the CANVAS (34%), EMPA-REG OUTCOME (21%) and VERTIS-CV (17%) trials, respectively. [123]

Nicolucci et al. conducted a study in 222 Italian clinics, with 455,662 adult patients with T2DM. Among the population identified, the proportion of patients meeting major eligibility criteria was 11.7% for EMPA-REG OUTCOME, 29.4% for CANVAS, 55.9% for DECLARE-TIMI 58, and 12.8% for VERTIS-CV. “Real-world” patients were older, had longer diabetes duration, lower BMI and HbA1c levels, lower prevalence of the established cardiovascular and cerebrovascular disease, and higher rates of microvascular complications and peripheral arterial disease than those participating in the CVOTs, eligible patients. [131]

III.2. Characteristics of the population registered in the Catalan SIDIAP database compared with CVOTs’ populations.

The Catalan Institute of Health (CIH) manages the main healthcare system of the autonomous community of Catalonia, Spain, at all its levels. 76% of the Catalan population attends at least once the primary care teams. General practitioners participating in SIDIAP (Information System for the Development of Research in Primary Care) record patients’ clinical, laboratory, diagnostic and medical procedures, referrals data and prescriptions, among others. The state CIH funds the prescribed

III. Generalizability of Cardiovascular Safety Trials 37

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 medicines in part (for workers or employees) or for full (for retired people). The population included in the SIDIAP database is approximately 74% of the total regional one; being representative of the Catalan population.

In the second part of this work, the design, methods and results of an observational study assessing the CV effects of the more used non-insulin blood-glucose-lowering agents during 2010-2015 in the SIDIAP database are presented. Basal characteristics of the total were calculated using descriptive statistics. Data for demographic (age and sex) and clinical variables (HbA1c, diabetes duration, history of CV disease, history of heart and renal failures) are shown in Table III.1.

This table presents the characteristics of the full cohort (any patient treated with at least one non-insulin blood-glucose-lowering agent), the cohort of drug-naïve patients treated with a single NIAD in monotherapy, the cohort of patients treated with two or more NIADs. To compare these populations’ characteristics with those of the cohorts included in CVOTs, Table 1 shows the basal characteristics of patients included in CVOTs of different classes of new NIADs.

When comparing basal characteristics of populations included in non-insulin blood- glucose-lowering therapies CVOTs and the general population registered in the primary care SIDIAP database, populations of five CVOTs have a substantially higher proportion of men (by 70%), in seven CVOTs the proportion of men ranges from 60.7% to 66.6%; in the PIONEER 6 trial (oral semaglutide) the proportion is slightly higher than the SIDIAP population prescribed two or more NIADs, and in the REWIND trial (dulaglutide) this proportion is lower. Except for the EXAMINE trial, the populations included in all CVOTs had a longer duration of diabetes mellitus. In five CVOTs all participants had established CVD, in the other studies the proportion of participants with established CVD were much higher than in SIDIAP populations, ranging from 31.5% (REWIND trial) to 84.6%. Participants of CVOTs had from 2-fold up to 5-fold- higher history of HF (from 9.9 to 26.8% vs 5.1%). (Table III.1)

III. Generalizability of Cardiovascular Safety Trials 38

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table III.1. Basal characteristics of cohorts of T2DM patients registered in SIDIAP database and cohorts of CVOTs assessing non-insulin blood-glucose-lowering drugs vs placebo

SIDIAP database 2010-2015 Men % Mean age, HbA1C T2DM BMI (SD) History CVD History HF Renal Failure years (SD) % (SD) duration % % % Years (SD) SIDIAP 55.5 % 66.1 (12.7) 7.9 (1.6) 5.4 (5.7) 30.8 (5.4) 21.2 % 4.8 % 2.6 % Full cohort SIDIAP 51.1 % 65.7 (13.2) 7.6 (1.6) 4.0 (5.3) 31 (5.4) 20.4 % 5.1 % 2.8 % Monotherapy users SIDIAP 56.2 % 66.8 (11.7) 8.3 (1.5) 7.8 (5.6) 30.5 (5.3) 22.4 % 4.3 % 2.3% ≥ 2 NIADs users Cardiovascular Outcomes Trials (CVOTs) Men % Mean age, HbA1C T2DM BMI (SD) History CVD History HF Renal Failure years (SD) % (SD) duration % % % Years (SD) EMPAREG 72 % 63.1 (8.6) 8.1 (0.8) ≤5 yrs: 18% 30.6±5.3 99% 10.1% 26% OUTCOMES ≥ 75 yrs: 9% >5-10 yrs: 25 eGFR <60: [105, 106] % 25.9% Empagliflozin >10 yrs 57 % CANVAS 64.2 % 63.3 ±8.3 8.2±0.9 13.5±7.8 32.0±5.9 65.6% 14.4 % 17.21% [107] eGFR 76.5± Canagliflozin 20.5 DECLARE 63.1 % 63.9±6.8 8.3±1.2% 11.0 32.1±6.0 40.6% CVD 9.9% eGFR TIMI 59.4% risk 85.4±15.8 [108, 109] factors Dapagliflozin ELIXA 69.7% 59.9 ± 9.7 7.6±1.3 9.4±8.3 30.2±5.8 100% 22.4% eGFR [94] 76.7±21.3 ≥65 yrs 33% Lixisenatide <60: 49.1%

III. Generalizability of Cardiovascular Safety Trials 39

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

LEADER 64.3% 64.3 ± 7.2 8.7 ± 1.5 12.7 ± 8.0 32.5 ± 6.3 85.1% CVD 17.12% 26.7 % [95] ≥60 yrs 75% 14.9 % risk eGFR <60: Liraglutide factors 21.8 % SUSTAIN-6 60.7% 64.6 ± 7.4 8.2±1.6 14.9 ± 8.5 32.3 ± 6.5 83% 23.6% NA [96] Semaglutide EXSCEL 62% 62.7 (56.4, 8.0 (7.3, 8.9) 12 (7, 18) 31.8 [28.2- 73.1% 16.2% eGFR <60: [97] 68.8 36.2] 18.6 % Exenatide LAR ≥65 yrs: 40.25% HARMONY 70 % 64.1 8.76 (1.5) 14.1 (8.6) 32.3 ±5.9 100% 20% eGFR 79.1 [98] (mean) Albiglutide PIONEER-6 58.4 % 66.1 (7.1) 8.2 (1.6) 14.9 (8.5) 32.3 ± 6.5 84.6% NR eGFR 74.2 ±21. [99] ≥75 yrs: 12.9% eGFR <60 : Oral semaglutide 27 % REWIND 53.7 % 66.2 7.3% (1.1) 10.5 (7.3) 32.3 ±5.7 31.5% CVD 8.5% eGFR 75.3 [100] 20.8% CV [61.6–91.8] Dulaglutide events TECOS 70,7% 66.0±8.0 7.3±0.7 9.4 [4.9-15.3] 30.2±5.7 100% 18% Mean eGFR [85] 74.9±21.1 Sitagliptin EXAMINE 69.7% 60.9 ± 10.0 8.0±1.1 7.1 median 28.7 median 100% 27.8% Median eGFR [86-87] 71.1 Alogliiptin <60: 28.6 % SAVOR-TIMI 66.6% 65.1±8.5 8.0±1.4 10.3[5.2–16.7] 31.1±5.5 78.4% 12.8% Mean 72.5±22.6 [88] ≤ 50: 15.7% Saxagliptin CARMELINA 62.9% 65.9 ±9.10 7.95 ±1.01 14.8 ±9.5 31.3 ±5.3 58.5% 26.8% eGFR 54.6 ±25.0 [93] <60: 62.3% Linagliptin

BMI: body mass index; CAD: coronary artery disease; CV: cardiovascular; CVD: cardiovascular disease; eGFR: (mean) estimated glomerular filtration rate ml/min/1.73 m2; HbA1C: glycated haemoglobin; HF: heart failure; IHD: ischemic heart disease; MA: microalbuminuria; MI: myocardial infarction; SD: standard deviation; SIDIAP: Information System for the Development of Research in Primary Care; T2DM: type 2 diabetes mellitus.

III. Generalizability of Cardiovascular Safety Trials 40

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Conclusion

The results of three CVOTs has led the FDA to approve indications for antihyperglycaemic medications to reduce the risk of CV death (empagliflozin) and to reduce the risk of MACE (liraglutide, canagliflozin), both indications specific to patients with established atherosclerotic cardiovascular disease. [115] The results of a meta- analysis of CVOTs show that the decreased risk of MACE in both glucagon-like peptide- 1 receptor agonists and sodium-glucose cotransporter-2 inhibitors was confined to patients with previous CVD (a significant 14% lower MACE) but patients without preexisting CVD had a nonsignificant 2% higher MACE risk. [119]

The studies mentioned in this section showed that, for SLT-2 i and GLP RAs large CVOTs, the percentage of unselected populations in European countries and US that could have met the inclusion criteria varies largely. In the mentioned studies, only 12.8%, 12.9% and 16.6% of the general population met the inclusion criteria for the LEADER trial (liraglutide) and 4.8% to 21% for the EMPA-REG trial (empagliflozin). Other trials with more inclusive participation’s criteria, such as EXCEL (exenatide) and DECLARE- TIMI (dapagliflozin) has obtained less stringent results. The percentage of patients with established CVD registered in the SIDIAP database were from 21.2 to 22.4%.

The assessment of CV outcomes of non-insulin blood glucose-lowering drugs in “real- world” populations adds complementary evidence of the effects of these drugs and helps to update clinical guidelines, in spite of the methodological limitations inherent to observational research.

III. Generalizability of Cardiovascular Safety Trials 41

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Part II Cardiovascular outcomes and mortality among type 2 diabetes mellitus patients prescribed first and second-line blood glucose-lowering drugs: a population-based cohort study in the Catalan electronic medical record database, SIDIAP, 2010-2015

IV. Protocol Rationale and Design

IV

Protocol for the assessment of major cardiovascular events (MACE) and mortality in new users of non-insulin glucose-lowering agents: observational longitudinal study in the Catalan population-based electronic health record

database, SIDIAP, 2010-2015

IV. 1 Introduction

Randomized clinical trials (RCTs) have revolutionised medical research and improved the quality of health care by clarifying the benefits and harms of many interventions. [132] Because of their randomized allocation, well designed and conducted RCTs are considered as the gold standard for assessing the effects of drugs on populations. Randomisation distributes equally known and unknown confounding factors in order to create a control group that is as similar as possible to the treatment group. However, in some cases clinical trials would be infeasible (e.g., when the safety outcome is very rare) or when the study of outcomes or exposures in an interventional or prospective study would be unethical. [133, 134]

It is often unclear to what extent results of clinical trials conducted in selected populations, as high CV risk, are applicable to unselected ones, composed mainly of low CV risk patients. Large RCTs select high-risk populations because the low rate of events would demand otherwise a larger size population - or much longer follow-up - to reach the intended difference and the statistical power. Conversely, in observational research, large cohorts and long-time follow-up can be obtained from longitudinal

IV. Protocol Rationale and Design 44

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 healthcare databases (LHDs) especially if the number of subjects treated and the time from marketing authorisation are large enough.

In recent years, Real-World Evidence has an increasingly important role in the generation of scientific evidence and the US “21 Century Cures Act” established a program to evaluate the potential regulatory use of real-world evidence. [135] Healthcare longitudinal databases are increasingly used in evaluating comparative drug efficacy and harmful effects.

Longitudinal healthcare databases (LHDs) encompass extensive records of medical practices, diagnostic and laboratory procedures, drug prescribing and/or dispensing, and diagnosis codes. They include electronic medical records (EMRs), administrative claims databases and registries. EMR data are generated in the course of routine clinical care provision and can contain clinical data such body mass index (BMI), smoking or alcohol consumption, or diabetes duration.

For comparative efficacy research and the study of relatively rare events, LHDs have many advantages over RCTs: they can provide large size data, include patients often underrepresented in clinical trials, are representative of routine clinical care and suitable to study real-world effectiveness and utilization patterns. [136] Retrospective research avoids long delays for collecting outcomes and are available at relatively low cost. On the other hand, observational studies almost always have bias because prognostic factors are unequally distributed between patients exposed or not exposed to an intervention. The non-randomized design inherent to observational research can produce misleading results due to a number of flaws arising from design, time-related bias, matching and analysis. [137] All studies conducted in large longitudinal databases are subject to information bias. [133] In claims administrative databases, the discontinuity of coverage can impact on follow-up periods for reasons other than the outcome of interest or the study period, EMRs can be more stable than administrative claims data. However, records from one part of the health system, such as primary care, may not capture health events occurring in other parts of the health system, such as hospital care. For instance, a study conducted between 2003-2009, found that UK

IV. Protocol Rationale and Design 45

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Clinical Practice Research Datalink (CPRD) captured most cases of MI, but missed about a 25% of cases, and hospital records alone underestimated the true incidence of myocardial infarction. [138]

On April 6th 2017, the following protocol for an observational study has been registered in ENCePP.

This protocol is in accordance with the ENCePP Guide on Methodological Standards in Pharmacoepidemiology [139] and the guidelines International Society of Pharmacoepidemiology (ISPE) guidelines, revision 3 dated June 2015. [140]

The protocol has been registered in ENCePP (18513) on April 6th, 2017. [141]

Title: Major cardiovascular events in new users of non-insulin glucose-lowering agents: observational longitudinal study in the Catalan population-based electronic health record database, SIDIAP, 2010-2015.

STUDY REFERENCE NUMBER: ENCePP 18510

AUTHORS: Raquel Herrera Comoglio1, Xavier Vidal Guitart2, Luisa Ibáñez Mora2, Mònica Sabaté Gallego2, Pili Ferrer Argelés2, Maria Elena Ballarin Alins2, Jean-Luc Faillie3

1 School of Medicine, National University of Cordoba, Argentina

2 Catalan Institut of Pharmacoepidemiology, Autonomous University of Barcelone, Barcelone, Spain

3 Département de Pharmacologie Médicale et Toxicologie, Centre Régional de PharmacoVigilance, CHRU Montpellier, 371 avenue du Doyen Gaston Giraud, 34295 Montpellier, France. [email protected]

IV. Protocol Rationale and Design 46

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

ABSTRACT:

Background:

Cardiovascular (CV) risk is the leading cause of morbidity and mortality in T2DM population. The effect of control serum glucose levels on macrovascular complications remains uncertain. Glucose-lowering agents are currently marketed based on results of clinical trials with subrogate variables, mainly the percentage of glycated haemoglobin and other glucose markers. In 2007, concerns about CV safety of rosiglitazone led to regulatory recommendations regarding CV risk of new hypoglycemic agents, which are in force since 2008 (FDA, US) and 2012 (EMA, EU). In order to fulfil these recommendations, since 2008 a number of large randomized clinical trials have been designed and conducted, with a non-inferiority design as basis. Among those, three recently published large RCTs showed beneficial effects on cardiovascular mortality, meanwhile five large RCTs have failed to demonstrate any beneficial effect on CV outcomes. Other large RCTs, on-going or recently completed, are currently assessing the CV effect of seven marketed agents– and are foreseen to be completed up to 2020- are currently unavailable. In spite of enrolling a large number of diabetic patients with established or at high risk of CVD and having long follow-up periods, these studies are not free of limitations of RCTs. In addition, concerns have risen about the effect of some therapeutic groups or agents on heart failure (HF).

Electronic healthcare data, collected in the course of actual clinical practice by physicians can provide information of drugs effects in a real-world setting. Electronic medical records (EMRs) contain demographic and clinical information tests, and can be linked with other databases (hospitalization or deaths). An increasing number of population-based observational research focuses on effect of glucose-lowering agents on CV outcomes in large cohorts of patients. Up to date, no such studies have been performed in Spain. The Catalan general practitioners (GP) database SIDIAP contains pharmacoepidemiological data of 80 % of the total regional population.

IV. Protocol Rationale and Design 47

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Aim: to evaluate the effect of currently marketed non-insulin glucose-lowering agents on major CV outcomes in cohorts of Spanish population based on records of population-based EMR SIDIAP.

Design: Longitudinal retrospective observational cohort study, period of observation of six years (1st January 2010- 31st Dec 2015)

Material and Methods Study population: Cohorts of patients aged 18 yrs. or older registered in the Catalan general practitioners (GP) database (SIDIAP), with an active diagnosis of type 2 diabetes mellitus. Exposures: approved blood glucose- lowering agents since their first prescription. Outcomes: The primary outcome (PCO) is the composite of all-cause death, non-fatal myocardium infarction (MI) and non-fatal stroke. Secondary outcomes are individual components of PCO, and diagnoses of HF and peripheral artery disease. Statistical analysis: For populations’ characteristics, a descriptive statistic report in percentages categorical variables and in mean (standard deviation) and median [interquartile rank]. Patients will be stratified by demographic and clinical variables and use of insulins. The crude incidence rates for each group will reported, and hazard ratios (HRs) for association.

Strengths and limitations: Analysis of these databases could provide an estimation of the effect of currently marketed glucose-lowering agents on CV outcomes in a sample of non-selected population. New-users design prevents survivors bias. Limitations are mainly derived from its observational design (no randomization and confounding by indication). Concerning exposure data, some more recently marketed agents will not reach a number of prescriptions or follow-up periods appropriate to make valid comparisons with older agents. Concerning outcomes data, CV fatal outcomes not occurring in healthcare setting can be misclassified. Concerning completeness of data, missing data about a number of patients have been reported in other studies with SIDIAP database. The study period (6 years, from 2010 to 2015), which has been chosen according to availability of prescription/dispensing data, results in a follow-up can be not enough long for new diagnosed T2DM patients or patients not at high cardiovascular risk.

IV. Protocol Rationale and Design 48

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Expected results: For the drugs and period’s study, adjusted results should be similar to those obtained in large randomized controlled trials evaluating the effect of glucose-lowering agents on cardiovascular outcomes. Thus, up to a 10% reduction in cardiovascular morbidity and mortality compared with the use of reference non- insulin glucose-lowering agents, metformin and sulfonylureas (SU) can be expected.

IV. 2. Protocol rationale

2.1 Research question:

Background:

Diabetes is an independent risk factor for CV disease. Cardiovascular disease has been estimated to be 2-4-fold higher in diabetic patients than non-diabetic, affecting 80% of patients with type 2 diabetes. The rates of CV disease, mortality and HF vary according T2DM patients’ characteristics.

T2DM is a multifactorial disease that is defined by its glucose-related manifestations, although its etiology is not fully understood. Glycaemic control has shown to reduce microvascular complications and to be related in part to macrovascular disease. The higher prevalence of traditional vascular risk factors in patients with type 2 diabetes (e.g. hypercholesterolemia, hypertension) can explain only partially the higher CV risk in patients with type 2 diabetes compared with patients without diabetes. Other non‐traditional risk factors, such as insulin resistance, micro‐albuminuria or inflammation, may be important in the pathophysiology of cardiovascular disease and vascular mortality. [142, 143]

Observational evidence has shown that the risk of CV and CV mortality in higher in patients with diabetes. An observational study with data of 431 579 patients with T2DM registered in the Swedish National Diabetes Register found that the excess risk of acute myocardial infarction (AMI), coronary heart disease (CHD) and death was higher for patients with type 2 diabetes compared with controls, overall hazard ratio adjusted for age, sex, level of education, country of birth, diabetes duration and comorbidities was

IV. Protocol Rationale and Design 49

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

1.42 (1.41-1.44). However, for T2DM patients with HA1c 6.9% or lower, normoalbuminuria and eGFR ≥60 mL/min, compared with controls was 0.95 (95% CI 0.92 to 0.98). [144]

Tancredi et al. studied the excess risks of death from any cause and death from cardiovascular causes among persons with type 2 diabetes and various levels of glycemic control and renal complications in the Swedish National Diabetes Register from January 1, 1998 to December 31, 2011. The adjusted HR of CV death was 1.14; 95% CI, 1.13 to 1.15). The excess risks of death from any cause and cardiovascular death was higher for younger patients, those with worse glycemic control, and greater severity of renal complications. In patients younger than 55 years who had a glycated hemoglobin level of 6.9% or less, the hazard ratio for death from any cause was 1.92 (95% CI, 1.75 to 2.11); for patients ≥ 75 years of age or older was 0.95 (95% CI, 0.94 to 0.96). T2DM patients ≤ 55 years of age with a glycated hemoglobin level of 6.9% or less and normoalbuminuria were at higher risk of death from any cause, HR 1.60 (95% CI, 1.40 to 1.82); for T2DM patients ≥ 75 years of age the risk of dying was 0.76 (95% CI, 0.75 to 0.78), and patients 65 to 75 years of age also had a significantly lower risk of death (hazard ratio, 0.87; 95% CI, 0.84 to 0.91). [145]

An observational study found that, although lower in the contemporary era than previously, diabetes mellitus remains significantly associated with all-cause and CVD mortality. A gradient of mortality risk with increasing HbA1c >6% to 6.9% was observed, suggesting that HbA1c remains useful predictor of outcomes, even if causality cannot be inferred. These authors retrospectively studied all-cause and CVD mortality among 963 648 adults in the US Veterans Affairs Healthcare System from 2002 to 2014 during a mean follow-up of 8 years, 34% had diabetes mellitus. Compared with nondiabetic individuals, patients with diabetes mellitus had higher all-cause and CVD mortality, respectively. The adjusted HR was 1.29 (95% CI, 1.28-1.31), declined with adjustment for CVD risk factors (HR 1.18 [95% CI, 1.16-1.19]) and glycemia (HR 1.03 [95% CI, 1.02-1.05]). In T2DM patients, CVD mortality increased as HbA1c exceeded 7%. For HbA1c 7%-7.9% HR was 1.11 [95% CI, 1.08-1.14]; for 8%-8.9%,

IV. Protocol Rationale and Design 50

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

HR was 1.25 [95% CI, 1.22-1.29]; for HA1c ≥9%, HR was 1.52 [95% CI, 1.48-1.56] relative to HbA1c 6%-6.9%). [146]

While the effect of effect of glucose-lowering agents on microvascular events has been established in different observational and interventional studies, the effect of glucose control on cardiovascular outcomes remains unclear.

An overview of the milestones in the diabetes research has been summarized in the section “Introduction”

Metformin is currently the primary drug of choice for type 2 diabetes patients. Sulfonylureas (SU) have been associated with weight gain and an increased rate of myocardial infarction, all-cause mortality, stroke, heart failure and CV death. A number of new drugs – including insulins analogs- and new drug classes have been developed in order to achieve the glucose control in type 2 diabetic patients. A meta-analysis showed an increased risk of MI for rosiglitazone; this led regulatory agencies to require new glucose-lowering agents to show absence of cardiovascular toxicity. Large RCTs, designed and conducted in order to comply regulatory requirements, including populations at high cardiovascular risk, showed results varying from a neutral effect (DPP-4 i) to an important reduction in the risk of CV mortality, hospitalization for HF and all-cause mortality with empaglifozin, and some agents among the GLP-1 RAs agents showed decreased risk of MACE and mortality (liraglutide, semaglutide, albiglutide) Concerns have arisen because of an increased incidence of heart failure with the use of some agents, particularly some DPP-4 inhibitors (saxagliptin, alogliptin) and glitazones. Liraglutide and empagliflozin received a US FDA approval for reducing the risk of cardiovascular death in adult patients with type 2 diabetes mellitus and cardiovascular disease prevention as a new indication; [113, 114] in the EMPA-REG study the decrease on CV mortality is probably related with an increased diuresis, and the volume depletion and subsequent effect on heart failure occurrence or worsening.

Rationale: Effects of non-insulin glucose-lowering agents on cardiovascular outcomes in Type 2 diabetic patients are still being assessed.

IV. Protocol Rationale and Design 51

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Experimental evidence: Two new glucose-lowering classes, GLP-1 RAs and SGLT-2 i, demonstrated a reduction on MACE events and all-cause mortality in large controlled- placebo trials conducted in high CV-risk T2DM patients. Cardiovascular outcomes trials (CVOTs) assessing the effects of dipeptidyl-peptidase-4 inhibitors have shown non-inferiority compared to placebo, but failed to show superiority.

Observational evidence: Previous epidemiological research has obtained variable results different regarding the association between cardiovascular disease events and exposure to different blood glucose-lowering agents. Observational research for each class obtained results of different magnitude in different settings.

The assessment of cardiovascular outcomes in cohorts of patients treated with hypoglycaemic agents in clinical settings could provide useful information if adequately collected and analysed, although it can be compromised by uncontrolled biases and confounding, in particular confounding by indication.

Health care databases’ pharmacoepidemiologic and pharmacoeconomic analysis has a growing role in supporting health care decision making and efficient management of health care organizations. [134, 147] While experimental research has well-established standards for assessing the effects of drug on clinical outcomes, in observational research results depend on the studies’ design regarding design-related bias or potential bias, comparator, selection of covariates for adjusting and strategies for minimising confounding.

Large datasets of healthcare records are increasingly used to study the clinical outcomes of health care interventions in a general population including older adults who are often under-represented in clinical trials. [148] Healthcare records often lack of detailed clinical information—in particular, frailty—that is central to the clinical management of older adults.

Objective: The aim of this study is to assess the effect in a clinical setting of non-insulin blood glucose-lowering agents on CV outcomes in cohorts of new users of Catalan general practitioners’ databases, in the period between Jan 1st 2010 to Dec 31st 2015.

IV. Protocol Rationale and Design 52

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

This database is linked with mortality registries through administrative healthcare database CatSalut, from the Catalan Institut of Health.

3. Research hypothesis: In the study period, non-insulin blood glucose-lowering agents introduced as add-on therapy (second or third-lines) or as monotherapy for blood- glucose control and diabetes-related complications’ prevention, do not provide a clinically relevant benefit defined as a 10% reduction in cardiovascular morbidity and mortality compared with the use of reference non-insulin glucose-lowering agents, metformin and sulfonylureas (SUs).

IV.3. Protocol design

Observational, retrospective, longitudinal, population-based cohorts study utilizing secondary data from electronic healthcare database (de-identified demographic, clinical, and prescription data of patients registered in general practitioner’s medical health records linked with databases hospitalizations’ records and death registries).

Setting: Catalonia (Spain, European Union), is a North-Eastern Spanish province of 32,000 km² and 7,5 million inhabitants.

Source population: All residents in Catalonia registered in the public Catalan Institut of Health, universal health care system, for at least one year, and recorded in the general practitioners’ “Information System for the Development of Research in Primary Care” (SIDIAP) healthcare database (https://www.sidiap.org/ )

The Catalan Institute of Health (CIH, http://ics.gencat.cat/es/inici/index.html ) is a state healthcare system, tax-funded, and mandatory for active workers and retirees. It is the main provider of health services in Catalonia, has 279 primary care teams (PCT) with data of 5.564.292 people, approximately 74% of the Catalan population. SIDIAP (Information System for the Development of Research in Primary Care) is a primary care population computerized database in Catalonia, Spain, containing anonymised patient’s records for the 5.8 million people attended by general practitioners in the Catalan Health Institute. SIDIAP includes data on demographic variables, diagnoses, clinical variables, prescriptions, specialist referrals, laboratory test results, and

IV. Protocol Rationale and Design 53

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 medications claimed from pharmacist offices, obtained from the administrative CatSalut general database.

SIDIAP has been created in 2006. Pharmacoepidemiologic cross-sectional studies conducted for T2DM assessed trends in prescription patterns, health care costs and time to intensification. [149 -1610]

Study population: Catalan population aged ≥18, attending primary care practitioners registered in the SIDIAP, between January 1st 2010 to December 31st 2015 with an active diagnosis of Type 2 diabetes mellitus and receiving a first prescription of a non- insulin glucose-lowering agent.

Study period: from January 1st 2010 until Dec 31 2015

Inclusion criteria (ICD-10 codes):

. all patients registered in the SIDIAP database for at least one year previous to the index date.

. aged ≥18,

. attending primary care practitioners registered in the SIDIAP, between January 1st 2010 to December 31st 2015

. diagnosed of Type 2 diabetes mellitus (Coded E11.0, E11.1, E11.2, E11.3, E11.4, E11.5, E11.6, E11.7, E11.8, and E11.9)

. receiving a first prescription of a non-insulin blood glucose-lowering agent. (NIAD)

Exclusion criteria (ICD-10 codes):

. people < 18 yrs,

. diagnosis of Type 1 diabetes mellitus, (Coded E10.0, E10.1, E10.2, E10.3, E10.4, E10.5, E10.6, E11.8, and E11.9.)

IV. Protocol Rationale and Design 54

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

. diabetes mellitus due to underlying condition, (E08); drug or chemical induced diabetes mellitus (E09); gestational diabetes (O24.4); post pancreatectomy diabetes mellitus (E13); postprocedural diabetes mellitus (E13); secondary diabetes mellitus NEC (E13)

. active prescription of the given non-insulin glucose-lowering agent in the 90 days previous to cohort entry.

SIDIAP provided a coded, de-identified dataset of new-users of NIADs. Patients on other drug(s) were considered “incident users” for the NIAD newly prescribed. In accordance, patients with ICD 10 code Z79.84 [long term (current use) of oral antidiabetic drugs] were included if received a first prescription on another one non-insulin glucose lowering agent.

. Patients registered in the SIDIAP database ≤ 365 days (washout period)

Restriction: In order to have a sample representative of the T2DM population treated with NIADs, no restrictions were applied other than age (<18) in the inclusion/exclusion criteria. Patients are not excluded for history of CV events, CVD or other risk factors for the CV outcomes. Restriction reduces the sample size, compromises the statistical power, limits the generalizability of the results, and precludes effect modification analyses by these risk factors. [133] Instead, subgroup analyses were performed in each of these sub-cohorts of patients (previous CVD yes/no, previous HF diagnosis yes/no, previous CKD diagnosis yes/no)

Risk of selection bias: The risk of selection bias is low. Patients are not excluded for future events (such as therapy duration, insulin dispensing during follow-up, number of visits to GPs or doses of drugs of interest)

Incident-users design: Patients should not have a prescription of the non-insulin blood glucose-lowering agent in the 90 days before the index date. This period could be too short of excluding prevalent users.

IV. Protocol Rationale and Design 55

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Patients with use of any NIAD or insulin were excluded for the study on first-line therapies; however, we cannot rule out that patients who entered in the cohort in early 2010 could have been exposed previously to the agent supposed to be first-prescribed.

For second-line therapies, patients had previous use of a NIAD agent, most patients were on metformin. The first prescription of the second-line agent added to metformin was considered the index date.

Risk of misclassification: The 90-days period with no prescriptions of a given NIAD could result too short if chronic patients have their prescription refilled for more than 3 months. Roumie et al. report that in the US National Veterans Health Administration databases, seventy percent of our population received 90-day prescriptions, and 93% and 94% of metformin and sulfonylurea users, respectively, refilled their prescriptions within 90 days. In our study, the unintended inclusion of prevalent users can be suspected by the prescription patterns in the monotherapy study, in which 2010 year were 16% higher than in 2011, and this pattern was not repeated in the following years. [162]

Washout period:

When assessing effects with a new-user design (i.e., patients newly prescribed a drug), the washout period warrants that patients have no prior prescription of the drug of interest in the healthcare system of the study. Washout period minimize the inclusion of prevalent users, who can be diagnosed of T2DM and treated outside the healthcare system used for the study. [133]

Primary objective is to compare the time from the first prescription of a given non- insulin blood glucose-lowering agent to the recorded first occurrence of any component of a composite of major cardiovascular (CV) outcomes in cohorts of users of non- insulin glucose-lowering drugs.

Primary outcome is a composite of three-components of mayor cardiovascular events (MACE): all-cause death, non-fatal myocardial infarction (MI) and non-fatal stroke.

IV. Protocol Rationale and Design 56

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Secondary objectives are to compare, in cohorts of users of non-insulin glucose- lowering drugs

1. the time from the first prescription of a A10B (ATC code) given agent to:

. individual components of PCO:

o all-cause death,

o the first occurrence of non-fatal MI,

o the first occurrence of non-fatal stroke

2. the time from the first prescription of a A10B (ATC code) given agent to:

. the record of heart failure diagnosis (HF) (either HF onset or worsening to another NYHA class)

. the first record of incident diagnosis of intermittent claudication, worsening of intermittent claudication or revascularization procedures for peripheral vasculopathy.

Secondary outcomes:

. All-cause death

. Recorded diagnosis of first occurrence of any of the following events:

o non-fatal MI,

o non-fatal stroke

o record of heart failure diagnosis,

o incident diagnosis of intermittent claudication, worsening of intermittent claudication or hospitalization for peripheral revascularization procedures.

Rational for the choice of outcomes:

IV. Protocol Rationale and Design 57

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Three points MACE: Major adverse cardiovascular events (MACE) is the composite endpoint mentioned in the FDA and EMA guidances for the assessment of cardiovascular safety in new antidiabetic drugs: “These events should include cardiovascular mortality, myocardial infarction, and stroke, and can include hospitalization for acute coronary syndrome, urgent revascularization procedures, and possibly other endpoints.” [163, 79] In RCTs, composite endpoints reduce the sample size requirements and allow to assess easily the net clinical benefit of an intervention, avoiding misinterpretations associated with competing risks and the challenge of using a single outcome to validate the study objectives. [164] All CVOTs assessing the CV safety of new blood glucose-lowering drugs have a three- or a four-points MACE as primary outcome. Three-points MACE (MI, stroke and CV death) has long been used for cardiac drug approvals by the US Food and Drug Administration and since late 2008 has become a primary safety outcome for non-insulin blood glucose-lowering drugs. Composite end points allow CVOTs to ensure that sample size and duration of follow- up remain reasonable. [117] The combination of clinical outcomes into a composite end point increases the numbers of events ascertained and thus statistical power and precision. [118] Composite CV end points in diabetes trials have included a larger number of components, while more recent CVOTs almost exclusively use a composite of CV death, nonfatal myocardial infarction (MI), and nonfatal stroke (three-point major adverse CV event,3P-MACE) composite-or add hospitalization for unstable angina (HUA) to these three outcomes (4P-MACE). Marx et al. pointed that “the primary outcome of 3P-MACE may offer a better balance than 4P-MACE between statistical efficiency, operational complexity, the likelihood of diagnostic precision (and therefore clinical relevance) for each of the component outcomes, clinical importance, and the aim to adequately capture any potential treatment effect of the intervention.”[165] However, mechanisms of actions of different classes of drugs often differ in their impact on CV outcomes, but also individual agents within a class – such it is the case for liraglutide and exenatide. Because of drugs’ differences, no particular individual or composite end point can be seen as a "gold standard" for CVOTs of all glucose- lowering drugs. [166]

IV. Protocol Rationale and Design 58

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

In the composite MACE, cardiac death is the top marker form benefit, but the rate of expected events usually makes necessary the addition of less definitive but more frequently occurring endpoints, such as MI and stroke, to enable outcome trials to be completed in a reasonable time and with reasonable costs.

In our study, the three-point MACE was chosen as a primary endpoint to emulate CVOTs and to make comparisons easier. MACE captures the first event of the composite, and also addresses competing risks. With some exposures, such as metformin and sulfonylureas, the high number of patients allows the assessment for each outcome separately. This cannot be the case, however, for exposures with a low number of patients, such as the ones newly introduced during the study period. The composite of MACE also allows the identification of the first event that occurred in the population at risk.

In our study, we included all-cause mortality. Diabetic patients, causes of death are mainly from CVD and cancer [25] All-cause mortality captures all the events and it reflect more accurately the benefit or harm of a treatment.

MI: Myocardial infarction is the component of the three-points MACE; and a major event of CVD, can be fatal or non-fatal; MI can be also silent, with no clinical manifestations. We included all the codes for MI, with no exclusions.

Stroke: We also included all the stroke events (ischemic, haemorragic and embolic) Embolic events can be manifestations of other cardiac diseases, namely atrial fibrillation. The inclusion of embolic stroke may introduce nondifferential outcome misclassification. [167]

All-cause mortality: All-cause mortality is the most important marker of benefit, although CV mortality is the end point included in MACE and most of the CVOTs. Mortality is always pertinent with regard to medications (such as lipid-lowering agents) used to treat asymptomatic patients, because the only benefit is prevention of future disease. [168] As SIDIAP has not access to complete registries coding the cause of death, CV mortality was excluded because of feasibility. SIDIAP database is linked

IV. Protocol Rationale and Design 59

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 with administrative database of the CIH and allows a reliable registry of deaths occurred both in in-hospital settings and outside a hospital.

Heart failure:

As many as 50% of patients with type 2 diabetes may develop heart failure. [169] Heart failure is a common complication of type 2 diabetes, and its prognostic implications including high mortality. [165]

Peripheral artery disease: In the Framingham study, the prevalence of PAD at study entry was 13.6% and the incidence of new PAD was 3.7 per 100 patient-years. Both prevalent and incident PAD was strongly and independently associated with increasing age, systolic blood pressure, total serum cholesterol, and especially smoking. [170]

Patients with low ankle-brachial index plus diabetes presented increased mortality, acute myocardial infarction, and ischemic stroke risk, all at rates similar to those with previous cardiovascular disease. A retrospective cohort study using the Database of the Catalan primary care system (SIDIAPQ), for 2006-2015, included 58,118 persons, mean (SD) age 66.6 (10.7) years, 53.4% were men, the median follow-up was 5.9 years. Compared to the reference group with no diabetes, no previous cardiovascular disease, and normal ankle-brachial index (ABI). Participants with low ABI showed increased mortality, acute myocardial infarction, and ischemic stroke incidence in all the subgroups. Patients with low ankle-brachial index plus diabetes presented increased mortality, acute myocardial infarction, and ischemic stroke risk, all at rates similar to those with previous cardiovascular disease. [171]

Risk of information bias:

SIDIAP contains diagnoses recorded by general practitioners. Although SIDIAP database can be linked with hospital discharges codes, in this dataset only GPs records are available. This is a significant source for information bias, that could be in principle non-differential, and non-fatal events (i.e., non-fatal MI and non-fatal stroke) could have been under recorded. The risk of information bias is much lower for all-cause mortality, because SIDIAP is linked with ICH administrative database.

IV. Protocol Rationale and Design 60

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Follow-up period: from January 1st 2010 to December 31st 2015

Data Sources

Clinical data for selecting study population

Selection of study population will be done on records of SIDIAP (Information System for the Development of Research in Primary Care), a primary care population computerized database. Data extraction (T2DM population ≥ 18 years, registered in SIDIAP ≥ 365 days, with a new claimed prescription of a new NIAD agent and no previous prescribing of this NIAD in the previous 90 days), coding and anonymization has been performed by SIDIAP.

Demographic and clinical data extraction:

In the coded, de-identified dataset provided by SIDIAP (Information System for the Development of Research in Primary Care), the following baseline characteristics have been retrieved for each cohort:

Covariates included in the adjustment were: demographic data, clinical variables,

Demographic data include:

o sex (men, women)

o age (further categorised into age subgroups: ≤44yrs, 45-64yrs; ≥ 65 years

(for sub-group analyses we considered the categories < 75 years and ≥ 75 years

94 o MEDEA deprivation index for socio-economic inequalities in health

Rationale for the inclusion of a deprivation index:

IV. Protocol Rationale and Design 61

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

The MEDEA score (Mortality in small Spanish areas and economic and environmental inequalities” summarises the geographic patterns of mortality in Spain cities related to socioeconomic and environmental characteristics.[172] A study found that in Barcelona, the mortality rate was 2.33 per 103 in men and 1.15 per 103 in women in Q1 and 3.49 per 103 in men and 1.52 per 103 in women living in Q4. Premature mortality rates showed higher premature mortality in the most deprived districts. [173]

Clinical variables include: o Type 2-diabetes related:

. standardised glycated haemoglobin (HbA1c) values; (registered up to 3 months before or on the index date [index prescription]

. diabetes duration (yrs) at the index date

. insulin treatment up to 3 months before or on the index date

. insulin treatment during the follow-up o body mass index (BMI); o other cardiovascular risk factors

. CV risk, according to REGICOR score

. smoking status (past, current, never, according to the most recent information recorded in the medical history)

. blood pressure (BP) (systolic [SBP] and diastolic [SBP]); (both SBP and DBP must be recorded in the same date in order to be included)

. lipid levels including total cholesterol (TC), low-density lipoproteins or LDL cholesterol (LDLc), high-density lipoproteins or HDL cholesterol (HDLc) and triglycerides (TG) (TG, LDLc and HDLc must be recorded in the same date in order to be included)

IV. Protocol Rationale and Design 62

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

We extracted the value of the local REGICOR index of CV risk during the 15 months prior the index date. The most recent recorded value of the body mass index (BMI) was retrieved from records in the last 15 months prior to the index date.

The prevalence of selected cardiovascular risk factors (CVRF) has been assessed in the North Catalonia. CVRF was: hypertension: 74.5%; dyslipidemia: 77.7%; smoking: 14.9%; obesity 44.9%, and familial CVD: 38.4%. Three or more CVRFs, including T2DM, were observed in 91.3%. Metabolic syndrome (MS) prevalence was 68.7%. Framingham score was 10.0%, higher in men than in women. CVD prevalence was related to: age, number of CVRFs, duration of diabetes, familial history of CVD, waist circumference, hypertension, lipid profile, kidney disease, and Framingham score, but not to MS by itself. Normal values of serum lipid and blood pressure were only observed in 18.9% and 24.0%, respectively. Platelet aggregation inhibitors were only recorded in 16.1% of the patient cohort. MS presence was not an independent risk factor of CVD in THIS study.[14]

Hypertension and type 2 diabetes are common comorbidities. Hypertension is twice as frequent in patients with diabetes compared with those who do not have diabetes. Moreover, patients with hypertension often exhibit insulin resistance and are at greater risk of diabetes developing than are normotensive individuals. [174]

The REGICOR index is a local score for cardiovascular risk [175, 176]

Clinical history [ICD-10 codes]

Past morbidities include registered diagnosis previously to the index date of the following:

. Cardiovascular disease

o Acute coronary syndrome (ACS) (I20, I24)

o Myocardial infarction (I21)

o Stroke (hospitalization for ischemic or haemorragic stroke, clinically evident transient ischemia) (I61, I62.9, I63, I64, I65, I66, I67, I69)

IV. Protocol Rationale and Design 63

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

o Diagnosis of heart failure, heart failure (I50, I51.7, I,51.9)

o Peripheric arteriopathy (excluding Buerger and Raynaud), intermittent claudication

. Renal failure (N17-N19)

Co-morbidities are those registered by general practitioners.

Risk of information bias:

For the recorded concomitant medication, there is a risk of information bias: events not recorded (such as MI or ACS treated in hospitals) or delays in recording.

For Renal failure, there is no data of eGFR creatinine levels or stages of disease. For heart failure, there is no data about hospitalization or NYHA classification categories.

Concomitant medication [ATC codes]

We extracted data of concomitant medication at baseline (up to 3 months before the index date)

Rationale for the inclusion of covariates:

Among the universally acknowledged risk factors for coronary artery disease: increased concentrations of low-density lipoprotein cholesterol, decreased concentrations of high- density lipoprotein cholesterol, and increased triglyceride concentration, haemoglobin A1c, systolic blood pressure, fasting plasma glucose concentration, and a history of smoking. [174, 177]

HbA1c has shown a direct association in the risk of stroke in T2DM patients, with a stepwise increased risk of death for every 10 mmol/mol categorical increment of HbA1c for the highest HbA1c category. [178]

Patients with history of myocardial infarction (MI) or stroke are at higher risk of a recurrent event, but also of other manifestations of cardiovascular (CV) disease such as stroke. [122] The annual death rate for survivors of MI is 5% six-fold than in people

IV. Protocol Rationale and Design 64

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 who do not have coronary heart disease.[179] Diabetes, peripheral artery disease, and history of stroke are strongly associated with subsequent MI, stroke, or death, additional risk factors are heart failure, renal disease, and chronic obstructive pulmonary. [180] Patients older, male, with diabetes, prior stroke, heart failure, unstable angina, and no revascularization have worse long-term prognosis. [122] After the first year, MI survivors remain at higher risk than the general population, particularly if there are additional risk factors such as older age, hypertension, or diabetes, all of which lead to worse outcomes. [121] Patients with a past stroke are an increased risk of a recurrent event (approximately 7%/year).

Concomitant medications

. Cardiovascular drugs

o Antihypertensives [C02, C03, C07, C08, C09]

o Antithrombotic agents [B01A]

o Lipid-modifying agents [C10]

. Drugs that potentially affect cardiovascular risk

o Antidepressants [N06A]

o NSAIDs [M01A]

. Insulin and analogues [A10A]

Concomitant medication is defined as every drug with an active prescription at the index date (the day of prescription of a given glucose-lowering agent)

Rationale for the choice of concomitant medications:

NSAIDs increase CV risk, especially selective cyclo-oxygenase-2 (COX-2) agents, but also non-selective agents not inhibiting enough the production of thromboxane A2. The higher cardiovascular risk is for diclofenac

IV. Protocol Rationale and Design 65

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Depression is very prevalent among patients with T2DM and is associated with several key diabetes-related outcomes. [181]

Exposures

Exposure definition

Subjects are considered exposed if a given non-insulin glucose lowering agent is prescribed. SIDIAP contains prescription records that are linked with administrative data of Catalan Institut of Health. However, as prescription is a proxy of dispensing, which is in turn a proxy of administration drug, treatment compliance could not be assessed.

SIDIAP also contains information about drug strength, number of units for package and the number of packages dispensed. This information has not been analysed in the present protocol.

A patient was considered as exposed to a given agent from the date of the filled prescription of a first new non-insulin glucose-lowering agent (index date) until the day of the prescription of a new agent (either addition or switching) or loss of follow-up.

To compare adequately the effect of two or more agents, similar disease severity’s populations should be selected, comparing similar lines of therapies (i.e., second-line to second-line) would allow to avoid bias.[182] When a given agent is added for the first time to a pre-existing glucose-lowering therapy (for instance, metformin or sulfonylurea), the exposure will be re-assigned to double-therapy (i.e., MET + X or SU + X). If a new agent is added to a double therapy, this addition will be compared to the existing double therapy, as an “add-on” therapy (i.e. “MET+SU+ X” will be compared to “MET+SU” OR “MET+SU+Y”

IV. Protocol Rationale and Design 66

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

For cohorts of new users of metformin and SU, patients will be reassigned to if another hypoglycemic agent is prescribed.

Initial date of observations: January 1st 2010

Final date of observations: December 31st 2015

Patients will be considered exposed to a non-insulin glucose-lowering agent after the first prescription of any of the following:

Substances [ATC code] (date of EU approval)

. Metformin[A10BA02]

. Sulfonylureas [A10BB]:

o First-generation SUs: chlorpropamide [A10BB02], tolbutamide [A10BB03], acetohexamide [A10BB31], carbutamide [A10BB06], tolazamide [A10BB05], metahexamide, [A10BB10].

o Second-generation SUs: glibenclamide [A10BB01], glibornuride [A10BB04], glipizide [A10BB07], gliquidone [A10BB08], gliclazide [A10BB09], glisoxepide [A10BB11], glimepiride [A10BB12]

. Meglitinides

. Repaglinide [A10BX02] (authorized in UE on August 17th 1998, authorized in Spain in 1999); nateglinide [A10BX03] (authorized in UE on April 3rd 2001)

. Alpha glucosidase inhibitors [A10BF]

. Acarbose[A10BF01] (authorized in Spain in 1999); Miglitol [A10BF02], voglibose [A10BF03]

. DPP-4 inhibitors:

IV. Protocol Rationale and Design 67

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

. vildagliptin [A10BH02] (authorized in EU in September 2007), saxagliptin, [A10BH03] (approved in October 2009 in EU), linagliptin[A10BH05] (approved in EU in August 2011), alogliptin [A10BH04] (approved in September 2013 in EU), sitagliptin [A10BH01] (approved in March 2007 in EU)

. GLP-1 Receptor Analogues:

. exenatide [A10BJ01] (approved in 2005/2012), liraglutide [A10BJ02] (approved in 2010), lixisenatide[A10BJ03] (approved in EU in 2013), albiglutide [A10BJ04] (approved in 2014), dulaglutide [A10BJ05] (approved in 2014).

Note: Semaglutide was not yet approved in EU or US FDA during the study period. Aliglutide was withdrawn from the market in July 26, 2017, therefore it was available during the study period.

. SGLT-2 inhibitors: . dapagliflozin [A10BK01] (approved in EU in November 2012), canaglifozin [A10BK02] approved in November 2013 in EU, empagliflozin [A10BK03] (approved in EU in May 2014).

. Thiazolidinediones:

. pioglitazone[A10BG03] (authorized in October 2000 in EU), rosiglitazone [A10BG02] (authorized in EU in and suspended in EU in 2010)

Combined drugs:

When fixed-combinations were prescribed, the exposure was considered as the addition of the compounds.

Exposure data source and extraction

IV. Protocol Rationale and Design 68

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

SIDIAP database contains coded data of individual patients and prescribing health professional, medicine’s National Code and ATC classification, description of presentation, number of units per package, daily dose, month of drug dispensing, monthly number of packages dispensed and dates of start and end of prescription.

We have used only data about the agent. No doses were extracted or analysed.

Exposure validation

The Catalan Health Institute (Servei Català de la Salut [CatSalut]) manages a database which contains information about the dispensed drug, patient, prescriber and primary care center, and funds a percentage (for work active population) or the total amount of dispensed medicines (for retired people). SIDIAP prescription records are linked with CatSalut dispensing records.

Further validation of exposure has not been performed.

Exposure ascertainment

Exposures have been identified by ATC codes. The exposure is defined though an “as treated” approach. Patients are considered at risk for a given agent from the date of prescription (index date) to the date of the prescription of another agent.

Information bias:

There are no data about adherence. Prescription and dispensing are both proxies of administration.

Low risk of nondifferential information bias. Low risk of nondifferential misclassification.

Outcomes

Outcomes data source and coding

IV. Protocol Rationale and Design 69

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

To obtain the data for the study’s endpoints, outcomes were extracted from SIDIAP database, with ICD-10 coded diagnoses recorded by general practitioners.

For the present study, incident CV events are identified by ICD-10-CM:

. myocardial infarction (I21) . Hospitalization for unstable angina (I20) . stroke (I61. I62, I63, I64), . Code for coronary revascularization procedure . Code for peripheral revascularization procedure or onset or worsening of intermittent claudication (I73.9) . Code for heart failure (I50.0)

Mortality: SIDIAP database is linked with the Catalan Institute of Health administrative database, which in turn is linked with mortality registry containing records of date and mortality causes for all Catalonia-residents’ deaths. Linkage between databases is based on the CatSalut unique identifier code for each Catalonia resident.

In our study, we do not have the codes for causes of death, and because of this, the initial aim to assess specifically CV mortality has given up.

In the initial version of the protocol, cardiovascular (CV) death was defined as:

. ICD-10 I46 code (cardiac arrest, I46.0, I46.1, I46.9) and

. any death occurred ≤ 30 days after following a recorded event of:

o acute myocardial infarction (I21, I23), stroke (I61. I62, I63), cardiac arrhythmia (I49.9), hearth failure, cardiovascular procedure, cardiovascular hemorrhage, non-stroke intracranial hemorrhage, non- procedural or non-traumatic vascular rupture (e.g., aortic aneurysm), or pulmonary hemorrhage from a pulmonary embolism, other cardiovascular causes, such as pulmonary embolism or peripheral arterial disease.

IV. Protocol Rationale and Design 70

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Outcomes measurement

Incidence rates of primary composite outcome (PCO)and secondary composite outcome (SCO) and secondary outcomes (SOs) will be retrieved from SIDIAP database.

For each cohort, incidence rate will be calculated from the index prescription (first prescription of a new non-insulin glucose-lowering agent) until the first event of the PCO, SCO, or events detailed in SOs, switching to or addition of another agent or end of the observation period.

Bias

Selection bias: SIDIAP includes data of the 80 % of total Catalan population. Since all new prescription of a non-insulin glucose-lowering agent defines selection, there is a low possibility of selection bias.

Differential misclassification of exposure in study and concomitant medications is unlikely due to the administrative nature of the system.

A possible outcomes’ misclassification can occur if the event occurred out of the Catalan health system and it hasn’t been recorded in GP records.

In studies comparing drugs that have similar indications, patients will be assigned to therapies for reasons that are difficult to measure (“channelling”). [168] The measurement of all factors plausibly related to prognosis is essential.

Analysis plan

Statistical Analyses (as described in the Section V: First-line monotherapies, XVG)

As descriptive statistics of the sample frequencies (percentage) were used for categorical variables and mean (standard deviation) or median (interquartile range) for quantitative variables.

IV. Protocol Rationale and Design 71

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Twenty multiple imputations by chained equations were obtained to replace missing baseline values for HbA1c, body mass index, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides, serum creatinine, and systolic and diastolic blood pressure.

To reduce the selection bias associated with the observational nature of the data, a propensity score as the conditional probability of being treated with SU drugs was estimated with the baseline covariates as age, gender, socioeconomic status, BMI, SBP, DBP, HbA1c, lipid profile, serum creatinine level, use of other medications (ACE inhibitors, aldosterone antagonists, antiplatelet drugs, beta-blockers, calcium channel blockers, diuretics, statins, other lipid-lowering drug, NSAIDs, antidepressants), history of acute coronary syndrome, myocardial infarction, ictus, heart failure, peripheral arterial disease, renal failure), and the duration of diabetes. Variables were considered well balanced if the standardised differences between both groups were <0.10 after adjusting by PS.

Inverse probability of treatment weighting (IPTW), was used to create pseudo- populations of SUs and metformin initiators with similar covariate distribution. This approach has been shown to be particularly suitable when the outcome is a time to event measure.

Weights were stabilised to avoid extreme values and increase precision in the estimates. Propensity scores and IPTWs were generated for each imputed data set.

Event rates per 10,000 person-years were calculated for events. For each event, time to follow-up we defined as the time between cohort entry (the first prescription of metformin or SUs alone) and the event. Patients were followed up until prescription of a different antidiabetic drug, death, transfer or end of study. Cox proportional hazard models were used to estimate the hazard ratios with 95% CIs.

All analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC).

IV. Protocol Rationale and Design 72

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Subgroups analysis

For exploring heterogeneity treatment’s effect, descriptive subgroup analysis will be performed:

. Use of insulin during monotherapy (for dual, second-lines only): yes/ no . Age: < 75 years and ≥ 75 years . Sex: male / female . Glycemic control

o HbA1c < 8%) o HbA1c 8 – 10 % o HbA1c > 10%

. Diabetes duration:

o 0-4 years o 5-9 years o 10-14 years o ≥ 15 years . Body Mass Index:

2 o < 25 kg/ m o 25-29.9 kg/ m2 o 30- 39.9 kg/m2 o ≥ 40 kg/m2 . History of cardiovascular disease: yes/no

. History of heart failure: yes/ no

. History of renal failure: yes/ no

IV. Protocol Rationale and Design 73

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

. MEDEA index:

o Rural o U1 (least deprived) o U5 (most deprived)

Exploratory subgroup analysis will be derived from data.

Quality assurance, feasibility and reporting

Quality assurance:

Data validation for outcomes is not foreseen due to feasibility reasons. For study’s exposures CatSalut, as an administrative database, performs validations.

Feasibility:

To conduct this study, an agreement has been signed between Institut Jordi Gol and the Fundacio Catalan Institut of Farmacologia.

