Glucose-lowering drugs or strategies and cardiovascular outcomes in patients with or at risk for type 2 diabetes: a meta-analysis of randomised controlled trials Jacob A Udell, Matthew A Cavender, Deepak L Bhatt, Saurav Chatterjee, Michael E Farkouh, Benjamin M Scirica Lancet Diabetes Endocrinol 2015; 3: 356 66 Published Online March 17, 2015 http://dx.doi.org/10.1016/ S2213-8587(15)00044-3 See Comment page 310 Women s College Research Institute and Cardiovascular Division, Department of Medicine, Women s College Hospital (J A Udell MD) and Peter Munk Cardiac Centre, University Health Network (J A Udell, Prof M E Farkouh MD), Heart and Stroke/Richard Lewar Centre of Excellence, University of Toronto, Toronto, ON, Canada; Cardiovascular Division, Brigham and Women s Hospital, Harvard Medical School, Boston, MA, USA (M A Cavender MD, Prof D L Bhatt MD, B M Scirica MD); and Division of Cardiology, St Luke s Roosevelt Hospital, Mount Sinai Health System, New York, NY, USA (S Chatterjee MD) Correspondence to: Dr Jacob A Udell, Division of Cardiology, Peter Munk Cardiac Centre, Toronto General Hospital and Women s College Hospital, Toronto, ON M5S 1B1, Canada jay.udell@utoronto.ca Summary Background Some glucose-lowering drugs or strategies adversely affect cardiovascular outcomes. We aimed to assess the extent to which glucose lowering by various drugs or strategies increases the risk of heart failure in patients with or at risk for type 2 diabetes, and to establish whether risk is associated with achieved differences in glycaemia or weight control. Methods We searched Ovid Medline, the Cochrane Library, and meeting abstracts up to Feb 20, 2015, for large randomised controlled trials of glucose-lowering drugs or strategies that assessed cardiovascular outcomes. The primary endpoint was incidence of heart failure. We derived pooled risk ratios (RRs) with random-effects models. Findings We included data from 14 trials, with mean duration 4 3 (2 3) years, comprising 95 502 patients, of whom 3907 (4%) patients developed a heart failure event. Glucose-lowering drugs or strategies were associated with a 0 50% (SD 0 33) reduction in HbA 1c and a 1 7 kg (2 8) weight gain. Overall, glucose-lowering drugs or strategies increased the risk of heart failure compared with standard care (RR 1 14, 95% CI 1 01 1 30; p=0 041). The magnitude of this effect varied dependent on the method of glucose lowering (p for interaction=0 00021). Across drug classes, risk was highest with peroxisome proliferator-activated receptor agonists (RR 1 42, 95% CI 1 15 1 76; six trials), intermediate with dipeptidyl peptidase-4 inhibitors (1 25, 1 08 1 45; two trials), and neutral with insulin glargine (0 90, 0 77 1 05; one trial). Target-based intensive glycaemic control strategies (RR 1 00, 95% CI 0 88 1 13; four trials) and intensive weight loss (0 80, 95% CI 0 62 1 04; one trial) were also not associated with development of heart failure. Metaregression analysis showed that for every 1 0 kg of weight gain associated with glucose-lowering drugs or strategies, there was a 7 1% (95% CI 1 0 13 6) relative increase in the risk of heart failure compared with standard care (p=0 022). Interpretation Compared with standard care, glycaemic lowering by various drugs or strategies might increase the risk of heart failure, with the magnitude of risk dependent on the method of glucose lowering and, potentially, weight gain. Funding None. Introduction Heart failure is common in patients with type 2 diabetes and greatly affects health-care use, quality of life, and prognosis. 1 6 Epidemiological data have shown an association between glycaemia and incident heart failure events in patients with or at risk for type 2 diabetes, 5 12 similar to the association between glycaemia and atherothrombotic cardiovascular events in this patient population. 13 Many glucose-lowering drugs or strategies have been tested in randomised controlled trials of patients with type 2 diabetes, and others are presently being studied, to establish whether improved glycaemic control translates into reductions in atherothrombotic cardiovascular outcomes. However, some glucose-lowering drugs or strategies might have properties that result in beneficial or adverse haemodynamic effects, with a resultant effect on cardiovascular outcomes. 1 For instance, the drug class of agonists of the peroxisome proliferator-activated receptor (PPAR)-γ subtype (thiazolidinediones) cause fluid retention and weight gain, and a resultant increase in the risk of heart failure, despite a modest beneficial effect on atherothrombotic events. 14 Additionally, trials of other glucose-lowering drugs or strategies, such as dipeptidyl peptidase (DPP)-4 inhibitors (gliptins), have shown an increased risk of hospital admission for heart failure. 15 18 Because potential mediators of risk of heart failure in patients with type 2 diabetes are obesity and insulin resistance, 7,19 it is unclear to what extent the degree of glycaemic control, the method of glucose lowering, a difference in weight gain, or other study design characteristics affect the risk of heart failure in trials of glucose-lowering drugs or strategies. Furthermore, randomised controlled trials of glycaemic treatments are underpowered to detect differences in individual cardiovascular outcomes, including heart failure, 20 and previous meta-analyses omitted or only partly explored the risk of heart failure with glucose-lowering drugs or strategies. 21 27 Therefore, we systemically assessed data 356 www.thelancet.com/diabetes-endocrinology Vol 3 May 2015
