The objective of this study was to compare the effectiveness of sodium glucose co-transporter 2 (SGLT2) inhibitors relative to metformin for reducing subsequent cardiovascular events in patients with type 2 diabetes mellitus.

The investigators will conduct a population-based, new-user, longitudinal-cohort study using a nationwide US commercial insurance claims database. The investigators will compare adults with diabetes mellitus type 2 over the age of 18 who were newly prescribed an SGLT2 inhibitor or metformin between March 29, 2013 (date of US approval of first SGLT2) and January 1st, 2017 (most recent available data). Patients with diabetes mellitus type 2 will be identified using the International Classification of Diseases, Ninth Revision (ICD-9) and ICD-10 codes. Cohort entry date will be the date of the first prescription for an SGLT2 or metformin. New users of SGLT2 or metformin will be defined as those without a prior prescription for either class of medications, or any other medication for diabetes, in the preceding 180 days.

Baseline Covariates: All covariates will be assessed prior to cohort entry. Covariates will reflect diagnoses and procedures recorded during health encounters, including chronic medical conditions (e.g., hypertension, coronary artery disease), diabetes severity (e.g., hemoglobin A1C, end-organ damage), overall healthcare utilization (e.g., recent hospitalization, emergency department visit), prescriber characteristics (e.g., endocrinologist, general practitioner), and medications (e.g., anti-hypertensives, diuretics).

Statistical analysis Propensity score matching will be used to adjust for confounding. The probability of initiating an SGLT2-inhibitor will be calculated through a multivariable logistic regression model containing all of the baseline covariates. Using this propensity score, patients prescribed an SGLT2 were matched 1:1 with patients prescribed metformin using a caliper of up to 0.1 on the probability scale. Covariate balance between the matched cohorts was assessed using standardized differences. Since laboratory data were not available for all patients, these were not included in the propensity score estimation.

After propensity score matching, proportional hazards models will be used to estimate the incidence rate, hazard ratios and 95% confidence intervals for the primary outcome without further adjustments. Schoenfeld residuals will be plotted to assess the proportional hazards assumption. Predefined sensitivity and subgroup analyses included an intention to treat analysis where the censoring criteria of drug discontinuation, switching or augmentation are removed. The investigators will also assess the primary risk in a cohort restricted to patients with a past-history of cardiovascular disease if our sample size allows it. To test the specificity of our findings, the investigators will also conduct a tracer analysis using cellulitis as an outcome, since cellulitis is not associated with SGLT2s or metformin.

INCLUSION:

- all patients newly prescribed an SGLT2 or metformin between March 29, 2013 to January 1st, 2017 with at least 6 months of continuous enrollment (1 year in a sensitivity analysis)

EXCLUSION:

• age < 18 years
• previous use of any diabetes medication
• lack of a diagnosis of type 2 diabetes mellitus
• history of malignant neoplasm
• dialysis
• type 1 diabetes

We will conduct a population-based, new-user, longitudinal-cohort study using the nationwide US commercial insurance claims database. This database provides patient demographics and longitudinal, individual-level data on healthcare utilization, inpatient and outpatient diagnoses, diagnostic tests, clinical procedures, outpatient laboratory results, and pharmacy dispensing of drugs.