Register:

The initial protocol has been registered in ENCePP with the number 18510. Results have been submitted to publication.

Reporting:

Study results are reported according to STROBE and RECORD-PE guidelines. [183, 184] Recommendations about graphical depiction were followed. [141]

Ethical issues

The protocol has been approved by the Clinical Investigations Ethics Review Board from the Investigation in Clinical Care Institut Jordi Gol. Local rules of confidentiality are respected (according to article 5, Ley Organica 15/1999, Regulación del Tratamiento de Datos de Carácter Personal). The Institut Jordi Gol provided a coded, de-identified data set of T2D patients ≥ 18 years registered in the SIDIAP database for

IV. Protocol Rationale and Design 74

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 at least 365 days. Patient Informed Consent is considered not necessary due to data anonymisation.

Strength & limitations of the study

Strengths

1. Healthcare data collected in clinical practice by physicians are representative of routine clinical practice in large populations. All age categories and all socio- economic areas are represented. Patients have long permanence in the database. 2. Some information about lifestyle is lacking (smoking habit, exercise) 3. Incident users “new-users” design reduces the risk of prevalent users’ bias 4. “As treated” approach reduces the risk of exposure misclassification 5. Adequate comparator: similar disease stage. We compared first-line monotherapy vs first-line monotherapy and second-line dual therapy vs second- line dual therapy. 6. Subgroup analysis are performed. 7. The database is linked with administrative database, data about mortality are reliable.

Limitations

1. Based only in general practitioners’ recorded diagnoses. Hospital data are not available. Risk of information bias 2. No data about adherence 3. No data about nursing homes residents. 4. The protocol didn’t include cancer or other comorbidities data in order to assess frailty. 5. Heart failure and renal failure diagnoses are not categorized. 6. The outcomes are comprehensive and encompass different levels of severity. 7. In the period of the study is a possible relatively reduced number of patients treated with non-insulin glucose-lowering agents that have been

IV. Protocol Rationale and Design 75

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

approved since 2008, especially for the newest agents (alogliptin, albiglutide, canagliflozin, empagliflozin, dapagliflozin). New users of rosiglitazone are expected to be recorded until 2010 (date of marketing suspension). These more recently marketed agents could not reach a number of prescriptions or follow-up periods appropriate to make valid comparisons with older agents 8. The study period (6 years, from 2010 to 2015), which has been chosen according to availability of prescription/dispensing data, could result in a follow-up not long enough for new diagnosed T2DM patients or patients not at high cardiovascular risk. 9. Missing data about 25-30% of patients have been reported in other studies with SIDIAP database.

We haven’t included antipsychotic agents in concomitant medications. Patients with diabetes are at higher risk of dementia and cognitive decline, antipychotics are used in dementia. Prospective observational studies in people with diabetes showed 73% increased risk of all types of dementia, 56% increased risk of Alzheimer dementia, and 127% increased risk of vascular dementia compared with individuals without diabetes. Conversely, patients with Alzheimer dementia are more likely to develop diabetes than people without Alzheimer dementia.

IV. Protocol Rationale and Design 76

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

V. Cardiovascular outcomes and mortality in type 2 diabetes mellitus patients prescribed first-line non-insulin blood-glucose-lowering agents as monotherapy

V.

Cardiovascular outcomes and mortality in type 2 diabetes mellitus patients prescribed first-line non- insulin blood-glucose-lowering agents as monotherapy

A population-based cohort study in the Catalan electronic medical record database, SIDIAP, 2010-2015

Raquel Herrera Comoglio – Xavier Vidal Guitart

Abstract

Aim: To assess cardiovascular (CV) events and all-cause mortality in type 2 diabetes mellitus (T2DM) patients treated with first-line monotherapies of non-insulin antidiabetic drugs.

Methods: Longitudinal retrospective cohort study in the Catalan database SIDIAP (Information System for the Development of Research in Primary Care). T2DM patients ≥18 years newly prescribed first-line monotherapies during 2010-2015 were followed since their first prescription until MACE (myocardium infarction MI, stroke and all-cause death), its components, heart failure (HF) and peripheral artery disease (PAD) or censoring. Cox proportional hazard models were used to estimate hazard ratios 95% confidence interval (HR [95%CI]).

Results: Compared with metformin, the use of sulfonylureas, dipeptidyl peptidase-4 inhibitors (DPP-4 i) and meglitinides were significantly associated with higher risk for MACE (1.55 [1.42 – 1.68]); 1.49 [1.22-1.84] and 2.01 [1.29-3.12]) and all-cause

V. First-line monotherapies 78

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 mortality (1.67 [1.52 – 1.84], 1.65 [1.30 -2.0] and 2.08 [1.26-3.42]. Sulfonylureas users had increased risk of non-fatal MI (1.38 [1.03 – 1.85]) non-fatal stroke (1.31 [1.11 – 1.54]), HF (1.49 [1.28-1.72]) and PAD (1.24 [1.02-1.51]). Meglitinides users were at increased risks of MI, HR 2.03 (1.10-3.74).

Conclusion: In first-line monotherapies, compared with metformin, sulfonylureas were associated with increased risks in all the outcomes; DPP-4 i showed increased risks of MACE and mortality compared with metformin. Residual confounding cannot be ruled out, deserving further research

Introduction

The goal of the T2DM treatment is the maintenance of glycaemic control to reduce the progression to long-term microvascular and macrovascular complications. When possible, diet and lifestyle modifications are recommended before prescribing drugs. International guidelines recommend metformin as a first-line treatment of choice in combination with lifestyle modifications. Sulfonylureas (SUs) are recommended if contraindications or intolerance exist. [185, 186] The rationale for the use of metformin as first-line glucose-lowering pharmacological therapy in type 2 diabetes is based on its glucose-lowering efficacy: as insulin-sensitiser, metformin counters insulin resistance and lowers basal hyperinsulinaemia, while avoiding significant hypoglycaemia or weight gain. [187]

Heart failure was a formal contraindication for metformin because of the risk of lactic acidosis, a very rare but potentially fatal adverse effect related to biguanides; this contraindication had been removed in 2006 based on observational studies. [188, 189] Severe renal impairment, heart failure and liver impairment are risk factors for lactic acidosis. [189, 190] In April 2016, the US Food&Drug Administration modified the labelling, the use of metformin is contraindicated if eGFR <30 mL/1.73m2 and not recommended if 30-45 mL/minute/1.73m2.[190]

In spite of the relatively low number of patients randomised to metformin, the United Kingdom Prospective Diabetes Study (UKPDS34) represents one of the milestones in T2DM treatment: it found a significant reduction of any diabetes-related end-point,

V. First-line monotherapies 79

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 diabetes-related mortality, and all-cause mortality in 342 obese, newly diagnosed T2DM people, compared with conventional therapy of insulin or first or second-generation sulfonylureas. [53] In the UKPDS post-trial monitoring, relative reductions in risk persisted at 10 years in the SU–insulin group for any diabetes-related endpoint (9%) and microvascular disease (24%), risk reductions for myocardial infarction (15%) and death from any cause (13%) emerged over time; in the metformin group, significant risk reductions were more than two-fold higher than those of the SUs-insulin group: 21%, for any diabetes-related endpoint, 33% myocardial infarction, and 27% for death from any cause. [55]

In 2005, the International Diabetes Federation recommended metformin as an initial glucose-lowering pharmacotherapy for T2DM. In 2011 metformin was included in WHO’s essential medicines list. [191, 192] Local guidelines in Catalonia, published in 2010, recommended metformin as initial diabetes therapy. [193]

Apart from the UKPDS study, there is scarce experimental evidence regarding cardiovascular (CV) and mortality outcomes for sulfonylureas (SUs) compared with metformin: a systematic review found six trials comparing second-generation sulphonylureas versus metformin. [194] A RCT assessing CV events in 304 high-risk Chinese T2DM patients obtained favourable results for metformin, compared with glipizide. [195] Observational studies have almost consistently reported higher risks for SUs compared with metformin. [196-200, 167, 162]

The CV benefits of metformin have been highlighted in recent publications [192]. Lamanna et al. suggested that the improved CV outcomes shown with metformin treatment appeared to be more beneficial in longer trials enrolling younger patients and that it is likely that metformin monotherapy is associated with improved survival; the concomitant use with sulphonylureas was associated with reduced survival. [201] Bailey et al. pointed out that metformin has been associated with reduced long-term CV risk in prospective controlled trials, observational studies and database analyses. [187]

Metformin is the agent of choice for the first-line T2DM treatment, while SUs are recommended to those patients who are intolerant to metformin. Approximately 5% of

V. First-line monotherapies 80

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 patients do not tolerate metformin, mainly due to diarrhoea and other gastrointestinal effects. Other agents used in first-line monotherapy are meglitinides, dipeptidyl peptidase-4 inhibitors (DPP-4 i), thiazolidinediones (TZD) and sodium-glucose co- transporter 2 inhibitors (SGLT-2 i).

We carried out a longitudinal cohort study to assess the risk of cardiovascular outcomes and all-cause mortality. The primary objective was to evaluate the risk of the time to the first event of the composite of major CV events (non-fatal myocardial infarction and non- fatal stroke) and all-cause mortality (MACE). Secondary objectives are to assess the risk of the time to the first event of all-cause mortality, myocardial infarction (MI), stroke, heart failure and peripheral artery disease (PAD) in cohorts of T2DM people registered in the general practitioner Catalan health care database, SIDIAP.

Cohorts of patients were defined by their exposures to non-insulin blood glucose- lowering drugs. In this paper, we first focused on monotherapies, metformin, sulfonylureas, dipeptidyl peptidase-4 inhibitors and meglitinides as first-line therapies in drug-naïve patients. Other first-line monotherapies were not assessed due to the reduced number of patients exposed. We also evaluated clinical outcomes in cohorts on the most used second-line combined treatments, metformin and SUs: we compared patients who were prescribed SUs added to metformin to patients who initially were treated with SUs and were prescribed metformin added to SUs. Based on the existing observational evidence, we hypothesise that patients had a lower risk of mortality and cardiovascular outcomes in patients newly treated with first-line metformin compared with those newly prescribed sulfonylureas as monotherapy.

The protocol has been evaluated by the Clinical Investigations Ethics Review Board from the Investigation in Clinical Care Institut Jordi Gol, and an agreement has been signed between the Institut Jordi Gol and the Catalan Institute of Pharmacology. According to local regulation, to maintain data confidentiality and patient anonymity, patient information was de-identified before data extraction and analysis. The protocol has been registered in the ENCePP registry (protocol 18510) and reported according to STROBE and RECORD-PE guidelines. [183, 184]

V. First-line monotherapies 81

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Design:

Analysis of longitudinal electronic health records data. We did a population-based cohort study of adult T2DM patients who started a first-line monotherapy. To reduce bias we used an incident user design. For patients who initiated these treatments in the study period, we analysed outcomes of patients who were prescribed a second-line therapy with the combination metformin+ SUs. Figure V.1 depicts the study design.

Study Population

People aged ≥18 years with an active T2DM diagnosis (ICD-10 codes E11) registered in the SIDIAP database ≥365 days before the cohort entry and without any blood-glucose- lowering prescription in the 90 days prior the index date (including insulin and non- insulin). The study period extended from January 1st 2010 to December 31st 2015.

Figure V.2 shows the selection flow.

Exposures:

Metformin and sulfonylureas, as first-line monotherapy or as a second-line dual combination therapy; and dipeptidyl peptidase-4 inhibitors and meglitinides in monotherapy. SUs are considered as a class: if a patient on a given SU switched to another agent, he/she was considered to continue in the same cohort. The index date (ID) is defined as the day of the first recorded claimed prescription. Patients were considered at risk from the day of the claimed prescription until the prescription to another non-insulin blood glucose-lowering agent or censoring.

Follow-up

Patients were followed until the first event of primary and secondary outcomes, the first prescription of a different NIAD group, or end of study 31st December 2015, whatever came first.

V. First-line monotherapies 82

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Figure V.1: Summary of design characteristics and design-related potential sources of bias

Data source Study design: Observational retrospective cohort study Risk of bias: confounding by indication Objective: To compare first-line agents in monotherapy vs first-line agents in SIDIAP database monotherapy Risk of time-lag bias: low Data source: SIDIAP (Information System for the Development of Research in Patient selection Primary Care) Patient selection: Risk of selection bias (selection of patient based on future events): No or very Patients ≥ 18 years registered in the low risk. Jordi Gol Institut SIDIAP database ≥ 365 days Risk of misclassification: Yes. There is a risk of misclassification of patients (prevalent users instead of incident users), mostly in the 2010 cohort entry, because of no access to previous periods’ data. Exposure: new-users of first-line monotherapies. Diagnosis of T2DM Excluded T1DM Data extraction: ATC codes from recorded prescriptions and linked with Catalan Health Institute administrative database dispensing. No records o previfous prescription of any Risk of information bias: low. Prescriptions and dispensing are proxies of NIAD ≥ 90 days (Whashout window) administration. No data on adherence Risk of survivors bias: No or low risk (incident-users design). Misclassification, as stated above, is possible. Exposure ascertainment: As-treated approach De-identified dataset Risk of exposure misclassification: lower than in the ITT approach Covariates ascertainment: previous to the index day. Outcomes ascertainment: clinical recorded diagnoses coded by International Catalan Institut First prescription of a non-insulin Excluded Classification of Diseases, Revision 10 prior use of of Pharmacology blood glucose-lowering as first-line Risk of information bias for non-fatal outcomes: Yes. GP SIDIAP database is insulins or monotherapy not linked with hospital databases any NIAD Mortality data linked with administrative database. Risk of information bias: low

V. First-line monotherapies 83 Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 Figure V.2: Flow chart of cohorts’ selection process from T2DM patients registered in the SIDIAP database from January 1st 2010 to December 31st 2015

269,792 T2DM patients prescribed one or more non-insulin blood glucose -lowering drug (NIAD)

102,616 ≥2 NIAD agents 167,176 patients on prevalent and incident monotherapy

123,260 patients newly prescribed monotherapy

614 patients with other NIADs

GLP-1 RA 110,535 patients on 7, 739 patients on SUs 2233 (0.13 %) 2139 S MET monoterapy monoterapy Meglitinides monotherapy DPP-4 i monotherapy SGLT-2 i (89.68 %) (6.54 %) (1.77 %) (1.73%) (0.02 %)

TZDs (0.18 %)

Others

SIDIAP (Information System for the Development of Research in Primary Care) NIAD: non-insulin blood glucose-lowering (“antidiabetic”) drug; MET: metformin; SUs: sulfonylureas; DPP-4 i: dipeptidyl-peptidase-4 inhibitors; GLP-1 RA: glucagon-like peptide receptor agonist; SGLT-2 i: sodium-glucose co-transporter 2 inhibitors; TZD: thiazolidinediones

V. First-line monotherapies 84 Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 Data extraction:

Selection: Patients with an active diagnosis code of T2DM and a new prescription of any NIAD in first-line monotherapy (either metformin, sulfonylurea (SU), dipeptidyl peptidase-4 inhibitors, meglitinides, thiazolidinediones, glucagon-like peptide 1 receptor agonists and sodium-glucose co-transporter 2 inhibitors). We excluded patients with T1DM, (Coded E10.0, E10.1, E10.2, E10.3, E10.4, E10.5, E10.6, E11.8, and E11.9.; patients with gestational diabetes; patients with diabetes mellitus due to underlying condition, (E08), drug or chemical induced diabetes mellitus (E09), post pancreatectomy diabetes mellitus (E13), postprocedural diabetes mellitus (E13) and secondary diabetes mellitus NEC (E13). No further exclusion criteria were applied. This study aimed to be representative of real-world practice; this is why patients with a history of cardiovascular disease at cohort entry were not excluded.

Washout period: We extracted data from patients with no prior exposure to any blood- glucose-lowering agent (including insulins) in the 90 days before the index date.

Covariates ascertainment: Age, sex and time from T2DM diagnosis were retrieved at the date of the first prescription of the NIAD prescribed in first-line. We extracted the following values up to 3 months before the ID: standardized glycated haemoglobin (HbA1c), history of CV disease (including coronary artery disease (ICD-10 codes I20, I21, I22, I23, or I24), stroke (ICD-10 codes I61. I62, I63, I64), heart failure (ICD-10 I50.0), peripheral artery disease (ICD-10 code I73.9). Values of cholesterol levels (total, low-density lipoproteins or LDL-cholesterol, and high-density lipoproteins or HDL- cholesterol), systolic and diastolic blood pressure (BP) were extracted up to 12 months before the index date. We extracted the value of the local REGICOR index of CV risk during the 15 months prior to the index date. The most recent recorded value of the body mass index (BMI) was retrieved from records in the last 15 months before the index date. We also extracted data of concomitant medication at baseline (any active prescription at the index date) of the following therapeutic groups: ACE inhibitors/ARBs, (C09) Aldosterone antagonists (C03), Antiplatelet drugs, (B01A) β-Blockers (C07), Calcium channel blockers (C08), Diuretics (C03), Statins (C10AA), Other lipid-lowering drugs

V. First-line monotherapies 85

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

(C10A), non-steroidal anti-inflammatory drugs (NSAIDs) (M01A), and Antidepressants (N06A). Besides the study’s exposures, we extracted the concomitant use of insulin (A10A) during the study period, to adjust for insulins use.

Outcomes:

The primary outcome is a composite of three-components of major cardiovascular events (MACE): all-cause death, non-fatal myocardial infarction (MI) and non-fatal stroke. Secondary outcomes are components of MACE: myocardial infarction [ICD-10 I21), stroke (I61. I62, I63, I64), a new diagnosis of heart failure (I50.0) and onset or worsening of intermittent claudication (I73.9)

Statistical analysis:

As descriptive statistics of the sample frequencies (percentage) were used for categorical variables and mean (standard deviation) or median (interquartile range) for quantitative variables.

Twenty multiple imputations by chained equations were obtained to replace missing baseline values for HbA1c, body mass index, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides, serum creatinine, and systolic and diastolic blood pressure.

To reduce the selection bias associated with the observational nature of the data, a propensity score as the conditional probability of being treated with each NIAD group other than metformin (i.e., SUs, DPP-4 i and meglitinides) was estimated with the baseline covariates as age, gender, socioeconomic status, BMI, SBP, DBP, HbA1c, lipid profile, serum creatinine level, use of other medications (ACE inhibitors, aldosterone antagonists, antiplatelet drugs, beta-blockers, calcium channel blockers, diuretics, statins, other lipid-lowering drug, NSAIDs, antidepressants), history of acute coronary syndrome, myocardial infarction, ictus, heart failure, peripheral arterial disease, renal failure), and the duration of diabetes. Variables were considered well balanced if the standardised differences between both groups were <0.10 after adjusting by PS.

V. First-line monotherapies 86

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Inverse probability of treatment weighting (IPTW), was used to create pseudo- populations of each NIAD group initiators with similar covariate distribution. This approach has been shown to be particularly suitable when the outcome is a time to event measure.

Weights were stabilized to avoid extreme values and increase precision in the estimates. Propensity scores and IPTWs were generated for each imputed data set.

Event rates per 1,000 person-years were calculated for events along with their 95% confidence intervals (CIs). For each event, time to follow-up we defined as the time between cohort entry (the first prescription of metformin or SUs alone) and the event, prescription of a different antidiabetic drug, death, or end of study. Cox proportional hazard models were used to estimate the hazard ratios with 95% CIs.

A sensitivity analysis for the 2011-2015 period was conducted to minimise in a higher extend the risk of misclassification of prevalent-users as incident-users.

All analyses were performed by Prof. Xavier Vidal using SAS 9.4 (SAS Institute Inc., Cary, NC).

Results:

123,260 patients were newly prescribed first-line monotherapy. The most used were metformin (89.68 %) and SUs (6.54%); 4900 patients (4.04%) received other agents: DPP-4 i (1.73%), repaglinide (1.77%), thiazolidinediones (0.18%), glucagon-like peptide 1 receptor agonists (0.13%), alpha-glucosidase inhibitors (0.15%), and a negligible proportion of patients were prescribed dapagliflozin and guar gum.

There were 110,535 patients newly prescribed metformin; 7,739 incident users received SUs; 2, 139 patients were treated with PPD-4 i and 2,233 patients received meglitinides as monotherapy.

Since 2011 to 2015, incident prescriptions for first-line monotherapies of metformin, sulfonylureas and meglitinides decreased gradually by 25%, 75% and 60%, respectively, while incident prescriptions of DPP-4 i as first-line monotherapy remained stable. The

V. First-line monotherapies 87

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 sharp fall in prescriptions of all therapeutic groups between 2010 and 2011 could be caused by misclassification of prevalent users, due to the short period (90 days) applied for the exclusion of prevalent users in the SIDIAP database. Trends in prescriptions are shown in Figure 3.

Almost all SUs used were second-generation compounds, being the most prescribed gliclazide (55%), glibenclamide (28%) and glimepiride (15%). For dipeptidyl peptidase- 4 inhibitors the most prescribed agent was sitagliptin (by 51%), followed by vildagliptin and linagliptin (by 22% and 20% respectively). Repaglinide was the only prescribed agent in the meglitinides class. Absolute and relative figures of prescriptions for first-line monotherapies are shown in Table 1.

V. First-line monotherapies 88

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table 1. Therapeutic classes and agents prescribed in incident first-line monotherapy for T2DM patients, SIDIAP database, 2010-2015

Therapeutic class n Class % Total % All classes 123, 260 100.00 Biguanides 110,535 89.68 metformin 110,535 100% Sulfonylureas 7,739 6.54 chlorpropamide 9 0.12 Glibenclamide/glyburide 2152 27.77 gliclazide 4219 54.45 glimepiride 1132 14.61 glipizide 133 1.72 gliquidone 97 1.25 glisentide 6 0.08 DPP-4 i 2,139 1.73 alogliptin 1 0.05 linagliptin 473 22.11 saxagliptin 142 6.64 sitagliptin 1102 51.52 vildagliptin 421 19.68 Meglitinides 2,233 1.77 repaglinide 2233 100 Thiazolidinediones 229 0.18 pioglitazone 202 90.58 rosiglitazone 21 9.42 Alpha-glucosidase inhibitors 189 0.15 acarbose 160 84.66 miglitol 29 15.34 Glucagon-like peptide 1 RA 164 0.13 exenatide 91 55.49 liraglutide 73 44.51 Guar gum 49 49 0.04 Sodium-glucose co-transporter 2 i 22 0.02 Dapagliflozin 22

V. First-line monotherapies 89

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

There were differences among patients’ basal characteristics in the four cohorts analysed. At baseline, patients prescribed meglitinides were the oldest cohort, mean age 74.0 (SD 12.4), the subgroup ≥ 65 years had the highest percentage (77.8%); the most prolonged diabetes duration (mean 5.2 years, SD 6.0); the lowest BMI (mean 29.8, SD 5.1) and the lowest mean total cholesterol (190, SD 41.0). Patients newly prescribed with repaglinide (the only meglitinides prescribed) had the high percentage of previous CVD (36.2%), in all its manifestations; had the highest percentage of renal failure history (14.4%) and heart failure (10.7%). Patients newly prescribed repaglinide were the group most treated with antiplatelet drugs (28.0%), diuretics (46.8%) and antidepressants (8.0%). Among the four cohorts analysed, they were among the less deprived, with 30.7% of patients living in the least deprived areas and 28% in the most deprived ones.

Patients newly prescribed metformin as first-line monotherapy were more likely to be men (57.0%), obese (mean BMI 31.4 kg/1.73m2, SD 5.3) and were the youngest cohort (mean age 63.7 years, SD 12.7). The metformin cohort had the shortest diabetes duration (2.5 years, SD 3.8), the highest total cholesterol levels (mean 206.2 mg/dL, SD 40.5), the lowest percentage of history of CVD (15.0%) in all its manifestations, the lowest percentage of recorded history of heart failure (1.4%) and renal failure (2.6%), had the lowest percentage of treated patients with aldosterone antagonists (1.2%), antiplatelet drugs (17.4%), beta-blockers (13..6%), calcium-channel blockers (12.6%) and, diuretics (31.1%), and the highest percentage of patients treated with NSAIDs (12.6%). Patients were more likely to live in deprived areas (34.8% in U4 and U5) than in the least deprived urban areas (26.6% in U1 and U2).

Compared with metformin, patients newly prescribed SUs were older (mean age 70.1, SD 12.8 vs 63.7, SD 12.7), had a longer diabetes duration (mean 5.0 SD 5.3 vs. 2.5 SD 3.8 years), were more likely to have a history of CV events (27.4% vs. 15.0%), chronic kidney disease (CKD) (4.1% vs. 1.4%) and heart failure (HF) (10.2% vs. 2.6%). Length of follow up also differed, median follow-up was 2.42 years [0.97-3.97] and 1.54 years [0.53-3.12] for new users of metformin and SUs, respectively. Concerning the socio-economic status, 31% of patients lived in the most deprived areas and 26.5% in the least deprived.

V. First-line monotherapies 90

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table 2: Characteristics of cohorts of new users of NIADs first-line monotherapy at baseline*, SIDIAP database, 2010-2015

n (%) n (%) n (%) n (%) Metformin Sulfonylureas DPP-4 i Repaglinide monotherapy monotherapy monotherapy monotherapy Participants (%) 110535 (100.0) 7739 (100.0) 2139 (100.0) 2233 (100.0) Sex Women 47509 (43.0) 3654 (47.2) 1011 (47.3) 1113 (49.8) Men 63026 (57.0) 4085 (52.8) 1128 (52.7) 1120 (50.2) Age, y mean (SD) 63.7 (12.7) 70.1 (12.8) 70.8 (12.4) 74.0 (12.4) Median [IR] 64.0 [55.0 - 73.0] 72.0 [61.0 - 80.0] 72.0 [62.0 - 80.0] 77.0 [66.0 - 83.0] 18-44 7662 (6.9) 243 (3.1) 48 (2.2) 43 (1.9) 45-64 49623 (44.9) 2280 (29.5) 582 (27.2) 453 (20.3) => 65 53250 (48.2) 5216 (67.4) 1509 (70.5) 1737 (77.8) Diabetes duration mean (SD) 2.5 (3.8) 5.0 (5.3) 4.6 (5.1) 5.2 (6.0) Median [IR] 0.5 [0.0 - 4.0] 4.2 [0.1 - 7.7] 3.2 [0.0 - 6.9] 3.8 [0.0 - 8.0] 0-4 y 88276 (79.9) 4251 (54.9) 1285 (60.1) 1246 (55.8) 5-9 y 16657 (15.1) 2350 (30.4) 586 (27.4) 612 (27.4) 10-14 y 4373 (4.0) 799 (10.3) 179 (8.4) 250 (11.2) => 15 y 1229 (1.1) 339 (4.4) 89 (4.2) 125 (5.6) Haemoglobin A1c, % Result not available n (%) 26477 (24.0) 2620 (33.9) 748 (35.0) 887 (39.7) Mean (SD) 7.5 (1.5) 7.8 (1.5) 7.4 (1.3) 7.4 (1.5) median [IQR] 7.1 [6.6 - 8.0] 7.5 [6.7 - 8.4] 7.1 [6.5 - 8.0] 7.2 [6.4 - 8.1] <8 n (%) 62603 (56.6) 3303 (42.7) 1032 (48.2) 976 (43.7) 8 to 10 n (%) 15428 (14.0) 1394 (18.0) 296 (13.8) 291 (13.0) >10 n (%) 6027 (5.5) 422 (5.5) 63 (2.9) 79 (3.5) Body Mass Index (BMI) Result not available n (%) 30891 (27.9) 2652 (34.3) 677 (31.7) 766 (34.3) Mean (SD) 31.4 (5.3) 30.0 (5.2) 30.4 (5.5) 29.8 (5.1) median [IQR] 30.7 [27.7 - 34.3] 29.3 [26.5 - 32.8] 29.7 [26.6 - 33.4] 29.4 [26.4 - 32.7] < 25 6633 (6.0) 743 (9.6) 211 (9.9) 234 (10.5) 25.0 to 29.9 (overweight) 28493 (25.8) 2116 (27.3) 556 (26.0) 566 (25.3) 30.0 to 39.9 (obese I-II) 39112 (35.4) 2019 (26.1) 607 (28.4) 616 (27.6) => 40 (obese III) 5406 (4.9) 209 (2.7) 88 (4.1) 51 (2.3) Cholesterol T-C mean (SD) 206.2 (40.5) 194.0 (42.1) 191.8 (41.6) 190.5 (41.0) 204.0 [179.0 - 191.5 [166.0 - 189.0 [164.0 - 188.0 [162.0 - T-C median [IQR] 232.0] 219.0] 218.0] 217.0] HDLC mean (SD) 48.6 (12.4) 48.6 (12.9) 48.6 (13.6) 49.2 (14.3) HDLC median [IQR] 47.0 [40.0 - 56.0] 47.0 [40.0 - 56.0] 47.0 [39.0 - 56.0] 47.0 [39.0 - 57.0]

V. First-line monotherapies 91

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

LDL-C mean (SD) 125.5 (35.2) 114.6 (36.1) 112.1 (35.0) 111.6 (34.9) 124.0 [101.0 - 112.0 [90.0 - 110.0 [88.0 - 108.0 (87.0 - LDL-C median [IQR] 148.0] 137.0] 134.0 133.0) Comorbidities Prior cardiovascular events 16566 (15.0) 2122 (27.4) 642 (30.0) 808 (36.2) Acute coronary syndrome (ACS) 6498 (5.9) 837 (10.8) 242 (11.3) 318 (14.2) Myocardial Infarctiona (MI) 3371 (3.0) 369 (4.8) 120 (5.6) 137 (6.1) Stroke 5287 (4.8) 596 (7.7) 178 (8.3) 253 (11.3) Peripheric artery disease (PAD) 2499 (2.3) 287 (3.7) 108 (5.0) 134 (6.0) Renal Failure (RF) 2898 (2.6) 791 (10.2) 244 (11.4) 321 (14.4) Heart Failure (HF) 1584 (1.4) 320 (4.1) 171 (8.0) 238 (10.7) REGICOR index Result not available n (%) 65371 (59.1) 5644 (72.9) 1576 (73.7) 1755 (78.6) Low CV risk 16791 (15.2) 658 (8.5) 191 (8.9) 180 (8.1) Medium CV risk 19501 (17.6) 977 (12.6) 252 (11.8) 206 (9.2) High CV risk 8872 (8.0) 460 (5.9) 120 (5.6) 92 (4.1) Concomitant medication use Cardiovascular drugs ACE inhibitors/ARBs 47535 (43.0) 3249 (42.0) 1006 (47.0) 1058 (47.4) Aldosterone antagonists 1359 (1.2) 304 (3.9) 116 (5.4) 115 (5.2) Antiplatelet drugs 19215 (17.4) 1745 (22.5) 503 (23.5) 625 (28.0) β-Blockers 14988 (13.6) 1248 (16.1) 445 (20.8) 455 (20.4) Calcium channel blockers 13926 (12.6) 1196 (15.5) 433 (20.2) 502 (22.5) Diuretics 34401 (31.1) 2636 (34.1) 862 (40.3) 1045 (46.8) Statins 44809 (40.5) 2663 (34.4) 938 (43.9) 855 (38.3) Other lipid-lowering drugs 6738 (6.1) 373 (4.8) 156 (7.3) 104 (4.7) NSAIDs 13967 (12.6) 862 (11.1) 226 (10.6) 212 (9.5) Antidepressants 6492 (5.9) 439 (5.7) 148 (6.9) 178 (8.0) MEDEA index Rural 19585 (17.7) 1561 (20.2) 464 (21.7) 446 (20.0) Urban 5436 (4.9) 569 (7.4) 156 (7.3) 178 (8.0) U1 – Less deprived 13379 (12.1) 1050 (13.6) 326 (15.2) 351 (15.7) U2 15979 (14.5) 996 (12.9) 311 (14.5) 336 (15.0) U3 17657 (16.0) 1159 (15.0) 331 (15.5) 296 (13.3) U4 18884 (17.1) 1163 (15.0) 304 (14.2) 315 (14.1) U5 – Most deprived 19615 (17.7) 1241 (16.0) 247 (11.5) 311 (13.9) Cohort entry (yr) 2010 23983 (21.7) 3185 (41.2) 583 (27.3) 721 (32.3) 2011 19977 (18.1) 1598 (20.6) 293 (13.7) 451 (20.2) 2012 18756 (17.0) 1090 (14.1) 321 (15.0) 367 (16.4)

V. First-line monotherapies 92

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

2013 17182 (15.5) 885 (11.4) 316 (14.8) 292 (13.1) 2014 15468 (14.0) 595 (7.7) 309 (14.4) 231 (10.3) 2015 15169 (13.7) 386 (5.0) 317 (14.8) 171 (7.7) Exitus at 31Dec2015 n (%) 6226 (5.6) 1351 (17.5) 298 (13.9) 550 (24.6)

SIDIAP: Information System for the Development of Research in Primary Care MEDEA: Mortality in small Spanish áreas and economic and environmental inequalities REGICOR index: Framingham-REGICOR (Registre Gironí del Cor) local index of cardiovascular risk NIAD: non-insulin blood-glucose-lowering “antidiabetic” drugs; DPP-4 i: dipeptidyl-peptidase-4 inhibitors T-C: Total cholesterol; HDLC: High-density lipoprotein cholesterol; LDL-C: Low-density lipoprotein cholesterol * Values for HbA1c, comorbidities, concomitant use of drugs: ≤ 3 months before the index date. Values of cholesterol levels and blood pressure were extracted up to 12 months before the index date. Values of REGICOR index of CV risk during the 15 months before the index date. Values of body mass index (BMI) were the most recent ≤15 months before the index date. Significant results are highlighted in bold.

There were 2139 patients newly prescribed DPP-4 i first-line monotherapy. This cohort had the highest percentage of patients treated with statins (43.9%) and other lipid- lowering drugs (7.3%). According to the MEDEA index, the cohort was well-balanced, 28.7 % and 29.7% lived in the most and least deprived areas, respectively.

Baseline REGICOR score is not commented, due to the high percentage of missing data for all the four cohorts (range 59.1% to 78.6%).

Table 2 shows the baseline characteristics of patients newly prescribed first-line monotherapies (only metformin, SUs, DPP-4 i and repaglinide, accounting for 99.72% of the total first-line monotherapy cohort).

In the four assessed cohorts, during the study period, there were 9588 first events of MACE. There were 1004 myocardial infarctions, 2,952 events of stroke and 6,123 patients died. There were also 2,779 new diagnoses of heart failure and 2,078 of PAD. After adjusting for all available demographic, clinical and concomitant medications data at baseline and for the use of insulin after the index date and before or on the date of the event, compared with MET monotherapy alone, all other classes (SUs, DPP-4 i and meglitinides) were at higher risk of MACE (for SUs, HR 1.55 [1.42 – 1.68]; for DPP-4 i HR 1.49 [1.22-1.84] and for meglitinides HR 2.01 [1.29 -3.12]), and all-cause mortality: for SUs, HR 1.67 (1.52-1.84); for DPP-4 i, HR 1.65 (1.30-2.09) and for meglitinides HR

V. First-line monotherapies 93

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

2.08 (1.26-3.42). Patients on secretagogues monotherapies were significantly at higher risk of non-fatal MI: for SUs, HR 1.38 (1.03 – 1.85) and for meglitinides, HR 2.03 (1.10- 3.74) while for DPP-4 i there was a trend for higher risk. For non-fatal stroke, SUs monotherapy showed an excess of risk (HR 1.31 [1.11 – 1.54]). Secretagogues were also associated with a higher risk of a new record of diagnoses of heart failure (SUs, HR 1.49 [1.28-1.72] and meglitinides HR 2.15 (1.39-3.32). SUs monotherapy was also associated with a higher risk of PAD (HR 1.24 [1.02-1.51]). Among the analysed cohorts, no class showed HR favouring the use of other therapeutic groups vs metformin, for any outcome.

Table 3 shows the number of events in each cohort, the person-years period, the crude incidence rate and the adjusted hazard ratios.

We also stratified the cohorts of metformin and SUs newly prescribed first-line monotherapy patients by the use of insulin during the follow-up, and by age, sex, HbA1c, diabetes duration, body mass index (BMI), history of CVD, history of HF, history of renal failure (RF) and MEDEA deprivation index.

Patients on SUs monotherapy were at higher risk of:

. MACE at any age (but at higher risk if they were ≥75years old) if they have not used insulin during the follow-up period, had no prior history of CVD, HF or RF, and lived in rural or most deprived areas.

. MI, if they had no use of insulin, patients were ≥ 75 years, were women, had HbA1c < 8%, any time of diabetes duration, but at higher risk if they had 0-4 years of >15 years, BMI ≥ 30, had no prior CVD, no prior HF and lived in the most deprived urban areas.

. Stroke, at any age, if patients had no use of insulins during the follow-up period, had HbA1c <8%, shortest diabetes duration, higher BMI, no history of CVD, HF or CKD, and living in rural areas.

V. First-line monotherapies 94

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table 3. Adjusted hazard ratio (95% CI) of MACE, all-cause mortality and CV outcomes in cohorts of T2DM patients treated with NIADs as first-line monotherapy, SIDIAP healthcare database, 2010-2015

Outcomes N° of events Person-years Crude Adjusted HR Incidence Rate (95% CI) per 1000 p-y Composite of MACE (AMI, stroke and all-cause mortality) MET 7764 279677,66 27,76 reference SUs 1073 16369,99 65,55 1.55 (1.42-1.68) DPP-4 i 246 3519,30 69,90 1.49 (1.22-1.84) Repaglinide 505 4710,21 107,21 2.01 (1.29-3.12) All-cause mortality MET 4696 285961,85 16,42 reference SUs 817 16899,92 48,34 1.67 (1.52-1.84) DPP-4 i 185 3616,43 51,16 1.65 (1.30-2.09) Repaglinide 425 4881,87 87,06 2.08 (1.26-3.42) MI MET 865 293664,23 2,95 reference SUs 82 18677,02 4,39 1.38 (1.03-1.85) DPP-4 i 26 3914,00 6,64 1.68 (0.94-2.99) Repaglinide 31 5815,38 5,33 2.03 (1.10-3.74) Stroke MET 2573 290015,76 8,87 reference SUs 251 18354,30 13,68 1.31 (1.11-1.54) DPP-4 i 49 3870,27 12,66 1.08 (0.70-1.65) Repaglinide 79 5712,07 13,83 1.63 (0.93-2.84) Heart failure MET 2478 290560,66 8,53 reference SUs 329 18148,97 18,13 1.49 (1.28-1.72) DPP-4 i 68 3839,36 17,71 1.03 (0.72 – 1.48) Repaglinide 163 5463,92 29,83 2.15 (1.39 – 3.32) PAD MET 1797 291814,75 6,16 reference SUs 172 18522,28 9,29 1.24 (1.02-1.51) DPP-4 i 31 3901,39 7,95 1.12 (0.64 -1.97) Repaglinide 78 5695,23 13,70 1.26 (0.64 – 2.51)

NIADs: non-insulin blood-glucose-lowering (“antidiabetic”) agents MACE: composite of major adverse cardiovascular events (MI, stroke and all-cause mortality); MI: myocardial infarction; PAD: peripheral artery disease; SIDIAP: Information System for the Development of Research in Primary Care; MET: metformin; SUs: sulfonylureas; DPP-4 i: dipeptidyl-peptidase 4 inhibitors.

V. First-line monotherapies 95

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

. HF with no insulin use, higher risks when older, male, had lower HbA1c, BMI from 25 to 39.9, no history of HF and rural or living in most deprived urban areas. The risks for HF were high irrespective of previous events of CV or CKD diagnosis, but the magnitude was greater for those with no prior events of CV or CKD diagnosis. . PAD with no use of insulin, elderly, HbA1c <8%, diabetes duration 5-9 years, BMI 30-39.9, and no history of HF or CKD. The risk of PAD for SUs users was higher than for metformin users irrespective of they had past diagnoses of CVD or not, or area of residence.

In general, SUs new users were at higher risk of for all the outcomes if they had no insulin therapy during the follow-up, HbA1c < 8%, shorter diabetes duration, older age, no prior history of CVD, HF or CKD and living in most deprived areas. Women using SUs were at higher risk of MI and men had a higher risk of HF. Diabetes duration > 15 years was positively associated with the occurrence of MI and PAD.

Living in deprived areas was positively associated with MACE, all-cause mortality, MI and HF, but not with stroke or PAD.

Results of subgroup analyses are shown in Table V.4.

Among patients on metformin monotherapy who received a second-line agent, 17,541 patients received SUs (52.68%); while among those who had been newly prescribed SUs, 3,217 patients (41.57%) were added MET as second-line combined treatment.

For second-line dual therapies, compared with the addition of SUs to metformin, the cohort of patients receiving metformin added to SUs did not present a statistically higher risk in any of the endpoints analyzed. Results are shown in Table V.5.

A sensitivity analysis was performed with drug-naïve patients who initiated their first- line monotherapy from January 1st 2011 to December 31st 2015. Results are shown in Table V.6.

V. First-line monotherapies 96

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table V.4. Adjusted HR (95% CI) of subgroup analyses for MACE, mortality, MI, stroke events and HF and PAD diagnosed in T2DM patients treated with first-line sulfonylureas (SUs) compared with metformin (MET) in the SIDIAP database, 2010-2015

MACE All-cause mortality MI Stroke HF PAD MET (total) reference reference reference reference reference Reference SUs (total) 1.55 (1.42-1.68) 1.67 (1.52-1.84) 1.38 (1.03-1.85) 1.31 (1.11-1.54) 1.49 (1.28-1.72) 1.24 (1.02-1.51) Use of insulins after the index date Ins Yes 1.23 (0.97 – 1.55) 1.35 (1.07 – 1.71) 0.70 (0.22- 2.20) 0.78 (0.35-1.72) 1.10 (0.67-1.82) 1.18 (0.67-2.08) Ins No 1.72 (1.57- 1.87) 1.90 (1.71 – 2.10) 1.63 (1.22-2.18) 1.45 (1.23-1.71) 1.69 (1.45-1.96) 1.40 (1.15-1.72) Age < 75 1.45 (1.32-1.59) 1.46 (1.32 – 1.62) 1.46 (0.99 - 2.14) 1.30 (1.07-1.58) 1.46 (1.24-1.71) 1.12 (0.84-1.51) ≥ 75 1.65 (1.43-1.91) 2.00 (1.66 – 2.41) 1.72 (1.16-2.54) 1.36 (1.05- 1.77) 1.55 (1.17- 2.06) 1.50 (1.17- 1.92) Sex Male 1.70 (1.52 – 1.89) 1.82 (1.61 – 2.07) 1.38 (0.98-1.94) 1.42 (1.13-1.77) 1.78 (1.45-2.20) 1.34 (1.07- 1.67) Female 1.74 (1.53 – 1.97) 1.74 (1.51 – 2.00) 2.13 (1.29- 3.51) 1.40 (1.11- 1.77) 1.48 (1.22-1.80) 1.53 (1.03- 2.26) HbA1c HbA1c < 8 1.73 (1.56-1.93) 1.91 (1.70-2.15) 1.65 (1.13-2.40) 1.42 (1.13-1.79) 1.79 (1.50-2.14) 1.37 (1.06-1.79) HbA1c 8-10 1.37 (1.12-1.69) 1.38 (1.10-1.75) 1.34 (0.74 -2.44) 1.31 (0.86-1.99) 1.14 (0.76 -1.70) 1.37 (0.95- 1.99) HbA1c > 10 1.71 (1.12-2.62) 2.04 (1.25- 3.32) 1.36 (0.40-4.66) 1.54 (0.70 - 3.38) 1.66 (0.74- 3.76) 1.36 (0.62 -3.01) Diabetes duration 0-4 y 1.92 (1.73 – 2.13) 2.03 (1.80- 2.30) 1.67 (1.18- 2.37) 1.62 (1.32-1.97) 1.83 (1.52-2.20) 1.27 (0.99-1.64) 5-9 y 1.32 (1.14 – 1.53) 1.46 (1.24-1.73) 1.22 (0.71-2.11) 0.97 (0.71-1.31) 1.21 (0.94 -1.56) 1.67 (1.17- 2.39) 10-14 y 1.30 (1.05 – 1.61) 1.21 (0.94 - 1.57) 1.37 (0.63- 2.95) 1.09 (0.70-1.70) 1.33 (0.89-1.99) 1.19 (0.71 -1.98) ≥ 15 y 0.92 (0.64 – 1.33) 0.95 (0.62- 1.43) 1.84 (0.51-6.64) 0.66 (0.29-1.51) 1.00 (0.52 -1.90) 2.07 (0.88- 4.87) BMI < 25 1.35 (1.08-1.68) 1.34 (1.04 -1.72) 1.51 (0.72-3.17) 1.26 (0.81 -1.94) 0.97 (0.56 -1.67) 1.20 (0.67- 2.15) 25 - 29.9 1.41 (1.21-1.64) 1.57 (1.32 -1.87) 1.15 (0.66 -2.00) 1.17 (0.88-1.56) 1.87 (1.44- 2.43) 1.22 (0.85- 1.75) 30 - 39.9 1.90 (1.63- 2.21) 2.13 (1.79- 2.53) 1.77 (1.07-2.95) 1.59 (1.21-2.09) 1.75 (1.39-2.19) 1.55 (1.15- 2.10) ≥ 40 1.94 (1.13-3.35) 1.74 (0.89- 3.42) 2.47 (0.48-12.69) 1.87 (0.73-4.76) 1.03 (0.42-2.56) 1.46 (0.38- 5.59) CVD history Prior CVD no 1.83 (1.65-2.03) 2.06 (1.82-2.32) 1.58 (1.10 - 2.27) 1.56 (1.29-1.89) 1.68 (1.37- 2.05) 1.37 (1.08 -1.75) Prior CVD yes 1.20 (1.05 -1.37) 1.25 (1.08 - 1.44) 1.39 (0.88 -2.18) 0.96 (0.72-1.29) 1.41 (1.15-1.72) 1.30 (0.95- 1.79) HF history Prior HF no 1.71 (1.57 – 1.87) 1.90 (1.72-2.10) 1.61 (1.20- 2.15) 1.42 (1.20- 1.69) 1.63 (1.40- 1.91) 1.41 (1.15- 1.73) Prior HF yes 0.92 (0.76- 1.12) 0.95 (0.77 - 1.18) 0.57 (0.22 - 1.52) 1.04 (0.63- 1.69) 1.37(0.95- 1.97) 0.92 (0.54- 1.58) CKD CKD no 1.66 (1.52-1.81) 1.81 (1.64-2.00) 1.54 (1.15 -2.07) 1.41 (1.20 - 1.67) 1.61 (1.38-1.88) 1.37 (1.12- 1.67) CKD yes 1.13 (0.83 -1.54) 1.02 (0.73 -1.42) 1.74 (0.62 -4.91) 1.10 (0.59-2.03 1.54 (1.01-2.33) 1.40 (0.73 - 2.68) MEDEA index Rural 1.68 (1.42-2.00) 1.89 (1.56-2.30) 1.23 (0.67- 2.25) 1.38 (0.98-1.95) 1.48(1.09-2.02) 0.98 (0.58-1.66) U1 (least deprived) 1.30 (1.02-1.66) 1.41 (1.04-1.91) 0.83 (0.36 -1.90) 1.41 (0.95 -2.09) 1.59 (1.00 -2.52) 1.31 (0.75-2.27) U5 (most deprived) 1.68 (1.3- 2.15) 2.09 (1.55- 2.80) 2.58 (1.25- 5.34) 0.89 (0.53 -1.50) 1.59 (1.08- 2.32) 0.94 (0.52-1.71)

SIDIAP: Information System for the Development of Research in Primary Care; MEDEA: Mortality in small Spanish areas and economic and environmental inequalities; MACE: composite of major adverse cardiovascular events; MI: myocardial infarction; PAD: peripheral artery disease MET: metformin; SUs: sulfonylureas

V. First-line monotherapies 97

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table V.5. Adjusted HRs for MACE, all-cause mortality, MI, stroke, HF and PAD in cohorts of T2DM patients treated with metformin and sulfonylureas, as monotherapy or as second-line dual therapy, SIDIAP healthcare database, 2010- 2015

Outcomes N° of events Person-years Crude Adjusted HR Incidence Rate per 1000 p-y Composite of MACE (AMI, stroke and all-cause mortality) MET 7764 279677,66 27,76 reference SUs 1073 16369,99 65,55 1.49 (1.37-1.62) MET + SUs 1053 32190,60 32,71 reference SU + MET 361 6340,48 56,94 1.13 (0.97-1.31) All-cause mortality MET 4696 285961,85 16,42 reference SUs 817 16899,92 48,34 1.56 (1.42-1.72) MET + SUs 751 32771,49 22,92 Reference SU + MET 277 6536,35 42,38 1.13 (0.95-1.35) MI MET 865 293664,23 2,95 Reference SUs 82 18677,02 4,39 1.46 (1.11-1.93) MET + SUs 91 32614,02 2,79 Reference SU + MET 28 6496,67 4,31 1.19 (0.78-1.83) Stroke MET 2573 290015,76 8,87 Reference SUs 251 18354,30 13,68 1.31 (1.11-1.54) MET + SUs 269 32341,79 8,32 Reference SU + MET 87 6373,20 13,65 1.13 (0.87-1.46) Heart failure MET 2478 290560,66 8,53 Reference SUs 329 18148,97 18,13 1.46 (1.26-1.69) MET + SUs 294 32337,17 9,09 Reference SU + MET 88 6417,06 13,71 1.01 (0.79-1.30) PAD MET 1797 291814,75 6,16 Reference SUs 172 18522,28 9,29 1.29 (1.07-1.55) MET + SUs 245 32381,58 7,57 Reference SU + MET 56 6450,40 8,68 0.98 (0.72-1.31)

MACE: composite of major adverse cardiovascular events (MI, stroke and all-cause mortality); MI: myocardial infarction; PAD: peripheral artery disease; SIDIAP: Information System for the Development of Research in Primary Care; MET: metformin; SUs: sulfonylureas; MET + SU: a sulfonylurea was added to previous metformin monotherapy; SU + MET: metformin was added to previous sulfonylurea monotherapy

V. First-line monotherapies 98

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 Table V.6. Adjusted hazard ratio (95% CI) of MACE, all-cause mortality and CV outcomes in cohorts of T2DM patients treated with NIADs as first-line monotherapy, SIDIAP healthcare database, 2011-2015.

Outcomes N° of events Person-years Crude Adjusted HR Incidence Rate (95% CI) per 1000 p-y 2011-2015 period Composite of MACE (AMI, stroke and all-cause mortality) MET 5436 212161.30 25.62 reference SUs 702 11506.55 61.01 1.49 (1.35 – 1.66) DPP-4 i 198 2786.45 71.06 1.58 (1.24 – 2.01) Repaglinide 356 3415.41 104.23 2.06 (1.32 – 3.23) All-cause mortality MET 3123 216308.05 14.44 reference SUs 506 11846.04 42.71 1.59 (1.41 – 1.80) DPP-4 i 150 2862.54 52.40 1.76 (1.30 – 2.38) Repaglinide 288 3528.49 81.62 2.20 (1.28 – 3.79) MI MET 674 220345.80 3.06 reference SUs 57 12750.93 4.47 1.36 (0.97 – 1.91) DPP-4 i 20 3073.88 6.51 1.82 (0.96 – 3.46) Repaglinide 25 4038.38 6.19 1.65 (0.89 – 3.05) Stroke MET 1929 218084.82 8.85 reference SUs 184 12547.63 14.66 1.33 (1.10 – 1.61) DPP-4 i 36 3041.05 11.84 1.07 (0.67 – 1.71) Repaglinide 62 3982.37 15.57 1.74 (0.94 – 3.19) Heart failure MET 1842 218454.12 8.43 reference SUs 234 12445.29 18.80 1.47 (1.24 – 1.76) DPP-4 i 53 3030.45 17.49 1.00 (0.63 – 1.58) Repaglinide 107 3859.18 27.73 1.94 (1.26 – 2.99) PAD MET 1396 219132.97 6.37 reference SUs 117 12648.87 9.25 1.22 (0.97 – 1.53) DPP-4 i 23 3074.43 7.48 1.07 (0.54 – 2.13) Repaglinide 57 3981.46 14.32 1.38 (0.72 – 2.66)

NIADs: non-insulin blood-glucose-lowering (“antidiabetic”) agents MACE: composite of major adverse cardiovascular events (MI, stroke and all-cause mortality); MI: myocardial infarction; PAD: peripheral artery disease; SIDIAP: Information System for the Development of Research in Primary Care; MET: metformin; SUs: sulfonylureas; DPP-4 i: dipeptidyl-peptidase 4 inhibitors. Covariates included estimating propensity-score are the same than the full-study period set.

V. First-line monotherapies 99

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Discussion

In our study, in first-line monotherapy, second-generation sulfonylureas were associated with a significant excess risk of MACE, all-cause mortality, MI and stroke events and risk of HF and PAD diagnoses, compared with metformin. Interestingly, SUs users with no history of CVD, HF or CKD had increased risk for all outcomes. These results are consistent in direction with those obtained in observational studies comparing SUs and MET in monotherapy, although the magnitude of estimates varied.

The meta-analysis of Lamanna et al. found that metformin was not associated with significant harm or benefit on cardiovascular events (MH-OR 0.94[0.82-1.07], p = 0.34); although a significant benefit was observed in trials versus placebo/no therapy (MH-OR 0.79[0.64-0.98], p = 0.031), but not in active-comparator trials (MH-OR 1.03[0.72-1.77], p = 0.89). There was a significant correlation of the effect of metformin on cardiovascular events with trial duration and with minimum and maximum age for inclusion.[201]

Except for a few prospective studies [202, 203.] and a meta-analysis [204], most of the observational evidence regarding the safety of newly treated T2DM patients, either with SUs or metformin, reported a higher risk of all-cause and CV mortality for SUs users. [196-200, 167, 162]

A meta-analysis of 82 RCTs and 26 observational studies found that SUs increase risks of all-cause mortality and cardiovascular-related mortality compared with all other treatments combined; the risk of myocardial infarction was significantly higher for SUs compared with DPP-4 inhibitors and sodium-glucose co-transporter-2 inhibitors, and the risk of stroke was significantly higher for SUs than for DPP-4 inhibitors, GLP-1 RAs, thiazolidinediones (TZD) and insulin. [205]

Schramm et al. found that monotherapy with the most used insulin secretagogues, including glimepiride, glibenclamide, glipizide, and tolbutamide was associated with increased mortality and cardiovascular risk compared with metformin; while gliclazide and repaglinide appear to be associated with a lower risk than other insulin secretagogues.[200]

V. First-line monotherapies 100

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

In second-line therapies combining metformin and sulfonylureas, there were no differences in any outcome between cohorts of patients previously treated with SUs compared with patients who were first treated with metformin and were added SUs as second-line treatment. These results suggest that patients on combined therapies including metformin can have lesser risks than those on SUs alone, irrespective of the initial monotherapy. To avoid immortal-time bias and time-lag bias, we didn’t do comparisons between patients on first-line and second-line therapies. Instead, Gulliford et al. reported [206] reported no differences in mortality between the second-lines of combined therapies metformin and SUs and their respective first-lines monotherapies; in this study, only patients who have survived to the first-line therapy could be included in the combination one, thus presenting with immortal-time bias. Fisman et al. in their prospective observational study, compare mortality in patients with metformin monotherapy and combination therapy, thus comparing different severity stages and presenting with lag-time bias, reported an increase of mortality with the combination therapy. [203]

Other monotherapies assessed in our study didn’t show any benefit compared with metformin: both meglitinides (repaglinide) and DPP-4 i increased risk of mortality and MACE. Evidence from CVOTs with DPP-4 inhibitors (TECOS [sitagliptin], SAVOR- TIMI [saxagliptin], and CARMELINA [linagliptin]) obtained neutral results on cardiovascular outcomes and mortality vs placebo.[85, 89, 93-] On the other hand, most of the observational evidence obtained favourable results for the use of DPP-4 i. (Appendix A ) However, the CAROLINA trial assessed a head-to-head comparison of linagliptin vs glimepiride and didn’t show differences between agents; this result should raise scepticism about the “no effect” of this DPP-4 i, given that glimepiride belongs to the SU class. [206] Of note, in a recently published observational study linagliptin was associated with a non-significant 9% decreased risk in the composite cardiovascular outcome (HR 0.91 [0.79- 1.05]) [207] In our study patients on DPP-4 i monotherapy were at a significantly increased risk of MACE and all-cause mortality; and an excess of risk of MI cannot ruled out, this finding deserves further research.

