Kanie, Takayoshi, Mizuno, Atsushi, Takaoka, Yoshimitsu, Suzuki, Takahiro, Yoneoka, Daisuke, Nishikawa, Yuri, Tam, Wilson Wai San, Morze, Jakub, Rynkiewicz, Andrzej, Xin, Yiqiao, Wu, Olivia, Providencia, Rui, and Kwong, Joey SW
Background:\ud Cardiovascular disease (CVD) is a leading cause of death globally. Recently, dipeptidyl peptidase‐4 inhibitors (DPP4i), glucagon‐like peptide‐1 receptor agonists (GLP‐1RA) and sodium‐glucose co‐transporter‐2 inhibitors (SGLT2i) were approved for treating people with type 2 diabetes mellitus. Although metformin remains the first‐line pharmacotherapy for people with type 2 diabetes mellitus, a body of evidence has recently emerged indicating that DPP4i, GLP‐1RA and SGLT2i may exert positive effects on patients with known CVD.\ud \ud Objectives:\ud To systematically review the available evidence on the benefits and harms of DPP4i, GLP‐1RA, and SGLT2i in people with established CVD, using network meta‐analysis.\ud \ud Search methods:\ud We searched CENTRAL, MEDLINE, Embase, and the Conference Proceedings Citation Index on 16 July 2020. We also searched clinical trials registers on 22 August 2020. We did not restrict by language or publication status.\ud \ud Selection criteria:\ud We searched for randomised controlled trials (RCTs) investigating DPP4i, GLP‐1RA, or SGLT2i that included participants with established CVD. Outcome measures of interest were CVD mortality, fatal and non‐fatal myocardial infarction, fatal and non‐fatal stroke, all‐cause mortality, hospitalisation for heart failure (HF), and safety outcomes.\ud \ud Data collection and analysis:\ud Three review authors independently screened the results of searches to identify eligible studies and extracted study data. We used the GRADE approach to assess the certainty of the evidence. We conducted standard pairwise meta‐analyses and network meta‐analyses by pooling studies that we assessed to be of substantial homogeneity; subgroup and sensitivity analyses were also pursued to explore how study characteristics and potential effect modifiers could affect the robustness of our review findings. We analysed study data using the odds ratios (ORs) and log odds ratios (LORs) with their respective 95% confidence intervals (CIs) and credible intervals (Crls), where appropriate. We also performed narrative synthesis for included studies that were of substantial heterogeneity and that did not report quantitative data in a usable format, in order to discuss their individual findings and relevance to our review scope.\ud \ud Main results:\ud We included 31 studies (287 records), of which we pooled data from 20 studies (129,465 participants) for our meta‐analysis. The majority of the included studies were at low risk of bias, using Cochrane's tool for assessing risk of bias. Among the 20 pooled studies, six investigated DPP4i, seven studied GLP‐1RA, and the remaining seven trials evaluated SGLT2i. All outcome data described below were reported at the longest follow‐up duration.\ud \ud 1. DPP4i versus placebo:\ud \ud Our review suggests that DPP4i do not reduce any risk of efficacy outcomes: CVD mortality (OR 1.00, 95% CI 0.91 to 1.09; high‐certainty evidence), myocardial infarction (OR 0.97, 95% CI 0.88 to 1.08; high‐certainty evidence), stroke (OR 1.00, 95% CI 0.87 to 1.14; high‐certainty evidence), and all‐cause mortality (OR 1.03, 95% CI 0.96 to 1.11; high‐certainty evidence). DPP4i probably do not reduce hospitalisation for HF (OR 0.99, 95% CI 0.80 to 1.23; moderate‐certainty evidence). DPP4i may not increase the likelihood of worsening renal function (OR 1.08, 95% CI 0.88 to 1.33; low‐certainty evidence) and probably do not increase the risk of bone fracture (OR 1.00, 95% CI 0.83 to 1.19; moderate‐certainty evidence) or hypoglycaemia (OR 1.11, 95% CI 0.95 to 1.29; moderate‐certainty evidence). They are likely to increase the risk of pancreatitis (OR 1.63, 95% CI 1.12 to 2.37; moderate‐certainty evidence).\ud \ud 2. GLP‐1RA versus placebo:\ud \ud Our findings indicate that GLP‐1RA reduce the risk of CV mortality (OR 0.87, 95% CI 0.79 to 0.95; high‐certainty evidence), all‐cause mortality (OR 0.88, 95% CI 0.82 to 0.95; high‐certainty evidence), and stroke (OR 0.87, 95% CI 0.77 to 0.98; high‐certainty evidence). GLP‐1RA probably do not reduce the risk of myocardial infarction (OR 0.89, 95% CI 0.78 to 1.01; moderate‐certainty evidence), and hospitalisation for HF (OR 0.95, 95% CI 0.85 to 1.06; high‐certainty evidence). GLP‐1RA may reduce the risk of worsening renal function (OR 0.61, 95% CI 0.44 to 0.84; low‐certainty evidence), but may have no impact on pancreatitis (OR 0.96, 95% CI 0.68 to 1.35; low‐certainty evidence). We are uncertain about the effect of GLP‐1RA on hypoglycaemia and bone fractures.\ud \ud 3. SGLT2i versus placebo:\ud \ud This review shows that SGLT2i probably reduce the risk of CV mortality (OR 0.82, 95% CI 0.70 to 0.95; moderate‐certainty evidence), all‐cause mortality (OR 0.84, 95% CI 0.74 to 0.96; moderate‐certainty evidence), and reduce the risk of HF hospitalisation (OR 0.65, 95% CI 0.59 to 0.71; high‐certainty evidence); they do not reduce the risk of myocardial infarction (OR 0.97, 95% CI 0.84 to 1.12; high‐certainty evidence) and probably do not reduce the risk of stroke (OR 1.12, 95% CI 0.92 to 1.36; moderate‐certainty evidence). In terms of treatment safety, SGLT2i probably reduce the incidence of worsening renal function (OR 0.59, 95% CI 0.43 to 0.82; moderate‐certainty evidence), and probably have no effect on hypoglycaemia (OR 0.90, 95% CI 0.75 to 1.07; moderate‐certainty evidence) or bone fracture (OR 1.02, 95% CI 0.88 to 1.18; high‐certainty evidence), and may have no impact on pancreatitis (OR 0.85, 95% CI 0.39 to 1.86; low‐certainty evidence).\ud \ud 4. Network meta‐analysis:\ud \ud Because we failed to identify direct comparisons between each class of the agents, findings from our network meta‐analysis provided limited novel insights. Almost all findings from our network meta‐analysis agree with those from the standard meta‐analysis. GLP‐1RA may not reduce the risk of stroke compared with placebo (OR 0.87, 95% CrI 0.75 to 1.0; moderate‐certainty evidence), which showed similar odds estimates and wider 95% Crl compared with standard pairwise meta‐analysis. Indirect estimates also supported comparison across all three classes. SGLT2i was ranked the best for CVD and all‐cause mortality.\ud \ud Authors' conclusions:\ud Findings from both standard and network meta‐analyses of moderate‐ to high‐certainty evidence suggest that GLP‐1RA and SGLT2i are likely to reduce the risk of CVD mortality and all‐cause mortality in people with established CVD; high‐certainty evidence demonstrates that treatment with SGLT2i reduce the risk of hospitalisation for HF, while moderate‐certainty evidence likely supports the use of GLP‐1RA to reduce fatal and non‐fatal stroke. Future studies conducted in the non‐diabetic CVD population will reveal the mechanisms behind how these agents improve clinical outcomes irrespective of their glucose‐lowering effects.