Diabetes
Discoveries and challenges in identifying GLP-1 RA-associated phenotypes using real-world data Maxwell Salvatore* Maxwell Salvatore Bingyu Zhang Yiwen Lu Dazheng Zhang Ting Zhou Marylyn Ritchie Yong Chen
Objective:
To investigate the utility and challenges of using real-world electronic health record (EHR) data to identify downstream phenotypes associated with GLP-1 RA use compared to other second-line diabetes therapies.
Methods:
We analyzed EHR data from the All of Us Research Program, including 15,146 participants with a type 2 diabetes diagnosis and one of three second-line diabetes therapies. We used phenome-wide propensity score-matched intention-to-treat (ITT) and per-protocol (PP) Cox proportional hazard models and restricted mean survival time (RMST) to identify phenotypes associated with GLP-1 RA use (n=6,020) versus sodium-glucose cotransporter-2 inhibitors (SGLT2i; n=3,897) and dipeptidyl peptidase-4 inhibitors (DPP4i; n=5,229).
Results:
We found one significant phenotype associated with GLP-1 RA use compared to SGLT2i:·
- ITT: Reduced risk of optic neuropathy (HR: 0.33, 95% CI: 0.20, 0.54)
- PP: Increased risk of vomiting (HR: 1.67, 95% CI: 1.31, 2.14)
In comparisons with DPP4i, 83 and 4 phenotypes were significantly associated under ITT and PP analyses, respectively. Comparing significant phenotypes with their 5-year RMST difference 95% CI, we found:
- ITT: 11% of significant HR associations had RMST differences >30 days.
- PP: All 4 significant phenotypes showed RMST differences >30 days.
A positive association of GLP-1 RA use with morbid obesity was an artifact likely due to confounding factors like under coding, billing practices, and drug indication, as this result contradicts the known action of GLP-1 RAs.
Conclusions:
Our study highlights phenotypes associated with GLP-1 RA use, generally in line with published findings. These findings demonstrate the benefits and pitfalls of using real-world EHR data for agnostic scanning for impacts of GLP-1 RAs. Future work will focus on validating these results across independent datasets and comparing GLP-1 RA-associated phenotypes with those of other anti-obesity medications in individuals with obesity without diabetes.