Scheller et al. didn’t find differences between sitagliptin monotherapy compared with metformin monotherapy in the risk of all-cause mortality or the composite endpoint

V. First-line monotherapies 101

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

MACE but was associated with an increased likelihood of changing glucose-lowering treatment. [209] Sitagliptin was the monotherapy used by almost 51% of patients treated with DPP-4 i in our study.

Ou et al. found that metformin users had significantly lower risks for composite CVD risk (aHR 0.87, 95 % CI 0.79-0.94), as well as those for MI, HF, and hypoglycemia, as compared to those of DPP4i users. [210]

Results of 19 observational studies assessing the use of DPP-4 i vs no use, metformin, SUs, insulin and other antidiabetic drugs (OADs) are summarised in the Appendix A. Only one cohort study [211] found an excess of risk of MI and percutaneous revascularisation with sitagliptin vs non-use in patients with chronic kidney disease and myocardial infarction. Another cohort study in Korean patients [212] found that, compared with metformin, sitagliptin was not associated with an elevated risk of CV complications including myocardial infarction, ischemic stroke, heart failure, and coronary revascularisation, compared to metformin. In patients with high CV risk, results were similar.

Repaglinide was the only meglitinide prescribed in our study. Repaglinide users were significantly at higher risk of MACE, all-cause mortality and MI. There are a few observational studies assessing meglitinides in first-line treatments. Huang et al. found an increased risk of MACE, HR 1.69 (1.25–2.59) and all-cause mortality, HR 1.88 (1.45– 2.43) in repaglinide users, compared with glimepiride users. [213] In the study conducted by Ou et al., compared to DPP4i users, meglitinides users were at higher risk of the composite of CV events (HR 1.3 [1.20-1.43]), as well as those for stroke, MI, HF, and hypoglycemia. [210] In our study, patients treated with meglitinides had a higher risk than those on SUs for MACE, all-cause mortality and MI. Basal characteristics of patients prescribed meglitinides show more basal comorbidities than all other groups. Although these unbalanced characteristics are controlled for confounding, other unassessed covariates could have contributed to residual confounding.

Our study has several strengths: it presents observational data from large cohorts of unselected patients with and without a history of past cardiovascular disease, heart and

V. First-line monotherapies 102

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 renal failure who attend routine clinical care, fully representative of the “real-world” use. The SIDIAP database of the nationwide healthcare system has extensively used for diabetes research. DM has been associated with increased mortality: all-cause mortality, involves cardiovascular, cancer and other morbidities’ death, being a proper measure to assess therapies’ benefits or harms. [25] Mortality records are linked with an administrative database and reflect all deaths, either within and outside hospitals. Another strength is that we assessed the first recorded diagnosis of PAD; a few observational studies included PAD in their outcomes. [214] PAD has been found to be the first manifestation of CVD in T2DM patients. [215]

As significant limitations, we have no data on the population’s smoking status, alcohol consumption or physical activity. The dataset is not linked with hospital registries, which is a source of non-differential information bias regarding events not recorded by GPs. In people with a history of HF or PAD at baseline, we can’t know if the new diagnosis refers to a worsening or hospitalisation. We have not retrieved data on hypoglycaemic events, that can be related to falls, hip fractures and CV and overall mortality. The number of patients in nursing homes was not available; elder people living in nursing homes are at higher risk of mortality.[216]. Data about the administered medication during hospitalisation, if any, were not available, and this can introduce risk of non-differential misclassification bias. Another limitation of our study is the lack of data on adherence, though prescriptions and dispensing are both proxies of drug administration. Additionally, we have not analysed different doses of the same agent, or stratified by agents in the cases of SUs and DPP-4 i. Another important remark is that prescriptions of first-line monotherapy agents fell sharply between 2010 and 2011, while steadily and slightly decrease since 2011, with differences among groups. As the coded, de-identified dataset which serves as a basis for our study excluded patients with a previous prescription of the given agent within the 90 days before the index day, it is plausible that this period was not long enough, and prevalent-users were included instead of incident-users only. (Fig. V.4) The exclusion of prescriptions initiated in 2009 could not have been enough to remove prevalent users entirely. A sensibility analysis was performed with data of 2011-2015 monotherapy cohorts to exclude more extensively possible prevalent patients. In the sensibility analysis, the

V. First-line monotherapies 103

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 results for the primary and secondary outcomes of the 2011-2015 period are similar in magnitude and direction to those of 2010-2015 period. (Table V.6)

High-quality observational studies have suggested an increased CV risk for SUs. One of the shortcomings highlighted in these studies is the use of a composite endpoint, [167] introducing possible masking of the associations with the individual components. We assessed both the composite – as primary - and individual components – as secondary outcomes, and results were consistent for an increased risk of SUs. However, our results showed higher estimates than those reported in other studies, but are close to the one reported by Filion et al. for all-causes mortality.[197, 167] Residual confounding and bias might explain these differences, but also they can be due to the choice of metformin as a comparator, [197] and the outcome all-cause mortality instead of CV mortality both in primary composite and in the individual secondary outcome. In all-cause mortality, hypoglycaemia events and possibly cancer or other causes could play a major role. [197, 25]

Excluding the data of 2010, trends in incident first-line monotherapies prescriptions could reflect a decrease in new diagnoses of T2DM, as highlighted by other authors. [3] Prescriptions of meglitinides and sulfonylureas decreased more than twice those of metformin (60% and 75% vs 25%), thus showing adherence to the local guidelines. [193]

A population-based study recently published compared major adverse cardiovascular events (MACE) among patients with diabetes who reached a reduced kidney function threshold and continued treatment with metformin or a sulfonylurea.[217] The study found better cardiovascular and mortality outcomes for patients treated with metformin monotherapy compared with SUs monotherapy. Patients who continued metformin monotherapy had reduced risk for the primary composite outcome of hospitalization for acute myocardial infarction (AMI), ischemic or hemorrhagic stroke, transient ischemic attack (TIA), or date of cardiovascular death, HR 0.80 (0.75 to 0.86) and cardiovascular death (HR 0.70 [0.63 - 0.78]).

V. First-line monotherapies 104

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

In our study, in patients with first-line treatment, 89,68% of patients were treated with metformin, according to local and international recommendations. A low percentage of patients who were newly prescribed metformin had a history of HF and or renal failure (4% adding the two proportions), suggesting that the contraindications amended in 2006 (for HF) and 2016 (for RF) could still be taken into account by prescribers. Among all the first-line newly prescribed patient, 7,739 patients (6.54%) were treated with SUs. In the SUs cohort, 10.2 % of patients had a history of RF, and 4.1% of HF, accounting for a total of 13.6% of patients with both conditions.

Conclusion

In this study, compared with metformin, the use of sulfonylureas in first-line monotherapy is associated with higher risks of all-cause mortality, MI, stroke, HF, PAD and MACE. This observational evidence strengths the concept of first-line use of metformin and adds confirmatory concerns about mortality and CV effects of sulfonylureas monotherapy. Additionally, in our study, the use of other agents (secretagogues such as repaglinide and DPP-4 i) didn’t show beneficial effects over metformin. The results of second-line comparison suggest that metformin has beneficial effects when added to SUs monotherapy.

In this study, roughly 86.4% of patients on SUs were free both of HF and RF at baseline could have benefited from metformin-based regimes.

If no sound contraindications exist, the use of metformin-based treatments should always be considered. Recommendations for metformin use should be strengthened in clinical practice.

V. First-line monotherapies 105

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

VI. Cardiovascular outcomes and mortality in type 2 diabetes mellitus patients prescribed second-line, metformin-based non-insulin blood- glucose-lowering agents dual therapies

VI. Second-line, metformin-based, dual therapies

Cardiovascular outcomes and mortality in type 2 diabetes mellitus patients prescribed second-line, metformin-based non-insulin blood-glucose-lowering agents dual therapies

Raquel Herrera Comoglio, Xavier Vidal Guitart

Background: Type 2 diabetes mellitus is a progressive disease and patients need intensification therapies as the condition worsens. Cardiovascular disease is the leading cause of morbidity and mortality among T2DM patients, the choice of non-insulin blood-glucose-lowering (“antidiabetic”) agents (NIAD) should consider potential cardiovascular (CV) effects.

Aim: To evaluate the risk of major adverse cardiovascular events (MACE), myocardial infarction (MI), stroke, all-cause mortality, heart failure (HF) and peripheral artery disease (PAD) in cohorts of type 2 diabetes patients newly prescribed a NIAD as a second-line agent added to metformin.

Methods: Longitudinal, observational study of cohorts of T2DM adult patients who received a second-line NIAD added to previous metformin monotherapy. Patients were followed since the date of the first prescription of the second-line agent (ID, index date) to the first event of the primary outcome (MACE) or secondary outcomes (all-cause mortality, MI, stroke, HF and PAD) or censoring. Patients’ basal characteristics were

VI. Second-line, metformin-based, dual therapies 107

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

ascertained before the ID. Cox proportional hazard models were used to estimate hazard ratios with 95% confidence interval (HR [95%CI]).

Results: Among 110,535 patients first-prescribed metformin monotherapy during the study period, 28,539 patients (25.81%) received a second-line NIAD as intensification therapy: SUs, Compared to MET+SUs, no difference was found in the risk of the primary and secondary outcomes in patients treated with DPP-4 i. Patients on MET + meglitinides were at an increased risk of MACE and all-cause mortality (HR 1.19 [1.01- 1.42] and all-cause mortality (HR1.38 [1.14-1.67])

Conclusion : With metformin plus sulfonylureas as reference, no significant differences were found between in the cohort treated with DPP-4 i. Patients with meglitinides were at higher risk of the composite MACE and all-cause mortality. Residual confounding cannot be ruled out, deserving further research.

Keywords : metformin, sulfonylureas, meglitinides dipeptidyl peptidase-4 inhibitors (DPP-4 i), glucagon-like peptide-1 receptor agonists (GLP-1 RA), sodium-glucose transporter 2 inhibitors (SGLT-2 i), thiazolidinediones.

Introduction

Type 2 Diabetes Mellitus is a chronic progressive condition disease characterised by elevated blood glucose levels. Diabetes can lead to cardiovascular disease, and microvascular and nerves damage, and premature death. [218] As the glycaemic control fails as the condition worsens, most patients need multiple therapies to attain these glycemic target levels in the longer term.[219]

Glycemic control of diabetes mellitus did not improve substantially in the past ten years; the proportion of diabetic patients achieving a haemoglobin A1c (A1C) target <7% is still around 50%.[220] The Catalan Health Institute has issued clinical

VI. Second-line, metformin-based, dual therapies 108

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

guidelines, current objectives of diabetes clinical care include an A1C value <7%, blood pressure (BP) < 140/90, and when possible, 130/80 mmHg24, LDL cholesterol (LDL- C) value ≤ 130 mg/dL (3,35 mmol/L), in primary prevention and ≤ 100 mg/dL (2,6 mmol/L) in secondary prevention.[193]

Metformin is currently the first drug of choice for T2DM. The Catalan guideline states that metformin should be used in all patients, unless contraindicated or not tolerated. Other blood-glucose-lowering drugs are recommended to be used in second-line, in combination with metformin. Other agents should be added in a third-line treatment if glycaemic targets are not achieved. In all stages, the control of cardiovascular risk factors includes lifestyle interventions and pharmacological treatment.

Sulfonylureas stimulate the insulin release by binding to the sulfonylurea receptor 1 (SUR1), which is part ATP-sensitive potassium channel. Agents within the class vary considerably in their pharmacokinetic properties, have high bioavailability and reach peak plasma concentrations within 1.5–4.0 h, have long half-lives, being their effect more extended. SUs bind to plasma proteins, which can lead to interactions with other drugs such as salicylates, sulfonamides and warfarin. SUs are metabolized in the liver and eliminated via the bile and urine, which restrains their use patients with hepatic and/or renal impairment.[221] Hypoglycaemia and weight gain are known adverse effects of this class of drugs. The cardiovascular safety of SUs has raised concerns since the early 1970s, though in the ten years follow-up analysis mortality and MI were reduced in the cohort initially treated with intensive therapy. [55, 187] (see also Section I, Introduction) Meta-analysis of RCTs have found SUs to increase mortality and CV mortality. [198, 205, 222,] Observational studies have also found more harm with the use of SUs than with other therapies. [205] [Appendix B]

As sulfonylureas (SUs), meglitinide analogues are insulin secretagogues that stimulate insulin release by inhibiting ATP-sensitive potassium channels of the beta-cell membrane, though binding to a receptor distinct from that of SUs. Both repaglinide and nateglinide are absorbed rapidly, stimulate insulin release within a few minutes, are rapidly metabolised in the liver and are mainly excreted in the bile. Therefore, following

VI. Second-line, metformin-based, dual therapies 109

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

the preprandial administration of these drugs, insulin is more readily available during and just after the meal. Pharmacokinetic characteristics allow a significant reduction in postprandial hyperglycaemia without the danger of hypoglycaemia between meals. The short action and biliary elimination make repaglinide and nateglinide especially suitable for geriatric patients or in whom one of the other first-line antidiabetic drugs, i.e. metformin, is strictly contraindicated (e.g. nephropathy with creatinine clearance < or = 50 ml/min). [223] A study found that glibenclamide and repaglinide have similar efficacy in reducing blood-glucose markers over an eight week period when these agents are used to treat patients with inadequate glycaemic control on metformin monotherapy; Repaglinide improved postprandial glycaemic control by stimulating early-phase insulin release. [224] Meglitinides have found to have similar blood- glucose-lowering potency than metformin. [225] Repaglinide, in addition to metformin, seems to be effective in reducing glucose-related markers. In a meta-analysis of 11 RCTs, repaglinide plus metformin was significantly more effective in decreasing two hours postprandial glucose-levels than glimepiride plus metformin; there were fewer events of hypoglycemia in the repaglinide plus metformin group, and no differences in HbA1c and fasting blood glucose.[226] However, there is scarce evidence concerning cardiovascular risk.

Dipeptidyl peptidase-4 inhibitors (DPP-4i) inactivates the proteolysis of the incretin hormones glucagon-like peptide 1 (GLP-1) and glucose dependent insulinotropic peptide (GIP), resulting in an increased glucose dependent insulin secretion. Native glucagon- like peptide-1 (GLP-1) and glucose-dependent insulinotropic polypeptide (GIP) account for up to 60% of postprandial insulin release, termed “incretin effect”. GLP-1 receptor’s activation potentiates insulin secretion from pancreatic beta cells and lower inappropriate high glucagon secretion in a glucose-dependent manner. GLP-1 receptor is also expressed in extrapancreatic tissues (gastrointestinal tract, heart, vasculature, and central and peripheral nervous system), what is thought to have pleiotropic effects. DPP-4 i stimulate both post-meal insulin secretion and inhibit glucagon secretion in a glucose-dependent manner, with low risk of hypoglycaemia. Their effect on weight is neutral. [227, 228].

VI. Second-line, metformin-based, dual therapies 110

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Four large CVOTs has shown no harmful or beneficial effects of DPP-4 i in CV outcomes [85, 87, 89, 93]

We conducted an observational, population-based cohort study to assess major cardiovascular outcomes and all-cause mortality in patients who were prescribed a second-line non-insulin blood glucose-lowering agent (“antidiabetic” agent, NIAD) added to metformin.

In chronic and evolving diseases, to differentiate clinical events as manifestations of the natural progression of the disease or as effects of treatments is challenging; moreover when treatment vary over time, and patients characteristics also change over the time. [182] We performed comparison in second-line treatment added to metformin, to avoid comparisons between different severity of the disease. hen assessing the effects of treatments. As diabetes is an established risk factor for cardiovascular disease, longer diabetes duration or more severe disease are both associated with higher cardiovascular risk, as well as history of cardiovascular disease. Patients with poor glycaemic control requiring drug intensification in second or third-line therapies are likely to experience more CV events. To reduce confounding, comparisons between cohorts of patients should be made in the same lines of treatment.

The general characteristics of the study design have been described in other sections, only specifically those related to this study are reported.

Design

Analysis of longitudinal electronic health records data. We did a population-based cohort study of adult T2DM patients who had started a first-line monotherapy with metformin during the study period (1st January 2010 to 31st December 2015) and were first prescribed a second-line non-insulin blood-glucose-lowering drug. Thus, we used a new- user design to reduce bias. Fig V.1 represents graphically the study design and risks of bias.

VI. Second-line, metformin-based, dual therapies 111

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Figure VI.1: Summary of design characteristics and design-related potential sources of bias Data source Study design: Observational retrospective cohort study SIDIAP database *: confounding by indication. Residual confoundingRisk of bias Objective: To compare metformin-based second-line combined dual therapies. Patient selection *Risk of time-lag bias: low. Intensification with a second NIAD was considered a proxy of disease severity Patients ≥ 18 years registered in the CIH database Data source: SIDIAP (Information System for the Development of Research in Primary Care) ≥ 365 days Selection process:

Institut *Risk of selection bias: very low risk. Patients initiating metformin monotherapy during the study period were selected only for new prescription of NIAD. Excluded Patients Diagnosed of T2DM Risk of misclassification: Low. T1DM Data extraction: Exposure: new-users of any NIAD prescribed as second-line therapy added to metformin Jordi Gol Gol Jordi No records o previfous prescription of any NIAD ≥ 90 days monotherapy. (Whashout window) Data extraction: ATC codes from recorded prescriptions and linked to the Catalan Institut of Health (CIH) administrative database dispensing *Risk of information bias: Low concerning exposures. De-identified dataset Prescriptions and dispensing are proxies of administration. No data on adherence * Risk of survivors bias: Low risk (incident-users design). Misclassification, as stated above, is possible. Excluded of Exposure ascertainment: As-treated approach Cohort of patients first prescription metformin as patients with *Risk of exposure misclassification: lower than in the ITT approach first-line monotherapy during the study period prior use of Covariates ascertainment: previous to the index day insulin Outcomes ascertainment: clinical recorded diagnoses coded by International Classification of Institut Study population Diseases, Revision 10 (ICD-10) *Risk of information bias: Yes. GP SIDIAP database is not linked with Hospital

Pharmacology databases Patients on metformin monotherapy first added a Catalan Mortality through administrative database linkage: low risk of information bias for second NIAD as second-line dual therapy mortality-

ATC: Anatomical-Therapeutic-Chemical code; AT: As-treated; CIH: Catalan Institut of Health; ITT: Intention-to-treat; NIAD: non-insulin blood-glucose lowering “antidiabetic” drug. SIDIAP: Information System for the Development of Research in Primay Care; T1DM: Type 1 diabetes mellitus; T2DM: Type 2 diabetes mellitus.

VI. Second-line, metformin-based, dual therapies 112 Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 Figure VI.2 : Flow chart of cohorts’ selection process from T2DM patients registered in the SIDIAP database from January 1st 2010 to December 31st 2015

269,792 T2DM patients prescribed one or more non-insulin blood glucose -lowering drug (NIAD)

123,260 patients newly prescribed monotherapy

110,535 drug-naïve patients started MET monoterapy during the study period

28,539 patients were added a second NIAD to metformin monotherapy

SUs DPP-4 i Meglitinides Others n=17,541 n= 8,035 n= 2,261 GLP-1 RAs:1.35%; SGLT- (61.46%) (28.15%), (7.92%), 2 i 0.48%; TZD:0.62%

SIDIAP (Information System for the Development of Research in Primary Care); NIAD: non-insulin blood glucose-lowering (“antidiabetic”) drug MET: metformin; SUs: suphfonylureas; DPP-4 i: dipeptidyl-peptidase-4 inhibitors; GLP-1 RA: glucagon-like peptide receptor agonist, SGLT-2 i: sodium-glucose co-transporter 2 inhibitors; TZD: thiazolidinediones

VI. Second-line, metformin-based, dual therapies 113 Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 Study Population

People aged ≥18 years with an active T2DM diagnosis (ICD-10 codes E11) who had been first-prescribed metformin during the study period and who were included in the cohort study assessing first-line monotherapies. The study period extended from January 1st 2010 to December 31st 2015. Patients on metformin monotherapy entered in the second- line cohorts as their condition worsened and needed treatment intensification. The flow of patients into the cohort increased as time passed, and the percentage in the 2010 second-line cohort is much lower than in the following years. Figure VI.2 shows the selection flow.

Exposures:

Any non-insulin blood-glucose-lowering drug added to metformin as a second-line dual combination therapy: sulfonylureas (A10BB), meglitinides (repaglinide [A10BX02], nateglinide [A10BX03]), DPP-4 inhibitors (vildagliptin [A10BH02], saxagliptin [A10BH03],linagliptin [A10BH05], alogliptin [A10BH04], and sitagliptin [A10BH01]; glucagon-like peptide 1 receptor agonists (GLP-1 RA) (exenatide [A10BJ01], liraglutide [A10BJ02], lixisenatide[A10BJ03], albiglutide [A10BJ04], dulaglutide [A10BJ05]); SGLT-2 inhibitors: dapagliflozin [A10BK01], canagliflozin [A10BK02] , empagliflozin [A10BK03]; thiazolidinediones (pioglitazone[A10BG03] and rosiglitazone [A10BG02]) and other agents. Switching between drugs of the same class doesn’t change the patient’s cohort. The index date (ID) is defined as the day of the first recorded claimed prescription of the NIADs first added to metformin. Patients were considered at risk from the day of the claimed prescription until the prescription to another non-insulin blood glucose- lowering class agent or censoring.

Follow-up

Patients were followed until the first event of primary and secondary outcomes, the first prescription of a different NIAD group, loss of follow-up (defined as no prescription of the drug during 12 months), or the end of study 31st December 2015.

VI. Second-line, metformin-based, dual therapies 114

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Data extraction:

Selection: Patients with an active diagnosis code of T2DM, who had been first prescribed metformin during the study period and who received a new prescription of any NIAD in second-line monotherapy. The previous use of insulin (after the first prescription of metformin monotherapy) was no exclusion criterium. No further exclusion criteria were applied.

We extracted data from patients with no prior exposure to the given second-line blood- glucose-lowering agent before the index date. We extracted data of the use of insulin (A10A) during the study period, and after the first prescription of insulin, to adjust for insulins use.

Covariates ascertainment: Age, sex and time from T2DM diagnosis were retrieved at the date of the first prescription of the second-line agent added to metformin, with the same criteria used in the first-line cohorts.

Outcomes:

The primary outcome is a composite of three-components of major cardiovascular events (MACE): all-cause death, non-fatal myocardial infarction (MI) and non-fatal stroke. Secondary outcomes are components of MACE: myocardial infarction [ICD-10 I21), stroke (I61. I62, I63, I64), a new diagnosis of heart failure (I50.0) and onset or worsening of intermittent claudication (I73.9)

Statistical analysis:

Statistical analyses have been described in the section “Firs-line therapies” All analyses were performed by Prof. Xavier Vidal using SAS 9.4 (SAS Institute Inc., Cary, NC).

Results

Among 110,535 patients initiating treatment with metformin monotherapy 28,539 patients received intensification with other therapies: 17,541 were on MET + SUs (61.46%), 8,035 were on MET+ DPP-4 i (28.15%), 2,261 were on meglitinides

VI. Second-line, metformin-based, dual therapies 115

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

(7.92%), 386 on MET+GLP-1 RA (1.35%), 138 on MET+SGLT-2 i (0.48%) and 178 on thiazolidinediones (0.62%). Table VI.1 shows the frequency of the different agents within a class.

Table VI.1. Therapeutic classes and agents prescribed in incident second-line therapies added to basal metformin, for T2DM patients, SIDIAP database, 2010- 2015

Therapeutic class n Class % Total % All classes 28,539 100.00 Sulfonylureas 17,541 61.46 Glibenclamide/glyburide 3313 18.87 Glipizide 196 1.12 Gliquidone 34 0.19 Gliclazide 11803 67.24 Glimepiride 2207 12.57 Glisentide 1 0.01 DPP-4 i 8,035 28.15 sitagliptin 4541 56.42 vildagliptin 2237 27.79 saxagliptin 235 2.92 alogliptin 3 0.04 linagliptin 1033 12.83 Meglitinides 2,261 7.81 Guar gum 33 1.46 Repaglinide 2233 98.50 Nateglinide 1 0.04 Thiazolidinediones 178 0.62 rosiglitazone 9 5.06 pioglitazone 169 94.94 Glucagon-like peptide 1 RA 164 1.35 exenatide 106 27.25 liraglutide 283 72.75 Sodium-glucose co-transporter 2 i 138 0.48 Dapagliflozin 138 100.00

VI. Second-line, metformin-based, dual therapies 116

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

As patients are selected from the cohort of first-line monotherapy new-users of metformin, the graphic shows that there is low risk of misclassification in prevalent and incident users. All curves increased over time, reflecting the incorporation of patients who needed intensification therapies. Fig. VI. 3 shows the patients’ entry to each cohort.

Figure VI.3. Yearly percent of second-line, metfomin-based dual therapies for each NIAD class, 2010-2015, SIDIAP database

Patients’ basal characteristics (i.e., at the time of the prescription of a second-line agent, in the temporal window already mentioned in the section of first-line therapies) are shown in Table VI.2

. The MET + SUs and the MET+DPP-4 i cohorts were similar in mean age (61.0, SD 12.4 and 61.4, SD 12.4 years). The MET + SUs cohort had the highest mean level of HbA1c (8.4, SD 1.6)

. The MET+ DPP-4 i cohort had the highest percentage of men and the highest use of statins.

VI. Second-line, metformin-based, dual therapies 117

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table VI. 2. Basal characteristics of patients newly treated with NIADs added to previous metformin as dual, second-line therapies SIDIAP database, 2010-2015

MET + SUs MET+DPP-4 i MET + meglitinides MET + GLP-1 RA MET + SGLT-2 i MET + TZD n (%) n (%) n (%) n (%) n (%) n (%) Total 17541 (100.0) 8035 (100.0) 2261 (100.0) 386 (100.0) 138 (100.0) 178 (100.0)

Sex Women 7613 (43.4) 3376 (42.0) 1068 (47.2) 211 (54.7) 70 (50.7) 80 (44.9) Men 9928 (56.6) 4659 (58.0) 1193 (52.8) 175 (45.3) 68 (49.3) 98 (55.1) Age Mean (SD) 61.0 (12.4) 61.4 (12.6) 66.9 (13.3) 51.4 (10.8) 56.6 (11.2) 62.6 (12.4) Median [IQR] 61 [52 – 70] 61 [53 – 70] 68 [57 – 78] 51 [45 – 59] 57 [49 – 65] 64 [53 – 72] 18-44 1654 (9.4) 731 (9.1) 128 (5.7) 96 (24.9) 22 (15.9) 12 (6.7) 45-64 9079 (51.8) 4089 (50.9) 807 (35.7) 246 (63.7) 79 (57.2) 84 (47.2) => 65 6808 (38.8) 3215 (40.0) 1326 (58.6) 44 (11.4) 37 (26.8) 82 (46.1) T2DM duration Mean (SD) 2.6 (3.8) 2.5 (3.9) 3.2 (4.3) 1.8 (3.4) 1.7 (2.8) 3.4 (4.6) Median [IQR] 0.8 [0.0 - 4.4] 0.4 [0.0 - 4.0] 1.1 [0.0 - 5.3] 0.0 [0.0 - 2.3] 0.1 ([0.0 - 2.3] 1.4 [0.0 - 5.6] 0-4 y 13706 (78.1) 6436 (80.1) 1658 (73.3) 334 (86.5) 119 (86.2) 126 (70.8) 5-9 y 2935 (16.7) 1230 (15.3) 431 (19.1) 41 (10.6) 18 (13.0) 37 (20.8) 10-14 y 697 (4.0) 283 (3.5) 128 (5.7) 7 (1.8) . (.) 9 (5.1) => 15 y 203 (1.2) 86 (1.1) 44 (1.9) 4 (1.0) 1 (0.7) 6 (3.4) HBA1c Missing 1626 (9.3) 1131 (14.1) 340 (15.0) 77 (19.9) 21 (15.2) 31 (17.4) Mean (SD) 8.4 (1.6) 8.2 (1.6) 7.8 (1.6) 8.1 (1.7) 8.1 (1.7) 8.0 (1.8) Median [IQR] 8.0 [7.2 - 9.1] 7.8 [7.1 - 8.9] 7.4 [6.7 - 8.5} 7.8 [6.9 - 9.0] 7.7 [6.8 - 8.8] 7.6 [6.8 - 9.0] < 8 7620 (43.4) 3692 (45.9) 1220 (54.0) 168 (43.5) 71 (51.4) 86 (48.3) 8 to 10 5986 (34.1) 2344 (29.2) 514 (22.7) 100 (25.9) 32 (23.2) 40 (22.5)

VI. Second-line, metformin-based, dual therapies 118

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

> 10 2309 (13.2) 868 (10.8) 187 (8.3) 41 (10.6) 14 (10.1) 21 (11.8) BMI Missing 1831 (10.4) 891 (11.1) 297 (13.1) 30 (7.8) 7 (5.1) 25 (14.0) Mean (SD) 30.9 (5.4) 31.1 (5.4) 30.4 (5.4) 40.5 (6.3) 35.5 (6.3) 31.2 (6.1) Median [IQR] 30.3 [27.2 - 34.0] 30.5 [27.4 - 34.2] 29.6 [26.7 - 33.3] 39.9 [35.8 - 44.5] 34.7 [31.4 - 38.9] 29.8 [27.2 - 34.9] < 25 1747 (10.0) 755 (9.4) 275 (12.2) 1 (0.3) 4 (2.9) 19 (10.7) 25.0 to 29.9 (overweight) 5767 (32.9) 2506 (31.2) 776 (34.3) 7 (1.8) 19 (13.8) 59 (33.1) 30.0 to 39.9 (obese I-II) 7224 (41.2) 3400 (42.3) 810 (35.8) 171 (44.3) 82 (59.4) 64 (36.0) => 40 (obese III) 972 (5.5) 483 (6.0) 103 (4.6) 177 (45.9) 26 (18.8) 11 (6.2) Comorbidities Prior cardiovascular events 2215 (12.6) 1196 (14.9) 473 (20.9) 36 (9.3) 12 (8.7) 23 (12.9) ACS 1029 (5.9) 585 (7.3) 222 (9.8) 15 (3.9) 11 (8.0) 10 (5.6) MI 531 (3.0) 345 (4.3) 116 (5.1) 14 (3.6) 4 (2.9) 4 (2.2) Stroke 809 (4.6) 444 (5.5) 178 (7.9) 7 (1.8) 3 (2.2) 10 (5.6) PAD 494 (2.8) 316 (3.9) 120 (5.3) 7 (1.8) 2 (1.4) 7 (3.9) Heart failure 593 (3.4) 366 (4.6) 200 (8.8) 11 (2.8) 4 (2.9) 3 (1.7) Renal failure 324 (1.8) 345 (4.3) 229 (10.1) 5 (1.3) 3 (2.2) 8 (4.5) REGICOR CV index Missing 8125 (46.3) 4083 (50.8) 1378 (60.9) 181 (46.9) 60 (43.5) 93 (52.2) Low CV risk 3232 (18.4) 1364 (17.0) 286 (12.6) 89 (23.1) 25 (18.1) 30 (16.9) Medium CV risk 4185 (23.9) 1731 (21.5) 415 (18.4) 98 (25.4) 41 (29.7) 40 (22.5) High CV risk 1999 (11.4) 857 (10.7) 182 (8.0) 18 (4.7) 12 (8.7) 15 (8.4) Concomitant medications Insulin use 155 (0.9) 261 (3.2) 127 (5.6) 23 (6.0) 5 (3.6) 11 (6.2) ACE inhibitors 8383 (47.8) 4107 (51.1) 1252 (55.4) 232 (60.1) 78 (56.5) 83 (46.6) Aldosterone antagonists 225 (1.3) 174 (2.2) 84 (3.7) 11 (2.8) 4 (2.9) 2 (1.1) Antiplatelet drugs 3362 (19.2) 1831 (22.8) 613 (27.1) 64 (16.6) 27 (19.6) 41 (23.0) Beta-Blockers 2628 (15.0) 1446 (18.0) 492 (21.8) 70 (18.1) 29 (21.0) 27 (15.2)

VI. Second-line, metformin-based, dual therapies 119

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Calcium channel blockers 2356 (13.4) 1419 (17.7) 501 (22.2) 65 (16.8) 23 (16.7) 24 (13.5) Diuretics 5870 (33.5) 2832 (35.2) 1018 (45.0) 155 (40.2) 52 (37.7) 45 (25.3) Statins 8030 (45.8) 4089 (50.9) 1095 (48.4) 169 (43.8) 78 (56.5) 76 (42.7) Other lipid-lowering drug 1431 (8.2) 908 (11.3) 137 (6.1) 45 (11.7) 13 (9.4) 15 (8.4) NSAIDs 1982 (11.3) 902 (11.2) 198 (8.8) 59 (15.3) 23 (16.7) 24 (13.5) Antidepressants 1091 (6.2) 541 (6.7) 202 (8.9) 43 (11.1) 16 (11.6) 17 (9.6) MEDEA deprivation index Rural 2999 (17.1) 1510 (18.8) 466 (20.6) 54 (14.0) 23 (16.7) 29 (16.3) Urban 820 (4.7) 362 (4.5) 142 (6.3) 20 (5.2) 6 (4.3) 8 (4.5) U1 (least deprived) 1931 (11.0) 1071 (13.3) 261 (11.5) 50 (13.0) 13 (9.4) 24 (13.5) U2 2232 (12.7) 1172 (14.6) 290 (12.8) 55 (14.2) 15 (10.9) 23 (12.9) U3 2807 (16.0) 1335 (16.6) 365 (16.1) 58 (15.0) 19 (13.8) 22 (12.4) U4 3179 (18.1) 1323 (16.5) 344 (15.2) 67 (17.4) 29 (21.0) 30 (16.9) U5 (most deprived) 3573 (20.4) 1262 (15.7) 393 (17.4) 82 (21.2) 33 (23.9) 42 (23.6)

Year Cohort entry 2010 1630 (9.3) 419 (5.2) 138 (6.1) 16 (4.1) 0 (0.) 43 (24.2) 2011 2644 (15.1) 793 (9.9) 359 (15.9) 30 (7.8) 0 (0) 38 (21.3) 2012 3127 (17.8) 1169 (14.5) 331 (14.6) 52 (13.5) 0 (0) 19 (10.7) 2013 3166 (18.0) 1408 (17.5) 451 (19.9) 77 (19.9) 2 (1.4) 21 (11.8) 2014 3389 (19.3) 1808 (22.5) 493 (21.8) 98 (25.4) 56 (40.6) 22 (12.4) 2015 3585 (20.4) 2438 (30.3) 489 (21.6) 113 (29.3) 80 (58.0) 35 (19.7) Vital Status at 31Dec2015 n (%) 883 (5.0) 358 (4.5) 250 (11.1) 3 (0.8) 3 (2.2) 9 (5.1)

NIAD: non-insulin blood glucose-lowering (“antidiabetic”) drugs; SIDIAP: Information System for the Development of Research in Primary Care. MET: metformin; SUs: sulfonylureas; DPP- 4 i: dipeptidyl-peptidase 4 inhibitors; GLP-1 RAs: glucagon-like peptide 1 receptor agonists; SGLT-2 i: sodium-glucose co-transporters 2 inhibitors; TZD: thiazolidinediones

MEDEA: Mortality in small Spanish áreas and economic and environmental inequalities Values for HbA1c and comorbidities: ≤ 3 months before the index date. Values of REGICOR index of CV risk during the 15 months prior the index date. Values of body mass index (BMI) were the most recent ≤15 months before the index date. Concomitant use of drugs refers to any active prescription at the index date.

VI. Second-line, metformin-based, dual therapies 120

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

. Mean age of patients of the MET+ meglitinides cohort was 66.9 years (SD 13.3). Patients of this cohort had the highest percentage of comorbidities, and also the highest percentage of aldosterone antagonists, antiplatelet drugs, beta- blockers, calcium-channel blockers and diuretics.

. 54.7 % of patients who were prescribed GLP-1 RA were women, mean age was 51.4 years, had the shortest mean diabetes duration, the highest BMI, and the highest percentage of use of lipid-lowering drugs other than statins.

. The cohort of patients on MET + TZD had the highest previous use of insulin.

The shortest mean follow-up was 0.75 years (SD 0.56) for the cohort of MET+SGLT-2 and the longer one of 2,2 years (SD 1.54) for MET + SUs. Mean follow-was 1.62 years (SD 1.38) for MET + DPP-4 i; 1.91 years (SD 1.44) for MET + meglitinides; 1.86 years (SD 1.66) for MET + TZD and 1.73 years (SD 1.31) for MET + GLP-1 RA.

In the six cohorts, during the study period, there were 1,834 first events of MACE. There were 174 myocardial infarctions, 484 events of stroke and 1,261 patients died. There were also 578 new diagnoses of heart failure and 465 of PAD. Cohorts of dual second-line metformin-based regimes were compared with MET + SUs as reference, because this combination is the recommended in local guidelines and the most widely used dual combined therapy in the study population.

After adjusting for all available demographic, clinical and concomitant medications data at baseline and for the use of insulin after the index date and before or on the date of the event, compared with MET + SUs dual therapy, only patients on MET + SUs dual therapy were at significant higher risk of MACE, HR 1.19 (1.01 – 1.42) and all-cause mortality, HR 1.38 (1.14 – 1.67). In all the other outcomes, meglitinides showed no difference compared with MET + SUs. The use of MET + DPP-4 i showed no statistical differences with the use of MET + SUs.Other combined therapies (MET + GLP-1 RA, MET + SGLT-2 i and MET + TZD) do not reach statistical significance or had 0 events.

VI. Second-line, metformin-based, dual therapies 121

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table VI.3. Adjusted HRs of first event of MACE, all-cause mortality, CV events and HF and PAD diagnoses and all-cause mortality in cohorts of T2DM patients treated with NIADs added to metformin as a dual second-line therapy, SIDIAP healthcare database, 2010-2015

Outcomes N° of events Person-years Crude Incidence Adjusted HR Rate per 1000 p/y Composite of MACE (AMI, stroke and all-cause mortality) MET + SU 1120 55269.56 20.26 reference MET + DPP-4 i 432 12754.07 33.87 1.10 (0.98-1.23) MET + meglitinides 267 4240.40 62.97 1.19 (1.01-1.42) MET + GLP-1 RA 6 662.51 9.06 MET + SGLT-2 i 2 104.13 19.21 MET + TZD 7 327.12 21.40 All-cause mortality MET + SU 751 56470.74 13.30 reference MET + DPP-4 i 283 12984.49 21.80 1.08 (0.94-1.25) MET + meglitinides 218 4313.03 50.54 1.38 (1.14-1.67) MET + GLP-1 RA 3 666.72 4.50 MET + SGLT-2 i 2 104.13 19.21 MET + TZD 4 331.33 12.07 MI MET + SU 113 56168.71 2.01 reference MET + DPP-4 i 49 12906.75 3.80 1.23 (0.87-1.75) MET + meglitinides 9 4304.33 2.09 0.70 (0.30-1.61) MET + GLP-1 RA 0 666.72 0.00 MET + SGLT-2 i 0 104.13 0.00 MET + TZD 3 327.12 9.17 Stroke MET + SU 315 55544.50 5.67 reference MET + DPP-4 i 117 12829.00 9.12 1.03 (0.83-1.29) MET + meglitinides 49 4249.10 11.53 0.86 (0.60-1.24) MET + GLP-1 RA 3 662.51 4.53 MET + SGLT-2 i 0 104.13 0.00 MET + TZD 0 331.33 0.00 Heart failure MET + SU 347 55565.41 6.24 reference MET + DPP-4 i 122 12856.76 9.49 0.91 (0.74-1.13) MET + meglitinides 100 4182.76 23.91 1.25 (0.96-1.64) MET + GLP-1 RA 5 660.51 7.57 MET + SGLT-2 i 1 103.25 9.69 MET + TZD 3 326.20 9.20

VI. Second-line, metformin-based, dual therapies 122

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

PAD MET + SU 307 55763.33 5.51 reference MET + DPP-4 i 104 12855.39 8.09 0.95 (0.76-1.20) MET + meglitinides 51 4256.88 11.98 1.35 (0.94-1.94) MET + GLP-1 RA 1 666.48 1.50 MET + SGLT-2 i 0 104.13 0.00 MET + TZD 2 330.79 6.05

HR: hazard ratio; (95% CI): 95% confidence interval MACE: composite of major adverse cardiovascular events (MI, stroke and all-cause mortality); MI: myocardial infarction; PAD: peripheral artery disease; SIDIAP: Information System for the Development of Research in Primary Care; MET: metformin; SUs: sulfonylureas; DPP-4 i: dipeptidyl-peptidase 4 inhibitors; GLP-1 RA: glucagon-like peptide 1 receptor agonists; SGLT-2 i: sodium glucose co-transport 2 inhibitors; TZD: thiazolidinediones

Table VI.3 shows the number of events in each cohort, the person-years period, the crude incidence rate/1000 p/y and the adjusted hazard ratios (95% CI)

We also stratified the cohorts of MET+SUs, MET+DPP-4 i; MET + meglitinides newly prescribed second-line dual patients by the use of insulin during the follow-up, and by age, sex, HbA1c, diabetes duration, body mass index (BMI), history of CVD, history of HF, history of renal failure (RF) and MEDEA deprivation index

Among patients on MET+DPP-4 i, compared with MET + SU dual therapy:

. For MACE, patients were at higher risk if they had previous history of H, were overweight and female. There was a trend toward higher risk for patients with previous history of CVD and for those with no previous use of insulin. . For all-cause mortality, patients were at higher risk if they had history of HF, there was a trends toward increased risk for women. . For MI, patients were at higher risk if they had history of CVD, an increased risk for patients with history of HF cannot ruled out. . For stroke, there was no differences between subgroups . For heart failure (HF) and peripheral artery disease (PAD), patients were at higher risk if they had history of HF. Results of the sub-group analysis for patients on MET+DPP-4 i are shown in Table VI.4. Results of subgroup analysis for MACE are shown in the Figure VI.4 a.

VI. Second-line, metformin-based, dual therapies 123

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table VI.4. Adjusted HRs Subgroup analyses for MACE, mortality, CV events and HF and PAD diagnoses in T2DM patients treated with dual second-line therapy of MET+ DPPP-4 i vs. MET+SUs in the SIDIAP database, 2010-2015

MACE All-cause mortality MI Stroke HF PAD MET+SU (total) reference reference reference reference reference reference MET+DPP-4I(total) 1.10 (0.98 – 1.23) 1.08 (0.94 – 1.25) 1.23 (0.87 – 1.75) 1.03 (0.83 – 1.29) 0.91 (0.74 – 1.13) 0.95 (0.76 – 1.20) Use of insulins after the index date Ins Yes 0.96 (0.69 – 1.34) 0.94 (0.65 – 1.36) 2.40 (0.55 – 10.35) 1.00 (0.45 – 2.21) 1.47 (0.78 – 2.79) 1.02 (0.45 – 2.30) Ins No 1.12 (0.99 – 1.26) 1.11 (0.96 – 1.30) 1.19 (0.83 – 1.70) 1.04 (0.83 – 1.30) 0.87 (0.78 – 2.79) 0.95 (0.75 – 1.20) Age < 75 1.09 (0.93 – 1.28) 1.05 (0.85 – 1.10) 1.31 (0.89 – 1.93) 1.03 (0.78 – 1.35) 0.95 (0.70 – 1.29) 0.89 (0.68 – 1.17) ≥ 75 1.17 (0.98 – 1.38) 1.18 (0.97 – 1.43) 1.00 (0.46 – 2.17) 1.10 (0.76 – 1.57) 0.92 (0.69 – 1.25) 1.19 (0.76 – 1.86) Sex Male 0.94 (0.78 – 1.14) 0.94 (0.75 – 1.18) 1.53 (0.77 – 3.04) 0.80 (0.57 – 1.13) 0.80 (0.58 – 1.11) 0.79 (0.47 – 1.33) Female 1.21 (1.04 – 1.39) 1.19 (1.00 – 1.43) 1.15 (0.77 – 1.72) 1.24 (0.94 – 1.65) 1.02 (0.77 – 1.35) 1.00 (0.77 – 1.33) HbA1c HbA1c < 8 1.02 (0.87 – 1.20) 1.01 (0.83 – 1.23) 1.19 (0.72 – 1.96) 0.96 (0.70 – 1.31) 0.89 (0.67 – 1.19) 0.96 (0.68 – 1.36) HbA1c 8-10 1.23 (0.99 – 1.53) 1.18 (0.91 – 1.55) 1.40 (0.76 – 2.58) 1.24 (0.83 – 1.83) 1.04 (0.71 – 1.54) 1.01 (0.67 – 1.51) HbA1c > 10 1.07 (0.75 – 1.54) 1.16 (0.73 – 1.83) 1.00 (0.37 – 2.72) 0.84 (0.38 – 1.83) 0.62 (0.28 – 1.39) 0.84 (0.47 – 1.51) Diabetes duration 0-4 y 1.09 (0.95 – 1.25) 1.02 (0.86 – 1.22) 1.39 (0.94 – 2.05) 1.08 (0.84 – 1.39) 0.91 (0.70 – 1.17) 0.92 (0.70 – 1.21) 5-9 y 1.12 (0.88 – 1.43) 1.23 (0.93 – 1.65) 1.02 (0.46 – 2.26) 0.99 (0.61 – 1.61) 1.02 (0.66 – 1.58) 0.85 (0.52 – 1.36) 10-14 y 1.35 (0.85 – 2.13) 1.47 (0.87 – 2.48) - 1.09 (0.46 – 2.57) 0.92 (0.35 – 2.42) 1.67 (0.58 – 4.80) ≥ 15 y 0.58 (0.27 – 1.21) 0.62 (0.27 – 1.46) 0.55 (0.05 – 5.66) 0.25 (0.03 – 1.99) 0.32 (0.07 – 1.50) 4.96 (0.88 – 28.06) BMI < 25 0.94 (0.70 – 1.26) 0.86 (0.61 – 1.22) 0.61 (0.15 – 2.53) 1.05 (0.58 – 1.89) 0.73 (0.34 – 1.60) 1.27 (0.75 – 2.13) 25 - 29.9 1.24 (1.03 – 1.49) 1.23 (0.98 – 1.55) 1.56 (0.93 – 2.61) 1.14 (0.80 – 1.64) 0.92 (0.63 – 1.35) 1.29 (0.91 – 1.82) 30 - 39.9 1.07 (0.88 – 1.29) 1.13 (0.88 – 1.44) 1.18 (0.67 – 2.08) 0.90 (0.64 – 1.28) 1.02 (0.75 – 1.39) 0.58 (0.37 – 0.92) ≥ 40 0.98 (0.54 – 1.76) 0.90 (0.44 – 1.87) 0.62 (0.06 – 6.21) 1.53 (0.52 – 4.50) 0.49 (0.19 – 1.23) 0.59 (0.12 – 2.86) CVD history Prior CVD no 1.07 (0.93 – 1.23) 1.05 (0.88 – 1.26) 1.00 (0.66 – 1.52) 1.09 (0.85 – 1.40) 0.84 (0.65 – 1.09) 0.88 (0.67 – 1.15) Prior CVD yes 1.22 (0.99 – 1.49) 1.20 (0.95 – 1.52) 2.38 (1.24 – 4.57) 0.91 (0.59 – 1.42) 1.15 (0.80 – 1.65) 1.26 (0.82 – 1.94) HF history Prior HF no 1.05 (0.92 – 1.19) 1.03 (0.88 – 1.21) 1.13 (0.78 – 1.65) 1.00 (0.79 – 1.25) 0.82 (0.65 – 1.04) 0.90 (0.71 – 1.14) Prior HF yes 1.59 (1.18 – 2.14) 1.51 (1.08 – 2.12) 2.66 (0.96 – 7.35) 1.60 (0.79 – 3.21) 1.82 (1.07 – 3.10) 2.69 (1.06 – 6.81)

VI. Second-line, metformin-based, dual therapies 124

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table VI.4. Adjusted HRs Subgroup analyses for MACE, mortality, CV events and HF and PAD diagnoses in T2DM patients treated with dual second-line therapy of MET+ DPPP-4 i vs. MET+SUs in the SIDIAP database, 2010-2015 (cont.)

MACE All-cause mortality MI Stroke HF PAD Renal failure history Prior RF no 1.11 (0.98 – 1.25) 1.07 (0.92 – 1.24) 1.25 (0.88 – 1.77) 1.09 (0.88 – 1.36) 0.91 (0.73 – 1.13) 0.97 (0.76 – 1.22) Prior RF yes 1.01 (0.63 – 1.61) 1.66 (0.98 – 2.80) 0.60 (0.05 – 6.60) 0.24 (0.07 – 0.85) 1.18 (0.59 – 2.36) 0.79 (0.31 – 2.04) MEDEA index Rural 0.89 (0.69 – 1.16) 0.78 (0.56 – 1.07) 0.95 (0.45 – 2.01) 1.21 (0.75 – 1.97) 0.95 (0.61 – 1.49) 0.88 (0.49 – 1.58) U1 (least deprived) 1.18 (0.85 – 1.65) 2.01 (1.35 – 3.01) 0.14 (0.02 – 1.08) 0.64 (0.33 – 1.24) 0.66 (0.35 – 1.26) 0.63 (0.31 – 1.26) U5 (most deprived) 1.10 (0.79 – 1.53) 0.99 (0.64 – 1.54) 1.02 (0.38 – 2.72) 1.18 (0.69 – 2.04) 1.68 (0.98 – 2.87) 1.15 (0.67 – 1.97)

Table VI.5. Adjusted HR of subgroup analyses for MACE, mortality, MI, stroke events and HF and PAD diagnosed in T2DM patients treated with dual second-line therapy of meglitinides added to metformin, SIDIAP database, 2010-2015

MACE All-cause mortality MI Stroke HF PAD MET+SU (total) reference reference reference reference reference reference MET+meglitinides 1.19 (1.01 – 1.42) 1.38 (1.14 – 1.67) 0.70 (0.30 – 1.61) 0.86 (0.60 – 1.24) 1.25 (0.96 – 1.64) 1.35 (0.94 – 1.94) (total) Use of insulins after the index date Ins Yes 1.74 (1.16 – 2.62) 1.80 (1.16 – 2.77) - 1.54 (0.55 – 4.31) 1.63 (0.80 – 3.35) 1.46 (0.49 – 4.38) Ins No 1.10 (0.91 – 1.33) 1.28 (1.04 – 1.58) 0.73 (0.31 – 1.68) 0.80 (0.54 – 1.18) 1.22 (0.91 – 1.62) 1.34 (0.91 – 1.96) Age < 75 1.24 (0.96 – 1.62) 1.46 (1.06 – 2.02) 0.88 (0.36 – 2.17) 0.96 (0.59 – 1.57) 1.11 (0.71 – 1.74) 1.36 (0.87 – 2.12) ≥ 75 1.07 (0.87 – 1.32) 1.24 (0.99 – 1.55) 0.19 (0.04 – 0.91) 0.68 (0.40 – 1.15) 1.28 (0.92 – 1.79) 1.29 (0.74 – 2.26) Sex Male 0.95 (0.73 – 1.24) 1.17 (0.87 – 1.58) – 0.45 (0.25 – 0.81) 1.37 (0.94 – 1.99) 1.20 (0.59 – 2.46) Female 1.38 (1.11 – 1.71) 1.55 (1.21 – 1.98) 0.89 (0.38 – 2.07) 1.24 (0.79 – 1.97) 1.14 (0.78 – 1.65) 1.39 (0.91 – 2.11) HbA1c HbA1c < 8 1.21 (0.98 – 1.50) 1.39 (1.10 – 1.75) 0.91 (0.34 – 2.44) 0.78 (0.50 – 1.24) 1.55 (1.12 – 2.16) 1.32 (0.81 – 2.15) HbA1c 8– 10 1.16 (0.79 – 1.70) 1.27 (0.84 – 1.93) 0.40 (0.03 – 5.55) 1.13 (0.57 – 2.24) 0.98 (0.54 – 1.77) 1.70 (0.93 – 3.11) HbA1c > 10 1.18 (0.59 – 2.34) 1.57 (0.76 – 3.23) – 0.57 (0.12 – 2.77) 0.53 (0.12 – 2.44) 0.83 (0.26 – 2.62)

VI. Second-line, metformin-based, dual therapies 125

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Table VI.5. Adjusted HR of subgroup analyses for MACE, mortality, MI, stroke events and HF and PAD diagnosed in T2DM patients treated with dual second-line therapy of meglitinides added to metformin, SIDIAP database, 2010-2015 (cont.)