from large randomised controlled trials of cardiovascular outcomes to define the risk of heart failure with glucoselowering drugs or strategies compared with standard care and to establish whether this risk is counterbalanced by any reduction in atherothrombotic events, with particular attention focused on effect modification by differential glycaemia or weight gain and other trial characteristics. Methods Search strategy and selection criteria We did a systematic review and meta-analysis to identify relevant published randomised controlled trials comparing experimental glucose-lowering drugs or modulation strategies with placebo or standard care according to the recommendations of the Cochrane Collaboration and the PRISMA guidelines. 28,29 We searched Ovid Medline and the Cochrane Library up to Feb 20, 2015, with MeSH and keyword search terms including diabetes mellitus, type 2, haemoglobin A1c, glycosylated, glycated, glucose, cardiovascular disease, heart failure, glycaemic control, glucose control, and intensive (appendix p 2). We restricted the search to randomised controlled trials of human adults (aged 19 years). We used no language restrictions. We screened titles and abstracts for eligibility and further searched and reviewed the supplemental appendices and reference lists of eligible papers for additional data and online cardiovascular conference abstracts between 2013 and 2014, and ClinicalTrials.gov, to ensure identification of relevant published and unpublished studies (appendix p 2). We included trials if they enrolled 1000 or more individuals with or at risk for type 2 diabetes (ie, prediabetes) and randomly assigned patients to investigational drugs or modulation strategies versus placebo or standard care that was exclusively focused on one risk factor (eg, glucose or weight) and resulted in an improvement in glycaemic control between treatment groups. We excluded trials that enrolled fewer than 1000 patients or patients with an acute cardiovascular event (eg, acute myocardial infarction, shock, or hospital admission for heart failure), or if they tested a multifactorial risk-factor intervention or non-glycaemic drug, or resulted in a mean difference of 0 1% or less in HbA 1c (or of 0 1 mmol/l in fasting plasma glucose if HbA 1c was not reported) between treatment groups. This design was intended to focus our analysis on large outcome trials with clinically relevant glycaemic differences between interventions that were powered to study composite cardiovascular endpoints to evaluate their atherothrombotic and heart failure effects over time in a stable population. Trials that compared three groups were reported as combined analyses comparing the experimental glucose-lowering drugs or strategies group with the two controls (metformin and glyburide). Outcomes The primary cardiovascular endpoint was incidence of heart failure. Secondary endpoints were major adverse cardiovascular events, which were included in preferential sequence if reported in the paper: the composite of cardiovascular death, myocardial infarction, or stroke; or cardiovascular death, myocardial infarction, or ischaemic stroke; or all-cause death, myocardial infarction, or stroke; or fatal and non-fatal myocardial infarction. Other secondary outcomes were an extended major adverse cardiovascular event endpoint, typically defined as a major adverse cardiovascular event plus hospital admission for unstable angina, coronary revascularisation, and, in some trials, heart failure; individual components of these composite endpoints besides heart failure; and allcause death (endpoint definitions for individual trials are listed in appendix p 5). Two investigators (JAU and MAC) independently extracted information about outcomes and trial characteristics from the published manuscripts and appendices; results were then compared and any disagreements were resolved by consensus. Statistical analysis Data for baseline characteristics were obtained with means (SDs), medians (IQRs), or rates from each study, and summary results were calculated weighted according to individual sample sizes. We used the originally reported hazard ratios (HRs) or risk ratios (RRs) from each study when available, or otherwise 3908 records identified 3899 through database searching 9 through other sources 3908 screened 26 randomised trials assessed for eligibility 14 randomised trials included in meta-analysis Figure 1: Study flow diagram 3882 excluded after screening titles and abstracts 12 excluded for not meeting inclusion criteria 3 enrolled <1000 patients 2 included patients with acute cardiovascular events 2 assessed multifactorial intervention 4 reported a 0 01% difference in fasting plasma glucose or HbA 1c 1 did not report cardiovascular outcomes See Online for appendix www.thelancet.com/diabetes-endocrinology Vol 3 May 2015 357
UK Prospective Diabetes Study 33 (1998) 35 Population N Intervention Control Age Outpatients with newly diagnosed type 2 diabetes receiving lifestyle management alone; excluded patients with myocardial infarction in the past year, present angina or heart failure, or polyvascular disease PROactive Outpatients with established (2005) 36 type 2 diabetes and atherothrombosis (coronary heart disease, stroke, or peripheral artery disease); excluded patients with recent acute coronary syndrome (<3 months), myocardial infarction (<6 months), stroke (<6 months), or coronary revascularisation (<6 months), or with symptomatic heart failure (NYHA class II) ADOPT Outpatients with recently (2006) 37 diagnosed type 2 diabetes receiving lifestyle management alone; excluded patients with unstable or severe angina, or known heart failure DREAM Outpatients with newly (2006) 38 diagnosed type 2 diabetes or prediabetes; excluded patients with previous cardiovascular disease (including heart failure) ACCORD Outpatients with established (2008) 39 type 2 diabetes and either cardiovascular disease or more than one risk factor ADVANCE Outpatients with established (2008) 40 type 2 diabetes and either atherothrombosis, microvascular disease, or multiple risk factors BARI2D Outpatients with established (2009) 41 type 2 diabetes and coronary heart disease candidates for revascularisation (angina or ischaemia); excluded patients with symptomatic heart failure (NYHA class III IV) or recent coronary revascularisation (<12 months) RECORD Outpatients with established (2009) 42 type 2 diabetes receiving metformin or sulfonylurea; excluded patients with recent admission to hospital for a cardiovascular event (<3 months) or known heart failure 3867 Intensive treatment with sulfonylurea or insulin (target fasting blood glucose <6 mmol/l) Standard care (diet; target fasting blood glucose <15 mmol/l) 53 3 (8 6) 5238 Pioglitazone Placebo 61 8 (7 7) 4351 Rosiglitazone (group 1) Metformin (group 2); glyburide (group 3) 56 9 (10 1) 5269 Rosiglitazone Placebo 54 7 (10 9) 10 251 Intensive treatment (target HbA 1c <6%) 11 140 Intensive treatment with gliclazide and other drugs as required (target HbA 1c 6%) 2368 Insulin sensitisation with oral treatment (metformin 75%, sulfonylurea 18%, thiazolidinedione 62%, insulin 29%) Standard care (HbA 1c 7 0 7 9%) Standard care (according to local guidelines) Insulin (metformin 10%, sulfonylurea 52%, thiazolidinedione 4%, insulin 61%) 4447 Rosiglitazone Metformin and sulfonylurea 62 2 (6 8) 66 (6) 62 4 (8 9) 58 4 (8 3) Followup 10 (7 7 12 4) 2 9 4 3 (2 5 4 7) 3 4 5 5 3 5 5 cardiovascular disease (%) 5238 (100%) 3608 (35%) 3590 (32%) 2368 (100%) 772 (17%) heart failure (%) duration of diabetes Change in fasting plasma glucose or HbA 1c (%)* NR NR 0 0 90 NR 8 0 (4 13) NR 0 1 5 0 50 0 13 (0 04) vs metformin; 0 42 (0 04) vs glyburide 0 0 0 0 50 497 (5%) 156 (7%) 10 NR 7 9 (6 4) 21 (0 5%) 10 4 (8 7) 7 1 (4 9) 1 10 0 67 (0 15) 0 50 0 27 (0 2) Achieved fasting plasma glucose or HbA 1c (%) 7 0 (6 2 8 2) 7 0 7 1 5 5 6 4 (6 1 7 0) 6 5 7 0 (1 2) 7 5 Change in weight (kg)* +3 10 (2 02) +4 00 +6 90 (0 28) vs metformin; +2 50 (0 28) vs glyburide +2 20 +3 10 +0 70 1 80 (8 6) +4 71 (0 45) (Table continues on next page) 358 www.thelancet.com/diabetes-endocrinology Vol 3 May 2015
