GLP-1 REceptor Agonists and Real World EvIdeNce

Introduction. Glucagon-like peptide-1 receptor agonists (GLP-1RA) are prioritized as a second-line therapy for the treatment of type 2 diabetes (T2D), especially in patients with a history of cardiovascular disease (CVD) or when there is a need to avoid weight gain and hypoglycemia. Randomized controlled trials (RCTs) have shown that addition of GLP-1RA is more effective in reducing HbA1c than addition of basal insulin (BI) in patients with uncontrolled T2D on oral therapy. GLP-1RA are associated with a low risk of hypoglycaemia and are provided with extra-glycemic effects, including reduction of body weight and blood pressure. Notably, GLP-1RA as a class improve cardiovascular outcomes in T2D patients with established CVD. These benefits justify the positioning of GLP-1RA as the first injectable therapy in most T2D patients and an alternative to insulin.

In the absence of data from RCTs, observational studies can provide medium-level evidence to inform clinical practice, if well-designed and carefully conducted. Real-world studies are hypothesis-generating and cannot substitute for RCTs, but they can guide the design of dedicated RCTs. Retrospective real-world studies are particularly attractive as they can rapidly gather data from large heterogeneous populations that are representative of those seen in routine clinical practice.

Design. The GLP-1REWIN study is a retrospective, multicenter, real-world study on T2D patients initiating GLP-1RA in the routine clinical practice of Italian diabetes outpatient clinics. The study will be conducted at 6 diabetes specialist outpatient clinics in the Veneto Region, North-East Italy.

Objective. The general objective of the study is to evaluate effectiveness of GLP-1RA on glycemic and extra-glycemic endpoints in the real world clinical practice from 2010 to 2018. The study will be conducted at diabetes Centers because only diabetologists were allowed to prescribe GLP-1RA in Italy during the study period.

Methods. Data will be collected retrospectively by automatically interrogating the same electronic chart at all Centers. A dedicated software was developed to extract all relevant anonymized patient information into a clinical research form without manual intervention. We will collect data on all patients aged 18-80 years, with a diagnosis of T2D since at least 1 year (as recorded in the chart), who initiated one GLP-1RA available on the market between 1st Jan 2010 to 31st Dec 2018. These included exenatide twice daily and once weekly, liraglutide, lixisenatide, dulaglutide and fixed ratio combinations of BI/GLP-1RA. The baseline visit date will be set as the date when the patients attended the outpatient clinic and received for the first time a new prescription of GLP-1RA. The following clinical characteristics and laboratory data will be collected from the electronic chart up to 90 days before baseline: age, sex, diabetes duration, body height and weight, body mass index (BMI), waist circumference, systolic and diastolic blood pressure (SBP and DBP), heart rate, fasting plasma glucose (FPG), HbA1c, total cholesterol, HDL cholesterol, triglycerides (LDL cholesterol was calculated using Friedwald's equation), liver enzymes, serum creatinine (eGFR was calculated using the chronic kidney diseae [CKD]-Epidemiology [EPI] equation), urinary albumin excretion rate (UAER, expressed as mg/g of urinary creatinine). Details on chronic complication, as reported by international classification of disease (ICD)-9 codes in the electronic charts wil be used to define the presence of micro- and macroangiopathy. Microangiopathy will be defined as any of the following: UAER >30 mg/g; eGFR<60 ml/min/1.73 m2; diabetic retinopathy (any stage) or diabetic macular edema; peripheral or autonomic neuropathy. Macroangiopathy will be defined as any of the following: peripheral arterial disease or peripheral revascularization; history of stroke/transient ischemic attack or carotid revascularization; ischemic heart disease, coronary artery disease, history of myocardial infarction, or coronary revascularization. Information on concomitant medications for the treatment of diabetes and of other cardiovascular risk factors will also be recorded. Detailed dose data will be collected for insulin and GLP-1RA. After having set the baseline date, we will identify the first available follow-up visit attended by the patients at the same Clinic at least 3 months after baseline. Updated values of HbA1c, FPG, SBP and body weight were recorded at the end of follow-up, along with updated information on medications and dosages of GLP-RA.

We will compare the changes in glycemic (HbA1c and fasting plasma glucose) and extra-glycemic (body weight and systolic blood pressure) effectiveness parameters between patients in two groups.

Different groups of GLP-1RA therapies will be compared:

• Long-acting (e.g. dulaglutide and exenatide once weekly) versus short acting (exenatide, liraglutide and lixisenatide)
• Fixed versus flexible combinations of GLP-1RA and basal insulin.
• GLP-1RA with similarities to human GLP-1 (e.g. liraglutide) versus exendin-based GLP-1RA (e.g. exenatide).

Statistics. Continuous variables will be expressed as mean ± standard deviation (SD) if normally distributed or as median (interquartile range) if non-normally distributed. Non-normal variables will be log-transformed before being analyzed with parametric tests. Categorical variables will be expressed as percentage. The comparison of baseline characteristics between two groups will be performed using unpaired 2-tail Student's t test for continuous variables and with chi square for categorical variables. To evaluate the balance between two groups, in addition to p-values, we will calculate the standardized mean difference (SMD). The intra-group change in effectiveness endpoint variables from baseline to end of follow-up will be analyzed using paired 2-tail Student's t test. We will then calculate the change in endpoint variables within each group, which will be compared using unpaired 2-tail Student's t test.

To address the issue of channeling bias (differences in baseline characteristics between the two groups that drive differential outcomes), we will use two different approaches. In the primary analysis, we will perform a propensity score matching (PSM), whereas multivariable adjustment (MVA) with linear regressions will be used as a second approach.

Statistical analyses will be performed using SAS version 9.4 (TS1M4) or higher and a 2-tail p-value <0.05 considered statistically significant.