MACE All-cause mortality MI Stroke HF PAD Diabetes duration 0– 4 y 1.20 (0.97 – 1.48) 1.44 (1.13 – 1.83) 0.63 (0.23 – 1.71) 0.73 (0.44 – 1.18) 1.09 (0.77 – 1.54) 1.56 (1.03 – 2.37) 5– 9 y 1.05 (0.73 – 1.51) 1.15 (0.78 – 1.70) 1.01 (0.25 – 4.06) 0.93 (0.45 – 1.89) 1.58 (0.94 – 2.66) 0.55 (0.21 – 1.44) 10– 14 y 1.80 (1.05 – 3.08) 1.55 (0.83 – 2.90) – 2.20 (0.84 – 5.72) 2.05 (0.89 – 4.70) 2.28 (0.76 – 6.85) ≥ 15 y 0.86 (0.42 – 1.76) 1.14 (0.54 – 2.39) – 0.13 (0.02 – 1.08) 0.41 (0.07 – 2.49) 8.86 (0.95 – 82.44) BMI < 25 1.04 (0.69 – 1.55) 1.17 (0.77 – 1.77) – 0.89 (0.36 – 2.17) 1.36 (0.60 – 3.07) 1.38 (0.55 – 3.43) 25 – 29.9 1.39 (1.06 – 1.82) 1.60 (1.18 – 2.17) 1.14 (0.37 – 3.46) 0.84 (0.46 – 1.56) 1.37 (0.87 – 2.15) 1.13 (0.60 – 2.14) 30 – 39.9 1.08 (0.80 – 1.45) 1.29 (0.92 – 1.79) 0.36 (0.08 – 1.63) 0.89 (0.49 – 1.63) 1.11 (0.72 – 1.71) 1.42 (0.79 – 2.56) ≥ 40 1.06 (0.39 – 2.88) 1.19 (0.40 – 3.53) 2.27 (0.23 – 22.24) – 1.42 (0.52 – 3.89) 2.69 (0.37 – 19.53) CVD history Prior CVD no 1.18 (0.95 – 1.46) 1.36 (1.06 – 1.74) 0.80 (0.32 – 1.97) 0.87 (0.56 – 1.34) 1.35 (0.98 – 1.86) 1.37 (0.90 – 2.09) Prior CVD yes 1.32 (1.01 – 1.73) 1.53 (1.15 – 2.06) 0.28 (0.06 – 1.42) 0.91 (0.47 – 1.77) 1.12 (0.70 – 1.79) 1.38 (0.70 – 2.73) HF history Prior HF no 1.19 (0.99 – 1.44) 1.37 (1.11 – 1.69) 0.73 (0.30 – 1.75) 0.94 (0.65 – 1.35) 1.20 (0.90 – 1.61) 1.26 (0.86 – 1.86) Prior HF yes 1.35 (0.92 – 1.97) 1.69 (1.13 – 2.52) 0.43 (0.05 – 4.10) 0.07 (0.01 – 0.54) 1.94 (1.03 – 3.63) 4.32 (1.57 – 11.86) Renal failure history Prior RF no 1.21 (1.01 – 1.44) 1.39 (1.14 – 1.69) 0.7 (0.3 – 1.64) 0.88 (0.6 – 1.29) 1.24 (0.94 – 1.66) 1.39 (0.96 – 2.02) Prior RF yes 0.94 (0.56 – 1.58) 1.19 (0.64 – 2.22) 0.74 (0.07 – 8.21) 0.64 (0.23 – 1.78) 1.33 (0.68 – 2.63) 0.61 (0.21 – 1.8) MEDEA index Rural 0.94 (0.65 – 1.36) 1.03 (0.68 – 1.55) 0.11 (0.01 – 0.82) 1.17 (0.56 – 2.42) 0.89 (0.50 – 1.60) 2.00 (0.91 – 4.42) U1 (least deprived) 1.63 (0.98 – 2.70) 2.09 (1.15 – 3.79) 1.66 (0.40 – 6.82) 1.08 (0.42 – 2.79) 2.05 (1.02 – 4.11) 0.70 (0.23 – 2.10) U5 (most deprived) 0.94 (0.56 – 1.60) 0.95 (0.50 – 1.80) 1.76 (0.50 – 6.21) 0.43 (0.13 – 1.46) 1.53 (0.74 – 3.17) 1.58 (0.67 – 3.72)

MET+SUs is reference. HR: hazard ratio; (95% CI): 95% confidence interval; Significant results in bold.

MACE: composite of major adverse cardiovascular events (MI, stroke and all-cause mortality); MI: myocardial infarction; PAD: peripheral artery disease; HF: heart failure; RF: renal failure; MET: metformin; SUs: sulfonylureas; DPP-4 i: dipeptidyl-peptidase 4 inhibitors; GLP-1 RA: glucagon-like peptide 1 receptor agonists; SGLT-2 i: sodium-glucose co-transporter 2 inhibitors; TZD: thiazolidinediones. SIDIAP: Information System for the Development of Research in Primary Care; MEDEA: Mortality in small Spanish áreas and economic and environmental inequalities

VI. Second-line, metformin-based, dual therapies 126

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Fig VI.4: Subgroup analyses of risk of MACE in cohorts of metformin patients who were added DPP-4 i (A) or meglitinides (B) compared to the addition of sulfonylureas (SUs).

A: MET + DPP-4 i vs MET + SUs B: MET + meglitinides vs MET + SUs

All estimators: adjusted HR: hazard ratio; (95% CI): 95% confidence interval. MET+SUs is reference. MACE: composite of major adverse cardiovascular events (MI, stroke and all-cause mortality); MI: myocardial infarction; HF: heart failure; RF: renal failure; MET: metformin; SUs: sulfonylureas; DPP-4 i: dipeptidyl- peptidase 4 inhibitors. SUs: sulphonylureas. SUs: Gliclazide 67.24%; Glibenclamide/glyburide 18.87%; Glimepiride 12.57%. DPP-4 i: dipeptidyl-peptidase 4 inhibitors. DPP-4 i: Sitagliptin 56.42%; Vildagliptin: 27.49%; Linagliptin 12.83%. Meglitinides: Repaglinide 98.50%

VI. Second-line, metformin-based, dual therapies 127 Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 Among patients on MET + meglitinides, compared with MET+ SUs dual therapy:

. For MACE, patients were at increased risk if they were women, overweighted, diabetes duration 10-14 years long, had previous history of CVD and no history of renal failure. . For all-cause mortality, patients were at higher risk irrespective of their previous use of insulin, CVD and HF history. Patients < 75 years, women, overweighted (BMI between 25 and 29.9 kg/1.73 m2), with glycated haemoglobin < 8%, shorter diabetes duration (0-4 years) and no history of renal failure were at increased risk of dying. Higher risk for people > 75 years cannot ruled out. . For MI and stroke, there was no differences between subgroups. . For heart failure (HF) and peripheral artery disease (PAD), patients were at increased risk if they had history of HF at baseline. Patients with HbA1c < 8% also showed higher risk of HF. In brief, in patients who were prescribed MET+ DPP-4 i and compared with the reference, MET+SUs, there was no difference in overall estimators for all the outcomes assessed. In subgroup analysis, history of HF was associated with increased risk of MACE, all-cause mortality, new onset or worsening of HF and PAD; female patients were also at higher risk of MACE.

For patients who were prescribed MET+meglitinides, compared with MET + SUs as a reference, overall estimators reflect a higher risk of MACE and all-cause mortality. In subgroup analysis, female patients were at increased risks both for MACE and all-cause mortality. For HF and PAD, a history of HF was associated with increased risks. Patients were at increased risk of all-cause mortality irrespective of their CVD or HF history, or the past use or not of insulins, but people with no history of renal failure also had increased risks.

Results of the subgroup analysis for patients on MET+meglitinides compared with MET+SUs are shown in Table VI.5. Results of subgroup analysis for MACE are shown in Figure VI.4 b.

VI. Second-line, metformin-based, dual therapies 128

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Interestingly, both for MET+SUs and MET+meglitines, patients living in the least deprived areas were at higher risks of all-cause mortality. Patients on MET + DPP-4 i who lived in the least deprived areas had a non-significant lower risk for MI (HR 0.14 [0.02-1.08]) All the patients included in these cohorts are within the primary care public health and medicines are funded partially or in full by the administration; this allows patients’ economic situation to have little impact on the choice of the agent. In this study, there were no differences in prescription on more expensive agents, such as GLP- 1 RA and, to a lesser extent, SGLT-2. 38% and 44% of patients prescribed GLP-1 RA and SGLT-2 lived in the most deprived urban areas.

Summary of the currently available evidence

RCTs and MA of RCTs

There is scarce RCTs evidence for comparisons between two metformin-based combination plus SUs or other NIADs assessing major clinical outcomes.

1. A systematic review of RCTs found inconclusive evidence whether metformin and SUs combination therapy compared with metformin plus another glucose-lowering intervention results in benefit or harm for most patient-important outcomes (mortality, SAEs, macrovascular and microvascular complications) except for hypoglycaemia. However, more adverse effects were found with the combination of metformin and SUs. [229]

. Five trials compared (n = 1194) with MET+GLP-1 RA, (n = 1675): all-cause mortality was 11/1057 (1%) versus 11/1537 (0.7%), risk ratio (RR) 1.15 (95% confidence interval (CI) 0.49 to 2.67).

. Nine trials compared MET + SUs ( n= 5414) with MET + DPP-4 i (n = 6346): all-cause mortality was 33/5387 (0.6%) versus 26/6307 (0.4%), RR 1.32 (95% CI 0.76 to 2.28). For CV mortality 11/2989 (0.4%) versus 9/3885 (0.2%), RR 1.54 (95% CI 0.63 to 3.79); 9 trials; 11,694 participants; non-fatal stroke 14/2098 (0.7%) versus 8/2995 (0.3%), RR 2.21 (95% CI 0.74 to 6.58); ); 4 trials; 5093 participants;; non-fatal MI 15/2989 (0.5%) versus 13/3885 (0.3%), RR 1.45 (95% CI 0.69 to 3.07); 6 trials; 6 trials; 6874 participants; low-certainty VI. Second-line, metformin-based, dual therapies 129

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

evidence; SAE 735/5387 (13.6%) versus 779/6307 (12.4%), RR 1.07 (95% CI 0.97 to 1.186874; one trial in 64 participants reported no microvascular complications were observed .

. Eleven trials compared MET + SUs (n = 3626) with MET + TZD (N = 3685): all-cause mortality was 123/3300 (3.7%) versus 114/3354 (3.4%), RR 1.09 (95% CI 0.85 to 1.40); 6 trials; 6654 participants; low-certainty evidence; cardiovascular mortality 37/2946 (1.3%) versus 41/2994 (1.4%), RR 0.78 (95% CI 0.36 to 1.67); 4 trials; 5940 participants; low-certainty evidence; SAE 666/3300 (20.2%) versus 671/3354 (20%), RR 1.01 (95% CI 0.93 to 1.11); 6 trials; 6654 participants; very low-certainty evidence; non-fatal stroke 20/1540 (1.3%) versus 16/1583 (1%), RR 1.29 (95% CI 0.67 to 2.47); P = 0.45; 2 trials; 3123 participants; very low-certainty evidence; non-fatal MI 25/1841 (1.4%) versus 21/1877 (1.1%), RR 1.21 (95% CI 0.68 to 2.14); P = 0.51; 3 trials; 3718 participants; very low-certainty evidence; three trials (3123 participants) reported no microvascular complications (very low-certainty evidence)

. Three trials compared MET + SUs (n = 462) with MET + meglitinides (n = 476): one person died in each intervention group (3 trials; 874 participants; low- certainty evidence); no cardiovascular mortality (2 trials; 446 participants; low- certainty evidence); SAE 34/424 (8%) versus 27/450 (6%), RR 1.68 (95% CI 0.54 to 5.21); P = 0.37; 3 trials; 874 participants; low-certainty evidence; no fatal stroke (1 trial; 233 participants; very low-certainty evidence); non-fatal MI 2/215 (0.9%) participants in the M+S group; 2 trials; 446 participants; low- certainty evidence; no microvascular complications (1 trial; 233 participants; low-certainty evidence)

. Four trials compared MET + SUs (n = 2109) with metformin plus a sodium- glucose co-transporter 2 inhibitor (SGLT-2 i) (n = 3032): all-cause mortality was 13/2107 (0.6%) versus 19/3027 (0.6%), RR 0.96 (95% CI 0.44 to 2.09); CV mortality 4/1327 (0.3%) versus 6/2262 (0.3%), RR 1.22 (95% CI 0.33 to 4.41); SAE 315/2107 (15.5%) versus 375/3027 (12.4%), RR 1.02 (95% CI 0.76 to 1.37); non-fatal stroke 3/919 (0.3%) versus 7/1856 (0.4%), RR 0.87 (95% CI

VI. Second-line, metformin-based, dual therapies 130

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

0.22 to 3.34); non-fatal MI 7/890 (0.8%) versus 8/1374 (0.6%), RR 1.43 (95% CI 0.49 to 4.18; amputation of lower extremity 1/437 (0.2%) versus 1/888 (0.1%)

SUs and DPP-i in combined metformin-based regimes:

1. RCTs

. The CAROLINA trial compared linagliptin vs glimepiride in standard care. Of 6042 participants randomized, 6033 (mean age, 64.0 years; 2414 [39.9%] women; mean glycated haemoglobin, 7.2%; median duration of diabetes, 6.3 years; 42% with macrovascular disease; 59% had undergone metformin monotherapy) did not find differences in the 3 points MACE (HR, 0.98 [95% CI, 0.84-1.14]). In the subgroup analyses, all patients with metformin 5005/6033, 83%) showed no differences either (HR,1.01 [95% CI 0.86, 1.19); there were not a separate subgroup analysis for patients with prior metformin monotherapy (Table S2. Hazard ratios for the primary outcome (3-point MACE) in subgroups) [207]

2. Observational evidence:

. Using the UK Clinical Practice Research Datalink, a cohort of patients newly treated with metformin or sulfonylurea monotherapy between January 1, 1988, and December 31, 2011, was identified and was followed until December 31, 2012. The cohort consisted of 11,807 patients that included 2286 on a DPP-4 inhibitor-metformin combination and 9521 on a sulfonylurea-metformin combination. The crude incidence rates (95% CIs) of the composite endpoint (myocardial infarction, stroke and all-cause mortality) were 1.2% (0.8% to 1.7%) and 2.2% (1.9% to 2.5%) per year for the DPP-4 inhibitor-metformin and sulfonylurea-metformin combinations, respectively. In the high-dimensional propensity score-adjusted model, the use of the DPP-4 inhibitor-metformin combination was associated with a 38% decreased risk for the composite endpoint (adjusted HR 0.62; 95% CI 0.40 to 0.98), compared with the sulfonylurea-metformin combination.[230]

VI. Second-line, metformin-based, dual therapies 131

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

. In a large Danish population-based observational cohort study, Mogensen et al. compared MET + DPP-4 i (n= 11,138) with MET+SUs (25,092) as a reference. Results showed a decreased risk for DPP-4 i users: RR of 0.65 (0.54–0.80) for mortality, RR of 0.57 (0.40–0.80) for CV mortality and RR of 0.70 (0.57–0.85) for the 3 point MACE of MI, stroke and CV mortality.[231]

. Seong et al. found decreased risks for MET+ DPP-4 i users. Compared with MET + DPP-4 i as a reference (n= 74,720), MET+SUs users (n=253,563) were at higher risks of CVD, MI and ischemic stroke. For total CVD, the result was HR 1.20 (1.09-1.32) ; for MI, HR 1.14 (1.04-1.91); for HF, HR 1.07 (0.71-1.62); and for ischemic stroke 1.51 (1.28-1.79). [232]

. Compared with MET + SUs (n=9,419) as a reference, Kannan et al. didn’t find differences in MET + DPP-4 i users (n=1,487) for all-cause mortality (HR 1.029 [0.81–1.31]) or risk of coronary artery disease, (HR1.056 [0.99– 1.13]), but found an increased risk of HF (HR 1.104 [1.04–1.17]), stroke (HR 1.28 [1.08-1.52]), ischemic heart disease (HR 1.35 [1.16-1.57]) and peripheral artery disease (HR 1.65 [1.16-2.36]) [233]

. Ekström et al. compared MET+ DPP-4 i with MET + SUs as a reference. The study didn’t find statistical differences in mortality (HR 0.79 [0.60, 1.04]) or stroke (HR 0.82 [0.53, 1.25]). Results for heart failure (HR 0.54 [0.38, 0.76]), fatal CVD (HR 0.34 [0.17, 0.68] , CVD (HR 0.70 [0.58, 0.84]), fatal coronary heart disease (HR 0.40 [0.22, 0.73]) or CHD (HR 0.66 [0.54, 0.80]) favoured the use of MET+ DPP-4 i. [234]

. O’Brien et al. compared MET + SUs vs. MET + DPP-4 i. Patients treated with MET + SUs were at higher risk of MACE (HR1.36 [1.23-1.49]) and congestive HF (HR 1.47 [1.23-1.75]) [235]

. In a Taiwan nationwide study using National Health Insurance Research Database, DPP-4 inhibitors added to metformin were associated with lower risks for all-cause death (hazard ratio [HR], 0.63 [95% CI, 0.55 to 0.72]),

VI. Second-line, metformin-based, dual therapies 132

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

MACEs (HR, 0.68 [CI, 0.55 to 0.83]), ischemic stroke (HR, 0.64 [CI, 0.51 to 0.81]), and hypoglycemia (HR, 0.43 [CI, 0.33 to 0.56]) compared with sulfonylureas as add-on therapy to metformin. There were no differences on risks for myocardial infarction and hospitalization for heart failure.[236]

Table VI.6 shows a summary of the current observational evidence concerning the effect of DPP-4 i added to metformin compared with metformin plus SUs.

Table VI.6 Observational evidence of cardiovascular outcomes with second-line DDP-4 i vs sulfonylureas added to metformin

Metformin + DPP-4 i vs metformin + SUs as reference

MACE Death CV CVD MI Stroke HF IHD PAD death Yu 0.62; 0.53 (0.29 - 0.74 (0.38 - [230] (0.40 - 0.97) 1.43) 0.98) Mogensen 0.70 0.65 (0.54 – 0.57 [231] (0.57 - 0.80) (0.40- 0.85) 0.80) Seong 0.83 0.88 0.66 0.93 [232] (0.76- (0.52- (0.56- (0.62- 0.92) 0.96) 0.78) 1.41) Kannan 1.03 (0.81- 1.06 1.28 1.104 1.35 1.65 [233] 1.31) (0.99- (1.08- (1.04- (1.16- (1.16- 1.13) 1.52) 1.17) 1.57) 2.36) Ekström 0.79 (0.60- 0.70 0.70 0.82 0.54 0.66 [234] 1.04) (0.34 (0.58- (0.53- (0.38- (0.54- – 0.84) 1.25) 0.76) 0.80) 0.68) 0.40 (0.22- 0.73) O’Brien 0.74 0.68 [235] (0.57- (0.57- 0.81) 0.81) 0.68 0.63 (0.55 - 0.64 Ou (0.55 - 0.72) (0.51 - 0.83) 0.81) [236]

VI. Second-line, metformin-based, dual therapies 133

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 MET + meglitinides compared with MET + SUs

. Using nationwide administrative Danish registries, Mogensen et al. found that users of MET+ repaglinide (n= 2,118) showed no differences in mortality compared with the use of MET + glimepiride as reference (n=40,026). For mortality were 0.81(0.62-1.05), CV mortality Hr 0.81 (0.56- 1.19) and for 3-points MACE 0.87 (0.68-1.10) [237]

. Ekström et al. compared MET+ meglitinides (n= 2,254) with MET + SUs (n= 8,801) as reference. The study didn’t find statistical differences for any outcome: mortality (HR 0.90 [0.75, 1.08]), stroke (HR 0.96 [0.70, 1.33]), heart failure (HR 1.00 [0.81, 1.24)]), fatal CVD (0.81 [0.56- 1.15]), CVD (1.08 [0.95- 1.23]), fatal coronary heart disease (0.76 [0.54 -1.06]) or CHD (1.08 [0.93-1.24]). [234]

Table VI.7 shows a summary of the current observational evidence concerning the effect of meglitinides added to metformin compared with metformin plus SUs.

Table VI.7 Observational evidence of cardiovascular outcomes with second-line meglitinides vs sulfonylureas added to metformin

Metformin + meglitinides vs Metformin + SUs as reference Mogensen 0.87 0.81 (0.62- 0.81 (0.68- 1.05) (0.56- 1.10) 1.19) Ekström 0.90 (0.70- 0.81 1.08 0.96 1.00 1.08 1.33) (0.56 (0.95- (0.70- (0.81- (0.93 - 1.23) 1.33) 1.24) - 1.15) 1.24)

VI. Second-line, metformin-based, dual therapies 134

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Discussion

In our study, 28,539 patients who had received the first-line treatment with monotherapy were added other NIADs as intensification second-line dual therapies: most on MET + SUs (61.46%), 28.15% on MET+DPP-4 i and by 8 % with MET + meglitinides; other treatments were used in a much lower proportion, and because of this they have not been included in our analysis. The cohorts of second-line therapy were not well balanced in basal characteristics. Patients who were prescribed meglitinides were the oldest ones and had more comorbidities. (Table VI.2) Compared with MET + SUs as a reference, the cohort of MET + DPP-4 i users did not show statistically significant differences in the overall HRs for any outcomes. For MET + meglitinides compared with MET + SUs, we found an increased risk of MACE and all-cause mortality; the higher risk of dying was consistently increased across many subgroups (Table VI.5).

We have not found statistical differences in the risks of MACE comparing the MET + DPP-4 i users vs the MET+SUs users. Our results are in line with the CAROLINA trial, though in this trial the only studied SUs was glimepiride. [207] in mortality and other outcomes, with observational research, [233, 234] although other studies observed improved outcomes for MET + DPP-4 i users [230, 231, 232, 235, 236] or in CV events [233, 234].

In our study, in the subgroup analyses of the combined second-line MET + DPP-4 i compared with MET + SUs, patients with a history of heart failure were at increased risk for MACE, all-cause mortality, HF and PAD, and there was also at a trend for increased risk of MI. In the CARMELINA trial, linagliptin did not affect the risk of hospitalisation for HF or other selected HF-related outcomes, including participants with and without a history of HF. [238] It has been hypothesised that other circulating peptides that are substrates for DPP4 that could have independent cardiovascular effects. A consensus is emerging that DPP4is might not be a preferred agent in patients with T2DM with increased risk for CVD with concomitant HF. [239] A history of HF is a risk factor for mortality; in the subgroup analyses of patients treated with meglitinides, patients were at increased risk of dying irrespective than they had or no previous history of HF (although the risk was higher in patients with previous HF). In patients treated with SUs

VI. Second-line, metformin-based, dual therapies 135

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

monotherapy, the risk of all-cause mortality, MACE and all the outcomes was higher in patients with no history of HF (Section V, Table V.4). The analyses of subgroups of other monotherapies and dual combined treatments could help to clarify whether there is or not an association with the use of DPP-4 i therapies. This is an area for further research.

Concerning the increased risk in MACE and all-cause mortality we have observed in MET + meglitinides users compared with MET + SUs users, the results are not consistent with those obtained in Swedish and Danish databases. Ekström et al. included cancer and smoking among basal covariates and Mogensen et al. the Charlson score, which is more comprehensive and includes cancer, AIDS and other comorbidities. In our study, patients on MET+ meglitinides were significantly older and sicker than those from different cohorts. Due to better pharmacodynamic and pharmacokinetic characteristics of meglitinides compared with SUs, general practitioners could have preferred meglitinides to treat frail patients. Thus, a channelling bias cannot be ruled out, and results deserve further reappraisal. If an unmeasured confounder is correlated with observed covariates, the potential for confounding can be mitigated through direct covariate adjustment and create a cohort balanced on observed covariates. [240] The propensity score matching (PS matching) produces well-balanced cohorts, at expenses of reducing the sample size; the characteristics of the PS matched cohorts could tend to be closer to one of the groups, in this case, it would be the MET+ meglitinides group, and results would be restricted to a population of this characteristics.

Other limitations and strengths have been already mentioned in the Section V, that report the results of a companion study about monotherapies.

Conclusion

In this study, no differences were found in mortality and CV outcomes between the dual second-line treatments for metformin-based regimes either with the addition of SUs or DPP-4 i. We also observed an increased risk of MACE and all-cause mortality in the MET+meglitinides cohort, channelling bias, and residual confounding cannot be ruled out and deserve further research.

VI. Second-line, metformin-based, dual therapies 136

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

VII. Discussion and Conclusion

VII. Discussion and Conclusion

The regulatory framework for non-insulin blood-glucose-lowering agents (“NIADs”) is based on surrogate blood glucose-related markers, mainly glycated haemoglobin. This surrogate marker has shown to be associated with microvascular complications of T2DM, but its relationship with the cardiovascular disease remains uncertain. In 2008 the FDA issued guidelines recommending the sponsors to demonstrate that the therapy will not result in an unacceptable increase in cardiovascular risk - as a result of the meta-analysis finding an increased risk of MI with rosiglitazone; the European Medicines Agency made a similar decision in 2012.[38, 39] The sponsor can get a marketing authorisation before cardiovascular safety is demonstrated; therefore, some new non-insulin blood glucose- lowering drugs are used in clinical practice before the results of cardiovascular safety become available. [163]

One decade later, fifteen large randomised controlled trials (RCTs) assessing CV outcomes have been published, as it has been summarised in Section II. In spite of some criticism raised by their non-inferiority design basis of these trials, the contribution of cardiovascular outcomes trials to the scientific knowledge and decision-making, both in clinical practice and public health, is outstanding. Instead, for older – and widely used - therapeutic groups, there is scarce RCTs evidence.

VII. Discussion and Conclusion 138

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Along with the conducting of these trials, the development of some agents have been terminated: this is the case of fasiglifam, in which events of hepatotoxicity led to an unfavourable balance risk/benefit; the hepatotoxicity has been later explained by the effects of the drug, TKD-875, on bile acid and bilirubin homeostasis. Other agents failed to show any beneficial effects, such as the aleglitazar, a dual peroxisome proliferator- activated receptors (PPARs). [241-243] The AleCardio (aleglitazar) trial was stopped due to futility for efficacy at an unplanned interim analysis and increased rates of safety endpoints, which included heart failure, gastrointestinal and renal dysfunction. [82]

In 2014, the first sentence of the paper reporting the AleCardio study results was: “No therapy directed against diabetes has been shown to unequivocally reduce the excess risk of cardiovascular complications”. At that time, this was true. Hereafter, some cardiovascular outcome trials (CVOTs) assessing some therapeutic groups of non-insulin blood glucose-lowering drugs have shown a reduction in mortality and cardiovascular outcomes. In terms of CV benefit of the absence of harm, the group of dipeptidyl peptidase-4 inhibitors showed no harmful or beneficial effects on CV outcomes, although in April 2016 added “Warnings and Precautions” to the labels of medicines that contain saxagliptin or alogliptin to inform of the potential increased risk of heart failure. [244]

In the group of glucagon-like peptide 1 receptor agonists, liraglutide (LEADER trial), semaglutide (SUSTAIN-6) and albiglutide (HARMONY) had shown a reduction of MACE, for liraglutide was driven by a decrease in all-cause and CV mortality, for semaglutide by a decreased risk of stroke and albiglutide by a reduction of the risk of MI. In the group of sodium-glucose co-transporter-2 inhibitors, CVOTs with empagliflozin (EMPA-REG trial), canagliflozin (CANVAS trial) and dapagliflozin (DECLARE-TIMI trial) achieved all significant reductions in hospitalization for heart failure, but only empagliflozin achieved a significant reduction in CV mortality and all-cause mortality.

Large RCTs also detected other safety concerns. The aggregated data of DPP-4 i trials have highlighted a safety signal regarding the risk of pancreatitis. [112] The example of fasiglifam highlights the importance of safety monitoring processes within the

VII. Discussion and Conclusion 139

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

randomised design of adequately sized clinical trials (7,600 patients in eight Phase III clinical trials)

However, the high CV risk populations included in CVOTs trials are substantially different from the “real-world” unselected ones - who are the people receiving medications, and, ultimately, pay for them- as summarized in the Section III “Generalizability of RCTs”. The observational research, especially the one conducted with data of thousands, hundred thousand or even million patients of routine clinical care has become increasingly important in terms of decision-making and public health policy. Observational research is included in the US 21st Century Cures Act as a source of complementary or confirmatory evidence. [247]

In spite of the obvious methodological limitations of the observational research due to non-randomised design and of restrictions in the availability of some clinical data – while they are retrieved in clinical trials, namely adherence and lifestyle habits – observational research plays now a more relevant role in the assessment of the clinical impact of therapies - either old and new agents used - in a given population. Individuals belonging to these “real-world” populations fund these therapies, by their own or through either private or public healthcare systems.

We conducted the present study with the data of a large, comprehensive, healthcare database that gather clinical, diagnoses and prescriptions and dispensing data of the 80% of the Catalan population. As this database serves to administrative purposes, it is linked with mortality registries; this allows a very reliable available information on deaths in individuals.

Once data have been analysed, our study’s results revealed some shortcomings in design and conducting; but also made evident some unique strengths. Respect to the all-cause mortality outcome, one of the more important limitations in the design is that we haven’t included data of cancer or other relevant morbidities diagnoses among covariates. Besides, data about people living in nursing homes were not available. The original protocol registered in ENCePP included CV mortality in the MACE (the primary

VII. Discussion and Conclusion 140

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

outcome) and all-cause mortality as one of the secondary outcomes; however, full data of mortality causes were unavailable, and only all-cause mortality has been assessed.

Frailty may be independently associated with antidiabetic medication exposure and risk of mortality, thus could introduce confounding.[240] In our study, unlikely to subgroups analyzed for DPP-4 i a and SUs, patients of many MET+ meglitinides cohort’s subgroups were at higher risk of dying; this could reveal a residual confounding by indication. Meglitinides are secretagogues that produce a rapid and short-lived insulin output; therefore, the potential for hypoglycaemia and other adverse effects are supposed to be reduced. There are no RCTs that report the effect of meglitinides on mortality or diabetes- related complications. [225] Frail patients might have been prescribed meglitinides because of their safer profile compared with SUs. The lack of cancer diagnosis or other covariates indicating frailty prevents us drawing any sound conclusion based on the results of all-cause mortality in the MET+ meglitinides group compared with the MET + SUs group (HR 1.38 [95% CI 1.14-1.67]) or the results of the monotherapy meglitinides vs metformin (HR 2.08 [1.26-3.42]). The measurement of all factors plausibly related to prognosis is essential. [168] Specific covariates allowing to control appropriately for confounding for indication were not available in our study. Even in studies that control for relevant measured confounders but have unmeasured or poorly measured frailty, the possibility of residual confounding by frailty exists. [240] Another strategy as the propensity score matching would have reduced the sample size and also produced a cohort representative of all MET + meglitinide users, but not fully representative of MET + SUs users, thus limiting its generalizability.

Another limitation of the design is that the washout time of 90 days before the first prescription is not enough long to avoid the inclusion of prevalent users: in the selection process performed by the Institut Jordi Gol (IJG), the subset of patients entering in the 2010 cohort is much larger than in the following years. For the first-line monotherapy analysis, patients having previous exposures to a blood glucose-lowering agent were not included, but it is likely that the process was not effective in the 2010 cohort, because data of the year 2009 were not available in the coded de-identified data set provided by

VII. Discussion and Conclusion 141

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

the IJG. A sensitivity analysis excluding the 2010 cohort was performed and showed results consistent with the ones of the primary analysis.

Other limitations have been already mentioned in the Section IV “Protocol design and rationale”, such as the inclusion of all events of stroke -which can retrieve embolic events not necessarily related with macrovascular atherosclerosis instead of only ischemic events. As data of hospitalisation are not available, the outcome referred to heart failure is based on recorded diagnoses (as well as those of peripheral arteriopathy) and encompasses both ambulatory diagnoses and hospitalisations. However, this outcome might be more sensitive in detecting the incidence of heart failure.

As it had been foreseen in the protocol submitted to ENCePP, we haven’t enough sample size to obtain significant results in patients treated with SGLT-2 i, GLP-1 RAs and TZD. An extension of this study, including 2016-2019 years, could overcome this limitation.

This study also has unique strengths. It is the first longitudinal study conducted with data of SIDIAP assessing cardiovascular outcomes and mortality in patients with T2DM treated with NIADs, in first and second-lines. The results of the first-line treatments are consistent with those obtained by observational research.

We also compared two cohorts, both treated with the combination metformin and SUs but with different previous monotherapies -either metformin or SUs, the analysis resulted in no differences. We found that combined metformin and sulfonylurea therapy had similar outcomes in patients previously treated with metformin o with SUs, suggesting a beneficial effect of metformin. Douros et al. found that, in patients on metformin who added or switching SUs, patients who switched to SUs were at increased risks of MI and all-cause mortality than those who added SUs (myocardial infarction, HR 1.5 95% CI 1.03 to 2.24; and all-cause mortality, HR 1.23, 95% CI 1.00 to 1.50). [245] No differences were observed for ischaemic stroke, cardiovascular death, or severe hypoglycaemia. These findings support the beneficial effect of metformin, [245] also showed in our study.

The results of some agents in the GLP-1 RAs and SGLT-2 i have risen enthusiasm regarding the potential of these new therapies. It should be kept in mind that SGLT-2 i,

VII. Discussion and Conclusion 142

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

while achieving a significant reduction in mortality, mainly due to their mechanism of action -volume depletion because of osmotic diuresis - have been associated to adverse effects such as amputations, especially canagliflozin, and ketoacidosis. Concerning GLP- 1 RA, oral formulations (semaglutide) can achieve better adherence, but the percentage of patients who are intolerant to their gastrointestinal adverse effects is much higher than the one of the patients intolerant to metformin. Besides, GLP-1 RAs are the more expensive agents among the NIADs, and CVOTs included mainly people with high CV risk, making it the results challenging to be extrapolated to the “real-world” populations. As shown in the Section II, populations’characteristics of published CVOTs and the one registered in SIDIAP differ, therefore, the results of CVTs should not be fully applicable to the latter.

In the meta-analysis of observational studies assessing CV outcomes and mortality in GLP-1 RAs users (see Annex 1) we found that the reduction in the risk of death had a larger size than those obtained in CVOTs. Crude incidence rate ratio observed in our study (second-line therapies) also suggest a reduced risk of mortality for GLP-1 users. This deserves further study, with a larger sample size and, probably, a longer follow-up.

Observational research is produced more and more and has obtained substantive evidence about the most used drugs; in some cases, prompted regulatory decision or modifications of clinical guidelines. The study design and analysis of observational data have evolved – as well as the RCTs- and are continuously improving. Standardisation of design and methods would allow the reproducibility and comparability among studies. Well- conducted observational research assessing the comparative effectiveness and safety of drugs used in clinical practice is now essential for an appropriate evaluation of the medicines, especially when RCTs are not available or their population do not represent the one that use the drug.

VII. Discussion and Conclusion 143

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

References References:

[1] http://www.who.int/mediacentre/factsheets/fs312/en/ Reviewed Nov 2016. Accessed on January 15th 2017

[2] International Diabetes Federation, http://www.idf.org/diabetesatlas/update-2014, accessed on January 17th 2017.

[3] Magliano DJ, Islam RM, Barr ELM, Gregg EW, Pavkov ME, Harding JL, Tabesh M Koye DN, Shaw JE. Trends in incidence of total or type 2 diabetes: systematic review. BMJ. 2019 Sep 11;366:l5003.

[4] Sharma M, Nazareth I, Petersen I. Trends in incidence, prevalence and prescribing in type 2 diabetes mellitus between 2000 and 2013 in primary care: a retrospective cohort study. BMJ Open. 2016 Jan 13;6(1):e010210.

[5] Vinagre I, Mata-Cases M, Hermosilla E, Morros R, Fina F, Rosell M, Castell C, Franch-Nadal J, Bolíbar B, Mauricio D. Control of Glycemia and Cardiovascular Risk Factors in Patients With Type 2 Diabetes in Primary Care in Catalonia (Spain). Diabetes Care 35:774–779, 2012

[6] Kirkman MS, Briscoe VJ, Clark N, Florez H, Haas LB, Halter JB, Huang ES, Korytkowski MT, Munshi MN, Odegard PS, Pratley RE, Swift CS. Diabetes in older adults. Diabetes Care. 2012 Dec;35(12):2650-64.

[7] Shahnam Sharif, Rolf H. H. Groenwold, Yolanda van der Graaf, Gijs F. N. Berkelmans, Maarten J. Cramer, Frank L. J. Visseren, Jan Westerink, On behalf of the SMART study group Mediation analysis of the relationship between type 2 diabetes and cardiovascular events and all‐cause mortality: Findings from the SMART cohort Diabetes Obes Metab. 2019 Aug; 21(8): 1935–1943.

[8] Tancredi M, Rosengren A, Svensson AM, Kosiborod M, Pivodic A, Gudbjörnsdottir S, Wedel H, Clements M, Dahlqvist S, Lind M. Excess Mortality among Persons with Type 2 Diabetes N Engl J Med. 2015 Oct 29;373(18):1720-32

[9] Skyler JS, Bergenstal R, Bonow RO, Buse J, Deedwania P, Gale EA, Howard BV, Kirkman MS, Kosiborod M, Reaven P, Sherwin RS; American DiabetesAssociation;

Bibliography 145

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

American College of Cardiology Foundation; American Heart Association. Intensive Glycemic Control and the Prevention of Cardiovascular Events: Implications of the ACCORD, ADVANCE, and VA Diabetes Trials. A Position Statement of the American Diabetes Association and a Scientific Statement of the American College of Cardiology Foundation and the American Heart Association 2009

[10] Leon BM, Maddox TM. Diabetes and cardiovascular disease: Epidemiology, biological mechanisms, treatment recommendations and future research. World J Diabetes. 2015 Oct 10;6(13):1246-58.

[11] Diabetes mellitus: a major risk factor for cardiovascular disease. A joint editorial statement by the American Diabetes Association; The National Heart, Lung, and Blood Institute; The Juvenile Diabetes Foundation International; The National Institute of Diabetes and Digestive and Kidney Diseases; and The American Heart Association. Circulation. 1999 Sep 7;100(10):1132-3.

[12] Buse JB, Ginsberg HN, Bakris GL, Clark NG, Costa F, Eckel R, Fonseca V, Gerstein HC, Grundy S, Nesto RW, Pignone MP, Plutzky J, Porte D, Redberg R, Stitzel KF, Stone NJ; American Heart Association; American Diabetes Association Primary Prevention of Cardiovascular Diseases in People With Diabetes Mellitus Circulation. 2007;115:114-126

[13] Einarson TR, Acs A, Ludwig C, Panton UH. Prevalence of cardiovascular disease in type 2 diabetes: a systematic literature review of scientific evidence from across the world in 2007-2017. Cardiovasc Diabetol. 2018 Jun 8; 17(1):83.

[14] Jurado J, Ybarra J, Solanas P, Caula J, Gich I, Pou JM, Romeo JH Prevalence of cardiovascular disease and risk factors in a type 2 diabetic population of the North Catalonia diabetes study. J Am Acad Nurse Pract. 2009 Mar;21(3):140-8. doi: 10.1111/j.1745-7599.2008.00377.x.

[15] Fox CS, Coady S, Sorlie PD, Levy D, Meigs JB, D'Agostino RB Sr, Wilson PW, Savage PJ. Trends in cardiovascular complications of diabetes. JAMA. 2004 Nov 24;292(20):2495-9.

Bibliography 146

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[16] Cheng YJ, Imperatore G, Geiss LS, Saydah SH, Albright AL, Ali MK, Gregg EW. Trends and Disparities in Cardiovascular Mortality Among U.S. Adults With and Without Self-Reported Diabetes, 1988-2015. Diabetes Care. 2018 Nov;41(11):2306-2315. doi: 10.2337/dc18-0831.

[17] Orozco-Beltrán D, Sánchez E, Garrido A, Quesada JA, Carratalá-Munuera MC, Gil-Guillén VF. Trends in Mortality From Diabetes Mellitus in Spain: 1998-2013. Rev Esp Cardiol (Engl Ed). 2017 Jun;70(6):433-443. doi: 10.1016/j.rec.2016.09.022. Epub 2016 Nov 5. [Article in English, Spanish]

[18] Bhatnagar P, Wickramasinghe K, Wilkins E, Townsend N. Trends in the epidemiology of cardiovascular disease in the UK. Heart. 2016 Dec 15;102(24):1945-1952. doi: 10.1136/heartjnl-2016-309573. Epub 2016 Aug 22.

[19] Gilbert RE, Krum H. Heart failure in diabetes: effects of anti-hyperglycaemic drug therapy. Lancet. 2015 May 23;385(9982):2107-17.

[20] McKee PA, Castelli WP, McNamara PM, Kannel WB. The natural history of congestive heart failure: the Framingham study. N Engl J Med. 1971 Dec 23;285(26):1441-6.

[21] Kannel WB, Hjortland M, Castelli WP. Role of diabetes in congestive heart failure: the Framingham study. Am J Cardiol. 1974 Jul;34(1):29-34

[22] Shah AD, Langenberg C, Rapsomaniki E, Denaxas S, Pujades-Rodriguez M, Gale CP, Deanfield J, Smeeth L, Timmis A, Hemingway H. Type 2 diabetes and incidence of cardiovascular diseases: a cohort study in 1·9 million people. Lancet Diabetes Endocrinol. 2015 Feb;3(2):105-13.

[23] Conrad N, Judge A, Tran J, Mohseni H, Hedgecott D, Crespillo AP, Allison M, Hemingway H, Cleland JG, McMurray JJV, Rahimi K. Temporal trends and patterns in heart failure incidence: a population-based study of 4 million individuals. Lancet. 2018 Feb 10;391(10120):572-580.

[24] Tran J, Norton R, Conrad N, Rahimian F, Canoy D, Nazarzadeh M, Rahimi K. Patterns and temporal trends of comorbidity among adult patients with incident

Bibliography 147

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

cardiovascular disease in the UK between 2000 and 2014: A population-based cohort study. PLoS Med. 2018 Mar 6;15(3):e1002513.

[25] Baena-Díez JM, Peñafiel J, Subirana I, Ramos R, Elosua R, Marín-Ibañez A, Guembe MJ, Rigo F, Tormo-Díaz MJ, Moreno-Iribas C, Cabré JJ, Segura A, García-Lareo M, Gómez de la Cámara A, Lapetra J, Quesada M, Marrugat J, Medrano MJ, Berjón J, Frontera G, Gavrila D, Barricarte A, Basora J, García JM, Pavone NC, Lora-Pablos D, Mayoral E, Franch J, Mata M, Castell C, Frances A, Grau M; FRESCO Investigators. Risk of Cause-Specific Death in Individuals With Diabetes: A Competing Risks Analysis. Diabetes Care. 2016 Nov;39(11):1987- 1995.

[26] Mata-Cases M, Casajuana M, Franch-Nadal J, Casellas A, Castell C, Vinagre I, Mauricio D, Bolíbar B. Direct medical costs attributable to type 2 diabetes mellitus: a population-based study in Catalonia, Spain. Eur J Health Econ. 2016 Nov;17(8):1001-1010.

[27] Gale EA. Is type 2 diabetes a category error? Lancet. 2013 Jun 1;381(9881):1956-7.

[28] Piarulli F, Sartore G, Lapolla A. Glyco-oxidation and cardiovascular complications in type 2 diabetes: a clinical update. Acta Diabetol. 2013 Apr;50(2):101-10.

[29] Goode KM, John J, Rigby AS, Kilpatrick ES, Atkin SL, Bragadeesh T, Clark AL, Cleland JG. Elevated glycated haemoglobin is a strong predictor of mortality in patients with left ventricular systolic dysfunction who are not receiving treatment for diabetes mellitus. Heart. 2009 Jun;95(11):917-23. doi: 10.1136/hrt.2008.156646. Epub 2009 Feb 19.

[30] Clark SL, Santin AE, Bryant PA, Holman R, Rodnick KJ. The initial noncovalent binding of glucose to human hemoglobin in nonenzymatic glycation. Glycobiology. 2013 Nov;23(11):1250-9.

[31] Kingwell A FDA eyes new diabetic endpoints. Nat Rev Drug Discov. 2016 Sep 29;15(10):666-7.

Bibliography 148

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[32] Selvaraj N, Bobby Z, Sridhar MG. Increased glycation of hemoglobin in chronic renal failure: [corrected] potential role of oxidative stress. Arch Med Res. 2008 Apr;39(3):277-84. doi: 10.1016/j.arcmed.2008.01.001.

[33] Diabetes Control and Complications Trial Research Group, Nathan DM, Genuth S, et al. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med. 1993;329(14):977–986. doi:10.1056/NEJM199309303291401

[34] Stratton IM, Adler AI, Neil HA, Matthews DR, Manley SE, Cull CA, Hadden D, Turner RC, Holman RR. Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ. 2000 Aug 12;321(7258):405-12.

[35] Sharif S, Groenwold RHH, van der Graaf Y, Berkelmans GFN, Cramer MJ, Visseren FLJ, Westerink J; SMART study group. Mediation analysis of the relationship between type 2 diabetes and cardiovascular events and all-cause mortality: Findings from the SMART cohort. Diabetes Obes Metab. 2019 Aug;21(8):1935-1943

[36] UK Prospective Diabetes Study Group (UKPDS) Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). UK Prospective Diabetes Study (UKPDS) Group. Lancet. 1998 Sep 12;352(9131):837-53.

[37] Rodríguez- Gutiérrez R, Montori VM. Glycemic Control for Patients With Type 2 Diabetes Mellitus: Our Evolving Faith in the Face of Evidence. Circ Cardiovasc Qual Outcomes. 2016 Sep;9(5):504-12.

[38] U.S. Department of Health and Human Services Food and Drug Administration Center for Drug Evaluation and Research (CDER). Guidance for Industry. “Diabetes Mellitus – Evaluating CV Risk in New Anti-diabetic Therapies to Treat Type 2 Diabetes”. Food &Drug Administration, Dec. 2008. http://www.fda.gov/ downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ ucm071627.pdf.

Bibliography 149

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[39] http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2 012/06/WC500129256.pdf

[40] Control Group, Turnbull FM, Abraira C, Anderson RJ, Byington RP, Chalmers JP, Duckworth WC, Evans GW, Gerstein HC, Holman RR, Moritz TE, Neal BC, Ninomiya T, Patel AA, Paul SK, Travert F, Woodward M. Intensive glucose control and macrovascular outcomes in type 2 diabetes. Diabetologia. 2009 Nov;52(11):2288-98

[41] Emerging Risk Factors Collaboration, Sarwar N, Gao P, Seshasai SR, Gobin R, Kaptoge S, Di Angelantonio E, Ingelsson E, Lawlor DA, Selvin E, Stampfer M, Stehouwer CD, Lewington S, Pennells L, Thompson A, Sattar N, White IR, Ray KK, Danesh J. Diabetes mellitus, fasting blood glucose concentration, and risk of vascular disease: a collaborative meta-analysis of 102 prospective studies. Lancet. 2010 Jun 26;375(9733):2215-22.

[42] Chen YY, Lin YJ, Chong E, Chen PC, Chao TF, Chen SA, Chien KL. The impact of diabetes mellitus and corresponding HbA1c levels on the future risks of cardiovascular disease and mortality: a representative cohort study in Taiwan. PLoS One. 2015 Apr 13;10(4):e0123116.

[43] Monesi L, Tettamanti M, Cortesi L, Baviera M, Marzona I, Avanzini F, Monesi G, Nobili A, Riva E, Fortino I, Bortolotti A, Fontana G, Merlino L, Trevisan R, Roncaglioni MC. Elevated risk of death and major cardiovascular events in subjects with newly diagnosed diabetes: findings from an administrative database. Nutr Metab Cardiovasc Dis. 2014 Mar;24(3):263-70

[44] Schöttker B, Rathmann W, Herder C, Thorand B, Wilsgaard T, Njølstad I, Siganos G, Mathiesen EB, Saum KU, Peasey A, Feskens E, Boffetta P, Trichopoulou A, Kuulasmaa K, Kee F Brenner H; CHANCES group. HbA1c levels in non-diabetic older adults -No J-shaped associations with primary vascular events, cardiovascular and all-causes mortality after adjustment for confounders in a meta- analysis of individual participant data from six cohort studies. BMC Med. 2016 Feb 11;14:26.

Bibliography 150

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[45] Gerstein HC, Pogue J, Mann JF, Lonn E, Dagenais GR, McQueen M, Yusuf S; HOPE investigators. The relationship between dysglycemia and cardiovascular and renal risk in diabetic and non-diabetic participants in the HOPE study: a prospective epidemiological analysis. Diabetologia. 2005 Sep;48(9):1749-55.

[46] Valenzuela-Garcia LF, Matsuzawa Y, Sara JD, Kwon TG, Lennon RJ, Lerman LO, Ruiz-Salmeron RJ, Lerman A. Lack of correlation between the optimal glycaemic control and coronary micro vascular dysfunction in patients with diabetes mellitus: a cross sectional study. Cardiovasc Diabetol. 2015 Aug 14;14:106.

[47] Kawata T, Daimon M, Miyazaki S, Ichikawa R, Maruyama M, Chiang SJ, Ito C, Sato F, Watada H, Daida H. Coronary microvascular function is independently associated with left ventricular filling pressure in patients with type 2 diabetes mellitus. Cardiovasc Diabetol. 2015 Aug 5;14:98.

[48] Frier BM, Schernthaner G, Heller SR. Hypoglycemia and cardiovascular risks. Diabetes Care. 2011 May;34 Suppl 2:S132-7

[49] Gerstein HC, Miller ME, Byington RP, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358:2545–2559

[50] Patel A, MacMahon S, Chalmers J, et al. Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes. N Engl J Med 2008;358:2560– 2572

[51] Duckworth W, Abraira C, Moritz T, et al. Glucose control and vascular complications in veterans with type 2 diabetes. N Engl J Med 2009;360:129–139

[52] Lachin JM, Orchard TJ, Nathan DM; DCCT/EDIC Research Group. Update on cardiovascular outcomes at 30 years of the diabetes control and complications trial/ epidemiology of diabetes interventions and complications study. Diabetes Care. 2014;37(1):39-43.

[53] UK Prospective Diabetes Study (UKPDS) Group. Effect of intensive blood- glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34). Lancet. 1998 Sep 12;352(9131):854-65.

Bibliography 151

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[54] Boussageon R, Gueyffier F, Cornu C. Metformin as firstline treatment for type 2 diabetes: are we sure? BMJ. 2016 Jan 8;352:h6748.

[55] Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HA. 10-year follow-up of intensive glucose control in type 2 diabetes. N Engl J Med. 2008 Oct 9;359(15):1577-89.

[56] Duckworth W, Abraira C, Moritz T, Reda D, Emanuele N, Reaven PD, Zieve FJ, Marks J, Davis SN, Hayward R, Warren SR, Goldman S, McCarren M, Vitek ME, Henderson WG, Huang GD; VADT Investigators Glucose control and vascular complications in veterans with type 2 diabetes. N Engl J Med. 2009 Jan 8;360(2):129-39. doi: 10.1056/NEJMoa0808431.

[57] Hayward RA, Reaven PD, Wiitala WL, Bahn GD, Reda DJ, Ge L, McCarren M, Duckworth WC, Emanuele NV; VADT Investigators. Follow-up of glycemic control and cardiovascular outcomes in type 2 diabetes.N Engl J Med. 2015 Jun 4;372(23):2197-206.

[58] ADVANCE Collaborative Group, Patel A, MacMahon S, Chalmers J, Neal B, Billot L, Woodward M, Marre M, Cooper M, Glasziou P, Grobbee D, Hamet P, Harrap S, Heller S, Liu L, Mancia G, Mogensen CE, Pan C, Poulter N, Rodgers A, Williams B, Bompoint S, de Galan BE, Joshi R, Travert F. Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes. N Engl J Med. 2008 Jun 12;358(24):2560-72.

[59] Action to Control Cardiovascular Risk in Diabetes Study Group., Gerstein HC, Miller ME, Byington RP, Goff DC Jr, Bigger JT, Buse JB, Cushman WC, Genuth S, Ismail-Beigi F, Grimm RH Jr, Probstfield JL, Simons-Morton DG, Friedewald WT. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med. 2008 Jun 12;358(24):2545-59

[60] ACCORD Study Group, Gerstein HC, Miller ME, Genuth S, Ismail-Beigi F, Buse JB, Goff DC Jr, Probstfield JL, Cushman WC, Ginsberg HN, Bigger JT, Grimm RH Jr, Byington RP, Rosenberg YD, Friedewald WT. Long-term effects of

Bibliography 152

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

intensive glucose lowering on cardiovascular outcomes. N Engl J Med. 2011 Mar 3;364(9):818-28. doi: 10.1056/NEJMoa1006524.

[61] Yudkin JS, Lipska KJ, Montori VM. The idolatry of the surrogate.BMJ. 2011 Dec 28;343:d7995. doi: 10.1136/bmj.d7995.

[62] Blackburn H, Jacobs DR Jr. The University Group Diabetes Program 1961-1978: pioneering randomized controlled trial. Int J Epidemiol. 2017 Oct 1;46(5):1354-1364. doi: 10.1093/ije/dyw168.

[63] .Meinert CL, Knatterud GL, Prout TE, Klimt CR. A study of the effects of hypoglycemic agents on vascular complications in patients with adult-onset diabetes. II. Mortality results. Diabetes. 1970;19:Suppl:789-830.

[64] King P, Peacock I, Donnelly R. The UK prospective diabetes study (UKPDS): clinical and therapeutic implications for type 2 diabetes.Br J Clin Pharmacol. 1999 Nov;48(5):643-8.

[65] Boussageon Control Group, Turnbull FM, Abraira C, Anderson RJ, Byington RP, Chalmers JP, Duckworth WC, Evans GW, Gerstein HC, Holman RR, Moritz TE, Neal BC, Ninomiya T, Patel AA, Paul SK, Travert F, Woodward M. Intensive glucose control and macrovascular outcomes in type 2 diabetes. Diabetologia. 2009 Nov;52(11):2288-98. doi: 10.1007/s00125-009-1470-0.

[66] Cefalu WT, Kaul S, Gerstein HC, Holman RR, Zinman B, Skyler JS, Green JB, Buse JB, Inzucchi SE, Leiter LA, Raz I, Rosenstock J, Riddle MC. Cardiovascular Outcomes Trials in Type 2 Diabetes: Where Do We Go From Here? Reflections From a Diabetes Care Editors' Expert Forum. Diabetes Care. 2018 Jan;41(1):14-31. doi: 10.2337/dci17-0057.

[67] Nissen SE, Wolski K, Topol EJ. Effect of muraglitazar on death and major adverse cardiovascular events in patients with type 2 diabetes mellitus. JAMA. 2005 Nov 23;294(20):2581-6.

[68] Dormandy JA, Charbonnel B, Eckland DJ, Erdmann E, Massi-Benedetti M et al, PROactive investigators. Secondary prevention of macrovascular events in patients with type 2 diabetes in the PROactive Study (PROspective pioglitAzone Clinical Trial

Bibliography 153

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

In macroVascular Events): a randomised controlled trial. Lancet. 2005 Oct 8;366(9493):1279-89.

[69] Erdmann E, Dormandy JA, Charbonnel B, Massi-Benedetti M, Moules IK, Skene AM; PROactive Investigators The effect of pioglitazone on recurrent myocardial infarction in 2,445 patients with type 2 diabetes and previous myocardial infarction: results from the PROactive (PROactive 05) Study. J Am Coll Cardiol. 2007 May 1;49(17):1772-80.

[70] Psaty BM, Furberg CD. The record on rosiglitazone and the risk of myocardial infarction. N Engl J Med. 2007 Jul 5;357(1):67-9.

[71] Nissen SE, Wolski K. Effect of rosiglitazone on the risk of myocardial infarction and death from cardiovascular causes. N Engl J Med 2007;356:2457-71.

[72] Home PD, Pocock SJ, Beck-Nielsen H, et al. Rosiglitazone evaluated for cardiac outcomes an interim analysis. N Engl J Med 2007;357:28-38.

[73] Lipscombe LL, Gomes T, Lévesque LE, Hux JE, Juurlink DN, Alter DA. Thiazolidinediones and cardiovascular outcomes in older patients with diabetes. JAMA. 2007 Dec 12;298(22):2634-43.

[74] Blind E, Janssen H, Dunder K2, de Graeff . he European Medicines Agency's approval of new medicines for type 2 diabetes.Diabetes Obes Metab. 2018 Sep;20(9):2059-2063. doi: 10.1111/dom.13349.