Population N Intervention Control Age Followup cardiovascular disease (%) heart failure (%) duration of diabetes Change in fasting plasma glucose or HbA 1c (%)* Achieved fasting plasma glucose or HbA 1c (%) Change in weight (kg)* (Continued from previous page) VADT (2009) 43 Outpatients with established type 2 diabetes; excluded patients with a recent cardiovascular event (<6 months), severe angina, or advanced heart failure 1791 Intensive therapy (treatment absolute difference in HbA 1c 1 5%) Standard care 60 4 (9 0) 5 6 723 (40%) NR 11 5 (7 5) 1 50 6 9 +4 05 ORIGIN (2012) 44 Outpatients with newly diagnosed or established type 2 diabetes or prediabetes and atherothrombosis; excluded patients with known heart failure 12 537 Insulin glargine Standard care 63 5 (7 9) 6 2 (5 8 6 7) 7378 (59%) 0 5 4 (6 0) 0 30 6 2 (5 8 6 8) +2 10 EXAMINE (2013) 16,17 Outpatients with established type 2 diabetes and stabilised within 15 90 days of an acute coronary syndrome (median 46 days); excluded patients with unstable cardiac disorders, including heart failure (NYHA class IV) 5380 Alogliptin Placebo 61 0 (10 0) 1 46 (0 77 2 06) 5380 (100%) 1501 (28%) 7 2 0 36 (0 04) 7 7 (1 1) +0 06 (0 16) Look AHEAD (2013) 45 SAVOR TIMI 53 (2013) 15 Outpatients with established type 2 diabetes; excluded patients with a recent cardiovascular event (<3 months) or unstable cardiac disorders, including symptomatic heart failure (NYHA class III) Outpatients with newly diagnosed or established type 2 diabetes and either atherothrombosis or more than one risk factor; excluded patients with a recent or unstable cardiovascular event (<2 months) AleCardio (2014) 46 Outpatients with newly diagnosed or established type 2 diabetes and stabilised within 8 weeks of an acute coronary syndrome (mean 29 days); excluded patients with symptomatic heart failure (NYHA class II), or recent hospital admission for heart failure (<12 months) 5145 Intensive lifestyle intervention for weight loss (caloric restriction and exercise) Standard care 58 8 (6 9) 16 492 Saxagliptin Placebo 65 1 (8 5) 7226 Aleglitazar Placebo 61 (10) Total 95 502 61 7 (9 1) 9 6 (8 9 10 3) 2 1 (1 8 2 3) 2 0 (1 6 2 2) 4 3 (2 3) 714 (14%) 12 959 (79%) 7226 (100%) 49 956 (57%) NR 5 0 (2 0 10) 2105 (13%) 759 (10%) 5039 (7%) 10 3 (5 3 16 7) 8 6 7 1 (4 9) 0 22 (0 03) 0 20 (0 01) 0 60 (0 03) 0 50 (0 33) 7 3 7 7 (1 4) 7 3 6 9 (0 92) 4 (1) 0 50 (0 15) +3 50 1 7 (2 8) Data are mean (SD), median (IQR), or %, unless otherwise indicated. NR=not reported. NYHA=New York Heart Association. *Absolute achieved difference in fasting blood glucose, HbA 1c, or bodyweight between treatment groups at either 1 year or end of treatment, whichever was longer and reported. Absolute achieved HbA 1c in the glucose-lowering strategy group. Reported as baseline history of ischaemic heart disease. Range reported only. Table: Characteristics of included trials calculated RRs and 95% CIs from the reported number of events and patients per treatment group. Data from each trial were considered as per the intention-to-treat principle and we derived pooled RRs and 95% CIs with inverse-variance random-effects models. Trials that did not report a specific endpoint that could not be estimated from other outcomes were excluded from the pooled analysis of only that specific endpoint. We assessed heterogeneity for treatment effect across individual trials with use of the Cochran s Q statistic and the I² measure. 30 We further stratified results by the type of glucose-lowering drug or strategy tested, in view www.thelancet.com/diabetes-endocrinology Vol 3 May 2015 359
of the greater effect of some drugs on fluid retention, weight gain, and risk of heart failure, and to establish whether summary effects persisted when a drug class was removed from investigation in sensitivity analyses. A test for interaction with a model term representing effects stratified by treatment categories was calculated. 31 In further sensitivity analyses, we sequentially removed each study result from the pooled effect estimate, or incorporated smaller trials, or incorporated trials with interventions that resulted in a mean difference of 0 1% or less in HbA 1c. We used meta-regression to examine the association of treatment effects with various trial characteristics, including the difference in weight gain and glycaemic control between treatment groups, the absolute glycaemic control achieved in the experimental treatment group, the baseline duration of diabetes, the proportion of patients with established cardiovascular disease or heart failure, and trial duration. We assessed publication bias by visual inspection of funnel plots, with ascertainment for potential asymmetry of published results by Egger s linear regression test and Duval and Tweedie s trim and fill method. 32 34 Two-sided p values were calculated with p<0 05 considered significant for all tests. We did statistical analyses with Review Manager (version 5.3.4), Comprehensive Meta-Analysis (version 2.2), and STATA (version 11SE). Role of the funding source There was no funding source for this study. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results Figure 1 shows the study flowchart. We identified 26 eligible randomised controlled trials 15 17,35 59 of glucoselowering drugs or strategies in patients with or at risk for type 2 diabetes. 12 trials were excluded, 47 59 leaving 14 trials that met inclusion criteria in the meta-analysis, comprising 95 502 patients (figure 1). 