(A-311) High Throughput Screening for GLP-1 RA-Induced Prescribing Cascades among Older Adults with and without Obstructive Sleep Apnea

Background: Recently, a dual glucose-dependent insulinotropic polypeptide/ glucagon-like peptide 1 (GLP-1) receptor co-agonist was approved as a pharmacological treatment for obstructive sleep apnea (OSA). As GLP-1 receptor agonists (RA) use continues to rise in those with OSA, it is important to understand the risk of GLP-1 RA-induced prescribing cascades in this population.

Objectives: To identify GLP-1 RA-induced prescribing cascades in patients with and without OSA.

Methods: We used Medicare administrative claims from 2011-2020 to identify new users of GLP-1 RAs aged ≥66 years with or without OSA. We used a high-throughput prescription sequence symmetry analysis to screen for initiation of 446 other drug classes (using Level 4 Anatomical Therapeutic Chemical codes) relative to initiation of GLP-1 RAs. Adjusted sequence ratios (aSRs) with 95% confidence intervals, accounting for secular prescribing trends over time, were calculated by comparing the initiation of marker drug use 180 days before and after GLP-1 RA initiation. We considered aSRs with 95% CI >1 to be prescribing cascade signal, and further calculated the number needed to harm (NNTH) within 6 months as inverse of the excess risk among exposed for these signals.

Results: We identified 58,990 GLP-1 RA users (mean ± SD age, 72.9 ± 5.5 years; 53.3% women; 24.4% with OSA). Dulaglutide (40.3% of GLP-1 RAs) and liraglutide (32.5%) were the most commonly initiated GLP-1 RAs. We identified 5 signals among patients with OSA (including 5-HT3 antagonists, enzyme preparations, imidazole derivatives, ophthalmic anticholinergics, and antidotes), and 10 signals among patients without OSA. Two signals—5-HT3 antagonists and imidazole derivatives—were detected in both groups, with aSRs of 1.32 (95% CI 1.17-1.49) and 1.66 (95% CI 1.54-1.79) among patients with and without OSA, respectively. These signals also had lowest NNTHs among patients with OSA (5-HT3 antagonists: 101; imidazole derivatives: 320). Comparatively, the NNTHs among patients without OSA for 5-HT3 antagonists and imidazole derivatives were 66 and 551, respectively.

Conclusions: Using a high-throughput sequence symmetry analysis approach, we identified 5 prescribing cascade signals in patients with OSA and 10 signals in patients without OSA. Amongst signals appearing in both groups, patients with OSA may be slightly less likely to have a GLP-1 RA-induced prescribing cascade with 5-HT3 antagonists but more likely to have a prescribing cascade with imidazole derivatives versus those without OSA. Clinicians should be aware of GLP-1 RA-induced adverse events and prescribing cascades in patients with OSA, but more research is needed to determine biological plausibility of these signals.