[75] PA3.https://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInformationf orPatientsandProviders/ucm226956.htm. Accessed on March 03 2017

[76] https://www.fda.gov/Safety/MedWatch/SafetyInformation/SafetyAlertsforHuma nMedicalProducts/ucm376683.htm Accessed on March 03 2017

[77] http://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm531517.h tm, accessed on Dec. 2nd 2016.

[78] http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2 016/03/WC500203804.pdf

Bibliography 154

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[79] EMA. Reflection paper on assessment of cardiovascular safety profile of medicinal products; 2016. https://www.ema.europa.eu/en/documents/scientific- guideline/reflection-paper-assessment-cardiovascular-safety-profile-medicinal- products_en.pdf Accessed 31 March 2018

[80] Rawlins M. De Testimonio: on the evidence for decisions about the use of therapeutic interventions. Clin Med (Lond). 2008 Dec;8(6):579-88..

[81] Cordoba G, Schwartz L, Woloshin S, Bae H, Gøtzsche PC. Definition, reporting, and interpretation of composite outcomes in clinical trials: systematic review. BMJ. 2010 Aug 18;341:c3920. Toxicol. 2016 Sep 19:1-5.

[82] Lincoff AM, Tardif JC, Neal B, Nicholls SJ, Rydén L, Schwartz GG, Malmberg K, Buse JB, Henry RR, Wedel H, Weichert A, Cannata R, Grobbee DE. Evaluation of the dual peroxisome proliferator-activated receptor α/γ agonist aleglitazar to reduce cardiovascular events in patients with acute coronary syndrome and type 2 diabetes mellitus: rationale and design of the AleCardio trial. Am Heart J. 2013 Sep;166(3):429-34. doi: 10.1016/j.ahj.2013.05.013.

[83] Lincoff AM, Tardif JC, Schwartz GG, Nicholls SJ, Rydén L, Neal B, Malmberg K, Wedel H, Buse JB, Henry RR, Weichert A, Cannata R, Svensson A, Volz D, Grobbee DE; AleCardio Investigators. Effect of aleglitazar on cardiovascular outcomes after acute coronary syndrome in patients with type 2 diabetes mellitus: the AleCardio randomized clinical trial. JAMA. 2014 Apr 16;311(15):1515-25. doi: 10.1001/jama.2014.3321

[84] http://www.worldpharmanews.com/roche/2488-roche-halts-investigation-of- aleglitazar

[85] Green JB, Bethel MA, Armstrong PW, Buse JB, Engel SS, Garg J, Josse R, Kaufman KD, Koglin J, Korn S, Lachin JM, McGuire DK, Pencina MJ, Standl E, Stein PP, Suryawanshi S, Van de Werf F, Peterson ED, Holman RR; TECOS Study Group. Effect of Sitagliptin on Cardiovascular Outcomes in Type 2 Diabetes. N Engl J Med. 2015 Jul 16;373(3):232-42.

Bibliography 155

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[86] Zannad F, Cannon CP, Cushman WC, Bakris GL, Menon V, Perez AT, Fleck PR, Mehta CR, Kupfer S, Wilson C, Lam H, White WB; EXAMINE Investigators. Heart failure and mortality outcomes in patients with type 2 diabetes taking alogliptin versus placebo in EXAMINE: a multicentre, randomized, double-blind trial. Lancet. 2015 May 23;385(9982):2067-76.

[87] White WB, Cannon CP, Heller SR, Nissen SE, Bergenstal RM, Bakris GL, Perez AT, Fleck PR, Mehta CR, Kupfer S, Wilson C, Cushman WC, Zannad F; EXAMINE Investigators. Alogliptin after acute coronary syndrome in patients with type 2 diabetes. N Engl J Med. 2013 Oct 3;369(14):1327-35.

[88] Scirica BM, Bhatt DL, Braunwald E, Steg PG, Davidson J, Hirshberg B, Ohman P, Price DL, Chen R, Udell J, Raz I. The design and rationale of the saxagliptin assessment of vascular outcomes recorded in patients with diabetes mellitus- thrombolysis in myocardial infarction (SAVOR-TIMI) 53 study. Am Heart J. 2011 Nov;162(5):818-825.e6

[89] Scirica BM, Bhatt DL, Braunwald E, Steg PG, Davidson J, Hirshberg B, Ohman P, Frederich R, Wiviott SD, Hoffman EB, Cavender MA, Udell JA, Desai NR, Mosenzon O, McGuire DK, Ray KK, Leiter LA, Raz I; SAVOR-TIMI 53 Steering Committee and Investigators. Saxagliptin and cardiovascular outcomes in patients with type 2 diabetes mellitus. N Engl J Med. 2013 Oct 3;369(14):1317-26.

[90] Udell JA, Bhatt DL, Braunwald E, Cavender MA, Mosenzon O, Steg PG, Davidson JA, Nicolau JC, Corbalan R, Hirshberg B, Frederich R, Im K, Umez- Eronini AA, He P, McGuire DK, Leiter LA, Raz I, Scirica BM; SAVOR-TIMI 53 Steering Committee and Investigators. Saxagliptin and cardiovascular outcomes in patients with type 2 diabetes and moderate or severe renal impairment: observations from the SAVOR-TIMI 53 Trial. Diabetes Care. 2015 Apr;38(4):696-705.

[91] Mosenzon O, Leibowitz G, Bhatt DL, Cahn A, Hirshberg B, Wei C, Im K, Rozenberg A, Yanuv I, Stahre C, Ray KK, Iqbal N, Braunwald E, Scirica BM, Raz I. Effect of Saxagliptin on Renal Outcomes in the SAVOR-TIMI 53 Trial. Diabetes Care. 2017 Jan;40(1):69-76.

Bibliography 156

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[92] Cavender MA, Scirica BM, Raz I, Steg PG, McGuire DK, Leiter LA, Hirshberg B, Davidson J, Cahn A, Mosenzon O, Im K, Braunwald E, Bhatt DL. Cardiovascular Outcomes of Patients in SAVOR-TIMI 53 by Baseline Hemoglobin A1c. Am J Med. 2016 Mar;129(3):340.e1-8.

[93] Rosenstock J, Perkovic V, Johansen OE, Cooper ME, Kahn SE, Marx N, Alexander JH, Pencina M, Toto RD, Wanner C, Zinman B, Woerle HJ, Baanstra D, Pfarr E, Schnaidt S, Meinicke T, George JT, von Eynatten M, McGuire DK; CARMELINA Investigators. Effect of Linagliptin vs Placebo on Major Cardiovascular Events in Adults With Type 2 Diabetes and High Cardiovascular and Renal Risk: The CARMELINA Randomized Clinical Trial. JAMA. 2019 Jan 1;321(1):69-79. doi: 10.1001/jama.2018.18269.

[94] Pfeffer MA, Claggett B, Diaz R, Dickstein K, Gerstein HC, Køber LV, Lawson FC, Ping L, Wei X, Lewis EF, Maggioni AP, McMurray JJ, Probstfield JL, Riddle MC, Solomon SD, Tardif JC; ELIXA Investigators. Lixisenatide in Patients with Type 2 Diabetes and Acute Coronary Syndrome. N Engl J Med. 2015 Dec 3;373(23):2247-57

[95] Marso SP, Daniels GH, Brown-Frandsen K, Kristensen P, Mann JF, Nauck MA, Nissen SE, Pocock S, Poulter NR, Ravn LS, Steinberg WM, Stockner M, Zinman B, Bergenstal RM, Buse JB; LEADER Steering Committee; LEADER Trial Investigators. Liraglutide and Cardiovascular Outcomes in Type 2 Diabetes. N Engl J Med. 2016 Jul 28;375(4):311-22.

[96] Marso SP, Bain SC, Consoli A, Eliaschewitz FG, Jódar E, Leiter LA, Lingvay I, Rosenstock J, Seufert J, Warren ML, Woo V, Hansen O, Holst AG, Pettersson J, Vilsbøll T; SUSTAIN-6 Investigators. Semaglutide and Cardiovascular Outcomes in Patients with Type 2 Diabetes. N Engl J Med. 2016 Nov 10;375(19):1834-184

[97] Holman RR, Bethel MA, Mentz RJ, Thompson VP, Lokhnygina Y, Buse JB, Chan JC, Choi J, Gustavson SM, Iqbal N, Maggioni AP, Marso SP, Öhman P, Pagidipati NJ, Poulter N, Ramachandran A, Zinman B, Hernandez AF; EXSCEL Study Group. Effects of Once-Weekly Exenatide on Cardiovascular Outcomes in Type 2 Diabetes. N Engl J Med. 2017 Sep 28;377(13):1228-1239.

Bibliography 157

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[98] Hernandez AF, Green JB, Janmohamed S, D'Agostino RB Sr, Granger CB, Jones NP, Leiter LA, Rosenberg AE, Sigmon KN, Somerville MC, Thorpe KM, McMurray JJV, Del Prato S; Harmony Outcomes committees and investigators. Albiglutide and cardiovascular outcomes in patients with type 2 diabetes and cardiovascular disease (Harmony Outcomes): a double-blind, randomised placebo-controlled trial. Lancet. 2018 Oct 1. pii: S0140-6736(18)32261-X

[99] Husain M, Birkenfeld AL, Donsmark M, Dungan K, Eliaschewitz FG, Franco DR, Jeppesen OK, Lingvay I, Mosenzon O, Pedersen SD, Tack CJ, Thomsen M, Vilsbøll T1, Warren ML1, Bain SC1; PIONEER 6 Investigators. Oral Semaglutide and Cardiovascular Outcomes in Patients with Type 2 Diabetes. N Engl J Med. 2019 Aug 29;381(9):841-851. doi: 10.1056/NEJMoa1901118.

[100] Gerstein HC, Colhoun HM, Dagenais GR, Diaz R, Lakshmanan M, Pais P, Probstfield J, Riddle MC,; REWIND Trial Investigators. Design and baseline characteristics of participants in the Researching cardiovascular Events with a Weekly INcretin in Diabetes (REWIND) trial on the cardiovascular effects of dulaglutide. Diabetes Obes Metab. 2018 Jan;20(1):42-49.

[101] Gerstein HC, Colhoun HM, Dagenais GR, Diaz R, Lakshmanan M, Pais P, Probstfield J, Botros FT, REWIND Investigators. Dulaglutide and renal outcomes in type 2 diabetes: an exploratory analysis of the REWIND randomised, placebo- controlled trial. Lancet. 2019 Jul 13;394(10193):131-138. doi: 10.1016/S0140- 6736(19)31150-X.

[102] Wu JH, Foote C, Blomster J, Toyama T, Perkovic V Sundström J, Neal B. Effects of sodium-glucose cotransporter-2 on cardiovascular events, death, and major safety outcomes in adults with type 2 diabetes: a systematic review and meta-analysis. Lancet Diabetes Endocrinol. 2016 May;4(5):411-9.

[103] Sano M, Takei M, Shiraishi Y, Suzuki Y. Increased Hematocrit During Sodium- Glucose Cotransporter 2 Inhibitor Therapy Indicates Recovery of Tubulointerstitial Function in Diabetic Kidneys. J Clin Med Res. 2016 Dec;8(12):844-847.

Bibliography 158

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[104] Imprialos KP, Boutari C, Stavropoulos K, Doumas M, Karagiannis AI. Stroke paradox with SGLT-2 inhibitors: a play of chance or a viscosity-mediated reality? J Neurol Neurosurg Psychiatry. 2017 Mar;88(3):249-253.

[105] Zinman B, Wanner C, Lachin JM, Fitchett D, Bluhmki E, Hantel S, Mattheus M, Devins T, Johansen OE, Woerle HJ, Broedl UC, Inzucchi SE; EMPA-REG OUTCOME Investigators. Empagliflozin, Cardiovascular Outcomes, and Mortality in Type 2 Diabetes. N Engl J Med 2015; 373:2117-2128November 26, 2015

[106] Fitchett D, Zinman B, Wanner C, Lachin JM, Hantel S, Salsali A, Johansen OE, Woerle HJ, Broedl UC, Inzucchi SE; EMPA-REG OUTCOME® trial investigators Heart failure outcomes with empagliflozin in patients with type 2 diabetes at highcardiovascular risk: results of the EMPA-REG OUTCOME® trial. Eur Heart J. 2016 May 14;37(19):1526-34

[107] Neal B, Perkovic V, Mahaffey KW, de Zeeuw D, Fulcher G, Erondu N, Shaw W, Law G, Desai M, Matthews DR; CANVAS Program Collaborative Group. Canagliflozin and Cardiovascular and Renal Events in Type 2 Diabetes. N Engl J Med. 2017 Aug 17;377(7):644-657. doi: 10.1056/NEJMoa1611925

[108] Wiviott SD, Raz I, Bonaca MP, Mosenzon O et al.; DECLARE–TIMI 58 Investigators. Dapagliflozin and Cardiovascular Outcomes in Type 2 Diabetes. N Engl J Med. 2019 Jan 24;380(4):347-357. doi: 10.1056/NEJMoa1812389. Epub 2018 Nov 10.

[109] McMurray JJV, Solomon SD, Inzucchi SEet al.; DAPA-HF Trial Committees and Investigators. Dapagliflozin in Patients with Heart Failure and Reduced Ejection Fraction. N Engl J Med. 2019 Sep 19. doi: 10.1056/NEJMoa1911303

[110] Tomlinson B, Hu M, Zhang Y, Chan P, Liu ZM. Effects of glucose-lowering drugs on cardiovascular outcomes in patients with type 2 diabetes. Expert Opin Drug Metab Toxicol. 2016 Sep 19:1-5.

Bibliography 159

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[111] Thompson PL, Davis TM. Cardiovascular Effects of Glucose-Lowering Therapies for Type 2 Diabetes: New Drugs in Perspective. Clin Ther. 2016 Nov 15. pii: S0149- 2918(16)30793-7.

[112] Home P. Cardiovascular outcome trials of glucose-lowering medications: an update. Diabetologia. 2019 Mar;62(3):357-369. doi: 10.1007/s00125-018-4801-1

[113] Press release Victoza® (liraglutide) is approved in the US as the only type 2 diabetes treatment indicated to reduce the risk of three major adverse cardiovascular events http://press.novonordisk-us.com/2017-08-25-Victoza-R-liraglutide-is- approved-in-the-US-as-the-only-type-2-diabetes-treatment-indicated-to-reduce-the- risk-of-three-major-adverse-cardiovascular-events accessed on 06 June 2018.

[114] https://www.fda.gov/news-events/press-announcements/fda-approves-jardiance- reduce-cardiovascular-death-adults-type-2-diabetes 02 December 2016

[115] McGuire DK, Marx N, Johansen OE, Inzucchi SE, Rosenstock J, George JT. FDA guidance on antihyperglyacemic therapies for type 2 diabetes: One decade later. Diabetes Obes Metab. 2019 May;21(5):1073-1078. doi: 10.1111/dom.13645.

[116] Marx N, McGuire DK, Perkovic V, Woerle HJ, Broedl UC, von Eynatten M, George JT, Rosenstock J. Composite Primary End Points in Cardiovascular Outcomes Trials Involving Type 2 Diabetes Patients: Should Unstable Angina Be Included in the Primary End Point? Diabetes Care. 2017 Sep;40(9):1144-1151. doi: 10.2337/dc17-0068.

[117] Freemantle N, Calvert M, Wood J, Eastaugh J, Griffin C. Composite outcomes in randomized trials: greater precision but with greater uncertainty? JAMA. 2003 May 21; 289(19):2554-9

[118] Montori VM, Permanyer-Miralda G, Ferreira-González I, Busse JW, Pacheco- Huergo V, Bryant D, Alonso J, Akl EA, Domingo-Salvany A, Mills E, Wu P, Schünemann HJ, Jaeschke R, Guyatt GH Validity of composite end points in clinical trials. BMJ. 2005 Mar 12; 330(7491):594-6.

Bibliography 160

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[119] Giugliano D, Maiorino MI, Bellastella G, Chiodini P, Esposito K. Glycemic Control, Preexisting Cardiovascular Disease, and Risk of Major Cardiovascular Events in Patients with Type 2 Diabetes Mellitus: Systematic Review With Meta- Analysis of Cardiovascular Outcome Trials and Intensive Glucose Control Trials. J Am Heart Assoc. 2019 Jun 18;8(12):e012356. doi: 10.1161/JAHA.119.012356.

[120] Jokhadar M, Jacobsen SJ, Reeder GS, Weston SA, Roger VL. Sudden death and recurrent ischemic events after myocardial infarction in the community. Am J Epidemiol. 2004 Jun 1;159(11):1040-6.

[121] Johansson S, Rosengren A, Young K, Jennings E. Mortality and morbidity trends after the first year in survivors of acute myocardial infarction: a systematic review. BMC Cardiovasc Disord. 2017 Feb 7;17(1):53. doi: 10.1186/s12872-017-0482-9.

[122] Tangri N, Ferguson TW, Whitlock RH, Rigatto C, Jassal DS, Kass M, Toleva O, Komenda P. Long term health outcomes in patients with a history of myocardial infarction: A population based cohort study. PLoS One. 2017 Jul 12;12(7):e0180010. doi: 10.1371/journal.pone.0180010.

[123] Birkeland KI, Bodegard J, Norhammar A, Kuiper JG, Georgiado E, Beekman- Hendriks WL, Thuresson M, Pignot M, Herings RMC, Kooy A. How representative of a general type 2 diabetes population are patients included in cardiovascular outcome trials with SGLT2 inhibitors? A large European observational study. Diabetes Obes Metab. 2018 Dec 9. doi: 10.1111/dom.13612

[124] .American Diabetes Association. 9. Pharmacologic Approaches to Glycemic Treatment: Standards of Medical Care in Diabetes-2019. Diabetes Care. 2019 Jan;42(Suppl 1):S90-S102. doi: 10.2337/dc19-S009.

[125] Barrull C, Massafrets M, Planellas L, Perulero N. Análisis de coste-efectividad de empagliflozina en pacientes con diabetes mellitus tipo 2 y enfermedad cardiovascular establecida en España http://www.aes.es/Jornadas2017/pdfs/posters/Aes_2017_23_1.pdf accessed on 22 October 2019.

Bibliography 161

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[126] Hinton W, Feher M, Munro N, Walker M, de Lusignan S . Real-world prevalence of the inclusion criteria for the LEADER trial: Data from a national general practice network. Diabetes Obes Metab. 2019 Jul;21(7):1661-1667

[127] Boye KS, Riddle MC, Gerstein HC, Mody R, Garcia-Perez LE, Karanikas CA, Lage MJ, Riesmeyer JS, Lakshmanan MC. Generalizability of glucagon-like peptide- 1 receptor agonist cardiovascular outcome trials to the overall type 2 diabetes population in the United States. Diabetes Obes Metab. 2019 Jun;21(6):1299-1304.

[128] Wittbrodt ET, Eudicone JM, Bell KF, Enhoffer DM, Latham K, Green JB. Generalizability of glucagon-like peptide-1 receptor agonist cardiovascular outcome trials enrollment criteria to the US type 2 diabetes population. Am J Manag Care. 2018 Apr;24(8 Suppl):S146-S155.

[129] Wittbrodt E, Chamberlain D, Arnold SV, Tang F, Kosiborod M. Eligibility of patients with type 2 diabetes for sodium-glucose co-transporter-2 inhibitor cardiovascular outcomes trials: An assessment using the Diabetes Collaborative Registry. Diabetes Obes Metab. 2019 Aug;21(8):1985-1989. doi: 10.1111/dom.13738. Epub 2019 Apr 24.

[130] Wittbrodt ET, Eudicone JM, Bell KF, Enhoffer DM, Latham K, Green JB. Eligibility varies among the 4 sodium-glucose cotransporter-2 inhibitor cardiovascular outcomes trials: implications for the general type 2 diabetes US population. Am J Manag Care. 2018 Apr;24(8 Suppl):S138-S145.

[131] .Nicolucci A, Candido R, Cucinotta D, Graziano G, Rocca A, Rossi MC, Tuccinardi F, Manicardi V. Generalizability of Cardiovascular Safety Trials on SGLT2 Inhibitors to the Real World: Implications for Clinical Practice. Adv Ther. 2019 Aug 13.

[132] Bothwell LE, Greene JA, Podolsky SH, Jones DS. Food and Drug Assessing the Gold Standard--Lessons from the History of RCTs. N Engl J Med. 2016 Jun 2;374(22):2175-81. doi: 10.1056/NEJMms1604593.

[133] Food and Drug Administration Best Practices for Conducting and Reporting Pharmacoepidemiologic Safety Studies Using Electronic Healthcare Data. May

Bibliography 162

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

2013 https://www.fda.gov/regulatory-information/search-fda-guidance- documents/best-practices-conducting-and-reporting-pharmacoepidemiologic-safety- studies-using-electronic Accessed on 25 May 2016

[134] Franklin JM, Schneeweiss S. When and How Can Real-World Data Analyses Substitute for Randomized Controlled Trials? Clin Pharmacol Ther. 2017 Dec;102(6):924-933. doi: 10.1002/cpt.857.

[135] https://www.congress.gov/bill/114th-congress/house-bill/34/text accessed on May 21st 2018

[136] Schneeweiss S, Avorn J. A review of uses of health care utilization databases for epidemiologic research on therapeutics. J Clin Epidemiol. 2005 Apr;58(4):323-37

[137] Najafzadeh M, Schneeweiss S. From Trial to Target Populations - Calibrating Real-World Data. N Engl J Med. 2017 Mar 30;376(13):1203-1205.

[138] Herrett E, Shah AD, Boggon R, Denaxas S, Smeeth L, van Staa T, Timmis A, Hemingway H. Completeness and diagnostic validity of recording acute myocardial infarction events in primary care, hospital care, disease registry, and national mortality records: cohort study. BMJ. 2013 May 20;346:f2350.

[139] http://www.encepp.eu/standards_and_guidances/documents/ENCePPGuideonM ethStandardsinPE_Rev7.pdf

[140] https://www.pharmacoepi.org/resources/policies/guidelines-08027/

[141] http://www.encepp.eu/encepp/openAttachment/fullProtocol/18513;jsessionid=m ClXTymXcRntbO4vtPMPVS82xGlCU52KWPDKwDWmkbizKC2P8AmR!119696 1961

[142] Fonseca V, Desouza C, Asnani S, Jialal I. Nontraditional risk factors for cardiovascular disease in diabetes. Endocr Rev. 2004;25(1):153–175. doi:10.1210/er.2002-0034

[143] Martín-Timón I, Sevillano-Collantes C, Segura-Galindo A, Del Cañizo-Gómez FJ. Type 2 diabetes and cardiovascular disease: Have all risk factors the same strength? World J Diabetes. 2014 Aug 15;5(4):444-70. doi: 10.4239/wjd.v5.i4.444.

Bibliography 163

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[144] Tancredi M, Rosengren A, Svensson AM, Pivodic A,5, Gudbjörnsdottir S, Wedel H, Lind M. Glycaemic control and excess risk of major coronary events in patients with type 2 diabetes: a population-based study. Open Heart. 2019 Sep 9;6(2):e000967.

[145] Tancredi M, Rosengren A, Svensson AM, Kosiborod M, Pivodic A, Gudbjörnsdottir S, Wedel H, Clements M, Dahlqvist S, Lind M. Excess Mortality among Persons with Type 2 Diabetes. N Engl J Med. 2015 Oct 29;373(18):1720-32.

[146] Raghavan S, Vassy JL, Ho Y, Song R, Gagnon DR, Cho K, Wilson PWF, Phillips LS. Diabetes Mellitus-Related All-Cause and Cardiovascular Mortality in a National Cohort of Adults. J Am Heart Assoc. 2019 Feb 19;8(4):e011295.

[147] Schneeweiss S, Rassen JA, Brown JS, Rothman KJ, Happe L, Arlett P, Dal Pan G, Goettsch W, Murk W, Wang SV Graphical Depiction of Longitudinal Study Designs in Health Care Databases. Ann Intern Med. 2019 Mar 19;170(6):398-406.

[148] Schneeweiss S, Patrick AR, Stürmer T, Brookhart MA, Avorn J, Maclure M, Rothman KJ, Glynn RJ. Increasing levels of restriction in pharmacoepidemiologic database studies of elderly and comparison with randomized trial results. Med Care. 2007 Oct;45(10 Supl 2):S131-42.

[149] Barrot-de la Puente J, Mata-Cases M, Franch-Nadal J, Mundet-Tudurí X, Casellas A, Fernandez-Real JM, Mauricio D. Older type 2 diabetic patients are more likely to achieve glycaemic and cardiovascular risk factors targets than younger patients: analysis of a primary care database. Int J Clin Pract. 2015 Dec;69(12):1486-95.

[150] Ramón Gomisa, Sara Artolab, Pedro Conthec, Josep Vidala, Ricard Casamord, Beatriu Fontd. Prevalence of type 2 diabetes mellitus in overweight or obese patients outpatients in Spain. Grupo de Estudio OBEDIA 14 May 2013.

[151] Bolíbar B, Fina Avilés F, Morros R, Garcia-Gil M del M, Hermosilla E, Ramos R, Rosell M, Rodríguez J, Medina M, Calero S, Prieto-Alhambra D; Grupo SIDIAP. [SIDIAP database: electronic clinical records in primary care as a source of

Bibliography 164

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

information for epidemiologic research]. [Article in Spanish] Med Clin (Barc). 2012 May 19;138(14):617-21.

[152] Canivell S, Mata-Cases M, Real J, Franch-Nadal J, Vlacho B, Khunti K, Gratacòs M, Mauricio D. Glycaemic control after treatment intensification in patients with type 2 diabetes uncontrolled on two or more non-insulin antidiabetic drugs in a real-world setting. Diabetes Obes Metab. 2019 Jun;21(6):1373-1380

[153] Losada E, Soldevila B, Ali MS, Martínez-Laguna D, Nogués X, Puig-Domingo M, Díez-Pérez A, Mauricio D, Prieto-Alhambra D. Real-world antidiabetic drug use and fracture risk in 12,277 patients with type 2 diabetes mellitus: a nested case- control study. Osteoporos Int. 2018 Sep;29(9):2079-2086.

[154] Suárez-Ortegón MF, McLachlan S, Price AH, Fernández-Balsells M, Franch- Nadal J, Mata-Cases M, Barrot-de la Puente J, Mundet-Tudurí X, Mauricio D, Ricart W, Wild SH, Strachan MWJ, Price JF, Fernández-Real JM. Decreased iron stores are associated with cardiovascular disease in patients with type 2 diabetes both cross-sectionally and longitudinally. Atherosclerosis. 2018 May;272:193-199.

[155] Tebé C, Martinez-Laguna D, Moreno V, Cooper C, Diez-Perez A, Collins GS, Prieto-Alhambra D. Differential Mortality and the Excess Rates of Hip Fracture Associated With Type 2 Diabetes: Accounting for Competing Risks in Fracture Prediction Matters. J Bone Miner Res. 2018 Aug;33(8):1417-1421. doi: 10.1002/jbmr.3435. Epub 2018 May 3.

[156] Pontes C, Marsal JR, Elorza JM, Aragón M, Prieto-Alhambra D, Morros R. Analgesic Use and Risk for Acute Coronary Events in Patients With Osteoarthritis: A Population-based, Nested Case-control Study. Clin Ther. 2018 Feb;40(2):270- 283.

[157] Ramos R, Comas-Cufí M, Martí-Lluch R, Balló E, Ponjoan A, Alves-Cabratosa L, Blanch J, Marrugat J, Elosua R, Grau M, Elosua-Bayes M, García-Ortiz L, Garcia-Gil M. Statins for primary prevention of cardiovascular events and mortality in old and very old adults with and without type 2 diabetes: retrospective cohort study. BMJ. 2018 Sep 5;362:k3359. doi: 10.1136/bmj.k3359.

Bibliography 165

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[158] Mata-Cases M, Franch-Nadal J, Real J, Mauricio D. Glycaemic control and antidiabetic treatment trends in primary care centres in patients with type 2 diabetes mellitus during 2007-2013 in Catalonia: a population-based study. BMJ Open. 2016 Oct 5;6(10):e012463.

[159] Bolíbar B, Pareja C, Astier-Peña MP, Morán J., Rodríguez-Blanco T, Rosell- Murphy M et al. Variability in the performance of preventive services and in the degree of control of identified health problems: A primary care study protocol.. BMC Public Health. 2008; 8: 281.

[160] http://www.sidiap.org/images/stories/docs/SIDIAP_angles_v2.pdf

[161] Mata-Cases M, Mauricio D, Franch-Nadal J. Clinical characteristics of type 2 diabetic patients on basal insulin therapy with adequate fasting glucose control who do not achieve HbA1c targets. J Diabetes. 2017 Jan;9(1):34-44

[162] Roumie CL, Min JY, D'Agostino McGowan L, Presley C, Grijalva CG, Hackstadt AJ, Hung AM, Greevy RA, Elasy T, Griffin MR. Comparative Safety of Sulfonylurea and Metformin Monotherapy on the Risk of Heart Failure: A Cohort Study. J Am Heart Assoc. 2017 Apr 19;6(4). pii: e005379. doi: 10.1161/JAHA.116.005379.

[163] U.S. Department of Health and Human Services Food and Drug Administration Center for Drug Evaluation and Research. Guidance for Industry. Diabetes Mellitus: Developing Drugs and Therapeutic Biologics for Treatment and Prevention [Internet]. Center for Drug Evaluation and Research. February 2008. http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/ Guidances/UCM071624.pdf. Accessed on June 22 2018.

[164] Chi GY . Some issues with composite endpoints in clinical trials. Fundam Clin Pharmacol. 2005 Dec;19 (6):609-19

[165] Marx N, McGuire DK, Perkovic V, Woerle HJ, Broedl UC, von Eynatten M, George JT, Rosenstock J.Composite Primary End Points in Cardiovascular Outcomes Trials Involving Type 2 Diabetes Patients: Should Unstable Angina Be

Bibliography 166

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Included in the Primary End Point? Diabetes Care. 2017 Sep;40(9):1144-1151. doi: 10.2337/dc17-0068.

[166] Ferreira-González I, Busse JW, Heels-Ansdell D, Montori VM, Akl EA, Bryant DM, Alonso-Coello P, Alonso J, Worster A, Upadhye S, Jaeschke R, Schünemann HJ, Permanyer-Miralda G, Pacheco-Huergo V, Domingo-Salvany A, Wu P, Mills EJ, Guyatt GH. Problems with use of composite end points in cardiovascular trials: systematic review of randomised controlled trials. BMJ. 2007 Apr 14;334(7597):786. Epub 2007 Apr 2.

[167] Filion KB, Douros A, Azoulay L, Yin H, Yu OH, Suissa S. Sulfonylureas as initial treatment for type 2 diabetes and the risk of adverse cardiovascular events: A population-based cohort study. Br J Clin Pharmacol. 2019 Oct;85(10):2378-2389. D

[168] Ray WA. Observational studies of drugs and mortality. N Engl J Med. 2005 Dec 1;353(22):2319-21.

[169] Rosano GM, Vitale C, Seferovic P. Heart Failure in Patients with Diabetes Mellitus. Card Fail Rev. 2017 Apr;3(1):52-55. doi: 10.15420/cfr.2016:20:2.

[170] Norman PE, Davis WA, Bruce DG, Davis TM. Peripheral arterial disease and risk of cardiac death in type 2 diabetes: the Fremantle Diabetes Study. Diabetes Care. 2006 Mar;29(3):575-80.

[171] Alves-Cabratosa L, Garcia-Gil M, Comas-Cufí M, et al. Role of Low Ankle- Brachial Index in Cardiovascular and Mortality Risk Compared with Major Risk Conditions. J Clin Med. 2019;8(6):870. Published 2019 Jun 18. doi:10.3390/jcm8060870

[172] Domínguez-Berjóna MF, Borrell C, Cano-Serral G, Esnaola S, Nolasco A, Pasarín MI, Ramis R, Saurina C, Escolar-Pujolar A (Constructing a deprivation index based on census data in large Spanish cities [the MEDEA project]) Construcción de un índice de privación a partir de datos censales en grandes ciudades españolas (Proyecto MEDEA) Gac Sanit. 2008;22(3):179-87

Bibliography 167

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[173] Álvarez A, Vicente Sánchez M, Alejos B , Pascual C, Regidor E. Process and Results of Constructing a Deprivation Index for the Districts of Madrid and Barcelona, Spain Rev Esp Salud Pública 2013; 87:317-329.

[174] Petrie JR, Guzik TJ, Touyz RM. Diabetes, Hypertension, and Cardiovascular Disease: Clinical Insights and Vascular Mechanisms Can J Cardiol. 2018 May; 34(5): 575–584. doi: 10.1016/j.cjca.2017.12.005

[175] Baena Díez JM, del Val Garcia JL, Héctor Salas Gaetgens L, Sánchez Pérez R, Altes Vaques E, Deixens Martínez B, Amatller Corominas M, Katia Núñez Casillas D. [Comparison of the SCORE and REGICOR models for calculating cardiovascular risk in cardiovascular disease-free individuals at a healthcare center in Barcelona, Spain]. Rev Esp Salud Publica. 2005 Jul-Aug;79(4):453-64.

[176] Buitrago F, Cañón-Barroso L, Díaz-Herrera N, Cruces-Muro E, Escobar- Fernández M, Serrano-Arias JM. [Comparison of the REGICOR and SCORE function charts for classifying cardiovascular risk and for selecting patients for hypolipidemic or antihypertensive treatment]. Rev Esp Cardiol. 2007 Feb;60(2):139-47. [Article in Spanish

[177] Turner RC, Millns H, Neil HA, Stratton IM, Manley SE, Matthews DR, Holman RR. Risk factors for coronary artery disease in non-insulin dependent diabetes mellitus: United Kingdom Prospective Diabetes Study (UKPDS: 23) BMJ. 1998 Mar 14;316(7134):823-8.

[178] Zabala A, Darsalia V, Holzmann MJ, Franzén S, Svensson AM, Eliasson B, Patrone C, Nyström T, Jonsson M. Risk of first stroke in patients with type 2 diabetes and the relation to glycaemic control: a nationwide observational study. Diabetes Obes Metab. 2019 Oct 1

[179] https://www.who.int/cardiovascular_diseases/priorities/secondary_prevention/co untry/en/index1.html Accessed on July 23rd 2019

[180] Rapsomaniki E, Thuresson M, Yang E, Blin P, Hunt P, Chung SC, Stogiannis D, Pujades-Rodriguez M, Timmis A, Denaxas SC, Danchin N, Stokes M, Thomas- Delecourt F, Emmas C, Hasvold P, Jennings E, Johansson S, Cohen DJ, Jernberg T,

Bibliography 168

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Moore N, Janzon M, Hemingway H. Using big data from health records from four countries to evaluate chronic disease outcomes: a study in 114 364 survivors of myocardial infarction. Eur Heart J Qual Care Clin Outcomes. 2016 Jul 1; 2(3):172- 183.

[181] Salinero-Fort MA, Gómez-Campelo P, San Andrés-Rebollo FJ, Cárdenas- Valladolid J, Abánades-Herranz JC, Carrillo de Santa Pau E, Chico-Moraleja RM, Beamud-Victoria D, de Miguel-Yanes JM, Jimenez-Garcia R, López-de-Andres A, Ramallo-Fariña Y, De Burgos-Lunar C; MADIABETES Research Group. Prevalence of depression in patients with type 2 diabetes mellitus in Spain (the DIADEMA Study): results from the MADIABETES cohort. BMJ Open. 2018 Sep 24;8(9):e020768. doi: 10.1136/bmjopen-2017-020768.

[182] Patorno E, Patrick AR, Garry EM, Schneeweiss S, Gillet VG, Bartels DB, Masso-Gonzalez E, Seeger JD. Observational studies of the association between glucose-lowering medications and cardiovascular outcomes: addressing methodological limitations. Diabetologia. 2014 Nov;57(11):2237-50. doi: 10.1007/s00125-014-3364-z. Epub 2014 Sep 12.

[183] von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370(9596):1453–1457. doi:10.1016/S0140-6736(07)61602-X

[184] Langan SM, Schmidt SA, Wing K, Ehrenstein V, Nicholls SG, Filion KB, Klungel O, Petersen I, Sorensen HT, Dixon WG1, Guttmann A, Harron K, Hemkens LG, Moher D, Schneeweiss S, Smeeth L, Sturkenboom M, von Elm E, Wang SV, Benchimol EI. The reporting of studies conducted using observational routinely collected health data statement for pharmacoepidemiology (RECORD-PE). BMJ. 2018 Nov 14;363:k3532

[185] International Diabetes Federation. Clinical Practice Recommendations for managing Type 2 Diabetes in Primary Care. 2017

Bibliography 169

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[186] Inzucchi SE, Bergenstal RM, Buse JB, Diamant M, Ferrannini E, Nauck M, Peters AL, Tsapas A, Wender R, Matthews DR; American Diabetes Association (ADA); European Association for the Study of Diabetes (EASD). Management of hyperglycemia in type 2 diabetes: a patient-centered approach: position statement of the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD).Diabetes Care. 2012 Jun;35(6):1364-79. doi: 10.2337/dc12-0413.

[187] Bailey CJ, Marx N. Cardiovascular protection in type 2 diabetes: Insights from recent outcome trials. Diabetes Obes Metab. 2019 Jan;21(1):3-14. doi: 10.1111/dom.13492.

[188] Inzucchi SE, Masoudi FA, McGuire DK. Metformin in heart failure. Diabetes Care. 2007 Dec;30(12):e129.

[189] Lipska KJ. Metformin Use in Patients With Historical Contraindications. Ann Intern Med. 2017 Feb 7;166(3):225-226. doi: 10.7326/M16-2712.

[190] https://www.fda.gov/drugs/drug-safety-and-availability/fda-drug-safety- communication-fda-revises-warnings-regarding-use-diabetes-medicine-metformin- certain and https://www.fda.gov/media/96771/download. Accessed on July 17th 2017.

[191] Bailey CJ. Metformin: historical overview. Diabetologia. 2017 Sep;60(9):1566- 1576. doi: 10.1007/s00125-017-4318-z.

[192] Scheen AJ, Paquot N. Metformin revisited: a critical review of the benefit-risk balance in at-risk patients with type 2 diabetes. Diabetes Metab. 2013 May;39(3):179-90. doi: 10.1016/j.diabet.2013.02.006.

[193] Manel Mata, Francesc Xavier Cos, Rosa Morros, Laura Diego, Joan Barrot, Maria Berengué, Montserrat Brugada, Teresa Carrera, Juan Francisco Cano, Montserrat Estruch, Josep Maria Garrido, Genoveva Mendoza, Jordi Mesa, Marifé Muñoz, Assumpció Recasens, Joan Antoni Vallés. http://ics.gencat.cat/web/.content/documents/assistencia/gpc/GuiaDiabetis2015.pdf

Bibliography 170

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[194] Hemmingsen B, Schroll JB, Lund SS, Wetterslev J, Gluud C, Vaag A, Sonne DP, Lundstrøm LH, Almdal T. Sulphonylurea monotherapy for patients with type 2 diabetes mellitus. Cochrane Database Syst Rev. 2013 Apr 30;(4):CD009008. doi: 10.1002/14651858.CD009008.pub2

[195] Han Y, Xie H, Liu Y, Gao P, Yang X, Shen Z. Effect of metformin on all-cause and cardiovascular mortality in patients with coronary artery diseases: a systematic review and an updated meta-analysis. Cardiovasc Diabetol. 2019 Jul 30;18(1):96. doi: 10.1186/s12933-019-0900-7.

[196] Powell WR, Christiansen CL, Miller DR. Meta-Analysis of Sulfonylurea Therapy on Long-Term Risk of Mortality and Cardiovascular Events Compared to Other Oral Glucose-Lowering Treatments. Diabetes Ther. 2018 Aug;9(4):1431- 1440. doi: 10.1007/s13300-018-0443-z.

[197] Azoulay L, Suissa S. Sulfonylureas and the Risks of Cardiovascular Events and Death: A Methodological Meta-Regression Analysis of the Observational Studies. Diabetes Care. 2017 May;40(5):706-714. doi: 10.2337/dc16-1943.

[198] Phung OJ, Schwartzman E, Allen RW, Engel SS, Rajpathak SN. Sulphonylureas and risk of cardiovascular disease: systematic review and meta-analysis. Diabet Med. 2013 Oct;30(10):1160-71. doi: 10.1111/dme.12232.

[199] Roumie CL, Hung AM, Greevy RA, et al. Comparative effectiveness of sulfonylurea and metformin monotherapy on cardiovascular events in type 2 diabetes mellitus: a cohort study. Ann Intern Med 2012;157:601–610

[200] Schramm TK, Gislason GH, Vaag A, Rasmussen JN, Folke F, Hansen ML, Fosbøl EL, Køber L, Norgaard ML, Madsen M, Hansen PR, Torp-Pedersen C. 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. 2011 Aug;32(15):1900-8. doi: 10.1093/eurheartj/ehr077.

[201] Lamanna C, Monami M, Marchionni N, Mannucci E. Effect of metformin on cardiovascular events and mortality: a meta-analysis of randomized clinical trials.

Bibliography 171

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Diabetes Obes Metab. 2011 Mar;13(3):221-8. doi: 10.1111/j.1463- 1326.2010.01349.x

[202] Fisman EZ, Tenenbaum A, Benderly M, Goldbourt U, Behar S, Motro M. Antihyperglycemic treatment in diabetics with coronary disease: increased metformin-associated mortality over a 5-year follow-up. Cardiology. 1999;91(3):195-202.

[203] Fisman EZ, Tenenbaum A, Boyko V, Benderly M, Adler Y, Friedensohn A, Kohanovski M, Rotzak R, Schneider H, Behar S, Motro M. Oral antidiabetic treatment in patients with coronary disease: time-related increased mortality on combined glyburide/metformin therapy over a 7.7-year follow-up.Clin Cardiol. 2001 Feb;24(2):151-8.

[204] Varvaki Rados D, Catani Pinto L Reck Remonti L, Bauermann Leitão C, Gross JL. The Association between Sulfonylurea Use and All-Cause and Cardiovascular Mortality: A Meta-Analysis with Trial Sequential Analysis of Randomized Clinical Trials. PLoS Med. 2016 Apr 12;13(4):e1001992

[205] Bain S, Druyts E, Balijepalli C, Baxter CA, Currie CJ, Das R, Donnelly R, Khunti K, Langerman H, Leigh P, Siliman G, Thorlund K, Toor K, Vora J, Mills EJ. Cardiovascular events and all-cause mortality associated with sulphonylureas compared with other antihyperglycaemic drugs: A Bayesian meta-analysis of survival data. Diabetes Obes Metab. 2017 Mar;19(3):329-335. doi: 10.1111/dom.12821.

[206] Gulliford M, Latinovic R. Mortality in type 2 diabetic subjects prescribed metformin and sulphonylurea drugs in combination: cohort study. Diabetes Metab Res Rev. 2004 May-Jun;20(3):239-45.

[207] Rosenstock J, Kahn SE, Johansen OE, Zinman B, Espeland MA, Woerle HJ, Pfarr E, Keller A, Mattheus M, Baanstra D, Meinicke T, George JT, von Eynatten M, McGuire DK, Marx N; CAROLINA Investigators. Effect of Linagliptin vs Glimepiride on Major Adverse Cardiovascular Outcomes in Patients With Type 2

Bibliography 172

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

Diabetes: The CAROLINA Randomized Clinical Trial. JAMA. 2019 Sep 19. doi: 10.1001/jama.2019.13772

[208] Patorno E, Schneeweiss S, Gopalakrishnan C, Martin D, Franklin JM. Using Real-World Data to Predict Findings of an Ongoing Phase IV Cardiovascular Outcome Trial: Cardiovascular Safety of Linagliptin Versus Glimepiride. Diabetes Care. 2019 Jun 25. pii: dc190069. doi: 10.2337/dc19-0069.

[209] Scheller NM, Mogensen UM, Andersson C, Vaag A, Torp-Pedersen C. All- cause mortality and cardiovascular effects associated with the DPP-IV inhibitor sitagliptin compared with metformin, a retrospective cohort study on the Danish population. Diabetes Obes Metab. 2014 Mar;16(3):231-6. doi: 10.1111/dom.12197.

[210] Ou HT1, Chang KC2, Li CY3,4, Wu JS5,6. Risks of cardiovascular diseases associated with dipeptidyl peptidase-4 inhibitors and other antidiabetic drugs in patients with type 2 diabetes: a nation-wide longitudinal study. Cardiovasc Diabetol. 2016 Mar 1;15:41. doi: 10.1186/s12933-016-0350-4.

[211] Chen DY, Wang SH, Mao CT, Tsai ML, Lin YS, Chou CC, Wen MS, Wang CC, Hsieh IC, Hung KC, Chen TH. Sitagliptin and cardiovascular outcomes in diabetic patients with chronic kidney disease and acute myocardial infarction: A nationwide cohort study. Int J Cardiol. 2015 Feb 15;181:200-6. doi: 10.1016/j.ijcard.2014.12.029.

[212] Shin S, Kim H. The effect of sitagliptin on cardiovascular risk profile in Korean patients with type 2 diabetes mellitus: a retrospective cohort study. Ther Clin Risk Manag. 2016 Mar 15;12:435-44. doi: 10.2147/TCRM.S105285.

[213] Huang HK, Yeh JI. Comparison of mortality and cardiovascular event risk associated with various insulin secretagogues: A nationwide real-world analysis. Diabetes Res Clin Pract. 2019 Jun;152:103-110. doi: 10.1016/j.diabres.2019.04.032.

[214] Yang TY, Liaw YP, Huang JY, Chang HR, Chang KW, Ueng KC. Association of Sitagliptin with cardiovascular outcome in diabetic patients: a nationwide cohort study. Acta Diabetol. 2016 Jun;53(3):461-8. doi: 10.1007/s00592-015-0817-x.

Bibliography 173

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[215] Shah AD, Langenberg C, Rapsomaniki E, Denaxas S, Pujades-Rodriguez M, Gale CP, Deanfield J, Smeeth L, Timmis A, Hemingway H. Type 2 diabetes and incidence of cardiovascular diseases: a cohort study in 1.9 million people. Lancet Diabetes Endocrinol. 2015 Feb;3(2):105-13.

[216] Szczerbińska K, Topinková E, Brzyski P, van der Roest HG, Richter T, Finne- Soveri H, Denkinger MD, Gindin J, Onder G, Bernabei R. The characteristics of diabetic residents in European nursing homes: results from the SHELTER study. J Am Med Dir Assoc. 2015 Apr;16(4):334-40. doi: 10.1016/j.jamda.2014.11.009. Epub 2014 Dec 19.

[217] Roumie CL, Chipman J, Min JY, Hackstadt AJ, Hung AM, Greevy RA Jr, Grijalva CG, Elasy T, Griffin MR. Association of Treatment With Metformin vs Sulfonylurea With Major Adverse Cardiovascular Events Among Patients With Diabetes and Reduced Kidney Function. JAMA. 2019 Sep 19:1-11. doi: 10.1001/jama.2019.13206.

[218] Guidelines on second- and third-line medicines and type of insulin for the control of blood glucose levels in non-pregnant adults with diabetes mellitus. Geneva: World Health Organization; 2018. Licence: CC BY-NCSA 3.0 IGO. https://apps.who.int/iris/bitstream/handle/10665/272433/9789241550284-eng.pdf

[219] Turner RC, Cull CA, Frighi V, Holman RR. Glycemic control with diet, sulfonylurea, metformin, or insulin in patients with type 2 diabetes mellitus: progressive requirement for multiple therapies (UKPDS 49). UK Prospective Diabetes Study (UKPDS) Group. JAMA. 1999 Jun 2;281(21):2005-12.

[220] Lipska KJ, Yao X, Herrin J, McCoy RG, Ross JS, Steinman MA, Inzucchi SE, Gill TM, Krumholz HM, Shah ND. Trends in drug utilization, glycemic control, and rates of severe hypoglycemia, 2006–2013. Diabetes Care. 2017;40:468–475.

[221] Tahrani AA, Barnett AH, Bailey CJ. Pharmacology and therapeutic implications of current drugs for type 2 diabetes mellitus. Nat Rev Endocrinol. 2016 Oct;12(10):566-92. doi: 10.1038/nrendo.2016.86.

Bibliography 174

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[222] Monami M, Genovese S, Mannucci E. Cardiovascular safety of sulfonylureas: a meta-analysis of randomized clinical trials. Diabetes Obes Metab. 2013;15:938-953.

[223] Landgraf R. Meglitinide analogues in the treatment of type 2 diabetes mellitus. Drugs Aging. 2000 Nov;17(5):411-25.

[224] Stephens JW, Bodvarsdottir TB, Wareham K, Prior SL, Bracken RM, Lowe GD, Rumley A, Dunseath G, Luzio S, Deacon CF, Holst JJ, Bain SC. Effects of short- term therapy with glibenclamide and repaglinide on incretin hormones and oxidative damage associated with postprandial hyperglycaemia in people with type 2 diabetes mellitus. Diabetes Res Clin Pract. 2011 Nov;94(2):199-206. doi: 10.1016/j.diabres.2011.07.014.

[225] Black C, Donnelly P, McIntyre L, Royle PL, Shepherd JP, Thomas S. Meglitinide analogues for type 2 diabetes mellitus Cochrane Database Syst Rev. 2007 Apr 18;(2):CD004654

[226] Xie J, Li N, Jiang X, Chai L, Chen JJ, Deng W. Short-term efficacy and safety of repaglinide versus glimepiride as augmentation of metformin in treating patients with type 2 diabetes mellitus. Diabetes Metab Syndr Obes. 2019 Apr 17;12:519- 526. doi: 10.2147/DMSO.S198154.

[227] Scheen AJ. Cardiovascular Effects of New Oral Glucose-Lowering Agents: DPP-4 and SGLT-2 Inhibitors.Circ Res. 2018 May 11;122(10):1439-1459. doi: 10.1161/CIRCRESAHA.117.311588

[228] Scheen AJ. Cardiovascular safety of DPP-4 inhibitors compared with sulphonylureas: Results of randomized controlled trials and observational studies. Diabetes Metab. 2018 Nov;44(5):386-392. doi: 10.1016/j.diabet.2018.05.007.

[229] Madsen KS, Kähler P, Kähler LKA, Madsbad S, Gnesin F, Metzendorf MI, Richter B, Hemmingsen B. Metformin and second- or third-generation sulphonylurea combination therapy for adults with type 2 diabetes mellitus. Cochrane Database Syst Rev. 2019 Apr 18;4:CD012368. doi: 10.1002/14651858.CD012368.pub2.

Bibliography 175

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[230] Yu OH, Yin H, Azoulay L. The combination of DPP-4 inhibitors versus sulfonylureas with metformin after failure of first-line treatment in the risk for major cardiovascular events and death. Can J Diabetes. 2015;39(5):383-389. doi:10.1016/j.jcjd.2015.02.002

[231] Mogensen UM, Andersson C, Fosbøl EL, Schramm TK, Vaag A, Scheller NM, Torp-Pedersen C, Gislason G, Køber L. Cardiovascular safety of combination therapies with incretin-based drugs and metformin compared with a combination of metformin and sulphonylurea in type 2 diabetes mellitus-a retrospective nationwide study. Diabetes Obes Metab. 2014 Oct;16(10):1001-8. doi: 10.1111/dom.12314..

[232] Seong JM, Choi NK, Shin JY, Chang Y, Kim YJ, Lee J, Kim JY, Park BJ. Differential cardiovascular outcomes after dipeptidyl peptidase-4 inhibitor, sulfonylurea, and pioglitazone therapy, all in combination with metformin, for type 2 diabetes: a population-based cohort study. PLoS One. 2015 May 20;10(5):e0124287. doi: 10.1371/journal.pone.0124287.

[233] Kannan S, Pantalone KM, Matsuda S, Wells BJ, Karafa M, Zimmerman RS. Risk of overall mortality and cardiovascular events in patients with type 2 diabetes on dual drug therapy including metformin: A large database study from the Cleveland Clinic. J Diabetes. 2016 Mar;8(2):279-85. doi: 10.1111/1753- 0407.12301.

[234] Ekström N, Svensson AM, Miftaraj M, Franzén S, Zethelius B, Eliasson B, Gudbjörnsdottir S. Cardiovascular safety of glucose-lowering agents as add-on medication to metformin treatment in type 2 diabetes: report from the Swedish National Diabetes Register. Diabetes Obes Metab. 2016 Oct;18(10):990-8. doi: 10.1111/dom.12704.

[235] O'Brien MJ, Karam SL, Wallia A, Kang RH, Cooper AJ, Lancki N, Moran MR, Liss DT, Prospect TA, Ackermann T. Association of Second-line Antidiabetic Medications With Cardiovascular Events Among Insured Adults With Type 2 Diabetes. JAMA Netw Open. 2018 Dec 7;1(8):e186125. doi: 10.1001/jamanetworkopen.2018.6125.

Bibliography 176

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[236] Ou SM, Shih CJ, Chao PW, et al. Effects on Clinical Outcomes of Adding Dipeptidyl Peptidase-4 Inhibitors Versus Sulfonylureas to Metformin Therapy in Patients With Type 2 Diabetes Mellitus. Ann Intern Med. 2015;163(9):663-672. doi:10.7326/M15-0308 [Abstract in PubMed]

[237] Mogensen UM, Andersson C, Fosbøl EL, Schramm TK, Vaag A, Scheller NM, Torp-Pedersen C, Gislason G, Køber L. Metformin in combination with various insulin secretagogues in type 2 diabetes and associated risk of cardiovascular morbidity and mortality--a retrospective nationwide study. Diabetes Res Clin Pract. 2015 Jan;107(1):104-12. doi: 10.1016/j.diabres.2014.09.047.

[238] McGuire DK, Alexander JH, Johansen OE, Perkovic V, Rosenstock J, Cooper ME, Wanner C, Kahn SE, Toto RD, Zinman B, Baanstra D, Pfarr E, Schnaidt S, Meinicke T, George JT, von Eynatten M, Marx N; CARMELINA Investigators. Linagliptin Effects on Heart Failure and Related Outcomes in Individuals With Type 2 Diabetes Mellitus at High Cardiovascular and Renal Risk in CARMELINA Circulation. 2019 Jan 15;139(3):351-361. doi: 10.1161/CIRCULATIONAHA.118.038352.

[239] Kenny HC, Abel ED. Heart Failure in Type 2 Diabetes Mellitus. Circ Res. 2019 Jan 4;124(1):121-141. doi: 10.1161/CIRCRESAHA.118.311371

[240] Presley CA, Chipman J, Min JY, Grijalva CG, Greevy RA, Griffin MR, Roumie CL. Evaluation of frailty as an unmeasured confounder in observational studies of antidiabetic medications. J Gerontol A Biol Sci Med Sci. 2018 Sep 26. doi: 10.1093/gerona/gly224.

[241] Menon V, Lincoff AM, Nicholls SJ, Jasper S, Wolski K, McGuire DK, Mehta CR, Rosenstock J, Lopez C, Marcinak J, Cao C, Nissen SE; GRAND 306 Investigators. Fasiglifam-Induced Liver Injury in Patients With Type 2 Diabetes: Results of a Randomized Controlled Cardiovascular Outcomes Safety Trial. Diabetes Care. 2018 Dec;41(12):2603-2609. doi: 10.2337/dc18-0755.