15 17,35 46 The table summarises individual trials and their characteristics. Six trials tested a PPAR agonist or PPAR-based strategy; two trials studied a DPP-4 inhibitor; one trial tested insulin glargine; four trials tested an intensive glycaemic target-based strategy; and one trial investigated an intensive weight-loss intervention. Six trials were unmasked open-label studies comparing interventions with routine standard care, five trials were double-blind placebo-controlled trials, two trials were unmasked trials with a specific active comparator, and one trial was a double-blind three-group active comparator trial. The double-blind trials tended to have higher rates of premature discontinuation of treatment than did the open-label studies, but overall, few patients were lost to follow-up or withdrew consent (appendix p 6). Definitions of cardiovascular endpoints were reported in each trial in accordance with standard diagnostic criteria, which allowed for comparisons across trials. Endpoints were available from all 14 trials for heart failure, major adverse cardiovascular events, expanded major adverse cardiovascular events, myocardial infarction, and stroke; from 12 trials for all-cause death alone; 15,16,35 37,39 41,43 46 12 trials for cardiovascular death alone; 15,16,35,36,38 40,42 46 seven trials for unstable angina; 15,36,38,43 46 and six trials for coronary revascularisation. 15,36,38,43,44,46 Outcome assessment was typically through clinical contact except for one trial that used administrative health-care service records. Outcome More glucose control Events Total Less glucose control Events Total Weight Heart failure risk ratio (95% CI) 1998 UK Prospective Diabetes Study 35 2005 PROactive 36 2006 ADOPT 37 2006 DREAM 38 2008 ACCORD 39 2008 ADVANCE 40 2009 BARI2D 41 2009 RECORD 42 2009 VADT 43 2012 ORIGIN 44 2013 EXAMINE 16,17 2013 Look-AHEAD 45 2013 SAVOR-TIMI 53 15 2014 AleCardio 46 80 281 22 14 152 220 248 61 76 310 106 99 289 122 2729 2605 1456 2635 5128 5571 1183 2220 892 6264 2701 2570 8280 3616 36 198 28 2 124 231 218 29 82 343 89 119 228 100 1138 2633 2895 2634 5123 5569 1185 2227 899 6273 2679 2575 8212 3610 5 5% 9 5% 3 6% 0 7% 8 3% 9 3% 9 7% 4 8% 6 7% 9 8% 7 3% 9 5% 0 91 (0 62 1 34) 1 43 (1 21 1 71) 1 56 (0 90 2 72) 7 03 (1 60 30 90) 1 18 (0 93 1 49) 0 95 (0 79 1 14) 1 14 (0 97 1 34) 2 10 (1 35 3 27) 0 91 (0 66 1 25) 0 90 (0 77 1 05) 1 19 (0 90 1 58) 0 80 (0 62 1 04) 1 27 (1 07 15 1) 1 22 (0 94 1 59) Total 2080 47 850 1827 47 652 100% 1 14 (1 01 1 30) Heterogeneity: Tau 2 =0 04; χ 2 =45 56, df=13; p<0 0001; I 2 =71% Test for overall effect: Z=2 04; p=0 041 0 2 0 5 1 2 5 Favours glucose-lowering Favours standard care Figure 2: Risk of heart failure events with glucose-lowering drugs or strategies versus standard care Determined with an inverse-variance random-effects model. 360 www.thelancet.com/diabetes-endocrinology Vol 3 May 2015
adjudication for the primary endpoint was done in most trials (appendix p 6). 3907 (4%) patients developed a heart failure event and 9378 (10%) patients had a major adverse cardiovascular event. Compared with standard care, glucose-lowering drugs or strategies increased the risk of heart failure (figure 2). This association represented an absolute risk increase of 0 51%; however, the magnitude of this effect varied dependent on the method of glucose lowering (I²=71%; p for interaction=0 00021; figure 3). Specifically, the risk was highest with PPAR agonists, intermediate with DPP-4 inhibitors, and neutral with insulin glargine (figure 3). Target-based intensive glycaemic control and intensive weight loss were also not associated with development of heart failure (figure 3). By contrast with findings for heart failure, glucoselowering drugs or strategies overall modestly lowered the risk of major adverse cardiovascular events (figure 4). This effect was driven by a reduction in myocardial infarction (p=0 017; figure 5). When we assessed an expanded major adverse cardiovascular events endpoint, the cardiovascular effect of glucose-lowering drugs or strategies was attenuated and no longer significant (p=0 10; figure 5). Consequently, glucose-lowering drugs or strategies had no significant effect on cardiovascular death, all-cause death, stroke, unstable angina, or coronary revascularisation (figure 5). Among atherothrombotic endpoints, these findings were consistent irrespective of strategy or drug class (all p for interactions 0 05; figure 5). In aggregate, the weighted mean difference in weight in patients randomly assigned to glucose-lowering drugs or strategies was +1 7 kg (SD 2 8), but weight change varied dependent on the method of glucose lowering. For instance, the mean difference in weight was +3 6 kg (SD 3 3) in trials of PPAR agonists, +2 2 kg (1 5) with intensive glycaemic control, +2 1 kg (SD not reported) PPAR agonists 2005 PROactive 36 2006 ADOPT 37 2006 DREAM 38 2009 BARI2D 41 2009 RECORD 42 2014 AleCardio 46 Subtotal Heterogeneity: Tau 2 =0 04; χ 2 =13 82, df=5; p=0 017; I 2 =64% Test for overall effect: Z=3 29; p=0 0010 More glucose control Events Total 281 22 14 248 61 122 2605 1456 2635 1183 2220 3616 13 715 Less glucose control Events Total 198 28 2 218 29 100 2633 2895 2634 1185 2227 3610 15 184 Weight 9 5% 3 6% 0 7% 9 7% 4 8% 35 9% Heart failure risk ratio (95% CI) 1 43 (1 21 1 71) 1 56 (0 90 2 72) 7 03 (1 60 30 90) 1 14 (0 97 1 34) 2 10 (1 35 3 27) 1 22 (0 94 1 59) 1 42 (1 15 1 76) DPP-4 inhibitors 2013 EXAMINE 16,17 2013 SAVOR-TIMI 53 15 Subtotal 106 289 2701 8280 10 981 89 228 2679 8212 10 891 7 3% 9 5% 16 8% 1 19 (0 90 1 58) 1 27 (1 07 1 51) 1 25 (1 08 1 45) Heterogeneity: Tau 2 =0 00; χ 2 =0 15, df=1; p=0 70; I 2 =0% Test for overall effect: Z=2 94; p=0 0033 Intensive control 1998 UK Prospective Diabetes Study 35 2008 ACCORD 39 2008 ADVANCE 40 2009 VADT 43 Subtotal 80 152 220 76 2729 5128 5571 892 14 320 36 124 231 82 1138 5123 5569 899 12 729 5 5% 8 3% 9 3% 6 7% 29 8% 0 91 (0 62 1 34) 1 18 (0 93 1 49) 0 95 (0 79 1 14) 0 91 (0 66 1 25) 1 00 (0 88 1 13) Heterogeneity: Tau 2 =0 00; χ 2 =2 80, df=3; p=0 42; I 2 =0% Test for overall effect: Z=0 01; p=0 99 Insulin glargine 2012 ORIGIN 44 Subtotal 310 6264 6264 343 6273 6273 9 8% 9 8% 0 90 (0 77 1 05) 0 90 (0 77 1 05) Heterogeneity: not applicable Test for overall effect: Z=1 34; p=0 18 Weight loss 2013 Look-AHEAD 45 Subtotal 99 2570 2570 119 2575 2575 0 80 (0 62 1 04) 0 80 (0 62 1 04) Heterogeneity: not applicable Test for overall effect: Z=1 67; p=0 10 Total 47 850 Heterogeneity: Tau 2 =0 04; χ 2 =45 56, df=13; p<0 0001; I 2 =71% Test for overall effect: Z=2 04; p=0 041 Test for subgroup differences: χ 2 =21 85, df=4; p=0.00021, I 2 =81 7% 47 652 100% 1 14 (1 01 1 30) 0 2 0 5 1 2 5 Favours glucose-lowering Favours standard care Figure 3: Risk of heart failure events with glucose-lowering drugs or strategies versus standard care, stratified by strategy or drug class Determined with an inverse-variance random-effects model. PPAR=peroxisome proliferator-activated receptor. DPP=dipeptidyl peptidase. www.thelancet.com/diabetes-endocrinology Vol 3 May 2015 361
