Individual Patient Data Meta-analysis of Randomised Controlled Trials Comparing Glycaemic Control During Continuous Subcutaneous Insulin Infusion vs. Multiple Daily Insulin Injections in Type 2 Diabetes Mellitus

The investigators will identify randomised controlled trials without language restriction that meet the inclusion criteria by searching the Cochrane database, Ovid Medline, Google Scholar and other sources. The investigators will exclude observational studies, short-term trials <2 months duration, studies in pregnant diabetic subjects, newly diagnosed type 2 diabetes, trials in type 1 diabetes and extensions of previous studies.

Data on individual participants will be obtained directly from trialists and from funding sponsors who hold the trial data and will include age, sex, duration of diabetes, treatment group, baseline and final HbA1c, baseline and final insulin dose, baseline and final BMI.

A 'two-step' meta-analysis will be performed to estimate overall mean differences in HbA1c, insulin dose and weight/BMI for the trials. Then, a 'one-step' meta-regression analysis will be conducted by creating a single large dataset from the individual patient data. Determinants of final HbA1c, BMI/weight and insulin dose will be explored using Bayesian approaches with covariates that include baseline HbA1c, age, diabetes duration, BMI, insulin dose and interactions between the covariates.