Bibliography 177

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

[242] Shavadia JS, Sharma A, Gu X, Neaton J, DeLeve L, Holmes D, Home P, Eckel RH, Watkins PB, Granger CB. Determination of fasiglifam-induced liver toxicity: Insights from the data monitoring committee of the fasiglifam clinical trials program. Clin Trials. 2019 Jun;16(3):253-262. doi: 10.1177/1740774519836766.

[243] Diabetes Care. 2018 Dec;41(12):2603-2609. doi: 10.2337/dc18-0755.Li X, Zhong K, Guo Z, Zhong D, Chen X. Fasiglifam (TAK-875) Inhibits Hepatobiliary Transporters: A Possible Factor Contributing to Fasiglifam-Induced Liver Injury. Drug Metab Dispos. 2015 Nov;43(11):1751-9. doi: 10.1124/dmd.115.064121.

[244] FDA Drug Safety Communication: FDA adds warnings about heart failure risk to labels of type 2 diabetes medicines containing saxagliptin and alogliptin https://www.fda.gov/drugs/drug-safety-and-availability/fda-drug-safety- communication-fda-adds-warnings-about-heart-failure-risk-labels-type-2-diabetes Accessed on May 13 2018

[245] Douros A, Dell’Aniello S, Yu OHY, Filion KB, Azoulay L, Suissa S. Sulfonylureas as second-line drugs in type 2 diabetes and the risk of cardiovascular and hypoglycaemic events: population-based cohort study. BMJ 2018;362:k2693.

[246] McGowan LD, Roumie CL. Sulfonylureas as second-line treatment for type 2 diabetes. BMJ. 2018 Jul 18;362:k3041. doi: 10.1136/bmj.k3041

[247] H.R.34 - 21st Century Cures Act https://www.congress.gov/114/plaws/publ255/PLAW-114publ255.pdf Accessed on May 21st 2018

Bibliography 178

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

VIII. Appendix A Appendix A: Results of retrospective observational studies conducted in LHDs assessing cardiovascular outcomes and mortality in DPP- 4 i users

Study drugs MACE All-cause mortality MI Stroke HF PAD [1] Scheller Sitagliptin vs 1.22 (0.92-1.61) 1.25 (0.92-1.71) metformin [2] Wang KL Sitagliptin vs no 0.87 (0.74–1.03) 1.21 (1.03–1.42) use [3] Chen DY Sitagliptin vs no 1.32 (0.97–1.79) 0.90 (0.67–1.23) 1.62 (1.05–2.52) use 1.73 (1.15–2.58) [4] Wang SH Sitagliptin vs no 0.97 (0.73–1.29) 0.65 (0.39–1.10)a 1.07 (0.72–1.59) 1.30 (0.75–2.26) Post-MI use [5] Chen DY Sitagliptin vs no 1.00 (0.82–1.22) 0.95 (0.78–1.16) Post-stroke use 1.07 (0.55– 2.11) b [6] Shih CJ DPP-4 i vs no 0.79 (0.75-0,83) 0.54 (0.52-0.56) 0.79 (0.72-0.87) 0.79 (0.75-0.84) users [7] Ou HT DPP-4 i vs no 0.83 (0.76-0.91) c users [8] Shin S Sitagliptin vs 0.831 (0.536– 1.863 (0.376–9.230) a 0.657 (0.237– 0.783 (0.470– 0.762 (0.389– metformin 1.289) 1.826) 1.306) 1.495) HUA 1.165 (0.791–1.715) [9] Fu AZ DPP-4 i vs SUs Prior CVD: 0.95 (0.78–1.15) No prior CVD: 0.59 (0.38–0.89)

Appendix A 181

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 [10] Toh S Saxagliptin vs 0.69 (0.54- 0.87) SUs DRS stratified: 0.86 (0.77-0.95) [11] Yang TY Sitagliptin vs no 0.56 (0.41-0.74) CVD: 0.75 (0.59-0.96) 0.86 (0.45-1.65) use 0.59 (0.48-0.72) d [12] Chan SY DPP-4 i vs non- 0.76 (0.65–0.90) 0.43 (0.39–0.47) No difference 0.77 (0.61–0.97) No difference users [13] Nyström T MET + insulin vs 0.59 (0.51-0.69) CVD (fatal and non-fatal): MET + DPP-4 i 0.71 (0.62-0.82) [14] Ou SM DPP-4 i vs DPP- 0.67 (0.64- 0.70) 0.81 (0.76 - 0.87) 0.96 (0.88-1.04) 4 i non-users. Patients prior HF 0.80 (0.71 - 0.89) 0.83 (0.76- 0.89) [15] Chin HJ DPP-4 i vs 0.87 (0.75-1.01) glimepiride [16] Lo Re V Saxagliptin vs 0.92 (0.86 - 0.98) OADs [17] Kim YG DPP-4 i vs SUs 0.76 (0.67–0.87) 0.63 (0.60–0.67) Full cohort: 0.78 (0.67–0.86) Baseline CVD: 0.77 (0.68–0.79) No baseline CVD: 0.71 (0.56–0.90) [18] Williams R Vildagliptin vs IRR: 0.22 to 1.02 IRR: 0.61 to 0.97 IRR: 0.02 to 0.77 CHF: 0.49 to NIADs ACS: 0.55 to 1.03 1.60 [19] Yu OH DPP-4 i alone or 0.88 (0.63–1.22) Case-control in combination

Appendix A 182

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 with OADs vs SUs [20] Filion K DPP-4 i vs 0.87 (0.63-1.21) Nested case- OADs control

a Cardiovascular mortality; b Haemorragic stroke; c Cardiovascular disease

LHDs: longitudinal healthcare databases; IRR: incident rate ratio DPP-4 i: dipeptidyl-peptidase inhibitors MET: metformin; SUs: sulfonylureas; ACS: acute coronary syndrome CHD: coronary heart disease; CHF: congestive heart failure; CVD: cardiovascular disease; IHD: ischemic heart disease; IS: ischemic stroke; MI: myocardial infarction

References

[1] Scheller NM, Mogensen UM, Andersson C, Vaag A, Torp-Pedersen C. All-cause mortality and cardiovascular effects associated with the DPP-IV inhibitor sitagliptin compared with metformin, a retrospective cohort study on the Danish population. Diabetes Obes Metab. 2014 Mar;16(3):231-6. doi: 10.1111/dom.12197. Epub 2013 Sep 10.

[2] Wang KL, Liu CJ, Chao TF, Huang CM, Wu CH, Chen SJ, Yeh CM, Chen TJ, Lin SJ, Chiang CE.Sitagliptin and the risk of hospitalization for heart failure: a population-based study. Int J Cardiol. 2014 Nov 15;177(1):86-90. doi: 10.1016/j.ijcard.2014.09.038.

[3] Chen DY, Wang SH, Mao CT, Tsai ML, Lin YS, Chou CC, Wen MS, Wang CC, Hsieh IC, Hung KC, Chen TH. Sitagliptin and cardiovascular outcomes in diabetic patients with chronic kidney disease and acute myocardial infarction: A nationwide cohort study. Int J Cardiol. 2015 Feb 15;181:200-6. doi: 10.1016/j.ijcard.2014.12.029.

Appendix A 183

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 [4] Wang SH, Chen DY, Lin YS, Mao CT, Tsai ML, Hsieh MJ, Chou CC, Wen MS, Wang CC, Hsieh IC, Hung KC, Chen TH.Cardiovascular Outcomes of Sitagliptin in Type 2 Diabetic Patients with Acute Myocardial Infarction, a Population-Based Cohort Study in Taiwan. PLoS One. 2015 Jun 26;10(6):e0131122. doi: 10.1371/journal.pone.0131122.

[5] Chen DY, Wang SH, Mao CT, Tsai ML, Lin YS, Chou CC, Wen MS, Wang CC, Hsieh IC, Hung KC, Chen TH. Cardiovascular Outcomes of Sitagliptin in Type 2 Diabetic Patients with Acute Myocardial Infarction, a Population-Based Cohort Study in Taiwan. Int J Cardiol. 2015 Feb 15;181:200-6. doi: 10.1016/j.ijcard.2014.12.029.

[6] Shih CJ, Chen HT, Kuo SC, Ou SM, Chen YT. Cardiovascular Outcomes of Dipeptidyl Peptidase-4 Inhibitors in Elderly Patients With Type 2 Diabetes: A Nationwide Study. J Am Med Dir Assoc. 2016 Jan;17(1):59-64. doi: 10.1016/j.jamda.2015.10.009.

[7] Ou HT, Chang KC, Li CY, Wu JS. Risks of cardiovascular diseases associated with dipeptidyl peptidase-4 inhibitors and other antidiabetic drugs in patients with type 2 diabetes: a nation-wide longitudinal study. Cardiovasc Diabetol. 2016 Mar 1;15:41. doi: 10.1186/s12933-016- 0350-4.

[8] Shin S, Kim H. The effect of sitagliptin on cardiovascular risk profile in Korean patients with type 2 diabetes mellitus: a retrospective cohort study. Ther Clin Risk Manag. 2016 Mar 15;12:435-44. doi: 10.2147/TCRM.S105285.

[9] Fu AZ, Johnston SS, Ghannam A, Tsai K, Cappell K, Fowler R, Riehle E, Cole AL, Kalsekar I, Sheehan J. Association Between Hospitalization for Heart Failure and Dipeptidyl Peptidase 4 Inhibitors in Patients With Type 2 Diabetes: An Observational Study. Diabetes Care. 2016 May;39(5):726-34. doi: 10.2337/dc15-0764

Appendix A 184

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 [10] Toh S, Hampp C, Reichman ME, Graham DJ, Balakrishnan S, Pucino F, Hamilton J, Lendle S, Iyer A, Rucker M, Pimentel M, Nathwani N, Griffin MR, Brown NJ, Fireman BH. Risk for Hospitalized Heart Failure Among New Users of Saxagliptin, Sitagliptin, and Other Antihyperglycemic Drugs: A Retrospective Cohort Study. Ann Intern Med. 2016 Jun 7;164(11):705-14. doi: 10.7326/M15-2568

[11] Yang TY, Liaw YP, Huang JY, Chang HR, Chang KW, Ueng KC. Association of Sitagliptin with cardiovascular outcome in diabetic patients: a nationwide cohort study. Acta Diabetol. 2016 Jun;53(3):461-8. doi: 10.1007/s00592-015-0817-x.

[12] Chan SY, Ou SM, Chen YT, Shih CJ. Effects of DPP-4 inhibitors on cardiovascular outcomes in patients with type 2 diabetes and end- stage renal disease. Int J Cardiol. 2016 Sep 1;218:170-175. doi: 10.1016/j.ijcard.2016.05.062.

[13] Nyström T, Bodegard J, Nathanson D, Thuresson M, Norhammar A, Eriksson JW. Second line initiation of insulin compared with DPP-4 inhibitors after metformin monotherapy is associated with increased risk of all-cause mortality, cardiovascular events, and severe hypoglycemia. Diabetes Res Clin Pract. 2017 Jan;123:199-208. doi: 10.1016/j.diabres.2016.12.004.

[14] Ou SM, Chen HT, Kuo SC, Chen TJ, Shih CJ, Chen YT. Dipeptidyl peptidase-4 inhibitors and cardiovascular risks in patients with pre- existing heart failure. Heart. 2017 Mar;103(6):414-420. doi: 10.1136/heartjnl-2016-309687.

[15] Chin HJ, Nam JH, Lee EK, Shin JY. Comparative safety for cardiovascular outcomes of DPP-4 inhibitors versus glimepiride in patients with type 2 diabetes: A retrospective cohort study. Medicine (Baltimore). 2017 Jun;96(25):e7213. doi: 10.1097/MD.0000000000007213.

[16] Lo Re V, Carbonari DM, Saine ME, Newcomb CW, Roy JA, Liu , Wu Q, Cardillo S, Haynes K, Kimmel SE, Reese PP, Margolis DJ, Apter AJ, Reddy KR, Hennessy S, Bhullar H, Gallagher AM, Esposito DB, Strom BL. Postauthorization safety study of the DPP-4 inhibitor saxagliptin: a large-scale multinational family of cohort studies of five outcomes. BMJ Open Diabetes Res Care. 2017 Jul 31;5(1):e000400. doi: 10.1136/bmjdrc-2017-000400.

Appendix A 185

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019 [17] Kim YG, Yoon D, Park S, Han SJ, Kim DJ, Lee KW, Park RW, Kim HJ. Dipeptidyl Peptidase-4 Inhibitors and Risk of Heart Failure in Patients With Type 2 Diabetes Mellitus: A Population-Based Cohort Study. Circ Heart Fail. 2017 Sep;10(9). pii: e003957. doi: 10.1161/CIRCHEARTFAILURE.117.003957.

[18] Williams R, de Vries F, Kothny W, Serban C, Lopez-Leon S, Chu C. Schlienger R4. Cardiovascular safety of vildagliptin in patients with type 2 diabetes: A European multi-database, non-interventional post-authorization safety study. Diabetes Obes Metab. 2017 Oct;19(10):1473- 1478. doi: 10.1111/dom.12951

[19] Yu OH, Filion KB, Azoulay L, Patenaude V, Majdan A, Suissa S. Incretin-based drugs and the risk of congestive heart failure. Diabetes Care. 2015 Feb;38(2):277-84. doi: 10.2337/dc14-1459.

[20] Filion KB, Azoulay L, Platt RW, Dahl M, Dormuth CR, Clemens KK, Hu N, Paterson JM, Targownik L, Turin TC, Udell JA, Ernst P; CNODES Investigators. A Multicenter Observational Study of Incretin-based Drugs and Heart Failure. N Engl J Med. 2016 Mar 24;374(12):1145-54. doi: 10.1056/NEJMoa1506115.

Appendix A 186

Nelly Raquel Herrera Comoglio Eu2P PhD December 2019

IX. Annexes

XI.1. Cardiovascular outcomes, heart failure and mortality in type 2 diabetic patients treated with glucagon-like peptide 1 receptor agonists (GLP-1 RAs): A systematic review and meta- analysis of observational cohort studies

Herrera Comoglio R, Vidal Guitart X.. Int J Clin Pract. 2020;74(9):e13553. doi:10.1111/ijcp.13553

XI.2. Linagliptin and Cardiac failure

WHO Pharmaceuticals Newsletter, N°1, 2015, World Health Organization, p. 20-26

XI.3. Glibenclamide/glyburide and palpitations in Asian population

WHO Pharmaceutical Newsletter, N° 2, 2019, World Health Organization, p. 17-22

IX. 1 Cardiovascular outcomes, heart failure and mortality in type 2 diabetic patients treated with glucagon-like peptide 1 receptor agonists (GLP-1 RAs): A systematic review and meta-analysis of observational cohort studies

IX. 1 Cardiovascular outcomes, heart failure and mortality in type 2 diabetic patients treated with glucagon-like peptide 1 receptor agonists (GLP-1 RAs): A systematic review and meta-analysis of observational cohort studies

Received: 19 September 2019 | Revised: 6 May 2020 | Accepted: 18 May 2020 DOI: 10.1111/ijcp.13553

META-ANALYSIS METABOLISM & ENDOCRINOLOGY Cardiovascular outcomes, heart failure and mortality in type 2 diabetic patients treated with glucagon-like peptide 1 receptor agonists (GLP-1 RAs): A systematic review and meta-analysis of observational cohort studies

Raquel Herrera Comoglio1,2 | Xavier Vidal Guitart2,3

1School of Medicine, Universidad Nacional de Córdoba, Córdoba, Argentina Abstract 2Eu2P European Programme Background: Cardiovascular outcomes trials (CVOTs) have assessed the effects of in Pharmacovigilance and glucagon-like peptide-1 receptor agonists (GLP-1 RAs) on major adverse cardio- Pharmacoepidemiology, University of Bordeaux Segalen, Bordeaux, France vascular events (MACE) and mortality in high cardiovascular (CV) risk populations. 3Fundacio Institut Catala de Farmacologia, Observational research can provide complementary evidence about these effects in Universitat Autonoma de Barcelona, Barcelona, Spain unselected populations. Aim: To systematically review retrospective observational cohort studies conducted Correspondence Raquel Herrera Comoglio, Hospital Nacional in electronic healthcare databases (EHDs) assessing GLP-1 RAs´ effects on MACE de Clínicas, Facultad de Ciencias Médicas, and/or hospitalisation for heart failure (HHF) and/or all-cause mortality in Type 2 Universidad Nacional de Córdoba, UNC – Santa Rosa 1564, 5000 Córdoba, Argentina. diabetes mellitus (T2DM) patients. Email: [email protected]; Methods: We systematically searched studies meeting inclusion criteria, compared [email protected] design, methods and population characteristics, assessed risk for bias and did a meta- analysis (MA) using a random-effects model to calculate overall hazard ratios (HRs) and 95% CI (confidence intervals). Results: Sixteen studies included 285,436 T2DM patients exposed to GLP-1 RAs (exenatide bid, liraglutide, lixisenatide, long-acting exenatide), n ranged from 219 to 160,803 patients. Comparators included: no exposure, other antidiabetic medica- tions (OADs), combined OADs, canagliflozin or multiple comparators. Ten studies estimated all-cause mortality, hazard ratios (HRs) ranged from 0.17 (95% CI 0.02- 1.22) to 1.29 (95% CI 0.54-3.13). Thirteen studies assessed cardiovascular events and/or MACE; HRs ranged from 0.27 (95% CI 0.14-0.53) to 1.11 (95% CI 0.99-1.24). Eight studies assessed HHF, HRs ranged from 0.12 (95% CI 0.02-0.66) to 1.64 (95% CI 1.28-2.13). Excluding two studies because of temporal bias, we obtained pooled estimates for all-cause mortality: HR 0.63 (0.44-0.89), CV outcomes HR 0.84 (0.75- 0.94) and HHF; HR 0.94 (0.78-1.14), (high between-study variability: I2 = 83.35%; I2 = 70.3%; and I2 = 90.1%, respectively). Conclusion: Pooled results of EHDs’ studies assessing GLP-1 RAs effects favoured GLP-1 RAs for all-cause mortality and MACE while were neutral for HHF. Results should be interpreted cautiously because of studies’ substantial heterogeneity and limitations of observational research.

Int J Clin Pract. 2020;74:e13553. wileyonlinelibrary.com/journal/ijcp © 2020 John Wiley & Sons Ltd | 1 of 18 https://doi.org/10.1111/ijcp.13553 2 of 18 | HERRERA COMOGLIO and VIDAL GUITART

1 | INTRODUCTION How did you gather the information you In patients with type 2 diabetes mellitus (T2DM), morbidity and mor- considered in your review? tality rate from macrovascular disease is up to 80%1,2; patients are Medline, Scopus and Web of Science were used to identify also at increased risk of heart failure and have a worse prognosis.3,4 abstracts and full texts published as of January 31, 2019. Blood-glucose levels control has shown to be effective to reduce Searches included individual and class terms for GLP-1 microvascular complications of diabetes, mainly diabetic nephropa- RAs, study design and sources of data. Inclusion and ex- thy, but the effect on cardiovascular disease (CVD) is unclear.5 Since clusion criteria were clearly defined. The process is sum- 2005, 15 new non-insulin blood-glucose-lowering drugs (non-insulin marised in Figure 1; and the list of retrieved and excluded “antidiabetic drugs,” NIADs) have been marketed, adding three new papers are shown in Tables S1 and S2, respectively. classes of non-insulin products: glucagon-like peptide-1 receptor agonists (GLP-1 RAs), sodium-glucose cotransporter-2 inhibitors What is the “take-home” message for the clinician? (SGLT2 i) and inhibitors of dipeptidyl-peptidase 4 (DPP-4 i).6 Native GLP-1 and glucose-dependent insulinotropic polypeptide • It is often unclear to what extent results of clinical tri- (GIP) account for up to 60% of post-prandial insulin release. The acti- als conducted in selected high-risk cardiovascular (CV) vation of GLP-1 receptor increases insulin secretion, lowers inappro- populations can apply to “real world” ones. priate high glucagon secretion in a glucose-dependent manner and • The effect of glucagon-like peptide 1 receptor agonists has effects in extrapancreatic tissues, decelerating gastric emptying, (GLP-1 RAs) on Type 2 diabetes mellitus (T2DM) unse- reducing food intake—and body weight—and lowering circulating li- lected populations, composed mainly of low CV risk pa- poproteins, inflammation and systolic blood pressure.7,8 GLP-1 RAs tients, has been assessed through observational cohort have longer T1/2 than native GLP-1 and are the only injectable class studies in Western healthcare databases. among the marketed non-insulin blood-glucose-lowering agents. • Pooled results of these observational studies suggest The route of administration and gastrointestinal adverse effects a beneficial effect of GLP-1 RAs on all-cause mortality, are the main reasons for the non-adherence to treatment, relatively and to a lesser extent, on CV events; results for the HHF frequent in patients initiating GLP—1 RAs.9 The first agent of the were neutral. GLP-1 RA class, exenatide b.i.d., is available in clinical practise since 2005. Semaglutide has been approved in late 2017-2018 and the only GLP-1 RA with an oral formulation, oral semaglutide, in 2019.10 to provide complementary or confirmatory evidence of the effects Non-insulin glucose-lowering agents are marketed based on re- of drugs in real-world populations, especially for those underrepre- sults of clinical trials with surrogate variables, mainly the percentage sented or excluded in RCTs. EHDs containing secondary informa- of glycated haemoglobin and other glucose markers.11 However, in tion—(EHDs)—are increasingly used as source data in observational spite of the clear correlation between diabetes and CVD, the effect research evaluating comparative drug effectiveness and safety and of glucose-lowering therapies and glycemic control on CVD is un- have an increasingly important role in the generation of scientific clear.5 Since 2008, glucose-lowering agents that received marketing evidence.21 For comparative effectiveness research and the study authorisation based on results of clinical trials with surrogate vari- of relatively rare events, EHDs have many advantages over RCTs: ables have to comply with regulatory recommendations regarding have large size data, include patients often underrepresented in CV safety.12,13 Because of the low rate of CV events, these cardio- clinical trials, are representative of routine clinical care and suitable vascular outcomes trials (CVOTs) enrol a large number of diabetic pa- to study real-world effectiveness and utilisation patterns.22 On the tients usually with established or at high risk of CVD, and have long contrary, observational studies almost always have a bias because follow-up (FU) periods. Up to date, seven CVOTs assessing GLP-1 of the unequal distribution of prognostic factors between patients RAs on cardiovascular outcomes in high CV risk populations are exposed or not exposed to an intervention. The non-randomised completed and published (Table 1).14–20 Liraglutide and semaglutide design inherent to observational research can produce misleading showed beneficial effects on major adverse cardiovascular events results because of several flaws arising from design, time-related (MACE); for liraglutide, these results were driven by the reduction bias, matching and analysis.23 Moreover, when assessing treatment in CV mortality and for semaglutide, by the reduction of non-fatal effects in chronic and progressive conditions, as T2DM is, research- strokes. In August 2017, the US FDA included as a new indication for ers have to deal with additional challenges, such as the addition or liraglutide the reduction of the risk of CV events in adults with type switching to other ant-diabetic therapies to control blood-glucose 2 diabetes who have established CVD. biomarkers.24 However, it is often unclear to what extent results of clini- The objective of this systematic review is to describe and cal trials conducted in selected high-risk CV populations apply to compare characteristics of designs, baseline study populations unselected ones, composed mainly of low CV risk patients. If ad- and results of observational cohort studies conducted in EHDs equately conducted, studies using data of electronic healthcare assessing MACE, hospitalisation for heart failure (HHF) and all- databases containing secondary information (EHDs) can be useful cause mortality in Type 2 diabetes patients treated with GLP-1 HERRERA COMOGLIO and VIDAL GUITART | 3 of 18 (0.73-1.05)

(0.54-0.90)] (0.54-0.90)] no history of HF [0.82 (0.69-0.98)] 0.87 0.94 (0.78-1.13) (0.77-1.61) 1.11 0.93 (0.77-1.12) 0.93 0.86 (0.48-1.55) HHF prior HF [0.70 0.96 (0.75 −1.23) (0.71-1.06) (0.66-1.14) (0.70-1.03) (0.62-0.94)

(0.38 −0.99) * 0.86 0.85 0.74 (0.350.74 to 1.57) Stroke 0.86 0.76 0.61 1.12 (0.79-1.58) (0.61-0.90) (0.73-1.00

(0.51-1.08)

*

0.86 0.97 (0.85-1.10) 0.74 1.18 (0.73 to 1.90) AMI 0.75 0.96 (0.79-1.15) 1.03 (0.87-1.22) (0.27 to 0.92) (0.66- 0.93) (0.76-1.02) (0.78-1.06)

* *

0.88 0.49 0.98; (0.65 1.48) CV M 0.93 (0.73-1.19) 0.93 0.78 0.91 0.98 (0.78-1.22) (0.77-0.97) (0.74 to 0.97) (0.31- 0.84) (0.80-1.01) * *

* 0.86 1.05 (0.74-1.50) 1.05 0.85 A-C M 0.95 (0.79-1.16) 0.90 0.51 0.94 (0.78 to 1.13) (0.68-0.90) (0.79-0.99)

(0.78- 0.97) (0.58- 0.95) * * * * 0.91 (0.83- 1.00) MACE Results 0.87 0.78 0.88 0.79; (0.57 to 1.11) 1.02 (0.89- 1.17) 0.74 16.2% History HF 17.12% 20% 8.5% 22.4% 23.6% 73.1% History CVD 100% 70% CAD; 47% IM; Stroke; 17% 20% HF 31.5% CVD 20.8% CVE 100% 83% IHD 33% y ≥65 40.25% 64.6 ± 7.4 62%; y ≥65 Men%; elderly Men%; Population characteristics Population 64.3%; ≥60 y 75% 70%; Mean age 64.1 y. 53.7% Mean 66.2 years 58.4% Median 66 y. 69.7%; 60.7%; mean age y y 3.2 y Follow-up 2.1 3.8 y 2.1 median 1.6 y median 5.4 y 5.1-5.9) (IQR (1:1)

­ n= (randomi sation) 6,068 (1:1) 9,340 (1:1) 3,297 (1:1) 14,752 9,463 (1:1) 9,901 (1:1) 3,183 (1:1) 17

18 19 20 16 15

21 Semaglutide Liraglutide

Lixisenatide

semaglutide PIONEER-6Oral SUSTAIN-6 EXSCEL exenatide LAR * ELIXA GLP-1 RA CVOT HARMONYAlbiglutide REWINDDulaglutide LEADER Abbreviations: A-C M, all-cause mortality; AMI, acute myocardial infarction; CV M, cardiovascular mortality;adverse cardiovascular cardiovascular CVOT, outcome events; trial; RCTs, randomised hospitalisation HHF, controlled for heart trials. failure; MACE, major All results Hazard ratio CI). (95% HARMONY trial reports the composite of death from cardiovascular causes or hospital admission for heart failure. The table shows*Significant results of the results subgroup analysis are highlighted for HHF. in bold. TABLE 1 PublishedTABLE assessing RCTs CV outcomes 2 diabetes in Type mellitus patients treated with RAs: GLP-1 characteristics, population and results 4 of 18 | HERRERA COMOGLIO and VIDAL GUITART

RAs, and to descriptively and quantitatively summarise their re- searched for some various aspects in study design and analysis sults. This systematic review has been registered in PROSPERO that could impact on study results: study populations, study and (CRD42019122102). To the best to our knowledge, no systematic comparator drugs, length of FU, exposure ascertainment, and the review (SR) has been performed exclusively for observational co- number and nature of covariates included in matching and/or ad- hort studies assessing the effect of GLP-1 RAs on CV outcomes, all- justment. We searched to identify major bias that can explain even- cause mortality and HF. This article has been reported according to tual differences in the results of individual studies. We summarised the MOOSE Guidelines for Meta-Analyses and Systematic Reviews analysis’ results (primary, secondary, subgroup and sensitivity of Observational Studies.25 analysis), and strategies to control for confounding (restriction, matching and analysis adjustment). Three separate meta-analyses were performed (all-cause mortality, CV events and HHF) using a 2 | METHODS random-effects model to calculate overall hazard ratios (HRs) and two-sided 95% CI (confidence intervals). Analyses were performed 2.1 | Data sources and searches using Stata, version 13.1 (Stata Corp.). Between-study heterogene- ity was assessed by I2 statistic, a threshold of I2 greater than 50% We used Medline, Scopus and Web of Science to identify abstracts indicates high heterogeneity. In the primary MA, we excluded the and full texts published as of January 31, 2019. Searches included spe- results of two studies with a temporal bias that could have affected cific and class terms for GLP-1 RAs (exenatide OR liraglutide OR lixi- estimates. A further sensitivity MA included these two studies. The senatide OR albiglutide OR dulaglutide OR semaglutide OR GLP-1 RA), MA was reported following the MOOSE checklist. study design (observational, observational cohort, population-based) and sources of data (healthcare databases). We performed additional manual searches in references or “related articles” up to 31 Jan 2019. 3 | RESULTS No filters for language or date were applied. We excluded nested case- control studies because of their different design and analysis. We identified sixteen observational cohort studies assessing CV outcomes, HHF and or all-cause mortality in T2MD patients treated with GLP-1 RAs: Best et al,27 Mogensen et al,28 Paul et al,29 Vélez 2.2 | Study selection and quality assessment et al,30 Patorno et al,31 Kannan et al,32 Ekström et al,33 Anyanwagu et al,34 Suissa et al,35 Zimmerman et al,36 Toulis et al,37 Anyanwagu We selected abstracts, read in-depth papers matching the in- et al,38 Patorno et al,39 Dawwas et al,40 Svanström et al 41 and clusion criteria and analysed full text and additional files if any. O'Brien et al.42 The total number of patients exposed to GLP-1 RAs Inclusion criteria: 1. Retrospective observational cohort studies was 285,436; ranging from 21933 to 160,803.40 (Table 2) Study peri- 2. Conducted using data of EHDs. 3. Including T2DM patients ods span from 200527 to 2015 37 (Figure S1). treated with GLP-1 RAs. 4. Assessing major cardiovascular events (ie, cardiovascular mortality, acute myocardial infarction (MI) and stroke—either in composite or individual outcomes), HHF and all- 3.1 | Data sources cause mortality. Exclusion criteria: 1. Studies not including GLP-R As 2. Assessing exclusively other clinical outcomes, such as glycemic All studies were conducted in Western databases validated for inves- control, body weight and/or hypoglycemic events, adherence, eco- tigational purposes for the condition and outcomes of interest: most nomic outcomes, different clinical outcomes (cancer, pancreatitis). in US27,29–32,36,39,40,42,43 and UK33,35,37,38; Denmark,28 Sweden,34 and 3. Nested case-control studies. The flow diagram (Figure 1) and Sweden and Denmark.41 European databases are electronic medi- the list of retrieved and excluded papers show the results of the cal records (EMRs), with partial (in the UK, the Health Improvement selection process. (Tables S1 and S2, respectively) Methodological Network [THIN] and the Clinical Practice Research Datalink [GPRD]) quality/risk of bias was assessed using the ROBINS-I assessment or universal general population coverage (Denmark and Sweden); tool.26 Funnel plot graphics for each outcome were used to evalu- in Danish databases28,41 information about clinical variables is lack- ate publication bias visually (Figure 2). ing. US studies were conducted in claims EHDs27,29–32,37,39,40,42 and EMRs.32,36

2.3 | Data extraction and analysis 3.2 | Study populations We descriptively synthesised data about data sources (country and type), study populations’ baseline characteristics, number of sub- All studies included T2DM patients aged ≥18 years and exclude pa- jects included in full and analysed cohorts, study period, patient se- tients with diagnosis codes related to Type 1 diabetes; two stud- lection, start, end and length of FU, comparator drug(s), outcomes, ies also excluded patients with a history of malignancies, end-stage covariates and the period in which covariates were assessed. We renal disease, human immunodeficiency virus, organ transplant31,39 HERRERA COMOGLIO and VIDAL GUITART | 5 of 18

ZĞĐŽƌĚƐŝĚĞŶƟĮĞĚƚŚƌŽƵŐŚ ĚĚŝƟŽŶĂůƌĞĐŽƌĚƐŝĚĞŶƟĮĞĚ ĚĂƚĂďĂƐĞƐĞĂƌĐŚŝŶŐ ƚŚƌŽƵŐŚŽƚŚĞƌƐŽƵƌĐĞƐ ;ŶсϴϯϳͿ ;ŶсϭϴͿ /ĚĞŶƟĮĐĂƟŽŶ

ZĞĐŽƌĚƐĞdžĐůƵĚĞĚ;ŶŽŶ ZĞĐŽƌĚƐƐĐƌĞĞŶĞĚĂŌĞƌĚƵƉůŝĐĂƚĞƐƌĞŵŽǀĞĚ ůŽŶŐŝƚƵĚŝŶĂůŚĞĂůƚŚĐĂƌĞ ;ŶсϳϱϴͿ ĚĂƚĂďĂƐĞƐƚƵĚŝĞƐͿ ;ŶсϲϮϭͿ

^ĐƌĞĞŶŝŶŐ ůŽŶŐŝƚƵĚŝŶĂůŚĞĂůƚŚĐĂƌĞ ZĞĐŽƌĚƐĞdžĐůƵĚĞĚ ĚĂƚĂďĂƐĞƐƚƵĚŝĞƐ ;ŶсϭϮϭͿ;^ƵƉƉůĞŵĞŶƚĂů ;ŶсϭϰϳͿƐĐƌĞĞŶĞĚ dĂďůĞϭͿ

&ƵůůͲƚĞdžƚĂƌƟĐůĞƐĂƐƐĞƐƐĞĚ ĨŽƌĞůŝŐŝďŝůŝƚLJ

ůŝŐŝďŝůŝƚLJ ;ŶсϭϲͿ

^ƚƵĚŝĞƐŝŶĐůƵĚĞĚŝŶ ƋƵĂůŝƚĂƟǀĞƐLJŶƚŚĞƐŝƐ ;ŶсϭϲͿ

/ŶĐůƵĚĞĚ ^ƚƵĚŝĞƐŝŶĐůƵĚĞĚŝŶ ƋƵĂŶƟƚĂƟǀĞƐLJŶƚŚĞƐŝƐ ;ŵĞƚĂͲĂŶĂůLJƐŝƐͿ ;ŶсϭϰͿ

FIGURE 1 Flowchart of the selection process (PRISMA statement) and nursing home admission at baseline.31 All study populations had liraglutide.41 In two studies assessing the effect of canagliflozin39 low CV risk; six studies excluded participants with a history of as- and DPP4 i40 on HF, GLP-1 RAs were the reference drugs. Three sessed outcomes27–29,33,38,40 Table 3 shows the populations’ baseline studies reported the percentage of use,36–38 being exenatide and characteristics. liraglutide the most used GLP-1 RAs; two studies37,38 made sub- group analysis for each agent. Two studies reported the use of dula- glutide—approved in September 2014—and albiglutide39,40 no study 3.3 | Study drug(s) included semaglutide (marketed in December 2017). Table S3 shows the reporting if any, and the number of individual GLP-1 RA used. Six studies did not report the individual GLP-1 RAs included in their analysis.30,32–35,42 One study focused on incretin thera- pies.30 Four studies studied GLP-1 RAs exclusively as second-line 3.4 | Comparator drug(s) therapies,28,31–33 and one study assessed only liraglutide added to metformin (MET).41 Three studies evaluated a single agent: exena- In five studies, comparators were other non-insulin antidia- tide b.i.d. (the only GLP-1 RA licenced at studies’ periods27,29) and betic agents.27,29,30,36–38 Eight studies compared GLP-1 RAs vs 6 of 18 | HERRERA COMOGLIO and VIDAL GUITART

FIGURE 2 Funnel plots of 16 observational longitudinal cohort studies assessing effects of GLP-1 RAs in Type 2 diabetes mellitus patients. A, All-cause mortality. B, Cardiovascular events. C, Hospitalization for heart failure

a well-defined class: vs SUs,35 vs DPP-4 i40,41 (two with prior use the comparator groups.39 In two studies, the comparator was insu- of MET at baseline) or evaluated GLP-1 RAs vs other second-line lin29,34: one assessed third-line therapies, GLP-1 RAs vs insulin (both therapies, all added to metformin: metformin with SUs as a refer- added to metformin + SUs) (Table S4).34 ence28,31–33 or DPP-4 i,31,42 thiazolidinediones or insulin.31 In one If prescribed according to existing guidelines, drugs and lines study, the studied drug was canagliflozin and GLP-1 RAs one of of treatment are proxies of disease severity. Studies comparing HERRERA COMOGLIO and VIDAL GUITART | 7 of 18 (Continues) estimator (0.22- 0.52) (0.22- 0.52 0.34 (0.22- & Renal D: 0.32 0.50) (0.21- (0.32, 0.50) CVD: CV& renal D: 0.35 (0.28- 0.45) Results HHF Results 1 event, no Full cohort: 0.34 No previous CVD: No previous CV Full cohort: 0.40 Without previous 0.40 (0.3-, 0.50) Without previous

hospitalisation: hospitalisation

1.68);(0.47- Non-fatal1.19); Stroke; 0.50 (0.28- 0.84) 0.99); non-fatal stroke 0.37 (0.18 0.75) 0.65 (0.44, 0.98) AMI 0.52 (0.31, 0.91) 0.45 (0.28, 0.85) CV mortality 0.89 CV-related hospitalisation: HR 0.88; (0.79 −0.98) All-cause Results CV events CV related hospitalisation HR 0.88 (0.79-0.98); All-cause Full cohort: Non-fatal MI: 0.52 (0.23, No previous CVD: Non-fatal AMI 0.24 (0.06- HR 0.94 (0.91- 0.97) Full cohort non-fatal MI: No previous CVD: Non-fatal No previous CV and renal D (0.74-0.86)

coronary revascularisation HR HR revascularisation coronary 0.80 0.33 (0.18, 0.63) 0.32 (0.22, 0.46) Non-fatal MI, ischemic stroke or RR 0.82 (0.55-1.21) M I + stroke: Full cohort: 0.50 (0.32, 0.79) Without previous CVD: 0.33 (0.18- 0.63) Without previous CV& renal D: CVD event: HR 0.86; 0.92 0.81- Results MACE RR 0.64 (0.34-1.20) Full cohort: 0.44 (0.34, 0.57) Without previous CVD: 0.33 (0.23, 0.47) Without previous CV& renal D: (0.51-1.17) (0.40-1.23) (0.02-1.22) RR 0.77 Results All-cause mortality RR 0.70 HR 0.17 days) (742 NR 2.1 y. 3.5 y. Follow-up Follow-up (years) 2.03 y. AT ITT ITT AT ITT Exposure Exposure analysis ITT GLP-1 agents Inc Inc Inc Incident Design

insulin insulin + 4345) 1901) = = MET + GLP-1 RAs = 1901) (n vs MET + SUs (n 092), = 7,870); (n = 28,551) (n 754) = 21 (n 771) = 361 (n 1(n BID + OADs = 2,804);(n vs exenatide RAs(GLP-1 n = 205) control group = 2,798)(n BID + insulin vs = 7,870); (n insulin + OADs = 28,551) (n vs MET + SU(n = 25 cohort: Matched insulin + OADs Exenatide b.i.d. b.i.d. Exenatide non-exenatide MET + GLP- exenatide agents Incretin vs not defined Study and drugs comparator exenatide exenatide 31

28 30 29 et al Mogensen Best et al Author, yr Paul et al Velez et al TABLE 2 ObservationalTABLE cohort studies assessing the effect of glucagon-like peptide receptor agonists RAs) (GLP-1 on cardiovascular outcomes, all-cause mortality and/or heart failure 8 of 18 | HERRERA COMOGLIO and VIDAL GUITART (Continues) failure: HR 1.11 (0.99-1.22) 0.57-0.93 Congestive Heart Congestive Results HHF Results HHF CI: 0.73 (95% : HR 0.48 (0.8- 2.8) : HR 0.45 1.72) (0.12- 2 2 ; HR 1.11 (0.99-1.24) HR 1.11 (0.65-0.99); (0.74- 0.91) 0.78 (0.54- 1.12; CI 0.63-0.93; Coronary artery disease: CVD 0.26 (0.10-0.67); fatal CVD 0.20 (0.03- 1.45): CHD 0.31 0.81); (0.12- fatal CHD (0.02- 0.17 1.22); Results CV events BMI ≥ 40 kg/m Full cohort: Full Non-fatal AMI 0.80 non-fatal stroke 0.82 Prior CVD: Non-fatal AMI non-fatal stroke 95% 0.76, MI Full cohort: HR 0.45 (0.12-1.69) BMI ≥ 30 kg/m2: HR 0.57 (0.15- 2.14) BMI ≥ 40 kg/m2: HR 0.33 (0.03-3.29) Stroke Full cohort: HR 0.39 (0.10- 1.44) BMI ≥ 30 kg/m : HR 0.31 CI (95% : HR 0.31 2 2 0.16- 0.61 0.13, 0.75) = 419) (n BMI ≥ 40 kg/m vs DPP-4 i 1.02 (0.84-1.24) vs SUs 0.86 (0.69-1.09) vs insulin 0.72 (0.62-0.84) After matching for PS that included HbA1: vs DPP-4 i HR 1:20 (0.76-1.89) vs SUs: HR 1.05 (0.63-1.74) (0.73-1.74) 1.01 HR insulin: vs vs DPP-4 i HR (0.75-1.18) 0.94 vs SUs HR 0.78 (0.59-1.04) vs insulin HR 0.88 (0.62-1.26) CVD: HR 0.26 (0.10-0.67) Results MACE MACE PS matched cohort HR 0.27 CI 0.28, (95% 0.97) BMI ≥ 30 kg/m Full cohort: Full HR 0.82 (0.74- 0.91) Prior CVD: (0.63-0.92; HR 0.76 All- HR All-cause : HR 0.24 : HR 0.33 2 2 (0.30-1.07) 2.25)(0.35- 0.21 (0.08 −0.51) m (0.10- 0.59) m (0.12-0.92) cause mortality HR 0.48 (0.41- Prior 0.57) CVD: mortality HR (0.50-0.99) 0.70 HR 0.57 HR 0.88 Results All-cause mortality BMI ≥ 30 kg/ BMI ≥ 40 kg/ Full cohort: Full Full cohort Full

† 3.74 y 5 years 0.8 y 8.7 y 4 y 3.6 y Full cohort Follow-up Follow-up (years) FU up to > 2 y 1.2 (0.4, 2.8) y AT AT ITT ITT ITT AT Exposure Exposure analysis ITT AT new users new incident incident incident incident incident Prevalent + Prevalent Design

466) = 14 (n = 114.480);(n 658 = 18 vsi (n 807);n = 69 (29 343 vs 982) 42 = 433)(n = 9419) (n agent(s) = 219 (n RA)GLP-1 RAs = 419) (n = 1584)(n each n = 419 2007) = 6196) (n = 79 682)(n = 8362)(n 500) = 97 (n MET + GLP1 RAs vs MET + SU vs MET + DPP-4 vs MET + insulin MET RAs + GLP-1 vs MET + SUs MET ra + GLP-1 MET + SU = 4293) (n + GLP-1 MET-SU MET + SU +insulin Matched cohorts: ra (firstGLP-1 use SUs (first use 2000) RAsGLP-1 Non-GLP-1 Study and drugs comparator

36 35 32 33 34 37 et al 2016 et al et al et al et al Anyanwagu Patorno Patorno Kannan Kannan Ekström Author, yr Suissa et al Zimmerman Zimmerman TABLE 2 (Continued) TABLE HERRERA COMOGLIO and VIDAL GUITART | 9 of 18

1.66

1.22 (Continues) (0.69-2.08) (0.66-2.24) (0.23-12.13) (0.76-1.94) HR: 1.47 (1.16-1.85) HR: (1.28- 2.13) Liraglutide: 1.20 1.22Exenatide: Lixisenatide: Results HHF Results Full cohort: Full (0.22-1.04)

(0.36-1.56)

0.48

0.75

0.93 (0.83-1.12). of nonfatal CV events: (0.41-1.42) 0.76 (0.54- 1.61) MI:HR (0.69-1.37) 0.97 Stroke: HR 0.93 (0.61-1.45) UA: HR 1.37 (0.85-2.17) CR: HR 0.99 (0.76-1.29) in low-risk patients (aIRR: Exenatide: Lixisenatide: - Lixisenatide: Results CV events AMI HR 0.53 (0.21, 1.30) Stroke HR 0.54 (0.38, 0.77) AMI HR 0.82 (0.66, 1.03) Stroke HR 0.20 (0.18, 0.23) Liraglutide: prior CV events: Composite AMI: 0.62 (0.06-6.53) Non-fatal stroke: 0.93 (0.32-0.89)

(0.50-1.31)

0.53 MACE: 0.64

0.81

ischemic stroke, or haemorrhagic stroke) (0.42-0.98 HR 1.03 (0.81-1.33) (Acute myocardial infarction, HR (0.74-1.27) 0.97 Results MACE HR 0.49 (0.34, 0.69) HR 0.86 (0.78, 0.96) Full cohort: Full Lixisenatide: - Lixisenatide: Exenatide: Liraglutide: ) (0.54-3.13) (0.56-0.74), (aIRR: 0.64 (0.53-0.76) 0.56 (46-0.67 (0.6-0.85) 1.34) 0.55) prior CV events: (0.17-0.73 0.35 (0.06-0.0.47) (0.24—0.99) HR 0.1.29 Full cohort: Full aIRR 0.64 low-risk patients Liraglutide aIRR 0.72 Exenatide Results All-cause mortality HR 0.80 (0.47, HR 0.46 (0.38, Cohort with no Liraglutide: 0.17 Liraglutide: 0.17 0.49 Exenatide: - Lixisenatide: 3.07) to 5 y 1 y 3.2 y (SD: Total FU up 0.6 y (0.5) 2.72 y (32.7 mo.) Follow-up Follow-up (years) ITT AT AT ITT Exposure Exposure analysis incident prevalent incident incident incident Design

GLP-1 GLP-1 + ra (exenatide, liraglutide, 1793 lixisenatide) vs insulin + OADs exposed to GLP-1 ra (exenatide, of canagliflozin or a comparator liraglutide, lixisenatide) = 8345)(n vs unexposed RAsto GLP-1 541). = 16 (n (PS matched cohort) matched (PS vs non-N = 1793 RAsGLP-1 insulin 20 539 pairs initiator Study and drugs comparator

38 40 39 et al et al Toulis et al Anyanwagu Patorno Patorno Author, yr TABLE 2 (Continued) TABLE 10 of 18 | HERRERA COMOGLIO and VIDAL GUITART (0.43-1.02)

(1.11- 1.20) (1.11- (1.12-1.22) (0.93-1.35) 1.11 Results HHF Results HR: 1.16 No HF history: 1.18 Prior HF history: HR 0.90 (0.80-1.03) 0.65 (0.68-0.91), Results CV events CV death HR 0.78 MI 0.94, (0.84-1.06) stroke 0.88 (0.77-1.01) Stroke 0.65 (0.44-0.97) IHD (0.67-1.24) 0.91 0.90PAD (0.42-1.95)

and

heart disease, revascularisation, 0.92 (0.90-1.22)

(0.89-1.01).

analysis instead of PS matching: HR 0.88, CI 0.82-0.94) 95% peripheral arterial disease.) (HR 0.95, 0.83-0.98) (0.71-0.92) ischaemic coronary (0.65-1.01) (0.50-1.10) (0.61-1.08) Results MACE HR 0.84 CI 0.75-0.94) (95% Full cohort with multivariate MACE: HR CI 0.90 (95% history of major CVD: 0.81 without CVD: (0.86-1.06) 0.96 Expanded MACE other (+ HR: 0.78 (0.63-0.96) Prior CV events: 0.78 (0.43-1.42) No history CV events 0.81 Prior MET No prior MET 0.66 (0.48-0.90) Adherent patients: 0.74 Time-varying model: 0.81 Results All-cause mortality HR 0.83, (0.77-0.90)

290 (± 170 d.) in DPP4 i and 159 (±285 d.) in RAsGLP-1 group 2.1) for liraglutide and 3.2 y for (2.0) inhibitors. liraglutide and 1.7 y (1.6) FU 3.5 y (SD DPP-4 Follow-up Follow-up (years) 2.1 y (SD 1.8) for For DPP-4 i Up to 2 y Mean 1.3 y.

(sensitivity (sensitivity analysis) AT ITT Exposure Exposure analysis AT ITT AT incident incident Design incident second- as RA RAs each matched cohort 160 803) i with previous/ basal MET line after taking MET alone OR reference as DPP-4 i vs GLP-1 (321,606 patients, Liraglutide vs DPP-4 Study and drugs comparator GLP-1 no ADM vs DPP-4 42 41 43 et al et al Dawwas Author, yr 351 n = 11 O’Brien Svanström for INS; y. 1.425.19 RAs. for GLP-1 TABLE 2 (Continued) TABLE † Note: All estimates CI). (95% Abbreviations: AMI, Acute myocardial infarction; As-treated AT, approach; BMI, Body mass index; CR, Coronary revascularisation;peptidase Cardiovascular; CV, 4 inhibitors; FU, Follow-up; CVD, RA, Cardiovascular GLP-1 Glucagon-like disease; peptide DPP-4 i, Dipeptidyl 1 receptor agonists; Heart HF, failure; Hospitalisations HHF, fortreat; heart MACE, failure; Major adverse HR, Hazard cardiovascular Ratio; INS, Insulins; events; ITT, MET, Metformin; Intention-to- MI, Myocardial infarction; PS, Propensity score; SU, Sulphonylureas; UA, Unstable angina. HERRERA COMOGLIO and VIDAL GUITART | 11 of 18 (Continues) 2 and EBID + insulin groups and 65% in group insulin exenatide and 9.1% non-exenatide SUs 30.7 (4.8) kg/m 37.6) 32.1 (27.9, median 36.98 (32.2, 43.0): MET + SUs 32.28 (28.2, 37.5) NA ≥30 89% in the EBID 93.2% Obese 37.4 (7.1) BMI kg/m2 Obese 16.2% NA 11.2 to 21.5% Mean 34.9 (5.1): mean BMI 39.6(7.1) median BMI (IQR) MET + GLP--1 RAs and 38.4% former 27% RAs: Current and9.5% Non- smokers 52.3% For MET + SU 15.8% 46.7% NA 51% Current 22.4% smoker Past 57.5% current 12%, 14.1% Current current or past smokers NA NA MET + GLP-1 5.6 to 7.7% current 18,1% † † † 0.2% 0.78% NR NR 1.7% NR % history MI 0% 0% 0% NR NR † † 0% 0% 0.50% NR 3.1% 5% 5.3% % history stroke 2.8% 0.4% NR 0%

† criterium) SUs 3.8% 2.6%GLP-1 0% 0% (exclusion 3.5% 1.71% raGLP-1 4.1%; 0.2% 4.6% NA 3.8% % history HF Total: 3.2; 1.2% anti-hyper hypertensives) hipertensives) 45% 42.5% 65.4% 83% (use of tensives) 75.3% (use of anti- 50.6% (use of anti- 63.0.6% 65% 63.8% % Hyper tension NA (use of ARA 44.4%

† 12.8%; SUs 15.2% other Coronary Heart Disease, 6,1%; GLP-1 5.6% 1.6% - 12.7% IHD 12.7% IHD 6.53% raGLP-1 1.6% for 23.2% 13.3% 16.5% % history of CVD IHD: total total IHD: to 13.1% 7.6 0%

77% 6.7% 15.6%

(3.4); SUs 7.7% (3.2) 7.1%- 8.0%: >8%: ± 1.77 7.98 NA 7.6% ra GLP-1 7.8% 9.4(2.0) % ≤7%: NA 8.4 (3.9) % median HbA1c NA to 9.9% 7.8 NR CKD)

y; SUs 5.0 (3.0, 8.0) y (IQR) 2.0 y (0.4-4.5 y); 3.0GLP-1 (1.4-5.1) diabetes indicators of severity) neuropathy, retinopathy, 2.47 y NA 1.0 (0.2, 1.8) y RAGLP-1 5.0 (2.0, 8.0) mean 2.7 y (3.0) 5.2 ± 3.0 y NA (5.7)9.8 T2DM duration (initiation monotherapy monotherapy (initiation NA (proxies as NR ( proxies: 56.1%GLP-1 60.0.6% MET + GLP-1 and 54.9% for MET + SU 45.4% 43.8% Total 59% total 59% to 47.2 55.7% 42%: for 54.2% 54.9% 51.8% 54.6% % men

(exenatide) y; GLP-1 54.3 ± 11.7; 53.5 (11.9) in matched cohort SU: 62.9 y MET + GLP-1 RAs y, 55 MET + SUs 62 y 10.5 y; 70) (51, 61 60.0 ± 11.4 52.7 ± 8.7 total 60.6 ± 13 GLP-1 58 y. 50.7 to 53.2 GLP--1:56.0 y; 60.6 ± 12.6 y; 49.60 ± 56.0 ± 10.5 Median (IQR) 57.5 (10.4) mean age 33 32 34

36

31 28 30 38 35 37 29 et al et al (refers to incretin therapies) et al Patorno et al Zimmerman Zimmerman Best et al Paul et al Velez et al Ekström et al Study Study identification: Author Suissa et al Mogensen Kannan et al et Kannan Toulis Anyanwagu TABLE 3 CharacteristicsTABLE of populations exposed RAs to GLP-1 in observational cohort studies 12 of 18 | HERRERA COMOGLIO and VIDAL GUITART

the same treatment lines (ie, first, second or third-lines therapies) compare treatments at the same stage, while in studies comparing the use of a drug vs non-use there is a source for time-lag bias. In our search, six studies compared second-line therapies: the use of MET + GLP-1 RAs vs MET + SUs,28,31–33 vs other agents,31 liraglu- BMI kg/m2 22.3% NR 32.4 (6.9) tide vs DPP-4 i41 and vs DPP4 i.42 One study compared third-line therapies,34 and other the intensification either with insulin or GLP-1 RA.38 former 38% current or past smokers Current 13%, 7.4% NR 3.5 | Outcomes ­

† 3.5.1 | All-cause mortality Revascul arization 2.3% % history MI Coronary 0% NR

Ten studies estimated all-cause mortality as primary or second- † ary endpoint,28,30,32–34,36–39,41: hazard ratios (HRs) ranged from 0% % history stroke 1.2% NR 0.17 (95% CI 0.02-1.22)30 to 1.29 (95% CI 0.54-3.13).39 Significant HRs spanned from 0.21 (95% CI 0.08-0.51) (third-line GLP-1 RAs vs insulin therapy33) to 0.83 (95% CI 0.77-0.90),/ liraglutide vs DPP-4 i41). In four studies comparing GLP-1 RAs vs non-GLP-1 60 d.) 6% % history HF 3.4% 0% previous the (in RAs30,36–38 significant estimators ranged from HR 0.35 (0.17- 0.73) to 0.64 (0.56-0.74): liraglutide subgroups37,38 and prevalent users36 showed more beneficial results than exenatide and in- cident users, respectively. Three studies comparing second-line therapies MET + GLP-1 RAs vs MET + SU28,32,33 obtained non- hypertensives 90% use of anti- % Hyper tension 53.5% 33.% ACEi 14.9% CCB significant results. The study comparing canagliflozin to GLP-1 RAs39 (mean follow-up (FU) 0.5 year), obtained the highest non- significant estimator benefiting canagliflozin. (Table 2) The pri- mary MA excluded two studies with lack of alignment of the start 20% % history of CVD IHD 11.8% 12.5% of exposure with start of FU.34,37 When pooling results, irrespec- tive of different study drugs and comparators, all-cause mortality favoured GLP-1 RAs, HR 0.63 (0.44-0.89) (with high between- study variability I2 = 83.3%). (Figure 3) In a sensitivity analysis, the addition of the two previously excluded studies, with a 6-mo. 8.6% (1.8%). median HbA1c 8.8 (1.8) % NR and 9-mo. delays, respectively, between the start of exposure and the start of treatment did not alter results (HR 0.61, [0.44- 0.77]) (Figure 4).