More glucose control Events Total Less glucose control Events Total Weight MACE risk ratio (95% CI) 1998 UK Prospective Diabetes Study 35 2005 PROactive 36 2006 ADOPT 37 2006 DREAM 38 2008 ACCORD 39 2008 ADVANCE 40 2009 BARI2D 41 2009 RECORD 42 2009 VADT 43 2012 ORIGIN 44 2013 EXAMINE 16,17 2013 Look-AHEAD 45 2013 SAVOR-TIMI 53 15 2014 AleCardio 46 387 301 27 32 352 557 261 154 64 1041 305 267 613 334 2729 2605 1456 2635 5128 5571 1183 2220 892 6264 2701 2570 8280 3616 188 358 41 23 371 590 288 165 78 1013 316 283 609 360 1138 2633 2895 2634 5123 5569 1185 2227 899 6273 2679 2575 8212 3610 5 7% 7 0% 0 7% 0 6% 0 6% 8 0% 12 0% 3 5% 1 6% 22 2% 4 7% 6 1% 12 3% 0 84 (0 71 1 00) 0 84 (0 72 0 98) 1 31 (0 81 2 38) 1 39 (0 81 2 38) 0 90 (0 78 1 04) 0 94 (0 84 1 06) 0 91 (0 78 1 05) 0 93 (0 75 1 16) 0 82 (0 59 1 14) 1 02 (0 94 1 11) 0 96 (0 79 1 16) 0 93 (0 79 1 10) 1 00 (0 89 1 12) 0 96 (0 83 1 11) Total 4695 47850 4683 47652 100% 0 95 (0 91 0 99) Heterogeneity: Tau 2 =0 00; χ 2 =13 48, df=13; p<0 0001; I 2 =4% Test for overall effect: Z=2 47; p=0 014 0 5 0 7 1 1 5 2 Favours glucose-lowering Favours standard care Figure 4: Risk of MACE with glucose-lowering drugs or strategies versus standard care Determined with an inverse-variance random-effects model. MACE=major adverse cardiovascular events. Major adverse cardiovascular events Expanded major adverse cardiovascular events All-cause death Cardiovascular death Myocardial infarction Stroke Heart failure Unstable angina Coronary revascularisation 0 6 0 8 1 1 2 1 4 Favours glucose lowering Favours standard care Risk ratio (95% CI) p interaction 0 95 (0 91 0 99) 0 23 0 96 (0 92 1 01) 0 17 0 99 (0 94 1 05) 0 67 0 98 (0 91 1 06) 0 92 0 92 (0 86 0 99) 0 15 0 99 (0 91 1 07) 0 27 1 14 (1 01 1 30) 0 0002 0 97 (0 84 1 11) 0 05 0 92 (0 91 1 04) 0 07 Figure 5: Major adverse cardiovascular events and individual cardiovascular events with glucose-lowering drugs or strategies versus standard care Determined with an inverse-variance random-effects model. with insulin glargine, 0 36 kg (0 29) with DPP-4 inhibitors, and 4 0 kg (1 0) with intensive weight loss. Overall, risk of heart failure was associated with weight gain (p=0 022 for the meta-regression model), with each 1 0 kg relative increase in weight between treatments associated with a 7 1% (95% CI 1 0 13 6) relative increase in the risk of heart failure (figure 6). Other trial characteristics associated with risk of heart failure with glucose-lowering drugs or strategies were mean duration of trial follow-up (a 13 7% [95% CI 7 9 19 2] greater relative risk reduction for heart failure for each 1 additional year of follow-up; p<0 0001), history of cardiovascular disease (each 1% increase in the proportion of patients with previous cardiovascular disease associated with a 0 7% [95% CI 0 2 1 2] relative increase in risk; p=0 0034), and the mean absolute HbA 1c achieved with glucose-lowering drugs or strategies (each 0 1% greater treatment-achieved HbA 1c associated with a 4 1% [95% CI 1 3 7 0] relative increase in risk; p=0 0041). The mean difference in HbA 1c achieved (p=0 49), baseline duration of diabetes (p=0 061), baseline proportion of patients with heart failure (p=0 16), and baseline age (p=0 44) had no significant effect modification on heart failure risk between glucoselowering drugs or strategies and controls. By contrast, we recorded a 3 5% (95% CI 0 4 6 5) greater reduction in relative risk for major adverse cardiovascular events with each 0 1% mean difference in HbA 1c achieved between glucose-lowering drugs or strategies and controls (p=0 027) and no association with weight gain (p=0 74). Visual inspection of funnel plots showed no evidence of publication bias affecting the results for heart failure or major adverse cardiovascular events (appendix pp 3, 4, 7). Results for the primary endpoint remained significant when we considered trials with interventions that resulted in a mean difference of 0 1% or less in HbA 1c 55 58 (RR 1 13, 95% CI 1 01 1 28; p=0 048), or those that enrolled fewer than 1000 individuals 47 49 (1 14, 1 001 1 28; p=0 038) with no additional evidence of publication bias (data not shown). However, with removal of the entire PPAR class of trials, the pooled effect of the other glucose-lowering drugs or strategies on the risk of heart failure was non-significant (RR 1 01, 95% CI 0 89 1 14; p=0 86), although heterogeneity in risk across the remaining glucose-lowering drugs or strategies remained significant (p for interaction=0 0045) 362 www.thelancet.com/diabetes-endocrinology Vol 3 May 2015
as the RRs for the other drug classes were unchanged. In addition, removal of the entire PPAR class of trials resulted in weight gain no longer being a significant modifier of the risk for heart failure by glucose-lowering drugs or strategies (p=0 97), but a significant association remained between heart failure and trial duration (12 6% [95% CI 6 0 18 7] greater relative risk reduction for heart failure for each 1 additional year of follow-up; p=0 00028), baseline duration of diabetes (each 1 year increase in diabetes duration associated with a 9 4% [2 3 16 9] relative increase in risk; p=0 0086), history of cardiovascular disease (each 1% increase in the proportion of patients with cardiovascular disease was associated with a 1 0% [0 2 1 7] relative increase in risk; p=0 011), history of heart failure (each 1% increase in the proportion of patients with a history of heart failure associated with a 2 9% [0 35 5 5] relative increase in risk; p=0 025), and the mean absolute HbA 1c achieved with glucose-lowering drugs or strategies (each 0 1% higher treatment achieved HbA 1c associated with a 3 4% [0 46 6 5%] relative increase in risk for heart failure; p=0 024). When considered in a multivariate metaregression model, no trial characteristic was an independent effect modifier of the risk of heart failure with glucose-lowering drugs or strategies (appendix p 8). Discussion This is the largest meta-analysis of the cardiovascular outcomes trials of glucose-lowering drugs or strategies tested in patients with or at risk for type 2 diabetes, and the first to focus on the risk of heart failure across various drugs or strategies. 