PDB2 - Bayesian Network Meta-Analysis to Assess the Relative Efficacy and Safety of Canagliflozin in Patients with Type 2 Diabetes Mellitus (T2DM) Inadequately Controlled on Metformin and Sulphonylurea (MET+SU)

Abstract

To assess the relative efficacy and safety of canagliflozin (CANA), a sodium-glucose co-transporter inhibitor, as an add-on to MET+SU, versus DPP-4 inhibitors, GLP-1 agonists and insulin, using Bayesian network meta-analysis (NMA). A systematic literature review was conducted according to NICE guidelines. Outcomes of interest included HbA1c, weight and hypoglycaemia. A Bayesian NMA using non-informative priors was conducted, based on linking trials with treatment and dose-specific common treatment arms. Assessment of model fit and selection of fixed versus random effects was based on the Deviance Information Criterion (DIC). Sensitivity analyses assessed the impact of individual trials and definition of priors. Consistency between direct and indirect evidence was assessed. Ten studies reporting results at 26 weeks +/- 4 weeks were identified. HbA1c-reduction (D) for CANA 100mg was comparable to DPP-4s (D between 0.05 and -0.14 versus sitagliptin and linagliptin respectively, with pairwise probabilities (P) of being more effective between 33-88%), and higher for CANA 300mg, which was comparable to GLP-1s (D=0.08;P=31% and 0.01;P=53%) versus liraglutide 1.8mg and exenatide 10µg respectively) and biphasic insulin (D=0.03;P=43%). CANA 300mg had the highest weight reduction with changes between 0.14kg;P=93% (vs. exenatide 10µg) and 5.13kg;P=100% (vs. biphasic insulin). The odds ratio for hypoglycaemia versus long-acting insulin were 0.31 and 0.39 for CANA 100mg and 300mg respectively, compared to 0.20-0.41 for other classes. NMA of add-on therapies to MET+SU suggests that glycaemic reductions for CANA at 26 weeks are at least as large for CANA 100 mg and greater for CANA 300 mg compared to DPP-4s. CANA 300mg was found to be comparable to liraglutide 1.8mg and biphasic insulin. Weight reduction was similar to GLP-1s and substantially higher compared to all other classes. All classes showed significantly lower hypoglycaemic event rates compared to insulin.