3.5.2 | Cardiovascular events 4.9 (4.5) 4.9 T2DM duration NA NR

Eleven studies estimated non-fatal cardiovascular events and/or MACE not including deaths; for individual CV events, HRs ranged from 0.24 (95% CI 0.06-0.99), with insulin as reference,29 to 0.82 53.2% % men 52.8% 44.1% (95% CI 0.74-0.91)37 (stroke in prevalent users vs non-GLP-1 use). MACE including non-fatal events ranged from 0.27 (0.28-0.97) vs insulin34 to 0.80 (95% CI 0.74-0.86) vs non- GLP-1 use.27 Six stud- ies assessed MACE including mortality, significant HR ranged from 62.7 (13.8) y mean age 56.7 (10.8) 53 (10.2) 0.26 (95% CI 0.1-0.67)32 to 0.90 (95% CI 0.83-0.98) for liraglutide

40 41

41 vs DPP-4 i. (Table 2) Pooled results of nine studies suggest a ben- efit of GLP-1 RAs on cardiovascular outcomes (HR 0.82, 0.73-0.92) 39 (high between-study variability: I2 = 75.5%) (Figure 3). The sensibility et al Anyanwagu Study Study identification: Author n = 20.539 n = 160.803 Dawas et al Patorno et al Abbreviations: AMI, Acute myocardial infarction; BMI, Body mass index; Cardiovascular; CV, CVD, Cardiovascular disease;RA, DPP-4 Glucagon-like i, Dipeptidyl peptide peptidase 1 receptor 4 inhibitors; agonists; EBID, Exenatide HbA1c, Glycated b.i.d.; haemoglobin; GLP-1 Heart HF, failure; IHD, Ischemic heart disease;infarction; INS, Insulins; NA, Not IQR, available; Interquartile NR, Not reported; range; MET, SU, Metformin; Sulphonylureas; MI, Myocardial T2DM, 2 diabetes Type mellitus. †It is an exclusion criterium. TABLE 3 (Continued) TABLE analysis did not vary results (Figure 4). HERRERA COMOGLIO and VIDAL GUITART | 13 of 18

Ŷс Ŷс

Ϯϭϳϱϰ ϰϯϰϱ ϰϯϰϱ ϮϬϱ ϮϴϬϰ ϰϯϯ ϴϯϲϮ Ϯϭϵ ϭϰϰϲϲ ϴϯϲϮ ϰϯϯ

ϭϳϵϯ Ϯϭϵ

ϮϬϱϯϵ ϭϳϵϯ

ϮϯϰϬϮ ϮϬϱϯϵ ϮϯϰϬϮ ϭϭϯϱϭ

(A) ĂůůͲĐĂƵƐĞŵŽƌƚĂůŝƚLJ (B) sĞǀĞŶƚƐ

FIGURE 3 Meta-analysis of observational cohort studies performed in LHDs assessing all-cause mortality (A) and CV events (B) in Type 2 DM patients treated with GLP-1 RAs. A, all-cause mortality. B, CV events

Ŷс Ŷс

Ϯϭϳϱϰ ϰϯϰϱ ϰϯϰϱ ϮϬϱ ϮϴϬϰ ϴϯϲϮ ϰϯϯ ϰϭϵ ϰϭϵ ϭϰϰϲϲ Ϯϭϵ ϰϯϯ ϴϯϲϮ Ϯϭϵ ϴϯϰϱ ϴϯϰϱ ϭϳϵϯ ϭϳϵϯ ϮϬϱϯϵ ϮϬϱϯϵ ϮϯϰϬϮ ϮϯϰϬϮ ϭϭϯϱϭ

(A) ^ĞŶƐŝƟǀŝƚLJ ĂŶĂůLJƐŝƐ ŽĨĂůůͲĐĂƵƐĞ (B) ^ĞŶƐŝƟǀŝƚLJ ĂŶĂůLJƐŝƐŽĨ ŵŽƌƚĂůŝƚLJ ĐĂƌĚŝŽǀĂƐĐƵůĂƌ;sͿĞǀĞŶƚƐ

FIGURE 4 Meta-analysis of observational cohort studies performed in LHDs assessing all-cause mortality (A) and CV events (B) in Type 2 DM patients treated with GLP-1 Ras. Sensitivity analysis including 16 studies. A, Sensitivity analysis of all-cause mortality. B, Sensitivity analysis of cardiovascular (CV) events

3.5.3 | Hospitalisation for heart failure 3.6 | User design

Eight studies assessed HHF, being the only outcome evaluated in The incident user design identifies patients starting a new treat- two studies (vs sulphonylureas35 and vs DPP-4 I40). HRs ranged from ment, FU begins after treatment initiation. The incident user's design 0.12 (95% CI 0.02-0.66) for MET + GLP-1 RAs vs MET + SUs 32 to reduces the study size, and thus, compromise the precision of the 1.64 (95% CI 1.28-2.13) for HF when compared with canagliflozin comparative estimates.43–45 (GLP-1 RAs were taken as reference, canagliflozin showed a risk re- Patients were incident new users in all studies but two.35,36 duction of 39%)39 Significant HRs ranged from 0.12 (0.02-0.66)32 to Zimmerman et al36 selected both current and incident users of GLP-1 1.16 (1.11-1.20) vs DPP-4 i.40 When pooling results, we obtained a RAs (vs non-GLP-1 RAs use). Suissa et al35 proposed the prevalent HR 0.94 (0.78-1.14) (between-study variability I2 = 90.1%) (Figure 5). new-user cohort design to compare GLP-1 RAs—irrespective to the 14 of 18 | HERRERA COMOGLIO and VIDAL GUITART

Covariate information should be assessed in the longitu- dinal health care data in the time preceding the cohort entry. Ŷс “Intermediates” covariates should be assessed a defined time be- fore the inclusion date because they can represent an effect of ϮϴϬϰ 47 ϰϯϯ treatment or a reason for the prescription. In different studies,

Ϯϭϵ covariates were derived from the pharmacy and medical claims ren- ϲϭϵϲ dered during the 9 months preceding cohort entry27 or during the ϭϳϵϯ 6 months preceding drug initiation31,39; Charlson co-morbidity score ϭϲϬϴϬϯ

ϮϬϱϯϵ 10 years prior the baseline, and concomitant CV therapies within the 28 ϮϯϰϬϮ last year. Paul et al included measures of HbA1c within a 3-month window before the index date.29 In others, baseline variables were retrieved closest to the date of dual therapy's start; comorbidities, including CAD and CHF, were screened back up to 10 years ear- lier.32,34 In Anyanwagu 201634 study covariates were collected at FIGURE 5 Meta-analysis of observational cohort studies least 180 days before intensification of MET + SU. Table S4 shows performed in LHDs assessing hospitalization for heart failure (HHF) covariates used in the different studies. in Type 2 DM patients treated with GLP-1 Ras previous use of the comparator drug—to SUs incident or prevalent 3.8.1 | Study duration users: a purely incident new-user design would have restricted the 6196 GLP-1 RAs users to only 1633 new users of GLP-1 RAs. The study periods ranged from 2.539 to 14 years.35 (Figure 5) Two stud- GLP-1 RAs users were matched to SUs users employing a time-con- ies30,35 started in 2000, while the first GLP-1 RA, exenatide b.i.d., ditional propensity score matching based both on time and number was marketed in 2005. The chart of studies’ periods and the most of prescriptions, to compare patients with the same severity of the relevant drug approvals are shown in Figure S1. disease.35

3.9 | Follow-up 3.7 | Exposure ascertainment For the primary analysis, the shortest mean FU periods were In the intention-to-treat approach (ITT), the initial drug exposure is as- 0.44 years40 (study duration of 7.5 yrs.) and 0.6 years.39 (study dura- sumed to be unchanged until the end of FU, irrespective of treatment's tion 2.5 years, with 55%-68% of patients censored because of the changes.46 The as-treated approach (AT) is considered more suitable end of study), to up to 4 years32 The most prolonged FU period was for the exposure ascertainment in observational cohort studies; the AT 8.7 years31 (AT approach, secondary analysis), and in two studies approach computes the specific drug exposure's time from exposure's was up to 5 years.34,38 start until the drug's discontinuation. For the main analysis, eight stud- ies used the ITT approach29–33,35,37,42 and eight studies used the AT approach.27,28,33,36,38–40 Seven studies used the opposite approach as 3.10 | Diabetes duration a sensitivity analysis, and results were consistent.27,28,31,32,39,41,42 Diabetes usually worsen with longer duration, and the risk for CV disease increases. Diabetes duration is reported in eight stud- 3.8 | Covariates ies28–30,32,34,35,37,38 and not-reported in the remaining ones. Proxies of disease severity, such as diabetes complications, are reported in In most studies, covariates for matching and/or adjustment in- two studies (Table 3).31,33 cluded demographics, clinical and laboratory data, concomitant medications (antidiabetic, anti-hypertensives, lipid-lowering therapies, antiplatelet therapy and other cardiovascular medica- 3.11 | Bias tions) co-morbidities. Some included socio-economic factors and duration of diabetes, an interaction term to account the effect of We used the ROBINS-I assessment tool to assess the risk of bias.26 statin use and LDL cholesterol.36 Best et al selected more than All studies conducted in large longitudinal databases are subject to 300 covariates, among them, the top 50 of each pre-exposure pa- information bias and have a risk of non-differential exposure mis- tient characteristics. In Patorno et al,31 three propensity score- classification and outcome misclassifications, that could be more matched cohorts included matching of more than 200 patients’ likely in claims databases than in EMRs.22 Studies with a specific risk characteristics. of bias are mentioned below. HERRERA COMOGLIO and VIDAL GUITART | 15 of 18

3.11.1 | Time-related bias since some studies used multivariable analyses or combinations of both. The differences that can probably be more relevant in terms Three studies with the new-user design did not start the FU at the same of impact on results refer to the use of time-dependent variables in time with the start of treatment. Ekström started the FU at the second the analyses and the management of missing values. Table S5 shows filled prescription of the agent added to MET; a delay not temporally de- covariates, methods and missing data approach used in each study fined, and considered not to be important. Toulis et al set the index date reporting. for each exposed patient as the date of the fourth quarter consecutive GLP-1 RAs prescriptions (at least 9 months after the start of GLP-1 RAs therapy): this would imply a selection bias for tolerant/responsiveness 4 | DISCUSSION or adherent patients and could have affected outcomes’ ascertainment. Besides, patients with less than four quarter consecutive prescriptions In our systematic search of retrospective cohort studies performed could be part of the non-exposed group (risk of misclassification).37 In in EHDs, there was a low risk of publication bias. Other risks of bias both studies, “immortal-time bias” (ie, person-time that is event-free by have been mentioned previously. All studies were performed in definition) was avoided by matching the unexposed cohort at the index Western EHDs (US, UK, Sweden and Denmark) and included low CV- date of their respective exposed patients, and assigning the same index risk patients, being widely representative of the “real-world” T2DM as their respective exposed patients. population. Six studies excluded patients with previous events, ap- In Anyanwagu 2016,34 outcomes must have occurred at least proximately 67%. The results of our MAs suggest a beneficial effect 180 days after the intensification of MET + SU with either INS or of GLP-1 RAs on all-cause mortality and, to a lesser extent, on CV GLP-1 RA; although there were no deaths in this period, the out- events and neutral results on the HHF. come's ascertainment could have been affected. In CVOTs assessing the effects of GLP-1 RAs on MACE in high In the study comparing canagliflozin to non-gliflozin, among CV-risk populations, liraglutide (LEADER) has shown significant them GLP-1 RAs, there is a risk of attrition bias because 55%-68% of results in reducing MACE, all-cause mortality and CV mortality; patients were censored because of the end of the study.39 semaglutide (SUSTAIN-6) and dulaglutide (REWIND) reduced MACE and stroke; albiglutide (HARMONY) significantly reduced MACE (included HHF); exenatide QW (EXSCEL) reduced all-cause 3.11.2 | Selection bias mortality, and oral semaglutide reduced CV mortality and all-cause mortality.14–20 No CVOT has been performed for exenatide b.i.d. If patients are excluded based on future events, there is a risk of Results of meta-analyses of RCTs—including results of SUSTAIN-6 selection bias. A study required the completeness of information on suggested 12% lower risk of all-cause mortality, 12%-16% of CV event dates for HF, MI and stroke following the index date,29 and mortality with the use of GLP-1 RAs in T2DM patients, and a 10% inclusion criteria required the availability of two prescriptions within risk reduction for MACE.48–52 (Table S6) A study pointed out that 60 days of index date, this potential temporal bias was controlled by only a small proportion of US population could be included in most adjusting the time to event for the survival analyses to avoid possible of the CVOTs assessing GLP-1 RAs, and because of this, the results immortal-time bias. In other studies, patients had to refill a prescrip- of CVOTs could not apply to US T2DM general populations.53 We tion of the second glucose-lowering agent within 180 days and main- found that pooled results of the observational studies included in tain continuous treatment with MET throughout the 180 days34; or this review are trended in the same direction than those of MAs of had to have at least two encounters after initiating the combination RCTs, although they had a much larger size for all-cause mortality therapy to be included.32 In Zimmerman et al,36 patients were in- and CV events. However, individual results vary in magnitude and, cluded if they received T2 diabetes medications during the study in some cases, also in direction. period and if they had to have at least two medical visits. In a study Differences of different GLP-1 RAs on CV outcomes have been that defined the cohort entry by insulin intensification, (ie, either the explained not only because of the respective T1/2 that could im- addition of GLP-1 RAs or OADs to baseline insulin), the analysis of pact on glycemic control, but also because of the action on other two points MACE (MI and stroke) was restricted to matched cohorts organs.7,54,55 In our search, study drug(s) varied over time, from excluding all patients with a CV event before or 180 days after in- exenatide b.i.d. in the oldest studies27,28 to agents with more pro- sulin initiation. longed action, mainly liraglutide and exenatide long-acting release (LAR) in the studies published more recently. Two studies—assess- ing only exenatide b.i.d.27,29—obtained HR 0.81 and HR 0.44 for 3.12 | Methods to control for confounding MACE not including deaths. The study evaluating liraglutide vs a neutral antidiabetic drug class, DPP-4 i,41 got HRs similar to those In observational studies, careful control of confounding is needed of LEADER trial, although characteristics of their populations were because physicians chose treatments based on patients’ character- not (100% high CV-risk patients −81% with established CVD—in istics, and therefore, cohorts’ characteristics differ. Propensity score the LEADER and 16% of patients with ischemic heart disease matching was the most frequently used method, but not exclusively, in the observational study). In two studies reporting subgroup 16 of 18 | HERRERA COMOGLIO and VIDAL GUITART analysis37,38 liraglutide has obtained better estimates than exen- size and number of events in some studies do not contribute to atide LAR (Table 3). None of the studies included in this systematic robust effect estimates, resulting in wide CIs. We have just sum- review assesses semaglutide, which has been marketed after the marised the differences in methods to control for confounding, end of studies periods. with no further analysis; 10 large cohorts analysed propensity The choice of comparator drug can also impact on results, either score-matched populations. The 37% risk difference in mortality because of their effects on CV outcomes and HF or the confounding obtained in our MA is threefold the one obtained in MAs of RCTs. for indication, not enough controlled. Study populations are only from three European countries and the It has been suggested that the decrease in major CV events ob- US, thus pooled results would be not generalisable across other served in RCTs with GLP-1 RAs could be as a result of an antiathero- health care systems. genic effect, and thus, related with duration of exposure.55 In the On the contrary, this set of studies represent the retrospective same way, a nested case-control study found the effect of liraglu- observational research conducted in Western EHDs available at pres- tide was dose and duration dependent.56 Zimmerman et al included ent. The pooled results, even if oversised, are clinically relevant and incident and prevalent users of GLP-1 RAs; for all-cause mortality, trend in the same direction found in the MAs of RCTs. We described the subgroup of prevalent-users obtained more benefit from the use and analysed similarities and differences in many aspects. Finally, this of GLP-1 RAs than one of the incident users; the overall HR 0.48 is the first SR and MA of observational cohort studies conducted in (0.41-0.57) was driven by prevalent users, what could be related to EHDs, whereas another one considered both RCTs and observational more prolonged exposure. Kannan et al, with the most extended research for HHF—57 and because of this add the “real-world evi- mean FU and an ITT approach, obtained a non-significant estimate dence” on the effects of GLP-1 RAs on low CV-risk populations. for all-cause mortality (HR 0.57 [0.30-1.07]). In one study with AT approach results for CVOs and all-cause mortality were similar at a year, but became significantly different after 5 years of FU.38 This 5 | CONCLUSION delayed benefit would be consistent with the finding of the LEADER study,15 with a median FU of 3.8 years, in which the beneficial effect In this review of observational cohort studies in non-selected T2DM of liraglutide on CV mortality and all-cause mortality became evident populations, pooled results suggest a beneficial effect of GLP-1 about 12-18 months after started the treatment 54 (a similar—although RAs on all-cause mortality and, to a lesser extent, on CV events, smaller—difference has become apparent in the EXSCEL trial—me- while the use of GLP-1 RA would have no impact on the HHF. These dian FU—12 months after exenatide LAR treatment started).17 results are in line with those obtained in some CVOTs and meta- Six studies analysed subgroups of prior o no history of CV events. analyses of RCTs. However, the magnitude of combined estimates Three studies estimated benefits for subgroups with no previous CV should be considered with caution, because of inherent limitations events, and three found that patients with a history of CV events of the non-randomised design, inter-study variability and residual had better outcomes than the no-history of CV events. confounding. Further research rigorously conducted is needed to In observational cohort studies assessing chronic and progres- add complementary evidence of the effects of GLP-1 RAs on low sive conditions, such diabetes, the ascertainment of exposure(s) CV-risk populations. period (s) during FU can be blurred by drug discontinuation, in- tensification with other drugs, and/or switching to other antidi- CONFLICT OF INTEREST abetic agents. The “as-treated” approach for the ascertainment The authors have no conflicts of interest relevant to this article to of exposure is more suitable because the “intention-to-treat” ap- disclose. proach can introduce exposure misclassification. For exposures’ ascertainment, in the studies included in this review, the choice AUTHORS' CONTRIBUTIONS of either the ITT or AT strategies seems not to have influenced Both authors contributed equally in the conception of work, search the results. Patorno et al31 used an ITT analysis of exposure and selection, analyses and interpretation of results. XVG made sta- with a short period of FU (up to 1 year, mean FU 0.8 years) in PS tistical analysis. The manuscript was drafted by RHC and reviewed matched cohorts; an AT approach and a FU of 8.7 years, obtained by XVG. a similar non-significant estimate. Another study with the ITT ap- proach28 that reappraised data in an AT analysis, the median FU ORCID was 2.1 years, results were consistent. Raquel Herrera Comoglio https://orcid. This systematic review has several limitations. Studies have org/0000-0002-2810-1749 substantial differences in size and number of covariates included Xavier Vidal Guitart https://orcid.org/0000-0001-6705-4298 in the matching and/or analysis. Study drugs and comparators also differ, some studies consider individual GLP-1 RAs agents, and REFERENCES others take the whole class; agents vary according to the study 1. Diabetes mellitus: a major risk factor for cardiovascular disease. A period, any study included patients treated with semaglutide be- joint editorial statement by the American Diabetes Association; The National Heart, Lung and Blood Institute; The Juvenile Diabetes cause of the date of its marketing authorisation. The small sample HERRERA COMOGLIO and VIDAL GUITART | 17 of 18

Foundation International; The National Institute of Diabetes 19. Gerstein HC, Colhoun HM, Dagenais GR, et al.; REWIND Trial and Digestive and Kidney Diseases; and The American Heart Investigators. Design and baseline characteristics of participants in Association. Circulation. 1999;100(10):1132-1133. the Researching cardiovascular Events with a Weekly INcretin in 2. Buse JB, Ginsberg HN, Bakris GL, et al.; American Heart Association, Diabetes (REWIND) trial on the cardiovascular effects of dulaglu- American Diabetes Association. Primary prevention of cardio- tide. Diabetes Obes Metab. 2018;20(1):42-49. vascular diseases in people with diabetes mellitus. Circulation. 20. Husain M, Birkenfeld AL, Donsmark M, et al.; PIONEER 6 2007;115:114-126. Investigators. Oral semaglutide and cardiovascular outcomes in pa- 3. Lehrke M, Marx N. Diabetes mellitus and heart failure. Am J Cardiol. tients with type 2 diabetes. N Engl J Med. 2019;381(9):841-851. 2017;120(1S):S37-S47. 21. Schneeweiss S, Avorn J. A review of uses of health care utiliza- 4. Gilbert RE, Krum H. Heart failure in diabetes: effects of anti-hyper- tion databases for epidemiologic research on therapeutics. J Clin glycaemic drug therapy. Lancet. 2015;385(9982):2107-2117. Epidemiol. 2005;58(4):323-337. 5. Bejan-Angoulvant T, Cornu C, Archambault P, et al. Is HbA1c a valid 22. Food and Drug Administration Best Practices for Conducting and surrogate for macrovascular and microvascular complications in Reporting Pharmacoepidemiologic Safety Studies Using Electronic type 2 diabetes? Diabetes Metab. 2015;41(3):195-201. Healthcare Data. 2013. https://www.fda.gov/regul​atory​-infor​matio​ 6. Blind E, Janssen H, Dunder K, de Graeff PA. The European n/searc​h-fda-guida​nce-docum​ents/best-pract​ices-condu​cting​ Medicines Agency's approval of new medicines for type 2 diabetes. and-repor​ting-pharm​acoep​idemi​ologi​c-safet​y-studi​es-using​-elect​ Diabetes Obes Metab. 2018;20(9):2059-2063. ronic. Accessed June 15, 2015. 7. Nauck MA, Meier JJ, Cavender MA, Abd El Aziz M, Drucker DJ. 23. Najafzadeh M, Schneeweiss S. From trial to target populations - cal- Cardiovascular actions and clinical outcomes with glucagon-like ibrating real-world data. N Engl J Med. 2017;376(13):1203-1205. peptide-1 receptor agonists and dipeptidyl peptidase-4 inhibitors. 24. Patorno E, Patrick AR, Garry EM, et al. Observational studies of Circulation. 2017;136(9):849-870. the association between glucose-lowering medications and car- 8. Bonnet F. GLP-1 receptor agonist confer target organ protection in diovascular outcomes: addressing methodological limitations. type 2 diabetes. Diabetes Metab. 2017;1:2S1-2S2. Diabetologia. 2014;57(11):2237-2250. 9. Sikirica MV, Martin AA, Wood R, Leith A, Piercy J, Higgins V. Reasons 25. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of obser- for discontinuation of GLP1 receptor agonists: data from a re- vational studies in epidemiology: a proposal for reporting. Meta- al-world cross-sectional survey of physicians and their patients with analysis Of Observational Studies in Epidemiology (MOOSE) group. type 2 diabetes. Diabetes Metab Syndr Obes. 2017;29(10):403-412. JAMA. 2000;283(15):2008-2012. 10. Davies M, Pieber TR, Hartoft-Nielsen ML, Hansen OKH, Jabbour 26. Sterne JA, Hernán MA, Reeves BC, et al. ROBINS-I: a tool for as- S, Rosenstock J. Effect of oral semaglutide compared with pla- sessing risk of bias in non-randomised studies of interventions. cebo and subcutaneous semaglutide on glycemic control in pa- BMJ. 2016;12(355):i4919. tients with type 2 diabetes: a randomized clinical trial. JAMA. 27. Best JH, Hoogwerf BJ, Herman WH, et al. Risk of cardiovascular 2017;318(15):1460-1470. disease events in patients with type 2 diabetes prescribed the glu- 11. U.S. Department of Health and Human Services Food and Drug cagon-like peptide 1 (GLP-1) receptor agonist exenatide twice daily Administration Center for Drug Evaluation and Research. Guidance or other glucose-lowering therapies: a retrospective analysis of the for Industry. Diabetes Mellitus: Developing Drugs and Therapeutic LifeLink database. Diabetes Care. 2011;34(1):90-95. Biologics for Treatment and Prevention [Internet]. Center for Drug 28. Mogensen UM, Andersson C, Fosbøl EL, et al. Cardiovascular safety Evaluation and Research. 2008. http://www.fda.gov/downl​oads/ of combination therapies with incretin-based drugs and metformin Drugs​/Guida​nceCo​mplia​nceRe​gulat​oryIn​forma​tion/Guida​nces/ compared with a combination of metformin and sulphonylurea in UCM07​1624.pdf. Accessed June 22, 2018 type 2 diabetes mellitus–a retrospective nationwide study. Diabetes 12. http://www.ema.europa.eu/docs/en_GB/docum​ent_libra​ry/Scien​ Obes Metab. 2014;16(10):1001-1008. tific_guide​line/2012/06/WC500​129256.pdf. Accessed June 22, 29. Paul SK, Klein K, Maggs D, Best JH. The association of the treat- 2018 ment with glucagon-like peptide-1 receptor agonist exenatide or 13. U.S. Food and Drug Administration. Guidance for Industry. Diabetes insulin with cardiovascular outcomes in patients with type 2 dia- Mellitus Evaluating Cardiovascular Risk in New Antidiabetic betes: a retrospective observational study. Cardiovasc Diabetol. Therapies to Treat Type 2 Diabetes, 2008. www.fda.gov/downl​ 2015;24(14):10. oads/Drugs​/Guida​nceCo​mplia​nceRe​gulat​oryIn​forma​tion/Guida​ 30. Velez M, Peterson EL, Wells K, et al. Association of antidiabetic med- nces/ucm07​1627.pdf. Accessed June 22, 2018 ications targeting the glucagon-like peptide 1 pathway and heart 14. Pfeffer MA, Claggett B, Diaz R, et al. Lixisenatide in patients failure events in patients with diabetes. J Card Fail. 2015;21(1):2-8. with type 2 diabetes and acute coronary syndrome. N Engl J Med. 31. Patorno E, Everett BM, Goldfine AB, et al. Comparative cardiovas- 2015;373(23):2247-2257. cular safety of glucagon-like peptide-1 receptor agonists versus 15. Marso SP, Daniels GH, Brown-Frandsen K, et al. Liraglutide other antidiabetic drugs in routine care: a cohort study. Diabetes and cardiovascular outcomes in type 2 diabetes. N Engl J Med. Obes Metab. 2016;18(8):755-765. 2016;375(4):311-322. 32. Kannan S, Pantalone KM, Matsuda S, Wells BJ, Karafa M, 16. Marso SP, Bain SC, Consoli A, et al. Semaglutide and cardiovas- Zimmerman RS. Risk of overall mortality and cardiovascular events cular outcomes in patients with type 2 diabetes. N Engl J Med. in patients with type 2 diabetes on dual drug therapy including met- 2016;375(19):1834-1844. formin: A large database study from the Cleveland Clinic. J Diabetes. 17. Holman RR, Bethel MA, Mentz RJ, et al. Effects of once-weekly 2016;8(2):279-285. exenatide on cardiovascular outcomes in type 2 diabetes. N Engl J 33. Ekström N, Svensson AM, Miftaraj M, et al. Cardiovascular safety of Med. 2017;377(13):1228-1239. glucose-lowering agents as add-on medication to metformin treat- 18. Hernandez AF, Green JB, Janmohamed S, et al.; Harmony Outcomes ment in type 2 diabetes: report from the Swedish National Diabetes committees and investigators. Albiglutide and cardiovascular out- Register. Diabetes Obes Metab. 2016;18(10):990-998. comes in patients with type 2 diabetes and cardiovascular disease 34. Anyanwagu U, Mamza J, Mehta R, Donnelly R, Idris I. Cardiovascular (Harmony Outcomes): a double-blind, randomised placebo-con- events and all-cause mortality with insulin versus glucagon-like pep- trolled trial. Lancet. 2018;. pii: S0140–6736(18)32261-X. tide-1 analogue in type 2 diabetes. Heart. 2016;102(19):1581-1587. 18 of 18 | HERRERA COMOGLIO and VIDAL GUITART

35. Suissa S, Moodie EE, Dell'Aniello S. Prevalent new-user cohort de- peptide 1 agonists, and dipeptidyl peptidase 4 inhibitors with all- signs for comparative drug effect studies by time-conditional pro- cause mortality in patients with type 2 diabetes: a systematic re- pensity scores. Pharmacoepidemiol Drug Saf. 2017;26(4):459-468. view and meta-analysis. JAMA. 2018;319(15):1580-1591. 36. Zimmerman RS, Hobbs TM, Wells BJ, et al. Association of gluca- 50. Bethel MA, Patel RA, Merrill P, et al.; EXSCEL Study Group. gon-like peptide-1 receptor agonist use and rates of acute myocar- Cardiovascular outcomes with glucagon-like peptide-1 receptor dial infarction, stroke and overall mortality in patients with type 2 agonists in patients with type 2 diabetes: a meta-analysis. Lancet diabetes mellitus in a large integrated health system. Diabetes Obes Diabetes Endocrinol. 2018;6(2):105-113. Metab. 2017;19(11):1555-1561. 51. Zhang Z, Chen X, Lu P, et al. Incretin-based agents in type 2 diabetic 37. Toulis KA, Hanif W, Saravanan P, et al. All-cause mortality in patients patients at cardiovascular risk: compare the effect of GLP-1 ago- with diabetes under glucagon-like peptide-1 agonists: a popula- nists and DPP-4 inhibitors on cardiovascular and pancreatic out- tion-based, open cohort study. Diabetes Metab. 2017;43(3):211-216. comes. Cardiovasc Diabetol. 2017;16(1):31. 38. Anyanwagu U, Mamza J, Donnelly R, Idris I. Effect of adding GLP- 52. Monami M, Zannoni S, Pala L, et al. Effects of glucagon-like pep- 1RA on mortality, cardiovascular events, and metabolic outcomes tide-1 receptor agonists on mortality and cardiovascular events: a among insulin-treated patients with type 2 diabetes: a large retro- comprehensive meta-analysis of randomized controlled trials. Int J spective UK cohort study. Am Heart J. 2018;196:18-27. Cardiol. 2017;1(240):414-421. 39. Patorno E, Goldfine AB, Schneeweiss S, et al. Cardiovascular out- 53. Boye KS, Riddle MC, Gerstein HC, et al. Generalizability of gluca- comes associated with canagliflozin versus other non-gliflozin antidi- gon-like peptide-1 receptor agonist cardiovascular outcome trials to abetic drugs: population based cohort study. BMJ. 2018;6(360):k119. the overall type 2 diabetes population in the United States. Diabetes 40. Dawwas GK, Smith SM, Park H. Risk of heart failure hospitaliza- Obes Metab. 2019;21(6):1299-1304. tion among users of dipeptidyl peptidase-4 inhibitors compared 54. Bailey CJ, Marx N. Cardiovascular protection in type 2 diabe- to glucagon-like peptide-1 receptor agonists. Cardiovasc Diabetol. tes: insights from recent outcome trials. Diabetes Obes Metab. 2018;17(1):102. 2019;21(1):3-14. 41. Svanström H, Ueda P, Melbye M, et al. Use of liraglutide and risk of 55. Vergès B, Charbonnel B. After the LEADER trial and SUSTAIN-6, major cardiovascular events: a register-based cohort study in Denmark how do we explain the cardiovascular benefits of some GLP-1 re- and Sweden. Lancet Diabetes Endocrinol. 2018;7(2):106-114. ceptor agonists? Diabetes Metab. 2017;1:2S3-2S12 42. O’Brien MJ, Karam SL, Wallia A, et al. Association of second-line 56. Gejl M, Starup-Linde J, Scheel-Thomsen J, Gregersen S, Vestergaard antidiabetic medications with cardiovascular events among insured P. Risk of cardiovascular disease: the effects of diabetes and an- adults with type 2 diabetes. JAMA Netw Open. 2018;1(8):e186125. ti-diabetic drugs - a nested case-control study. Int J Cardiol. 43. Schneeweiss S. A basic study design for expedited safety signal 2015;15(178):292-296. evaluation based on electronic healthcare data. Pharmacoepidemiol 57. Li L, Li S, Liu J, et al. Glucagon-like peptide-1 receptor agonists and Drug Saf. 2010;19(8):858-868. heart failure in type 2 diabetes: systematic review and meta-anal- 44. Johnson ES, Bartman BA, Briesacher BA, et al. The incident user ysis of randomized and observational studies. BMC Cardiovasc design in comparative effectiveness research. Pharmacoepidemiol Disord. 2016;11(16):91. Drug Saf. 2013;22(1):1-6. 45. Ray WA. Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol. 2003;158(9):915-920. SUPPORTING INFORMATION 46. Kiri VA, MacKenzie G. How real is intention-to-treat (ITT) analysis Additional supporting information may be found online in the in non-interventional post authorization safety studies? We can do Supporting Information section. better. Curr Drug Saf. 2009;4(2):137-142. 47. Patorno E, Garry EM, Patrick AR, et al. Addressing limitations in observational studies of the association between glucose-low- ering medications and all-cause mortality: a review. Drug Saf. How to cite this article: Herrera Comoglio R, Vidal Guitart X. 2015;38(3):295-310. Cardiovascular outcomes, heart failure and mortality in type 48. Kristensen SL, Rørth R, Jhund PS, et al. Cardiovascular, mortality, and 2 diabetic patients treated with glucagon-like peptide 1 kidney outcomes with GLP-1 receptor agonists in patients with type receptor agonists (GLP-1 RAs): A systematic review and 2 diabetes: a systematic review and meta-analysis of cardiovascular meta-analysis of observational cohort studies. Int J Clin Pract. outcome trials. Lancet Diabetes Endocrinol. 2019;7(10):776-785. 49. Zheng SL, Roddick AJ, Aghar-Jaffar R, et al. association between 2020;74:e13553. https://doi.org/10.1111/ijcp.13553 use of sodium-glucose cotransporter 2 inhibitors, glucagon-like

IX. 2 Linagliptin and Cardiac Failure WHO Pharmaceuticals 2015 NEWSLETTER No.1

Prepared in collaboration with the WHO Collaborating Centre for International Drug Monitoring, Uppsala Sweden

The aim of the Newsletter is to The WHO Pharmaceuticals Newsletter provides you disseminate information on the safety and efficacy of with the latest information on the safety of medicines pharmaceutical products, based on and legal actions taken by regulatory authorities across communications received from our the world. It also provides signals based on information network of "drug information officers" and other sources such as derived from Individual Case Safety Reports (ICSRs) specialized bulletins and journals, available in the WHO Global ICSR database, as well as partners in WHO. VigiBase®.

The information is produced in the form of résumés in English, full texts of which may be obtained on request from: Safety and Vigilance, EMP-HIS, World Health Organization, 1211 Geneva 27, Switzerland, E-mail address: [email protected]

This Newsletter is also available on our Internet website: http://www.who.int/medicines

Further information on adverse reactions may be obtained from the WHO Collaborating Centre for Contents International Drug Monitoring Box 1051 751 40 Uppsala Regulatory matters Tel: +46-18-65.60.60 Fax: +46-18-65.60.80 E-mail: [email protected] Safety of medicines Internet: http://www.who-umc.org Signal

© World Health Organization 2015

All rights reserved. Publications of the World Health Organization can be obtained from WHO Press, World Health Organization, 20 Avenue Appia, 1211 Geneva 27, Switzerland (tel.: +41 22 791 3264; fax: +41 22 791 4857; e-mail: [email protected]). Requests for permission to reproduce or translate WHO publications – whether for sale or for non-commercial distribution – should be addressed to WHO Press, at the above address (fax: +41 22 791 4806; e-mail: [email protected]).

The designations employed and the presentation of the material in this publication do not imply the expression of any opinion whatsoever on the part of the World Health Organization concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Dotted lines on maps represent approximate border lines for which there may not yet be full agreement.

The mention of specific companies or of certain manufacturers’ products does not imply that they are endorsed or recommended by the World Health Organization in preference to others of a similar nature that are not mentioned. Errors and omissions excepted, the names of proprietary products are distinguished by initial capital letters.

All reasonable precautions have been taken by the World Health Organization to verify the information contained in this publication. However, the published material is being distributed without warranty of any kind, either expressed or implied. The responsibility for the interpretation and use of the material lies with the reader. In no event shall the World Health Organization be liable for damages arising from its use.

Printed by the WHO Document Production Services, Geneva, Switzerland

Table of Contents

Regulatory Matters Amoxicillin containing products ...... 4 Cabazitaxel acetonate ...... 4 Combined hormonal contraceptives ...... 4 Dimethyl fumarate ...... 5 Epoetin alfa ...... 5 Freeze-dried live attenuated mumps virus vaccine ...... 6 Galantamine hydrobromide ...... 6 Hydroxychloroquine or chloroquine ...... 6 Hydroxyethyl starch intravenous infusions ...... 7 Interferon beta products ...... 8 Ivabradine ...... 9 Levetiracetam ...... 9 Metoclopramide ...... 9 Simeprevir sodium ...... 10 Sodium-glucose co-transporter 2 inhibitors (ipragliflozin, dapagliflozin, tofogliflozin, luseogliflozin, canagliflozin and empagliflozin) ...... 11 Tofogliflozin hydrate ...... 11 Topiramate ...... 11 Ustekinumab ...... 12 Ziprasidone ...... 12

Safety of medicines Agomelatine ...... 13 Bromocriptine ...... 13 Hydrogen peroxide ...... 14 Ipilimumab ...... 14 Isotretinoin ...... 15 Nonsteroidal anti-inflammatory drugs (NSAIDs), opioids, and acetaminophen ...... 16 Valproate-containing medicines ...... 16 Vascular endothelial growth factor receptor inhibitors ...... 16

Signal Hexetidine and Severe hypersensitivity reactions ...... 18 Linagliptin and Cardiac failure ...... 20 Temsirolimus and Myocardial infarction ...... 29

WHO Pharmaceuticals Newsletter No. 1, 2015 • 3

SIGNAL

Discussion and Conclusion References The available information on hexetidine and E122, 1. MHRA, SPC for hexetidine (Oraldene). URL: the significant number of reports listing hexetidine http://www.mhra.gov.uk/home/groups/spcpil/ as only or only suspected drug, the range, severity documents/spcpil/con1384326269327.pdf and consistent pattern of adverse reactions Accessed: 17 July 2014. reported, the plausible time to onset and positive 2. Gell PGH, Coombs RRA, eds. Clinical aspects of dechallenge (and rechallenge in one case) support immunology. 1st ed. Oxford, England: the causal association between hexetidine Blackwell; 1963. preparations and severe hypersensitivity reactions. Three reports list co-administered medications 3. Luskin AT, Luskin SS, Anaphylaxis and also known to elicit allergic reactions; these might anaphylactoid reactions: Diagnosis and also have contributed to the reactions reported. management. Am J Ther. 1996 Jul;3(7):515- 20. Given the available information from VigiBase®, it is not possible to determine if the reactions were 4. Merk H, Ebert L, Goerz G, Allergic contact elicited by hexetidine itself or E122. Individuals dermatitis due to the fungicide hexetidine. with a history of salicylate intolerance seem to be Contact Dermatitis. 1982 May;8(3):216. at a higher risk of developing an allergic reaction to products containing azorubin. The information 5. Irish Medicines Board, SPC for hexetidine currently available to prescribers and patients (Oraldene). URL: https://www.hpra.ie/img/ does not list severe, potentially life-threatening uploaded/swedocuments/LicenseSPC_PA0823- hypersensitivity reactions adequately. 026-001_18062013164055. pdf Accessed: 17 July 2014.

6. Swissmedic, SPC for hexetidine (Hextril). URL: http://www.swissmedicinfo.ch/default.aspx Accessed 17 July 2014.

Linagliptin and Cardiac failure Dr Raquel Herrera Comoglio, Argentina

Dipeptidyl peptidase-4 inhibitors (DPP4i) are Summary expected to have beneficial effects on cardiac Linagliptin is a reversible, selective inhibitor of the outcomes, mainly through the prolonged effect of enzyme dipeptidyl peptidase-4 (DPP4), which is GLP-1. However, two large trials assessing the responsible for the metabolic inactivation of the impact of saxagliptin and alogliptin on incretin glucagon-like peptide 1 (GLP-1), thus cardiovascular death, non-fatal myocardial extending the GLP-1 half-life. GLP-1 acts on infarction and non-fatal stroke failed to show any glucose control by stimulating glucose-dependent beneficial effect of these drugs on the composite insulin secretion and suppressing glucagon of major cardiovascular outcomes. In addition, release. Linagliptin was approved in 2011 for concerns arose about the effect of saxagliptin and patients with type 2 diabetes mellitus as alogliptin on cardiac failure. A meta-analysis monotherapy or in combination with other suggested that cardiac failure could be a DPP4i antidiabetic agents in the United States of America class effect; if it is a class effect, its mechanisms (US), the European Union, in Australia and in are unknown. other countries. In spite of their inherent limitations, spontaneous Heart failure can be caused by structural or reports from VigiBase® add observational data in functional abnormalities of the heart. Up to 6 May support of the association cardiac failure - 2014, 15 ISCRs associating cardiac failure with linagliptin/ (DPP4i) as a drug-related effect in linagliptin had been received in the WHO Global some patients with risk factors (e.g. old age), ICSR Database, VigiBase®. All but two of these underlying concomitant conditions or pre-existing ISCRs were reported as serious, and there was cardiac failure and/or other concomitant one death reported. Linagliptin was reported as medication. The pair linagliptin-cardiac failure the only suspected drug in 13 cases. Age, reported should be considered as a signal, and deserves in nine cases, ranged from 60 to 88 years. Five further investigation. patients were 83 years and older.

WHO Pharmaceuticals Newsletter No. 1, 2015 • 20

SIGNAL

Introduction (Table 1). All but one (a case from literature) were Linagliptin is a synthetic, reversible inhibitor of spontaneous reports. All but two cases were dipeptidyl peptidase-4 (DPP4). Acting through serious (one case doesn’t state serious- ness, inhibition of DPP4, linagliptin inhibits the another one was reported as not serious), and proteolytic degradation of the incretin hormones there was one death reported. The patients were glucagon-like peptide 1 (GLP-1) and glucose- 11 women (73%) and three men (20%), while the dependent insulinotropic peptide (GIP), resulting literature case doesn’t report the patient’s sex. in increased glucose dependent insulin secretion. Age was reported in nine cases (60%), being all Linagliptin is indicated in the treatment of type 2 patients 60 years or older; age ranged from 83 to diabetes mellitus (T2DM) to improve glycaemic 88 years in four out of nine patients (44%). control in adults, as monotherapy or as Another patient’s age (case 11) was estimated to combination therapy with metformin, or 85-90 years. sulphonylurea and metformin, or in combination Relevant medical history was reported in six cases 1-3 with insulin with or without metformin. (cases 8, 9, 10, 11, 12 and 13). In case 8, Cardiac failure, or heart failure (HF), is a concomitant diseases included renal artery pathophysiologically complex clinical syndrome, stenosis, renal insufficiency, renal disease not fully understood, which results from an (shrinked kidneys), renal anaemia, renal impaired function of the heart as a pump osteodystrophy, hyperuricaemia, dyslipidaemia, supporting physiological circulation. Symptoms are hypertension, and chronic obstructive pulmonary dyspnoea, exercise intolerance, and sodium and diseases (COPD). In case 9, relevant medical water retention, often manifested as oedema. history mentions cardiovascular disease (CVD) Cardiac failure can be caused by any abnormality (percutaneous coronary intervention, angina of the structure, mechanical function or electrical pectoris and myocardial ischaemia), hypertension, activity of the heart, or as a secondary dysfunction hyperlipidaemia, chronic renal failure and diabetes of other organs and tissues, e.g. kidneys, liver or mellitus. Of note, both patients had low weight (43 muscles; other systemic processes, as kg and 40 kg). Case 10 reports hypertension neurohumoral activation, are also involved. 4,5 pulmonale, cor pulmonale and lung fibrosis. Case 11 reports unspecified heart disease, case 12 Receptors of GLP-1 (GLP-1R) are expressed in mentions hypertension and atrial fibrillation and pancreas and extrapancreatic tissues (lung, case 13 renal insufficiency and hypertension. kidney, central, enteric and peripheral nervous systems, lymphocytes, blood vessels, and heart). Reports come from Europe (six cases), the US GLP-1 exerts direct actions on the cardiovascular (three), Canada (two), Australia (two) and Japan system, the heart, vessels and kidney, mainly via (one). ICSRs were sent by physicians (ten cases), GLP-1R. In preclinical studies, incretin-based manufacturers (two), pharmacist (one) agents control body weight, improve glycaemic consumers/non health professionals (one case) control with a low risk of hypoglycaemia, decrease and other health professionals (one case). The blood pressure, inhibit the secretion of intestinal completeness score of the ICSRs was low (0.17- chylomicrons, and reduce inflammation. 6 0.27) in 31%, medium (0.33-0.53) in 44% and high (0.75-0.95) in 25% of the reports. One report GIP and GLP-1 are rapidly inactivated by DPP4, 1- mentions a recently published article about the 3, 6-8 a transmembrane protein that removes N- effect of saxagliptin on cardiac outcomes. terminal dipeptides from various substrate hormones, chemokines, neuropeptides, growth Seven reports provide time to onset, which varied factors and incretins. Other cardioactive peptides from 6 days (two reports) to 295 days. cleaved by DPP4, are brain natriuretic peptide In all cases but two, linagliptin is the only (BNP) and neuropeptide Y. BNP is a cardiac suspected drug. In the case from literature, neurohormone with natriuretic and vasodilatory metformin 2 g/day (part of the investigational actions, secreted into the plasma from the product) was also suspected. In case 9, cilostazol ventricles in response to ventricular volume was started nine days before the adverse event 6-8 expansion and pressure overload. BNP has been and was also mentioned as a suspected drug. established as a diagnostic and prognostic marker of left ventricular (LV) systolic and diastolic Co-administered medication was reported in nine dysfunction. 9 BNP plasma levels have been shown cases, although it was not specified in one case. to be significantly higher in patients with Three cases included calcium-channel blockers decompensated chronic HF. 10 (amlodipine in two cases and nifedipine in one). T2DM medication was reported in six cases, metformin was the only antidiabetic drug concomitantly administrated in four (in the case Reports in VigiBase® from a clinical trial it was reported as a suspected Fifteen ICSRs were retrieved from the WHO Global drug), there was one case with metformin and ICSR Database, VigiBase®, up to 6 May 2014 pioglitazone, and another with insulin.

WHO Pharmaceuticals Newsletter No. 1, 2015 • 21

SIGNAL

Dechallenge action was reported in nine cases (SAEs) in the same patient on day 295 after (60%): linagliptin was reported as withdrawn in starting the study treatment (2.5 mg linagliptin + eight of these, and in one linagliptin dose was 1000 mg metformin). The patient experienced reported as not changed, the patient recovered cardiogenic shock and supraventricular tachycardia with sequelae. within one hour of administration of medication. The patient required hospitalisation and study The fatal case (case 8) refers to a 66 year-old medication was discontinued; the patient woman with renal insufficiency (renal artery recovered. 12 stenosis, renal anaemia, renal dystrophia), hyperuricemia, dyslipemia, hypertension and COPD, who presented with acute dyspnoea and cardiac decompensation 16 days after having Literature and Labelling started therapy with linagliptin. The patient died The EMA Summary of Product Characteristics for 15 days later, and linagliptin was considered linagliptin, the Australian Public Assessment report “implied”. and the product label for linagliptin (US FDA) do The first spontaneous reporting (case 1) referred not mention heart/cardiac failure as an event to a woman presenting with cardiac failure 44 associated with or described for linagliptin 1-3 days after having started her treatment with therapy. linagliptin 5 mg. All the other adverse events In a published clinical trial report, cardiac failure is (urinary infection, pulmonary infection) occurred mentioned as having occurred in a patient after at least 11 days after the HF onset, and were 295 days of linagliptin 5 mg/metformin 2 g probably related to complications derived from treatment. 12 This case has been retrieved in hospitalization. Linagliptin was withdrawn after the VigiBase® and already mentioned in the “Reports patient had recovered from her cardiac failure, and in VigiBase®” section. hypothetically, after hospitalization. A meta-analysis, published in February 2013, In the case reported by a manufacturer (case 2), a found that treatment with DPP4i reduces the risk female patient was also treated with pioglitazone, of cardiovascular events (particularly myocardial amlodipine, metformin and irbesartan. Case 9 infarction) and all-cause mortality in patients with refers to a Japanese 87 year-old woman, weight type 2 diabetes. Although HF was a pre-specified 40 kg, treated with several drugs including component of major cardiovascular events nifedipine for hypertension, hyperlipidaemia, (MACE), meta-analysis’s results don’t mention chronic renal failure and myocardial ischaemia. HF. 13 A trial in older patients does not mention The patient presented with congestive HF and cardiac failure among adverse events. 14 Results aggravated renal failure six days after having posted on Clinicaltrials.gov mention only rhythm started linagliptin 5 mg/day and nine days after abnormalities and coronary artery diseases as start of cilostazol treatment with 200 mg/day. In serious cardiac adverse events. 15 case 10, a 67 year-old woman with a history of scleroderma, lung fibrosis and pulmonary A meta-analysis of 50 DPP4-inhibitors trials, hypertension presented with atrial fibrillation and enrolling 55,141 participants, found a statistically cardiac failure after approximately six months in significant trend towards increased risk of HF treatment with linagliptin. The reporter, a outcomes with no increase in risk with regards to physician, mentions an article of Scirica et al., all-cause mortality, cardiovascular mortality, acute published the previous month. 11 coronary syndrome (ACS) or stroke. Most of the HF cases were retrieved from results of SAVOR- In case 12, an 88 year-old woman who had TIMI-53 (saxagliptin, 66.2% of the data of HF), started linagliptin 45 days earlier, went to a EXAMINE (alogliptin, 21.3%) and the VIVIDD trial hospital for routine pacemaker battery (vildagliptin, 6.9%), the latter enrolled only replacement, and cardiac insufficiency was patients with left ventricular fraction <40%. Seven detected; the patient lost 13 kg with appropriate clinical trials for linagliptin with 5,260 participants therapy (this indicates the amount of fluid were included in the analysis. 16 retention), and her ejection fraction was 45%. A meta-analysis of clinical trials with vildagliptin, Case 13 reports that an 83 year-old man with sitagliptin, saxagliptin, alogliptin, linagliptin, and renal insufficiency and hypertension presented dutogliptin found that the overall risk of acute HF with cardiac failure. Linagliptin was withdrawn, but was higher in patients treated with DPP4i in there is no information about the outcome. comparison with those treated with placebo/active Concomitant poly-medication is mentioned but comparators, and suggested that DPP4i could be details are not provided. associated with an increased risk of HF.17 The case from literature (case 14) was extracted A recently published analysis of pooled data of 22 from a 52 weeks multifactorial design study, and placebo-controlled trials found a negative reports two drug related serious adverse events relationship, with an incidence of HF adverse

WHO Pharmaceuticals Newsletter No. 1, 2015 • 22

SIGNAL events (AEs) for linagliptin- and placebo-treated respectively nine days and five days before the HF patients of 0.2% (n = 11) for linagliptin and 0.3% onset. Cilostazol is a reversible phosphodiesterase (n = 7) for placebo. 18 III inhibitor with anti-platelet, vasodilatory and antithrombotic properties, metabolized by CYP3A4 Discussion and Conclusion and CYP2C19. Cilostazol is formally Ageing, dyslipidaemia, hypertension, renal contraindicated in patients with pre-existing HF. insufficiency and diabetes are risk factors for Nifedipine is metabolized by CYP2C19 (interaction developing cardiac failure. As stated, all patients described in product information). with a reported age were 60 years or older: four For the patient treated with pioglitazone/ ISCRs refer to patients 83 year-old and older, and metformin (case 2), no time to onset was for the other five the patient’s age was between reported. Pioglitazone is a thiazolidinedione, which 60 to 71 years. Another patient’s age was selectively ligands the nuclear transcription factor estimated to 85-90 years. peroxisome proliferator-activated receptor-γ Co-morbidities predisposing to HF (dyslipidaemia, (PPAR-γ). Thiazolidinediones improve glycaemic hypertension, prior cardiac disease, renal control by increasing insulin sensitivity. Fluid insufficiency) are mentioned in six cases. retention, that can cause or exacerbate HF in some patients, is a known effect of PPAR-γ, and Concomitant medication was reported in nine pioglitazone can cause or exacerbate congestive cases (although not specified in one case). HF in some patients. 21 Linagliptin was the only suspected drug in thirteen ISCRs, in two cases another drug was reported as In eight cases, linagliptin was withdrawn. Except suspected (cilostazol and metformin, respectively). for the fatal case and one with unknown outcome, In three cases (cases 7, 8, and 9), cholesterol dechallenge was positive. No cases with lowering agents, i.e. statins or ezetimibe or both, rechallenge were reported. In case 3, the dose are reported among concomitant medication. was not changed and the patient recovered with Three patients (3, 8 and 9) were under diuretic hospitalization. treatment (furosemide/amiloride, Time to onset (reported in seven cases) was <6 hydrochlorotiazide and furosemide respectively). months in six patients (44 days, 6 days, 147 days, Two patients (7 and 9) were treated with 16 days, 6 days and 164 days respectively). In a angiotensin converting enzyme inhibitors large study with saxagliptin, the risk for HF (quinapril and enalapril), two other patients with hospitalization associated with the use of angiotensin receptors antagonists (2 and 9, saxagliptin was highest in the first six months and irbesartan and losartan respectively). Three declined thereafter. patients were treated with calcium channel blockers and two patients with doxazocin ( α- DDP4 inhibitors have been expected to have blocker). beneficial effects on cardiac outcomes, both due to GLP-1 actions and to other peptide hormones with Calcium channel blockers can lead to worsening HF direct cardiorenal effects. Preclinical data and and have been associated with an increased risk of mechanistic studies suggested a possible cardiovascular events, especially the non additional non-glycaemic beneficial action on blood 5 vasoselective ones. In three cases, patients were vessels and the heart, via both GLP-1 dependent also treated with calcium channel blockers, two and GLP-1-independent effects. 22 It has been with amlodipine and one patient with nifedipine. suggested that DPP4 inhibitors reduce the risk for Time to onset was reported in two out of these the multiple co-morbidities associated with three cases (6 days and 16 days). obesity/T2DM including hypertension, 23 In case 8 (fatal outcome), a 66 year-old woman cardiovascular and kidney disease. presented with HF 16 days after having started linagliptin. She was also treated with amlodipine 10 mg, quinapril 5 mg, bisoprolol 5 mg, gliquidone 30 mg, and simvastatin 20 mg. Recent studies have found that amlodipine does not exert favourable effects on the clinical course of patients with HF;19,20 other not significant interactions for HF seem unlikely, such as simvastatin with amlodipine (amlodipine increases the systemic exposure of simvastatin, this patient being on amlodipine low-dose). In case 9 (an 87 year-old woman treated with nifedipine and other cardiovascular therapies), cilostazol and linagliptin 5 mg had been added

WHO Pharmaceuticals Newsletter No. 1, 2015 • 23

SIGNAL Table 1. Characteristics of reports in VigiBase ® for linagliptin and cardiac failure

Case Age/ Time to Duration of Other suspected (S) or Reported adverse reactions/ Dechallenge Outcome Sex onset treatment concomitant (C) drugs adverse events action 1 -/F 44 days 68 days - Urinary tract infection, Drug withdrawn Recovered pulmonary congestion, (before drug hypotension, pain, nausea, withdrawal) pleural fibrosis, hypoxia, cardiac congestive failure 2* -/F - - Pioglitazone, metformin, Cardiac failure - Unknown amlodipine, irbesartan (all C) 3 84/M 6 days - Metformin, clarithromycin, Oedema, respiratory rate Dose not Recovered doxazocin, fluticasone, increased, wheezes, changed with sequelae lactulose, senna, latanoprost, orthopnoea, condition furosemide/amiloride, aggravated, congestive heart clopidogrel (allC) failure 4 71/F - - - Heart failure - - 5 -/F 147 days 147 days - Anaemia, congestive heart Drug withdrawn Recovered failure, hypertension with sequelae pulmonary 6 60/F - - - Congestive heart failure - - 7* 69/F - - - Dyspnoea, swelling, cardiac - - failure 8** 66/F 16 days 16 days Amlodipine, quinapril , Vomiting, renal failure acute, Drug withdrawn Death colecalciferol, sodium acute dyspnoea, myocardial ascorbate/ferrous sulfate, decompensation, general calcium acetate, calcium physical health deterioration carbonate, sodium bicarbonate, cloxazolam, acetylsalicylic acid, bisoprolol, gliquidone , simvastatin (all C) 9 87/F 6 days 6 days Cilostazol (S) Renal failure aggravated, Drug withdrawn Recovered cardiac failure congestive Zopiclone doxazocin, haloperidol, ezetimibe, famotidine, nifedipine, hydrochlorotiazide/losartan potassium, acetylsalicylic acid (all C) 10 67/F 164 days 176 days Estradiol, terbutaline, formoterol Disease progression, oedema, Drug withdrawn Recovered fumarate/ budesonide, cardiac arrest, atrial prednisolone, insulin aspart, fibrillation, myocardial insulin glargine, acetylsalicylic decompensation acid, magnesium oxide, simvastatin, ezetimibe, enalapril, potassium, furosemide, sildenafil, cyclophosmamide (all C)

11 ***/F - - Metformin (C) Myocardial decompensation Drug withdrawn Recovered 12 88/F - 45 days Metformin (C) Oedema, cardiac failure - Recovered 13 83/M - - Polymedication (not further Cardiac failure Drug withdrawn Unknown specified) 14 -/- 295 days 295 days Metformin (S) Supraventricular tachycardia, Drug withdrawn Recovered cardiogenic shock 15 -/M - - - Asthenia, hyperglycaemia, Unknown Unknown tremor, nocturnal dyspnoea, abdominal discomfort *causality reported as possible, ** causality reported as ‘implied’, *** estimated age: 85-90 years

WHO Pharmaceuticals Newsletter No. 1, 2015 • 24

SIGNAL

However, a recently published meta-analysis Linagliptin also has significant inhibitory activity on suggested that HF is a class effect of DPP4i. 17 Two the human M1, M2 and M3 muscarinic receptors, large trials specifically designed to assess with half minimal inhibitory concentration (IC 50 ) composite cardiovascular outcomes contributed values of 295 to 1000 nM (more than 22 times the largely to this conclusion. A large trial found a clinical peak concentration (Cmax), which has not higher statistically significant risk of hospitalization been considered clinically relevant. 1,2 In animal due to HF in the saxagliptin group than in the models a relationship between B-adrenergic and placebo group. Patients had T2DM and established M2 muscarinic receptors and diminished cardiovascular disease or multiple cardiovascular ventricular contractility has been suggested. 30 risks factors, and were followed for a median of In spite of their inherent limitations, spontaneous 2.1 years; HF was included in a composite reports from VigiBase® add observational data in secondary endpoint. 12.8% of participants had support of the association cardiac failure- prior HF.11 The events were more frequent in linagliptin/ (DPP4i) as a drug-related effect in patients with diabetes and HF. 24 In a large trial some patients with risk factors (e.g. old age), assessing the effect of alogliptin on cardiovascular underlying concomitant conditions or pre-existing outcomes in 5,380 patients followed for a median cardiac failure and/or other concomitant of 18 months, 28% of the participants had HF at medication. The pair linagliptin-cardiac failure baseline; HF was not part of the primary should be considered as a signal, and deserves composite outcome or secondary outcomes. further clinical and pharmacoepidemiological Although the heterogeneity of sub-groups is investigation. mentioned, no specific details are provided. 25

Prior HF was the strongest predictor of hospitalization during the study, followed by impaired kidney function. In the EXAMINE study, References the risk increase for hospitalization due to HF 1. Australian Public Assessment Report for associated with alogliptin was apparently less clear Linagliptin. Department of Health and Ageing, in spite of the percentage of patients with prior Therapeutic Goods Administration (TGA). HF, the higher use of ß-blocking agents, and the December 2011. more frequent medical controls with treatment 2. Summary of Product Characteristics for adaptations in EXAMINE might be one of the Linagliptin. European Medicines Agency. URL: potential explanations. 26 http://www.ema. europa.eu/ema/… Mechanisms of the hypothesized effect of DDP4 Accessed: June 2014 inhibitors on HF are unknown. As previously 3. Product Label for linagliptin (Tradjenta®). US mentioned, DPP4 cleaves not only GLP-1 and GIP, Food & Drug Administration. URL: but also other cardioactive peptides, such as http://www. substance P, brain natriuretic peptide (BNP), accessdata.fda.gov/scripts/cder/drugsatfda/in neuropeptide Y, CXCL12, bradykinin, and related dex.cfm?fuseaction=Search.Search_Drug_Na peptides. 6-7 BNP is increased in HF, being both a me. Accessed: June 2014. diagnostic and prognostic marker. 9,10 It has also been suggested that DPP4 is abnormally increased 4. Braunwald E. Heart failure. JACC Heart Fail. in patients with T2DM and these increased DPP4 2013 Feb;1(1):1-20. doi: levels are independently associated with 10.1016/j.jchf.2012.10.002. Epub 2013 Feb asymptomatic left ventricular both diastolic and 4. systolic dysfunction in T2DM patients which have a 5. Jessup M, Abraham WT, Casey DE, Feldman higher risk of presenting left ventricular AM, Francis GS, Ganiats TG et al. 2009 focu- dysfunction. 27 Neuropeptide Y (NPY1-36) is sed update: ACCF/AHA Guidelines for the released from sympathetic neurons; DPP4 diagnosis and Management of Heart Failure in removes the N-terminal from NPY1-36 to generate Adults: a report of the American College of NPY3-36, which binds to Y2 receptors that have Cardiology Foundation/American Heart relative antagonist properties to Y1 receptor Association Task Force on Practice Guidelines: activation. Any decrease in the DPP4 mediated developed in collaboration with the generation of NPY3-36 would decrease the activity International Society for Heart and Lung of Y2 inhibitory autoreceptors; and so augment Transplantation. Circulation. 2009 Apr sympathetic and parasympathetic 14;119(14):1977-2016. neurotransmitter release. 28 A clinical study in 53 patients found that peptide Y is augmented in 6. Ussher JR, Drucker DJ. Cardiovascular actions diabetic patients. 29 of incretin-based therapies. Circulation research. 2014; 114:1788-1803.

