Exploration of regional heterogeneity in trials assessing the efficacy of recent non-insulin glucose lowering drugs on cardiovascular outcomes

-Data extractionData extraction will be conducted independently by two reviewers. Any discrepancy will be solved by a third reviewer. Data per region are usually available only for the primary outcome. We therefore address a request to the sponsors of these trials to obtain individual patient data, or aggregated data (to ensure confidentiality) per region for all the variables of interest. The following data are requested, from most to least preferred:1.   Data by country. In order to avoid any breach of confidentiality regarding personal data, countries for which 15 patients or less were included in a single trial will be excluded. Data by country will allow addressing whether variability is related to geographic variations or development status.2.   Alternatively, aggregated data by region, and by development status will be sought. There will be 14 regions, following the UN definition: Northern America, Latin America and the Caribbean, Northern Africa, Sub-Saharan Africa, Northern Europe, Southern Europe, Eastern Europe, Western Europe, Central Asia, Eastern Asia, South-eastern Asia, Southern Asia, Western Asia and Oceania; and there will be 2 development status: developed and developing (Table 1).The following variables will be extracted: o   Characteristics of the study (authors, year of publication, regions, number of participants by country, follow-up duration, study design); o   Cardiovascular outcomes (main outcomes and additional outcomes listed above).o   Characteristics of participants at baseline (age, sex, ethnicity, BMI, heart rate, systolic and diastolic pressure, HbA1c, fasting glucose, eGFR, total cholesterol, HDL cholesterol, LDL cholesterol, triglycerides, current smoking, family history of cardiovascular disease, hypertension, coronary artery disease, peripheral artery disease, cerebrovascular disease, heart failure, stroke, myocardial infarction, atrial fibrillation, duration of diabetes, retinopathy, neuropathy, nephropathy); o   Concomitant medication at baseline (insulin, metformin, sulfonylurea, DPP4 inhibitor, SGLT-2 inhibitor, GLP-1 analog, thiazolidinediones, antiplatelet agents, angiotensin converting enzyme inhibitor or angiotensin II receptor blockers, beta-blocker, statin, ezetimibe, diuretics)- Data aggregation per region or countryIf we are granted access to individual patient data, we will first recalculate Cox proportional-hazards models for the primary and secondary outcomes, with treatment group only as an explanatory variable, for each country or region included in the trial (according to the data provided, see above). The primary analysis will be conducted following the intent-to-treat principle. We will use the time to the first occurrence of a MACE over the full duration of follow-up. We will also conduct a descriptive analysis of all others variables (listed above) at baseline, and per country or region.- Meta-analysisWe will perform multilevel meta-analyses with recalculated hazard ratios for each outcome. Then subgroup meta-analyses according to the region of inclusion will be conducted. . A test of subgroup difference <0.05 will be considered significant, as previously done in our preliminary assessment (Roustit et al, N Engl J Med 2017, 376;12:1196-97) (8). A random effect model will be used in case of important heterogeneity (I2>50%). Lastly, several meta-regressions will be performed to assess the potential influence of the potential variables of interest (type of glucose-lowering drug, regional development status, baseline characteristics of patients, study characteristics) on the outcomes. Regions and development status will be defined according to the UN classification (11). All statistical analyses will be performed using R statistical software (version 3.2.4), with the metafor and meta packages.