21 25,60,61 In this meta-analysis of more than 95 000 patients, there were four major findings. Overall, compared with standard care, treatment with glucose-lowering drugs or strategies resulted in a 14% relative increase in the risk of heart failure. There was significant heterogeneity in the magnitude of this effect, with the highest risk for PPAR agonists, intermediate risk with DPP-4 inhibitors, and a neutral risk with insulin glargine, target-based intensive glycaemic control, and intensive weight-loss strategies. Several other trial characteristics correlated with the risk of heart failure with glucose-lowering drugs or strategies, including weight gain, trial duration, and absolute glycaemic concentration achieved. The relative increase in the risk of heart failure seemed to outweigh a 5% relative risk reduction in major adverse cardiovascular events that was mainly driven by a reduction in myocardial infarction, although the absolute increase in risk of heart failure with glucose-lowering drugs or strategies was only 0 51%. The mechanism of action linking the increased risk of heart failure with glucose lowering remains controversial. 1,62 Our findings extend those of previous reports and confirm the substantial heterogeneity in risk of heart failure across drug classes and strategies of glucose lowering. With PPAR agonist-based interventions, the mechanism seems to be driven by an increase in Log risk ratio for heart failure 2 00 1 76 1 52 1 28 2009 RECORD 42 1 04 2013 Look-AHEAD 45 0 80 2014 AleCardio 46 2006 ADOPT 37 0 56 2013 SAVOR-TIMI 53 15 2013 EXAMINE 16,17 2005 PROactive 36 0 32 2009 BARI2D 41 2008 ACCORD 39 0 08 1998 UK Prospective Diabetes Study 35 0 16 2008 ADVANCE 40-0 40 2012 ORIGIN 44 2009 VADT 43 5 09 3 78 2 47 1 17 0 14 1 45 2 76 4 07 5 37 6 68 7 99 Change in weight (kg) 2006 DREAM 38 Figure 6: Relation between weight gain and heart failure Slope of the regression line is +0 0298, which after exponentiation, translates into a 7 1% (95% CI 1 0 13 6) relative greater risk for heart failure for every 1 0 kg increase in weight between glucose-lowering drugs or strategies versus standard care (p=0 022). weight, partly due to fluid retention, resulting in adverse haemodynamic consequences. 14 Our meta-regression analysis suggests that other glucose-lowering drugs or strategies might also be associated with an increased risk of heart failure despite a neutral or slight lowering effect on weight (eg, DPP-4 inhibitors). Besides weight gain, which did not show a consistent correlation with risk of heart failure when the PPAR-based trials were excluded, trial duration and baseline proportion of patients with established cardiovascular risk were strongly associated with heart failure. The association of baseline proportion of patients with established cardiovascular risk might be expected to indicate detection of potential harm in a secondary prevention diabetes population more sensitive to adverse haemodynamic effects or presence of nephropathy. 63 By contrast, the effect of trial duration was a surprising finding and supports calls for trials of cardiovascular outcomes in patients with type 2 diabetes to have extended follow-up to enable better assessment of the association between glucose lowering and heart failure. 1,20 Short trial duration might simply be a marker of a trial design that tested an intervention potentially associated with heart failure, or a marker of a potentially underpowered trial resulting in the chance for a falsepositive finding (type 1 error) when reporting a significantly harmful (or protective) result. Further results from ongoing trials can be compared with the data we analysed to confirm or refute these hypotheses. The mean difference in HbA 1c achieved between treatment strategies and duration of diabetes did not significantly affect the risk of heart failure with glucoselowering drugs or strategies. Combined with our finding of an association between the mean absolute HbA 1c achieved with glucose-lowering drugs or strategies and reduced risk of heart failure, these observations do not www.thelancet.com/diabetes-endocrinology Vol 3 May 2015 363
seem to support the myocardial defuelling hypothesis, which posits that any pharmacological reduction in blood glucose concentrations (and subsequently insulin concentrations) might be a mechanism of adverse haemodynamic effect in patients with type 2 diabetes with prolonged myocardial exposure to elevated concentrations of these substrates. 64,65 Our study has some limitations that are generally inherent to meta-analyses. We combined data from trials that varied in study design, intervention and controls, extent of glucose lowering, and definitions of cardiovascular endpoints including heart failure. However, we grouped interventions that were similar in the main mechanism of action or intended surrogate marker for modification, control therapies were contemporary standards of care with or without the use of a masked placebo, and endpoints involved common diagnoses, albeit of potentially varied severity and prognostic importance. Nevertheless, our approach might attenuate the summary effect of glucose lowering on major cardiovascular adverse events compared with previous, smaller, meta-analyses; 24 similarly, our results for the effect of glucose lowering on heart failure might also be an underestimate. Second, no individual patientlevel data were available. However, the primary results of this analysis would not be expected to substantially change with individual patient-level data, although such an approach might further delineate the independent effect of effect-modifying variables. Third, several studies had design limitations, including absence of masking and premature treatment discontinuation, that might restrict interpretation of the true effect of glucose lowering on heart failure. Nevertheless, masked adjudication of this outcome was done in all but two trials, and, when done in retrospect in one trial, confirmed the accuracy of investigator-reported serious events of heart failure. 