Comparing insulin vs GLP-1, DPP-4, SGLT-2 on 5-year incident heart failure for patients with type 2 diabetes mellitus: a real-world evidence study using insurance claims (Preprint)

BACKGROUND Type 2 diabetes mellitus (T2D) is a common health issue, with heart failure (HF) being the common and lethal long-term complication. Although insulin is widely used for the treatment of T2D, evidence regarding the efficacy of insulin compared to non-insulin therapies on incident heart failure risk is missing among randomized clinical trials. Real-world evidence on insulin’s effect on long-term heart failure may supplement existing guidelines on the management of T2D. OBJECTIVE This study compared the insulin therapy versus other medications on heart failure (HF) among T2D patients using real-world data (RWD) extracted from insurance claims. METHODS We employed doubly robust Augmented Inverse Probability Weighted estimation that extensively adjusted for high-dimensional confounding factors in both the propensity score and outcome regression models, using a data-driven approach for feature selection implemented through a LASSO sparsity penalty in each model. RESULTS After adjusting for broad list of confounders, insulin was found to be associated with 11.8% [95% CI: 11.0%-12.7%] higher 5-year HF rate compared to patients receiving GLP-1, 12.0% [95% CI: 11.5%-12.4%] higher 5-year HF rate compared to DPP-4, and 15.1% [95% CI: 14.3%-16.0%] higher 5-year HF rate compared to SGLT-2. Subgroup analysis shows the insulin effect with a higher HF rate is significant in subgroup with high baseline HF risk but not significant in subgroup with low baseline HF risk. CONCLUSIONS This study generated real-world evidence on the association with higher 5-year heart failure rate of insulin therapy compared to GLP-1, DPP-4, and SGLT-2 based on claims data. These findings also demonstrated the value of real-world data for comparative effectiveness studies to complement established guidelines. On the other hand, the study shares the common limitation of observational studies. Even though high-dimensional confounders are adjusted, remaining confounding may exist and induce bias in analysis.