WHO Pharmaceuticals Newsletter No. 1, 2015 • 25

SIGNAL

7. Ussher JR, Drucker DJ. Cardiovascular biology Comprehensive Pooled Analysis of 22 Placebo- of the incretin system. Endocr Rev. 2012 Controlled Studies. Clin Ther. 2014 Jul 8. pii: Apr;33(2):187-215. S0149-2918(14)00371-3. 8. Drucker DJ. Dipeptidyl peptidase-4 inhibition 19. Lee SA, Choi HM, Park HJ, Ko SK, Lee HY. and the treatment of type 2 diabetes: Amlodipine and cardiovascular outcomes in preclinical bio- logy and mechanism of action. hypertensive patients: meta-analysis Diabetes care. 2007 Jun;30(6):1335-43. comparing amlodipine-based versus other antihypertensive therapy. Korean J Intern 9. Braunwald E. Biomarkers in heart failure. N Med. 2014 May;29(3):315-24. Engl J Med. 2008 May 15;358(20):2148-59. 20. Packer M, Carson P, Elkayam U, Konstam MA, 10. Gong H, Wang X, Ling Y, Shi Y, Shi H. Moe G, O’Connor C et al. Effect of amlodipine Prognostic value of brain natriuretic peptide in on the survival of patients with severe chronic patients with heart failure and reserved left heart failure due to a nonischemic ventricular systolic function. Exp Ther Med. cardiomyopathy: results of the PRAISE-2 2014 Jun;7(6):1506-1512. study (prospective randomized amlodipine 11. Scirica BM, Bhatt DL, Braunwald E, Steg PG, survival evaluation 2). JACC Heart Fail. 2013 Davidson J, Hirshberg B et al. Cardiovascular Aug;1(4):308-14. out- comes in patients with type 2 diabetes 21. Information for Healthcare Professionals: mellitus. N Engl J Med. 2013 Oct Pioglitazone HCl. URL: 3;369(14):1317-26. http://www.fda.gov/Drugs/DrugSafety/ost- 12. Haak T, Meinicke T, Jones R, Weber S, von marketDrugSafetyInformationforPatientsandP Eynatten M, Woerle HJ. Initial combination of roviders/ ucm124178.htm. Accessed: 2014. linagliptin and metformin in patients with type 22. Scheen AJ. Cardiovascular effects of 2 diabetes: efficacy and safety in a dipeptidyl peptidase-4 inhibitors: from risk randomised, double-blind 1-year extension factors to clinical outcomes. Postgrad Med. study. Int J Clin Pract. 2013 2013 May;125(3):7-20. Dec;67(12):1283-93. 9. 23. Aroor AR, Sowers JR, Jia G, DeMarco VG. 13. Monami M, Ahrén B, Dicembrini I, Mannucci E. Pleiotropic Effects of the Dipeptidylpeptidase- Dipeptidyl peptidase-4 inhibitors and 4 Inhibitors on the Cardiovascular System. cardiovascular risk: a meta-analysis of Am J Physiol Heart Circ Physiol. 2014 Jun 13 randomized clinical trials. Diabetes Obes Metab. 2013 Feb;15(2):112-20. 24. Scirica BM, Raz I, Cavender MA, Steg PG, Hirshberg B, Davidson J et al. Abstract 14. Barnett AH, Huisman H, Jones R, von 17503: Outcomes of Patients With Type 2 Eynatten M, Patel S, Woerle HJ. Linagliptin for Diabetes and Known Congestive Heart Failure patients aged 70 years or older with type 2 Treated With Saxagliptin: Analyses of the diabetes inadequately controlled with SAVORTIMI 53 Study. Cardiovasc Ther. 2014 common antidiabetes treatments: a Apr 21. randomised, double-blind, placebo-controlled trial. Lancet. 2013 Oct 26;382(9902):1413- 25. White WB, Cannon CP, Heller SR, Nissen SE, 23. Bergenstal RM, Bakris GL, et al. Alogliptin after acute coronary syndrome in patients 15. Clinicaltrials.gov (identifier NCT01084005). with type 2 diabetes. N Engl J Med. 2013 Oct URL: http://www.clinicaltrials.gov. Accessed: 3;369(14):1327-35. June 2014. 26. Schernthaner G, Sattar N. Lessons from 16. Wu S, Hopper I, Skiba M, Krum H. Dipeptidyl SAVOR and EXAMINE: Some important peptidase-4 inhibitors and cardiovascular answers, but many open questions. J outcomes: Meta-analysis of randomized Diabetes Complications. 2014 Jul- clinical trials with 55,141 participants. Aug;28(4):430-3. doi: 10.1016/j. Cardiovasc Ther. 2014 Aug;32(4):147-58. jdiacomp.2014.02.011. 17. Monami M, Dicembrini I, Mannucci E. 27. Ravassa S, Barba J, Coma-Canella I, Huerta Dipeptidyl peptidase-4 inhibitors and heart A, López B, González A, et al. The activity of failure: A meta-analysis of randomized clinical circulating dipeptidyl peptidase-4 is trials. Nutr Metab Cardiovasc Dis. 2014 associated with subclinical left ventricular Jul;24(7):689-97. dysfunction in patients with type 2 diabetes 18. Lehrke M, Marx N, Patel S, Seck T, Crowe S, mellitus. Cardiovasc Diabetol. 2013 Oct Cheng K et al. Safety and Tolerability of 7;12:143. Linagliptin in Patients With Type 2 Diabetes: A

WHO Pharmaceuticals Newsletter No. 1, 2015 • 26

SIGNAL

28. Baraniuk JN, Jamieson MJ. Rhinorrhea, cough A dechallenge is described in 28 reports: 12 and fatigue in patients taking sitagliptin. positive, 7 negative and 2 unknown outcomes. Allergy, Asthma & Clinical Immunology. Four rechallenges are described: 2 negative, one 2010;6(1):8- unknown outcome. Only one positive rechallenge was described, in a patient with end stage renal 29. Matyal R, Mahmood F, Robich M, Glazer H, and chronic cardiac failure with critical coronary Khabbaz K, Hess P, et al. Chronic type II occlusion. In 4 reports the patient recovered while diabetes mellitus leads to changes in linagliptin continued unchanged. Six reports were neuropeptide Y receptor expression and not temporally related to linagliptin; one patient distribution in human myocardial tissue. Eur J had viral cardiomyopathy 5 months after Pharmacol. 2011 Aug 31;665(1-3):19-28. discontinuation, another with pre-existing chronic 30. Kashihara T, Hirose M, Shimojo H, Nakada T, renal failure experienced HF 3 months after Gomi S, Hongo M, et al. β(2)-Adrenergic and discontinuation, 3 had evidence of worsening of M(2)-muscarinic receptors decrease basal t- pre-existing HF prior to starting linagliptin and one tubular L-type Ca2+ channel activity and event occurred on the day that linagliptin was suppress ventricular contractility in heart started, when hyperthyroidism was also failure. Eur J Pharmacol. 2014 Feb 5 diagnosed. ;724:122-31. Time to onset was reported in 71 of the 92 events. Thirty eight events (54%) were <6 months, and 11 (15%) events occurred more than 1 year after Response from Boehringer starting linagliptin. Ingelheim Of the 15 ICSRs identified on the Vigibase database, 14 are included in the search results. Thank you for the opportunity to provide our Case 15 is not because the Preferred Term comments. Boehringer Ingelheim completed a Nocturnal dyspnoea is not in the narrow cardiac cumulative assessment of HF on 14 February 2014 failure SMQ definition. In addition to the following an internally detected signal after information presented, case 1 reported a patient in publication of the results for the saxagliptin and a Boehringer Ingelheim sponsored trial in patients alogliptin cardiovascular outcome trials (CVOT) in with renal impairment. The patient was taking September 2013. furosemide for congestive heart failure (CHF) and A cumulative review of linagliptin and spironolactone and metolazone was added 3 linagliptin+metformin fixed dose combination months before linagliptin was started indicating (FDC) reports entered in the Boehringer Ingelheim worsening or instability of the disease prior to global drug safety database until 27AUG2014 was linagliptin. undertaken using the narrow MedDRA v17.0 SMQ The evaluation of HF is complex in the type 2 cardiac failure. Data from clinical trials, Diabetes Mellitus (T2DM) population. As stated in epidemiologic studies and the published literature the article above, the risk factors for HF would be were also reviewed. expected to be higher in the T2DM population 92 events (87 serious) in 83 reports were compared with non- diabetics. An assessment was identified in association with linagliptin; no reports conducted of past or concomitant conditions that were identified in association with the are known risk factors for HF. Seventy seven linagliptin+metformin FDC. Forty-six events in 42 patients (93%) had at least one risk factor and 55 reports were from clinical trials, 32 events in 28 patients (66%) had 3 or more risk factors. Six reports were spontaneous, 3 events in 3 reports patients did not have at least one risk factor for HF were from Health Authorities and 11 events in 10 and all 6 reports had limited or no medical history reports were from observational studies. The main provided. reason for seriousness was hospitalization alone or Hypertension may be the single most important associated with life-threatening or fatal reports. modifiable risk factor for HF. 1 Fifty eight patients Forty six patients were female (55%), 35 were (70%) reported hypertension. male (42%) and 2 did not report gender. Obesity and insulin resistance are also important The 78 reports which provide age range between risk factors for cardiac failure. 1 Using the 39 and 92 years with most (75%) over 60. Twenty international classification of body mass index one patients (25%) were older than 80. (BMI) 2, of 54 patients reporting a BMI, 12 (22%) All of the reports were individually assessed and were pre-obese and 27 (50%) were obese. Only had a plausible alternative explanation, factors 10 reports described patients with normal BMI and (e.g. temporality) that make a causal relationship the 5 underweight patients had renal failure. to linagliptin unlikely or insufficient information for Linagliptin does not require dose alteration in a more full causal assessment. patients with renal insufficiency and may be more

WHO Pharmaceuticals Newsletter No. 1, 2015 • 27

SIGNAL likely to be used in this patient population than References other DPP-4 inhibitors. Thirty six patients (43%) 1. Yancy CW, Jessup M, Bozkurt B, Butler J, reported renal insufficiency, acute or chronic renal Casey Jr DE, Drazner MH, et al. 2013 failure or diabetic nephropathy which are ACCF/AHA Guideline for the Management of 1 recognised risk factors for HF. Heart Failure: A Report of the American The reported incidence rate of HF in patients with College of Cardiology Foundation/American T2DM varies across studies largely reflecting Heart Association Task Force on Practice differences in ascertainment and adjustment Guidelines. Circulation. 2013;128:e240-e327. 3 approaches. A study conducted in multiple 2. http://apps.who.int/bmi/index.jsp?introPage= countries across Europe and North Africa intro_3. html Accessed 8 September 2014. estimated an annual incidence rate of CHF requiring hospitalisation of 10 per 1000 persons.4 3. Roger VL. Epidemiology of heart failure. Circ In the US, the CDC reported the annual age- Res 2013. 113(6):646-659. adjusted hospital discharge rate with HF as first- 4. Vaur L, Gueret P, Lievre M, Chabaud S, Passa listed diagnosis in diabetes patients to be 13.4 per P, DIABHYCAR Study Group. Development of 1000 in 2006. 5 Another US study estimated a 6 congestive heart failure in type 2 diabetic crude incidence rate of about 11.8 per 1000 PY. patients with microalbuminuria or proteinuria: The incidence rate of patients with unadjudicated observations from the DIABHYCAR (type 2 narrow MedDRA SMQ cardiac failure events for DIABetes, Hypertension, CArdiovascular linagliptin is 5.8 per 1000PY (N=9060). 7 Pooled Events and Ramipril) study. Diabetes Care analysis of safety data from 23 randomized clinical 2003. 26(3):855-860. trials (N=5488 linagliptin, 3290 placebo) showed 5. Centres for Disease Control and Prevention same overall incidence of cardiac disorders (3.3% 7 Diabetes Public Health Resource. and 3.3%, respectively). Using the narrow http://www.cdc.gov/diabetes/ statistics/ MedDRA SMQ cardiac failure, the frequency was 8 cvdhosp/hf/fig3.htm Accessed 10 September 0.5% (linagliptin) and 0.2% (placebo). This 2014. frequency in patients with a history of cardiac failure was 5.1% (linagliptin) and 5.5% 6. Kanaya AM, Adler N, Moffet HH, Liu J, (placebo). 7 External adjudication of events of Schillinger D, Adams A, Ahmed AT, Karter AJ. hospitalization for HF in 8 randomized, double- Heterogeneity of diabetes outcomes among blind studies, (N=2039 linagliptin, 1275 placebo) Asians and Pacific Islanders in the US: the showed 9 (0.4%) and 5 (0.4%) patients Diabetes Study of Northern California respectively were adjudicated to have HF. Further (DISTANCE). Diabetes Care 2011. 34(4):930- adjudicated analyses will be available when the 937. large ongoing CVOTs CAROLINA and CARMELINA 7. Internal Boehringer Ingelheim data. complete. 8. Schernthaner G, Khunti K, Patel S, Cheng K, It is not possible to demonstrate a direct causal Mattheus M, Woerle HJ. Safety of linagliptin in effect of linagliptin with HF due to the confounding 8778 patients with type 2 diabetes mellitus: of the reports with risk factors for the condition pooled analysis of 23 placebo-controlled and the known relatively high background randomized clinical trials. 74th Sci Sess of the incidence in the T2DM population. In addition, the American Diabetes Association (ADA), San observed incidence rate in the clinical trials Francisco, 13 - 17 Jun 2014 (Poster) 2014. appears to be within the published range. Data does not demonstrate an increased frequency in patients with previous HF. Boehringer Ingelheim concluded that linagliptin is not casually associated with HF however this topic will continue to be closely monitored and the linagliptin CVOTs CAROLINA and CARMELINA will provide important further information.

WHO Pharmaceuticals Newsletter No. 1, 2015 • 28

IX. 3 Glibenclamide/glyburide and palpitations in Asian population

2019 WHO Pharmaceuticals

No. NEWSLETTER 2

The WHO Pharmaceuticals Newsletter provides you

with the latest information on the safety of medicines WHO Vision for Medicines Safety and legal actions taken by regulatory authorities around No country left behind: the world. It also provides signals based on information worldwide pharmacovigilance for safer medicines, safer patients derived from the WHO global database of individual case safety reports, VigiBase.

This newsletter also includes a brief report on WHO The aim of the Newsletter is to disseminate regulatory missions to Lebanon and Ethiopia, for strengthening information on the safety of the national pharmacovigilance systems. pharmaceutical products, based on communications received from our network of national pharmacovigilance centres and other sources such as specialized bulletins and journals, as well as partners in WHO.

The information is produced in the form of résumés in English, full texts of which may be obtained on request from: Safety and Vigilance: Medicines, EMP-HIS, World Health Organization, 1211 Geneva 27, Switzerland, Contents E-mail address: [email protected] This Newsletter is also available at: Regulatory matters http://www.who.int/medicines Safety of medicines

Signal

Feature

© World Health Organization 2019

Some rights reserved. This work is available under the Creative Commons Attribution-NonCommercial-ShareAlike 3.0 IGO licence (CC BY-NC-SA 3.0 IGO; https://creativecommons.org/licenses/by-nc-sa/3.0/igo). Under the terms of this licence, you may copy, redistribute and adapt the work for non-commercial purposes, provided the work is appropriately cited, as indicated below. In any use of this work, there should be no suggestion that WHO endorses any specific organization, products or services. The use of the WHO logo is not permitted. If you adapt the work, then you must license your work under the same or equivalent Creative Commons licence. If you create a translation of this work, you should add the following disclaimer along with the suggested citation: “This translation was not created by the World Health Organization (WHO). WHO is not responsible for the content or accuracy of this translation. The original English edition shall be the binding and authentic edition”. Any mediation relating to disputes arising under the licence shall be conducted in accordance with the mediation rules of the World Intellectual Property Organization. Suggested citation. WHO Pharmaceuticals Newsletter No.2, 2019: World Health Organization; 2019. Licence: CC BY-NC- SA 3.0 IGO. Cataloguing-in-Publication (CIP) data. CIP data are available at http://apps.who.int/iris.

Sales, rights and licensing. To purchase WHO publications, see http://apps.who.int/bookorders. To submit requests for commercial use and queries on rights and licensing, see http://www.who.int/about/licensing. Third-party materials. If you wish to reuse material from this work that is attributed to a third party, such as tables, figures or images, it is your responsibility to determine whether permission is needed for that reuse and to obtain permission from the copyright holder. The risk of claims resulting from infringement of any third-party-owned component in the work rests solely with the user. General disclaimers. The designations employed and the presentation of the material in this publication do not imply the expression of any opinion whatsoever on the part of WHO concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Dotted and dashed lines on maps represent approximate border lines for which there may not yet be full agreement. The mention of specific companies or of certain manufacturers’ products does not imply that they are endorsed or recommended by WHO in preference to others of a similar nature that are not mentioned. Errors and omissions excepted, the names of proprietary products are distinguished by initial capital letters. All reasonable precautions have been taken by WHO to verify the information contained in this publication. However, the published material is being distributed without warranty of any kind, either expressed or implied. The responsibility for the interpretation and use of the material lies with the reader. In no event shall WHO be liable for damages arising from its use.

Printed in Switzerland

Table of Contents

Regulatory Matters Baloxavir marboxil ...... 5 Carbimazole ...... 5 Deferiprone ...... 5 Eliglustat ...... 5 Febuxostat ...... 6 Fenspiride ...... 6 Finasteride ...... 6 Fingolimod ...... 6 Glecaprevir hydrate/pibrentasvir combination ...... 7 Hydrochlorothiazide ...... 7 Lithium ...... 7 Macitentan ...... 8 Nivolumab (genetical recombination) ...... 8 Opioids ...... 8 Oseltamivir ...... 9 Palbociclib ...... 9 Pembrolizumab ...... 9 Sodium-glucose co-transporter 2 (SGLT2) inhibitors ...... 9 Trastuzumab (genetical recombination) ...... 10

Safety of medicines Carbimazole ...... 11 Chlorhexidine digluconate ...... 11 Dipeptidyl peptidase-4 (DPP-4) inhibitors ...... 12 Fluoroquinolones ...... 12 Paracetamol (modified-release) ...... 12 Rivaroxaban ...... 12 Tofacitinib ...... 13

Signal Esomeprazole and gynaecomastia in obese adults ...... 14 Glibenclamide/Glyburide and palpitations in the Asian population ...... 17 Phenprocoumon – Accidental overdose ...... 23

WHO Pharmaceuticals Newsletter No. 2, 2019  3 Table of Contents

Selegiline and hypoglycaemia in underweight adults ...... 23

Feature Strengthening Pharmacovigilance in Lebanon and Ethiopia ...... 27

WHO Pharmaceuticals Newsletter No. 2, 2019  4 Signal

References induced gynaecomastia on the oxidation of estradiol at C-2 or C-17. Biol Pharm Bull. 1. Meyboom, RH et al. Proton-pump inhibitors 2003;26(5):695-700. and gynecomastia. Nederlands Bijwerkingen Centrum Lareb, August 2005. 7. Carvajal A, Macias D, Gutierrez A, Ortega S, Sáinz M, Martín Arias LH, et al. Gynaecomastia 2. electronic Medicines Compendium: Summary associated with proton pump inhibitors: a case of Product Characteristics for esomeprazole series from the Spanish Pharmacovigilance (Emozul®). Available from: System. Drug Saf. 2007;30(6):527-31. https://www.medicines.org.uk/emc/product/47 27. Accessed: 2018-03-29. 8. Goldstein JA, Ishizaki T, Chiba K, de Morais SM, Bell D, Krahn PM, et al. Frequencies of the 3. US Food and Drug Administration: Product defective CYP2C19 alleles responsible for the label for esomeprazole (Nexium®). Available mephenytoin poor metabolizer phenotype in from: various Oriental, Caucasian, Saudi Arabian and https://dailymed.nlm.nih.gov/dailymed/drugIn American black populations. fo. cfm?setid=f4853677-1622-4037-688b- Pharmacogenetics. 1997;7:59-64. fdf533a11d96. Accessed: 2018-03-29. 9. Desta Z, Zhao X, Shin JG, Flockhart DA. 4. Johnson RE, Murad MH. Gynecomastia: Clinical significance of the cytochrome Pathology, Evaluation and Management. Mayo P4502C19 genetic polymorphism. Clin Clin Proc. 2009;84(11):1010-15. Pharmacokinet. 2002;41(12):913-58. 5. Roberto G, Melis M, Biagi C. Drug-induced 10. Lardinois CK, Mazzaferri EL. Cimetidine blocks gynecomastia. Focus Farmacovigilanza. testosterone synthesis. Arch Intern Med. 2013;77(7):2. 1985;145(5):920-2. 6. Satoh T, Munakata H, Fujita K, Itoh S, 11. Nexium and Gynaecomastia - from FDA reports Kamataki T, Yoshizawa I. Studies on the [histogram]. 2018. Available from: interactions between drug and estrogen: II. On https://www.ehealthme.com/ds/nexium/gynae the inhibitory effect of 29 drugs reported to comastia/. Accessed: 2018-01-30.

Glibenclamide/glyburide and palpitations in the Asian population Raquel Herrera Comoglio, Argentina

Summary hepatic metabolism of sulfonylureas (SUs). It has been found that pharmacokinetics of glibenclamide Glibenclamide (glyburide in the US) is an oral blood depends significantly on CYP2C9 genotypes. glucose lowering drug (BGLD), a second-generation, CYP2C9 pharmacogenetic variants are more long-acting sulfonylurea (SU) approved for use in frequent in South-Asian populations (10-25%) than the US in 1984, following its introduction in Europe in Caucasian ones (2%–6%). by several years; in other countries it was approved later (in Singapore in 1990). SUs are currently first- Hypoglycaemia, a very well-known type A adverse line agents or an add-on therapy to other oral effect of glibenclamide and other BGLDs, can hypoglycaemic agents (OHAs), usually metformin. manifest through a variety of symptoms. About a quarter of newly-diagnosed patients initiate Palpitations are an unspecific symptom that can be therapy with SUs. the expression of the adrenergic counter-regulation to hypoglycaemia. Sulfonylureas stimulate insulin release by binding to specific sites on the beta cells (insulin secretagogue One hundred individual case safety reports (ICSRs) agent). As other SUs, glibenclamide is metabolized with the combination glibenclamide and palpitations in the liver and excreted by the kidney. The were retrieved from VigiBase, the WHO global genetically polymorphic cytochrome P450 (CYP), database of ICSRs on 15 January 2018. Many also enzyme CYP2C9 is mainly responsible for the mention other related terms, including WHO Pharmaceuticals Newsletter No. 2, 2019  17 Signal hypoglycaemia, sweating and blurred vision. Half of interactions (agents with high protein-binding), the set came from South-Asian countries. In the 59 antibiotics, substandard or fake medicines, reports in which time to onset (TTO) could be comorbidities (infection, pancreatic cancer, cancer, calculated, it ranged from 0 days to 20 years, with renal, hepatic and cardiovascular disease) and the majority (30 reports) in the 0 to 10 days group. patient characteristics (elderly, BMI, poor Of these 30 reports, 21 originated from South-Asian nutrition).2,3 In addition, published case reports and countries. case series have highlighted the association between CYP2C9 genetic polymorphisms and To the best of our knowledge, for glibenclamide hypoglycaemic events. there is no regulatory labelling or warning that refers to an increased risk of hypoglycaemia in Glibenclamide’s daily dosage is between 1.25 and patients with defective isoenzyme genetics. 20 mg. The usual starting dose of glibenclamide tablets is 2.5 to 5 mg daily.4 The defined daily dose ICSRs contain limited information, and a patient’s (DDD) is 7 mg or 10 mg (micronized and non- pharmacogenetic status is usually never stated. In micronized, respectively).5 Glibenclamide is not spite of these known limitations, we have identified recommended in the elderly or in individuals with a characteristics of ICSRs mentioning glibenclamide glomerular filtration rate (GFR)<50 mL/min. The as a suspected drug, and palpitations as an early combination of glibenclamide and metformin may hypoglycaemic symptom that could be suggestive have a synergistic effect, since both agents act to of a defective metabolism (short time-to-onset, improve glucose tolerance by different, but geographic region). We hypothesize that patients complementary mechanisms.4 presenting with hypoglycaemia and with a short TTO could have defective genetic variants, and Several studies which evaluated the safety of SUs therefore this is considered a signal. consistently showed that glibenclamide is associated with a higher risk of hypoglycaemia

when compared to other SUs, including glipizide, Introduction gliclazide and glimepiride.6-9 Glibenclamide was associated with a 52% greater risk of experiencing Glibenclamide is an oral blood glucose lowering at least one episode of hypoglycaemia compared drug (BGLD), a second-generation sulfonylurea with other secretagogues (relative risk 1.52 [95% (SU). SUs induce glucose-independent insulin CI 1.21-1.92]) and with an 83% greater risk release from the pancreatic β-cells by binding to the compared with other sulfonylureas (1.83 [1.35- ATP-sensitive potassium (KATP) channel. The 2.49]).10 polymorphic enzyme cytochrome P450 (CYP) 2C9 is the main enzyme catalysing the biotransformation According to the National List of Essential Medicines of SUs. (NLEMs), glibenclamide (2.5 and 5.0 mg) is an essential SU in five countries, India, Pakistan, Hypoglycaemia is a pharmacological, dose- Nepal, Sri Lanka and Bangladesh. In Sri Lanka, dependent (type A) adverse effect of blood glucose glibenclamide is the most commonly prescribed SU lowering agents, especially insulin and insulin by GPs.11 secretagogues. Mild hypoglycaemia is associated with adrenergic and neurogenic symptoms, such as The polymorphic CYP2C9 isoenzyme catalyses the tremor, palpitations and perspiration. Severe biotransformation of SUs in the liver. The mutant hypoglycaemia is characterized by symptoms alleles CYP2C9*2 and CYP2C9*3 are known to have related to reduced glucose to the brain, such as a reduced drug-metabolizing activity than the weakness, poor concentration, slurred speech, wildtype CYP2C9*1, the decrease in catalysing confusion or even seizure or coma. activity of the *3 allele being more pronounced. The *3 variant is most common in Asians with a In normal individuals, hypoglycaemic counter frequency of 10%–25% compared to that of 2%– regulation is a multifactorial process that involves 6% in Caucasians. Involvement of CYP2C19 in the reduction of insulin secretion, increasing glucagon metabolism of SUs is also reported. CYP2C19*2 and secretion, adrenergic activation, increased growth CYP2C19*3 are variants that encode a non- hormone and cortisol secretion. Hypoglycaemia functional CYP2C19 enzyme. Individuals with either increases plasma levels of both epinephrine and of the variants are labelled as poor metabolizers.12 norepinephrine, released primarily from the adrenal medulla. Recovery from hypoglycaemia is In Asian populations, genetic variability of CYP2C9 dependent on the adrenergic response. Individuals is dominated by the less functioning allele *3 (3.4% with preserved autonomic neurological response in East Asians and 11.3% in South Asians) while in manifest these higher levels of catecholamines Europeans, genetic variability expresses mainly the through palpitations, increased heart rate, *2 variant (Europeans 11.7% and admixed piloerection, etc.1 Americans 6.6%), Other allelic variants are also present in South Asian and African populations. 13 A number of factors can increase the possibility of hypoglycaemic events: over-prescribing, inappropriate dosing, changes in exercise or diet, pharmacodynamic (PD) interactions (other blood glucose lowering agents),2 pharmacokinetic (PK)

WHO Pharmaceuticals Newsletter No. 2, 2019  18 Signal

Reports in VigiBase Table 2. Age distribution A total of 100 Individual Case Safety Reports (ICSRs), with the combination glibenclamide and Age Asian Other N Total (%) countries n/n countries n/n palpitations were retrieved from VigiBase, the WHO total (%) total (%) global database of ICSRs, on 15 January 2018, and Reported 47/47 (100%) 39/48 (81%) were reviewed case by case. Four repeated reports were identified, and there was a likely duplicate 6 years 0 1/48 (2%) 1 (1.1%) from the US, therefore only 95 cases were 24-44 years 11/47 (23%) 0 11 (11.6%) considered. Of these, 47 reports were from Asian countries and 48 reports from non-Asian countries. 45-64 years 25/47 (53%) 22/48 (46%) 47 (49.5%) Most reports included hypoglycaemic symptoms 65-74 years 6/47 (13%) 7/48 (15%) 13 (13.7%)

(dizziness, sweating, vision blurred, etc.), and 18 above 75 years 5/47 (11%) 9/48 (19%) 14 (14.7%) explicitly reported hypoglycaemia. Most of the Asian ICSRs (34/47, 72%) had a completeness score Unknown 0 9 (19%) 9 (9.5%) (level of documentation) >0.50. Case series Total ≥65 11/47 (23%) 16/48 (31%) 26/95 (27.4%) distribution of gender, region and completeness years score is set out in Table 1. Total ≤ 64 36/47 (77%) 23/48 (52%) 61/95 (64.2%) years

Table 1. Case series characteristics: gender, Onset date and at least start date of glibenclamide region and completeness score therapy were stated in 59 out of 95 ICSRs (57.9%); so it was possible to calculate time-to-onset (TTO) Gender in this subset of reports. In three ICSRs, the onset Total n (%) date was stated as previous to the reported start of Female 60 (63.16) drug administration. Male 30 (31.6) In 28 Asian reports, TTO ranged from 0 to 60 days, Unknown 5 (5.3) and in 21 reports, TTO was up to 10 days. In six ICSRs symptoms manifested the same day, and in Number of reports (%) 11 ICSRs TTO was 1 day. Time to onset is set out in Asian countries Other countries Table 3. n 47 (49.5%) n 48 (50.5%) Thailand 26 (27.3%) US 29 (30.5%) Table 3. Time to onset India 8 (8.4%) Germany 4 (4.2%)

Singapore 6 (6.3%) Peru 3 (3.2%) Time to onset (TTO) n (total) n (Asia) China, 6 (6.3%) Canada, 4 (4.2%) Malaysia 3 each Sweden 2 each 0-10 days 30 21 Japan 1 (1.1%) Australia, 8 (8.4%) 12-28 days 4 4 Italy, 1 each Within a month* 4 2 Denmark, Eritrea, 33-36 days 2 1 Namibia, Oman, Spain 43-61 days 3 2 and United 90-270 days 9 4 Kingdom Completeness Score 1-20 years 7 2 Total n (%) Total 59 36 ≥ 0.70 23 (24.2) * Reports with the same year and month reported for 0.31 - 0.69 43 (45.3) drug start and reaction onset

0.1 - 0.28 29 (30.5) The dose for glibenclamide was reported in 65 cases (68.4%) and ranged from 1.8 mg to 20 mg daily. In Asian ICSRs, the reporter is given in 35 reports Fifty ICSRs reported low doses: one 1.8 mg, 13 (74.5%), and all but one were physicians, the cases between 2-2.5 mg, 33 a dose of 5 mg, two remaining one, a pharmacist. In 32 ICSRs, the with 3.5 mg and one with 7.5 mg. Nine patients sender’s comment highlighted the Asian origin. were treated with 10 mg, three with 15 mg and another three patients with 20 mg. The percentage Patients were relatively younger in Asia than in the of patients with doses ≤ 5 mg was similar in Asian other countries: 36/47 (76.6%) were ≤ 64 years countries and in non-Asian countries (77% and 76% old. The age distribution is described in Table 2. respectively). Thirteen reports (14%) were classed as serious, two of them were fatal, and 18 (19%) were classed as non-serious. In the two fatal cases, other BGLDs WHO Pharmaceuticals Newsletter No. 2, 2019  19 Signal were suspected (ipragliflozin and sitagliptin); Health Sciences Authority of Singapore issued a glibenclamide is reported as not withdrawn. Two warning about glibenclamide in older patients and ICSRs reported syncope as an adverse reaction, renal impairment. However, there is no mention of which indicates a more serious state. pharmacogenetic variability.14 A statement from the Royal Dutch Pharmacists Association Working Group concluded that there are Alternative causes of hypoglycaemia no dose recommendations based on patients’ Possible interactions: An ICSR from a non-Asian pharmacogenetic status to give at this time country reports hypoglycaemia in a 78-year-old (2011).15 woman treated with hydroxychloroquine because of “Place of sulfonylureas in the management of type 2 a rheumatoid arthritis. Three Asian country reports diabetes mellitus in South Asia: A consensus mention drugs possibly interacting: one of a 75- statement”, an initiative of the South Asian year-old woman, who started a triple therapy with Federation of Endocrine Societies (SAFES), metformin 1 g, pioglitazone 30 mg and developed in accordance with the American glibenclamide 5 mg, and presented with palpitations Association of Clinical Endocrinologists/ American and dizziness on the seventh day; another one College of Endocrinology (AACE/ACE) doesn’t mentions phenytoin administration with confusing mention pharmacogenetic variability as a cause of TTO dates; another report mentions hypoglycaemic sulfonylurea intolerance.11 symptoms the same day that amoxicillin was administered, and 10 days after the therapy with According to a paper published in 2014, gefitinib glibenclamide was started. product information is the only EMA label containing a warning about CYP2C9 metabolization.16 Changes in physical activity: An ICSR from a non-Asian country reports the case of a 62-year-old woman under therapy with glibenclamide and metformin who developed bouts of palpitations, SUs Metabolism and isoenzymes genetic confusion, and chest discomfort, particularly at polymorphisms times of unpredicted physical activity. A PK study performed in healthy male volunteers Incident comorbidities and incident showed that in homozygous carriers of the concomitant therapies: An ICSR from an Asian genotype *3/*3, total oral clearance was less than country reports hypoglycaemia in a 52-year-old half of that of the wildtype genotype *1/*1 (P woman who was started on dual therapy with <.001). Correspondingly, insulin secretion metformin 1 g daily and glibenclamide 5 mg; the measured within 12 hours after glyburide ingestion TTO was 10 days. On day 8, a prescription of 2 g of was higher in carriers of the genotype *3/*3 amoxicillin is reported (only one-day of treatment), compared with the other genotypes (P =.028), with 17 and doxycycline 200 mg daily, given for three days, no clinical effects. because of an acute upper respiratory infection. In a case-control study of 20 diabetic patients Some anti-infective agents may enhance the admitted to the emergency department with severe hypoglycaemic effect of glibenclamide and infection hypoglycaemia during SU drug treatment, it was itself can trigger hypoglycaemia. found that the CYP2C9 genotypes *3/*3 and *2/*3 An ICSR from a non-Asian country reports that are predictive of low enzyme activity were hypoglycaemia and palpitations in a 59-year-old more common in the hypoglycaemic group than in woman treated with glibenclamide 5 mg over nine the comparison groups (10% vs < 2%, months. Two days before the hypoglycaemic event respectively). Other factors in the group with severe occurred, the patient had received medication for a hypoglycaemia were lower body mass indexes, cardiac event (morphine, nifedipine, isosorbide higher rates of renal failure, older age, and higher 18 dinitrate, furosemide, enalapril). doses of glibenclamide. Other alternative causes: In two reports from a In a study assessing the frequency of CYP2C9 non-Asian country (TTO 170 days and 215 days), genetic variants in Type 2 diabetes mellitus (T2DM) palpitations are more likely related with other patients receiving sulfonylureas (92 reporting drug- morbidities (atrial fibrillation, reduced left associated hypoglycaemia, and 84 having never ventricular ejection fraction). In another ICSR from experienced hypoglycaemia), it was found that the an Asian country (TTO 270 days), other medications presence of the allele CYP2C9*3 increased the risk suggesting an acute coronary syndrome are of hypoglycaemia (OR: 1.687, adjusted for age, reported. BMI, mean daily dose of SU, duration of T2DM and renal function; p = 0.011).19

A study performed in Chinese healthy male Literature and labelling volunteers found that CYP2C9 polymorphism appears to exert a dominant influence on The US FDA label only mentions CYP2C9 as induced glibenclamide pharmacokinetics and by rifampicin, and potentially reducing pharmacodynamics in vivo; hypoglycaemia glibenclamide plasma levels as a consequence. developed in 3 of 6 CYP2C9*1/*3 carriers and 2 of There is no mention of poor metabolizers.4 The 12 CYP2C9*1/*1 carriers.20

WHO Pharmaceuticals Newsletter No. 2, 2019  20 Signal

A prospective population-based study did not To date, no regulatory labelling or warning has observe over-representation of the CYP2C9 slow highlighted the contribution of poor metabolizer metabolizer genotypes in the hypoglycaemic status on higher frequency or severity of patients group. However, in the control group, hypoglycaemia in patients treated with patients with CYP2C9 genotypes predicting slower glibenclamide. metabolism of SU drugs were treated with Even though ICSRs provide no information on the significantly lower doses than were extensive pharmacogenetic status of patients, short TTO and, metabolizers.21 to a lesser extent source countries may suggest a A study performed in Turkey with 108 diabetic pharmacogenetic cause. This can also be a class patients treated for ≥ 3 months with SUs characteristic. (glimepiride, gliclazide, glipizide) found that In this set of VigiBase reports, the proportion of heterozygosity and homozygosity for CYP2C9 cases with short TTO, and the number of good variant alleles (*2 or *3) tended to be more quality reports coming from Asian countries in frequent among patients who reported relatively young patients might be considered as a hypoglycaemic attacks.22 signal of the potential association of an increased Interactions: Several published cases report frequency of palpitations as symptoms of hypoglycaemia induced by hydroxychloroquine, and hypoglycaemia in patients treated with observational studies suggest a dose-dependent glibenclamide who are poor metabolisers. Also, in protective effect of hydroxychloroquine on drug- 32 ICSRs, the sender’s comments highlighted the induced diabetes in rheumatic patients treated with Asian origin. This hypothesis would need specific corticosteroids.23-26 Co-administration of anti- pharmacogenetic studies in patients treated with infective agents that are CYP2C9 inhibitors can glibenclamide and experiencing hypoglycaemia with increase the risk of hypoglycaemia in glipizide and short TTO or interacting medications. glyburide users.26 Therefore, co-administration of other CYP2C9 inhibitors might also increase the risk of hypoglycaemia, although an increased risk of References hypoglycaemia might not be present with less strong (non-clinically-relevant) CYP2C9 inhibitors. 1. Hoffman RP. Sympathetic mechanisms of P-Glycoprotein inhibitors might also increase the hypoglycemic counterregulation. Curr Diabetes risk of hypoglycaemia.26 A few ICSRs of this set Rev. 2007 Aug;3(3):185-93. mentioned suspected or concomitant drugs that 2. Scheen AJ. Drug Interactions of Clinical could interact with glibenclamide. It is worth noting Importance with Antihyperglycaemic Agents: that an interacting agent can theoretically exert An Update. Drug Saf. 2005 Jul;28(7):601-631. additive effects on an isoenzyme with genetically reduced functionality. 3. Thulé PM, Umpierrez G. Sulfonylureas: a new look at old therapy. Curr Diab Rep. 2014 Apr;14(4):473. Discussion and conclusion 4. Glyburide FDA label: Hypoglycaemia is a sulfonylureas dose-related type https://www.accessdata.fda. A adverse effect. In subjects with preserved gov/drugsatfda_docs/label/2009/017532s030l autonomic function, the fall in blood glucose levels bl.pdf triggers the adrenergic counter regulation, which 5. ATC/DDD Index 2018 2018 manifests through palpitations. Glibenclamide is https://www.whocc.no/ metabolized by CYP2C9 isoenzyme, which is highly atc_ddd_index/?code=A10BB01 polymorphic. The frequency of polymorphism of defective CYP2C9 alleles (CYP2C9*2 and CYP2C9*3 6. Leonard CE, Han X, Brensinger CM, Bilker WB, variants) in South-Asian populations is reported to Cardillo S, Flory JH, Hennessy S. Comparative be 10-25% of a total population; defective isoforms risk of serious hypoglycemia with oral are also present – but much less frequently - in antidiabetic monotherapy: A retrospective Caucasian populations. Poor metabolizers cohort study. Pharmacoepidemiol Drug Saf. (CYP2C9*2 and CYP2C9*3 alleles carriers) can show 2018 Jan;27(1):9-18. higher plasmatic levels of glibenclamide, leading to 7. Douros A, Yin H, Yu OHY, Filion KB, Azoulay L, low blood glucose levels and triggering counter Suissa S. Pharmacologic Differences of regulation mechanisms, such as adrenergic Sulfonylureas and the Risk of Adverse response. Cardiovascular and Hypoglycemic Events. Even though observational studies have not shown Diabetes Care. 2017 Nov;40(11):1506-1513. conclusive results, a literature review supports a 8. Zhou M, Wang SV, Leonard CE, Gagne JJ, relationship between clinical effects and CYP2C9 Fuller C, Hampp C, Archdeacon P, Toh S, Iyer polymorphic variants. A PK study conducted in A, Woodworth TS, Cavagnaro E, Panozzo CA, Caucasian volunteers saw no clinical effects of the Axtman S, Carnahan RM, Chrischilles EA, difference in metabolization, but hypoglycaemia Hennessy S. Sentinel Modular Program for was more frequent in one conducted with Chinese Propensity Score-Matched Cohort Analyses: volunteers. WHO Pharmaceuticals Newsletter No. 2, 2019  21 Signal

Application to Glyburide, Glipizide, and Serious CYP2C9 amino acid polymorphisms on Hypoglycemia. Epidemiology. 2017 glyburide kinetics and on the insulin and Nov;28(6):838-846. glucose response in healthy volunteers. Clin Pharmacol Ther 2002; 71: 286–96. 9. Shorr RI, Ray WA, Daugherty JR, Griffin MR. Individual sulfonylureas and serious 18. Holstein A, Plaschke A, Ptak M, Egberts EH, El- hypoglycemia in older people. J Am Geriatr Din J, Brockmöller J, Kirchheiner J Association Soc. 1996 Jul;44(7):751-5. between CYP2C9 slow metabolizer genotypes and severe hypoglycaemia on medication with 10. Gangji AS, Cukierman T, Gerstein HC, sulphonylurea hypoglycaemic agents. Br J Clin Goldsmith CH, Clase CM. A systematic review Pharmacol. 2005 Jul;60(1):103-6. and meta-analysis of hypoglycemia and cardiovascular events: a comparison of 19. Ragia G, Petridis I, Tavridou A, Christakidis D, glyburide with other secretagogues and with Manolopoulos VG.Presence of CYP2C9*3 allele insulin. Diabetes Care. 2007 Feb;30(2):389- increases risk for hypoglycemia in Type 2 94. diabetic patients treated with sulfonylureas. Pharmacogenomics. 2009 Nov;10(11):1781-7. 11. Kalra S, Aamir AH, Raza A, Das AK, Azad Khan AK, Shrestha D, Qureshi MF, Md Fariduddin, 20. Yin OQ, Tomlinson B, Chow MS. CYP2C9, but Pathan MF, Jawad F, Bhattarai J, Tandon N, not CYP2C19, polymorphisms affect the Somasundaram N, Katulanda P, Sahay R, pharmacokinetics and pharmacodynamics of Dhungel S, Bajaj S, Chowdhury S, Ghosh S, glyburide in Chinese subjects. Clin Pharmacol Madhu SV, Ahmed T, Bulughapitiya U2.Place of Ther. 2005 Oct;78(4):370-7. sulfonylureas in the management of type 2 21. Holstein A, Hahn M, Patzer O, Seeringer A, diabetes mellitus in South Asia: A consensus Kovacs P, Stingl J. Impact of clinical factors statement. Indian J Endocrinol Metab. 2015 and CYP2C9 variants for the risk of severe Sep-Oct;19(5):577-96. sulfonylurea-induced hypoglycemia Eur J Clin 12. Dawed AY, Zhou K, Pearson ER. Pharmacol. 2011 May;67(5):471-6. Pharmacogenetics in type 2 diabetes: influence 22. Gökalp O, Gunes A, Cam H, Cure E, Aydın O, on response to oral hypoglycemic agents. Tamer MN, Scordo MG, Dahl ML. Mild Pharmgenomics Pers Med. 2016 Apr 6;9:17- hypoglycaemic attacks induced by 29. sulphonylureas related to CYP2C9, CYP2C19 13. Zhou Y, Ingelman-Sundberg M, Lauschke VM. and CYP2C8 polymorphisms in routine clinical Worldwide Distribution of Cytochrome P450 setting. Eur J Clin Pharmacol. 2011 Alleles: A Meta-analysis of Population-scale Dec;67(12):1223-9. Sequencing Projects. Clin Pharmacol Ther. 23. Winter EM, Schrander-van der Meer A, 2017 Oct;102(4):688-700. Eustatia-Rutten C, Janssen M. 14. Singapore government, Health Science Hydroxychloroquine as a glucose lowering Authority HSA Recommendations to avoid use drug. BMJ Case Rep. 2011 Oct 28;2011. of glibenclamide in the elderly and renal- 24. Shojania K, Koehler BE, Elliott T. Hypoglycemia impaired. 31 Dec 2013: induced by hydroxychloroquine in a type II http://www.hsa.gov.sg/content/hsa/en/ diabetic treated for polyarthritis. J Rheumatol. Health_Products_Regulation/Safety_Informatio 1999 Jan;26(1):195-6. n_ and_Product_Recalls/Product_Safety_Alerts/20 25. Chen YM, Lin CH, Lan TH, Chen HH, Chang SN, 13/ recommendations_to.html Chen YH, Wang JS, Hung WT, Lan JL, Chen DY. Hydroxychloroquine reduces risk of incident 15. Annotation of DPWG Guideline for diabetes mellitus in lupus patients in a dose- glibenclamide and CYP2C9. dependent manner: a population-based cohort https://www.pharmgkb.org/guideline/ study. Rheumatology (Oxford). 2015 PA166104953 Accessed 18 February 2018. Jul;54(7):1244-9 16. Ehmann F, Caneva L, Papaluca M. European 26. Schelleman H, Bilker WB, Brensinger CM, Wan Medicines Agency initiatives and perspectives F, Hennessy S. Anti-infectives and the risk of on pharmacogenomics. Br J Clin Pharmacol. severe hypoglycemia in users of glipizide or 2014 Apr;77(4):612-7. glyburide. Clin Pharmacol Ther. 2010 17. Kirchheiner J, Brockmöller J, Meineke I, Bauer Aug;88(2):214-22. S, Rohde W, Meisel C, Roots I. Impact of

WHO Pharmaceuticals Newsletter No. 2, 2019  22

Type 2 diabetes mellitus (T2DM) is a multifactorial, chronic, progressive disease, affecting more than 422 million people over the world, and having a significant societal and economic impact. Cardiovascular disease is the leading cause of morbidity and mortality in T2DM patients, who have higher rates of mortality than the non-diabetic population.

T2DM is defined by its metabolic -mainly glucose-related- manifestations which serve as markers for controlling the evolution of disease. However, while the effect of control serum glucose levels on microvascular complications is acknowledged, its impact on macrovascular complications remains uncertain.

Since 2008, new blood glucose-lowering agents have to demonstrate cardiovascular safety, and some have shown to reduce cardiovascular outcomes and mortality. However, the populations included in these large cardiovascular outcome trials differ from the general population, making results no fully generalizable.

While randomised controlled trials are the gold standard for generating scientific evidence, observational studies conducted with secondary data of Electronic medical records (EMRs) are increasingly used as a source of complementary or confirmatory evidence, especially when RCTs are not feasible or unavailable.

This work report an observational, population-based cohort study conducted in SIDIAP, a large Catalan general practitioners database that contains health data of 5,5 million people. We assessed cardiovascular outcomes and mortality in general, unselected T2DM population treated with non-insulin blood-glucose-lowering agents. The results are expected to be useful both for clinical and public health decision-making.