66 Finally, despite the power of meta-regression to provide additional insight into causes of heterogeneity in meta-analysis, inferences from results must be considered hypothesis-generating. In summary, glucose lowering by various drugs or strategies might increase the risk of heart failure compared with standard care in patients with or at risk for type 2 diabetes. The magnitude of this risk seems to be driven by specific drug classes and correlated with other trial characteristics, including weight gain and trial duration. Conversely, glucose lowering by various drugs or strategies modestly decreased the risk of myocardial infarction. Clinicians should consider the trade-off between ischaemic and haemodynamic cardiovascular events when choosing between different drugs or strategies for lowering blood glucose. Contributors JAU did the literature search and data analysis, and drafted the figures. JAU and MAC did data collection. All authors analysed the data and interpreted the findings. JAU wrote the first draft of the manuscript and all authors provided critical revision of the manuscript for important intellectual content. Declaration of interests JAU reports that Brigham and Women s Hospital and the TIMI Study Group, with whom he was a trainee, received grant support from AstraZeneca and Bristol-Myers Squibb to undertake SAVOR-TIMI 53 before doing the present study; grants from the American College of Cardiology Foundation; and fees for consultancy from Cisbio Bioassays and Novartis outside the submitted work. MAC reports that Brigham and Women s Hospital and the TIMI Study Group received grant support from AstraZeneca and Bristol-Myers Squibb to conduct SAVOR-TIMI 53 prior to the conduct of the present study; and that he has been a consultant for AstraZeneca and Merck. DLB has received research grant support from Amarin, AstraZeneca, Bristol-Myers Squibb, Eisai, Ethicon, Forest Laboratories, Ischemix, Medtronic, Pfizer, Roche, Sanofi-Aventis, and The Medicines Company; unfunded research for FlowCo, PLx Pharma, Takeda; has been an advisory board consultant for Cardax, Elsevier Practice Update Cardiology, Medscape Cardiology, Regado Biosciences; was a member of the board of directors of the Boston VA Research Institute, Society of Cardiovascular Patient Care; was Chair of the American Heart Association Get With The Guidelines Steering Committee; was a member of the data monitoring committee of the Duke Clinical Research Institute, Harvard Clinical Research Institute, Mayo Clinic, Population Health Research Institute; and has received honoraria from American College of Cardiology (Senior Associate Editor, Clinical Trials and News), Belvoir Publications (Editor in Chief, Harvard Heart Letter), Duke Clinical Research Institute (clinical trial steering committees), Harvard Clinical Research Institute (clinical trial steering committee), HMP Communications (Editor in Chief, Journal of Invasive Cardiology), Journal of the American College of Cardiology (Associate Editor; Section Editor, Pharmacology), Population Health Research Institute (clinical trial steering committee), Slack Publications (Chief Medical Editor, Cardiology Today s Intervention), WebMD (CME steering committees), and Clinical Cardiology (Deputy Editor). MEF has received honoraria from the American College of Cardiology (Senior Associate Editor, Journal of the American College of Cardiology). BMS has received research grant support through the TIMI Study Group and Brigham and Women s Hospital from AstraZeneca and Bristol-Myers Squibb, Daiichi-Sankyo, GlaxoSmithKline, Johnson and Johnson, Bayer Healthcare, Gilead, Eisai, Merck; and fees for consultancy from Arena, AstraZeneca, Boehringer-Ingelheim, Decision Resources, GlaxoSmithKline, GE Healthcare, Gilead, Lexicon, Merck, Eisai, St Jude s Medical, Forest Pharmaceuticals, Bristol-Myers Squibb, Boston Clinical Research Institute, Covance, University of Calgary, and Elsevier Practice Update Cardiology. SC declares no competing interests. Acknowledgments JAU is supported by the Heart and Stroke Foundation of Canada; Canadian Institutes for Health Research; Women s College Research Institute and Department of Medicine, Women s College Hospital; Peter Munk Cardiac Centre, University Health Network; Department of Medicine and Heart and Stroke/Richard Lewar Centre of Excellence in Cardiovascular Research, University of Toronto. References 1 McMurray JV, Gerstein HC, Holman RR, Pfeffer MA. Heart failure: a cardiovascular outcome in diabetes that can no longer be ignored. Lancet Diabetes Endocrinol 2014; 2: 843 51. 2 Kannel WB, Hjortland M, Castelli WP. Role of diabetes in congestive heart failure: the Framingham study. Am J Cardiol 1974; 34: 29 34. 3 Stratton IM, Adler AI, Neil HA, et al. Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ 2000; 321: 405 12. 4 Nichols GA, Gullion CM, Koro CE, Ephross SA, Brown JB. The incidence of congestive heart failure in type 2 diabetes: an update. Diabetes Care 2004; 27: 1879 84. 5 Iribarren C, Karter AJ, Go AS, et al. Glycemic control and heart failure among adult patients with diabetes. Circulation 2001; 103: 2668 73. 6 van Melle JP, Bot M, de Jonge P, de Boer RA, van Veldhuisen DJ, Whooley MA. Diabetes, glycemic control, and new-onset heart failure in patients with stable coronary artery disease: data from the heart and soul study. Diabetes Care 2010; 33: 2084 89. 364 www.thelancet.com/diabetes-endocrinology Vol 3 May 2015
7 Ingelsson E, Sundström J, Arnlöv J, Zethelius B, Lind L. Insulin resistance and risk of congestive heart failure. JAMA 2005; 294: 334 41. 8 Nielson C, Lange T. Blood glucose and heart failure in nondiabetic patients. Diabetes Care 2005; 28: 607 11. 9 Thrainsdottir IS, Aspelund T, Thorgeirsson G, et al. The association between glucose abnormalities and heart failure in the population-based Reykjavik study. Diabetes Care 2005; 28: 612 16. 10 Held C, Gerstein HC, Yusuf S, et al, and the ONTARGET/ TRANSCEND Investigators. Glucose levels predict hospitalization for congestive heart failure in patients at high cardiovascular risk. Circulation 2007; 115: 1371 75. 11 Matsushita K, Blecker S, Pazin-Filho A, et al. The association of hemoglobin a1c with incident heart failure among people without diabetes: the atherosclerosis risk in communities study. Diabetes 2010; 59: 2020 26. 12 Zhang Y, Hu G, Yuan Z, Chen L. Glycosylated hemoglobin in relationship to cardiovascular outcomes and death in patients with type 2 diabetes: a systematic review and meta-analysis. PLoS One 2012; 7: e42551. 13 Seshasai SR, Kaptoge S, Thompson A, et al, and the Emerging Risk Factors Collaboration. Diabetes mellitus, fasting glucose, and risk of cause-specific death. N Engl J Med 2011; 364: 829 41. 