An observational study of electronic patient data to compare diabetes medications and to determine which ones offer the best balance of risks and benefits.

Type 2 diabetes is an increasingly common disease that affects more than 10 percent of adults in the United States. Multiple medications are available to treat this condition. However, many of these medications have never been directly compared against one another, which makes it unclear which medication is the best to use.

Investigators will use a large database of electronic patient data to compare diabetes medications to determine which ones offer the best balance of risks and benefits. This database will draw from several large healthcare institutions and health insurance companies that collectively pull from a patient population of over 130 million across all major regions of the United States. Investigators will study adults aged 30 or older who have type 2 diabetes and are at moderate risk of heart attacks and strokes, and who are starting a second diabetes medication (after metformin). Investigators will use clinical trial emulation, a cutting-edge statistical technique, to compare the following classes of diabetes medications: (1) DPP4 inhibitors (alogliptin, linagliptin, sitagliptin, or saxagliptin); (2) GLP1 receptor agonists (dulaglutide, exenatide, liraglutide, or semaglutide); (3) basal insulin (degludec, detemir, glargine, or NPH); (4) SGLT2 inhibitors (canagliflozin, dapagliflozin, empagliflozin, or ertugliflozin); and (5) sulfonylureas (glimepiride, glipizide, or glyburide).

To determine which diabetes medications provide the greatest benefit to patients, investigators will compare how much they reduce the risks of the following \conditions: (1) heart attack; (2) heart failure; (3) stroke; (4) kidney disease; (5) eye disease; and (6) liver disease. To allow patients with diabetes and their doctors to make fully informed decisions about which diabetes medication to take, investigators will also compare these medications' risks of the following possible side effects: (1) low blood sugar; (2) kidney infection; (3) severe skin infections in the genital area; (4) foot/leg amputations; (5) broken bones; (6) pancreatitis (i.e., severe inflammation of the pancreas); (7) pancreatic cancer; (8) thyroid cancer; and (9) death from causes other than heart attacks or strokes.

In order to minimize the risk of bias and confounding, the analysis will be conducted using causal inference techniques, by emulating a randomized clinical trial (including both intention-to-treat and per-protocol analyses), while incorporating rigorous data quality checks.

Individuals meeting the following criteria between January 1, 2015 and December 31, 2021:

• Diabetes mellitus (DM) type II

• HbA1c of 7-11% within the past year

• Monotherapy with metformin for at least 3 months

• No prior non-metformin outpatient diabetes therapy

• Aged ≥30y

• At "moderate" risk of ASCVD

  • Men aged ≥35y and women aged ≥45y with no history* of stroke, myocardial infarction, revascularization, or heart failure hospitalization
  • Men aged 30-34 and women aged 30-44 with history* of hypertension, hyperlipidemia, retinopathy, kidney disease or neuropathy

• estimated glomerular filtration rate (eGFR) ≥ 45 ml/min/1.73m2 within the past 3 years

• Not pregnant at time 0

• No history* of institutionalization with a diagnosis of dementia, metastatic cancer, end stage lung disease, end stage liver disease, pancreatitis, medullary thyroid cancer or severe UTIs

• Engagement with the healthcare system: enrollment for at least 12 months and attendance of at least one outpatient encounter in the prior 12 months

(*) History will be derived from at least 12 months of EHR and claims prior to time zero and will use all available EHR and claims data between January 1, 2014 and time zero

The study population is adults aged 30 or older who have type 2 diabetes and are at moderate risk of Atherosclerotic Cardiovascular Disease (ASCVD), and who are starting a second diabetes medication (after metformin). Patients from collaborating institutions will be included in the analysis based on study population inclusion and exclusion criteria.