14 Nesto RW, Bell D, Bonow RO, et al, and the American Heart Association, and the American Diabetes Association. Thiazolidinedione use, fluid retention, and congestive heart failure: a consensus statement from the American Heart Association and American Diabetes Association. October 7, 2003. Circulation 2003; 108: 2941 48. 15 Scirica BM, Bhatt DL, Braunwald E, et al, and the SAVOR-TIMI 53 Steering Committee and Investigators. Saxagliptin and cardiovascular outcomes in patients with type 2 diabetes mellitus. N Engl J Med 2013; 369: 1317 26. 16 White WB, Cannon CP, Heller SR, et al, and the EXAMINE Investigators. Alogliptin after acute coronary syndrome in patients with type 2 diabetes. N Engl J Med 2013; 369: 1327 35. 17 Zannad F, Cannon CP, Cushman WC, et al. Heart failure and mortality outcomes in patients with type 2 diabetes taking alogliptin versus placebo in EXAMINE: a multicentre randomised double-blind trial. Lancet 2015; published online March 10. http:// dx.doi.org/10.1016/s0140-6736(14)62225-x. 18 Scirica BM, Braunwald E, Raz I, et al, and the SAVOR-TIMI 53 Steering Committee and Investigators*. Heart failure, saxagliptin, and diabetes mellitus: observations from the SAVOR-TIMI 53 randomized trial. Circulation 2014; 130: 1579 88. 19 Wong YW, Thomas L, Sun JL, et al. Predictors of incident heart failure hospitalizations among patients with impaired glucose tolerance: insight from the Nateglinide And Valsartan in Impaired Glucose Tolerance Outcomes Research study. Circ Heart Fail 2013; 6: 203 10. 20 Holman RR, Sourij H, Califf RM. Cardiovascular outcome trials of glucose-lowering drugs or strategies in type 2 diabetes. Lancet 2014; 383: 2008 17. 21 Lago RM, Singh PP, Nesto RW. Congestive heart failure and cardiovascular death in patients with prediabetes and type 2 diabetes given thiazolidinediones: a meta-analysis of randomised clinical trials. Lancet 2007; 370: 1129 36. 22 Selvin E, Bolen S, Yeh HC, et al. Cardiovascular outcomes in trials of oral diabetes medications: a systematic review. Arch Intern Med 2008; 168: 2070 80. 23 Kelly TN, Bazzano LA, Fonseca VA, Thethi TK, Reynolds K, He J. Systematic review: glucose control and cardiovascular disease in type 2 diabetes. Ann Intern Med 2009; 151: 394 403. 24 Ray KK, Seshasai SR, Wijesuriya S, et al. Effect of intensive control of glucose on cardiovascular outcomes and death in patients with diabetes mellitus: a meta-analysis of randomised controlled trials. Lancet 2009; 373: 1765 72. 25 Castagno D, Baird-Gunning J, Jhund PS, et al. Intensive glycemic control has no impact on the risk of heart failure in type 2 diabetic patients: evidence from a 37,229 patient meta-analysis. Am Heart J 2011; 162: 938-48. 26 Monami M, Dicembrini I, Mannucci E. Dipeptidyl peptidase-4 inhibitors and heart failure: a meta-analysis of randomized clinical trials. Nutr Metab Cardiovasc Dis 2014; 24: 689 97. 27 Wu S, Hopper I, Skiba M, Krum H. Dipeptidyl peptidase-4 inhibitors and cardiovascular outcomes: meta-analysis of randomized clinical trials with 55,141 participants. Cardiovasc Ther 2014; 32: 147 58. 28 Higgins JPT, Green S. Cochrane Collaboration. Cochrane handbook for systematic reviews of interventions. Chichester, England; Hoboken, NJ: Wiley-Blackwell; 2008. 29 Moher D, Liberati A, Tetzlaff J, Altman DG, and the PRISMA Group. Preferred reporting items for systematic reviews and metaanalyses: the PRISMA statement. BMJ 2009; 339: b2535. 30 Higgins JP, Thompson SG. Quantifying heterogeneity in a metaanalysis. Stat Med 2002; 21: 1539 58. 31 Altman DG, Bland JM. Interaction revisited: the difference between two estimates. BMJ 2003; 326: 219. 32 Sterne JA, Egger M. Funnel plots for detecting bias in meta-analysis: guidelines on choice of axis. J Clin Epidemiol 2001; 54: 1046 55. 33 Sterne JA, Sutton AJ, Ioannidis JP, et al. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. BMJ 2011; 343: d4002. 34 Duval S, Tweedie R. Trim and fill: A simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics 2000; 56: 455 63. 35 UK Prospective Diabetes Study (UKPDS) Group. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet 1998; 352: 837 53. 36 Dormandy JA, Charbonnel B, Eckland DJ, et al, and the PROactive investigators. Secondary prevention of macrovascular events in patients with type 2 diabetes in the PROactive Study (PROspective pioglitazone Clinical Trial In macrovascular Events): a randomised controlled trial. Lancet 2005; 366: 1279 89. 37 Kahn SE, Haffner SM, Heise MA, et al, and the ADOPT Study Group. Glycemic durability of rosiglitazone, metformin, or glyburide monotherapy. N Engl J Med 2006; 355: 2427 43. 38 Gerstein HC, Yusuf S, Bosch J, et al, and the DREAM (Diabetes REduction Assessment with ramipril and rosiglitazone Medication) Trial Investigators. Effect of rosiglitazone on the frequency of diabetes in patients with impaired glucose tolerance or impaired fasting glucose: a randomised controlled trial. Lancet 2006; 368: 1096 105. 39 Gerstein HC, Miller ME, Byington RP, et al, and the Action to Control Cardiovascular Risk in Diabetes Study Group. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008; 358: 2545 59. 40 Patel A, MacMahon S, Chalmers J, et al, and the ADVANCE Collaborative Group. Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes. N Engl J Med 2008; 358: 2560 72. 41 Frye RL, August P, Brooks MM, et al, and the BARI 2D Study Group. A randomized trial of therapies for type 2 diabetes and coronary artery disease. N Engl J Med 2009; 360: 2503 15. 42 Home PD, Pocock SJ, Beck-Nielsen H, et al, and the RECORD Study Team. Rosiglitazone evaluated for cardiovascular outcomes in oral agent combination therapy for type 2 diabetes (RECORD): a multicentre, randomised, open-label trial. Lancet 2009; 373: 2125 35. 43 Duckworth W, Abraira C, Moritz T, et al, and the VADT Investigators. Glucose control and vascular complications in veterans with type 2 diabetes. N Engl J Med 2009; 360: 129 39. 44 Gerstein HC, Bosch J, Dagenais GR, et al, and the ORIGIN Trial Investigators. Basal insulin and cardiovascular and other outcomes in dysglycemia. N Engl J Med 2012; 367: 319 28. 45 Wing RR, Bolin P, Brancati FL, et al, and the Look AHEAD Research Group. Cardiovascular effects of intensive lifestyle intervention in type 2 diabetes. N Engl J Med 2013; 369: 145 54. 46 Lincoff AM, Tardif JC, Schwartz GG, et al, and the 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; 311: 1515 25. 47 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; 352: 854 65. www.thelancet.com/diabetes-endocrinology Vol 3 May 2015 365