High-throughput screening for prescribing cascades among real world statin initiators.

ABSTRACT

PURPOSE: Statins are among the most prevalent medications prescribed and associated with adverse events that may prompt additional treatment (i.e., a prescribing cascade). No comprehensive assessment of statin-related prescribing cascades has been performed to our knowledge. METHODS: We utilized sequence symmetry analysis to iteratively screen prescribing sequences of all therapeutic classes ("marker" classes) based on Level 4 Anatomical Therapeutic Chemical codes among adult statin initiators, using IBM Marketscan commercial and Medicare supplemental claims databases (2005-2019). Order of initiation and secular trend-adjusted sequence ratios were calculated for each statin-marker class dyad, among marker class initiators ±90 days of statin initiation. Among signals classified as prescribing cascades, we calculated naturalistic number needed to harm (NNTH) within 1 year as the inverse of the excess risk among exposed. RESULTS: We identified 2 265 519 statin initiators (mean ± SD age, 56.4 ± 12.0 years; 48.7% women; 7.5% with cardiovascular disease). Simvastatin (34.4% of statin initiators) and atorvastatin (33.9%) were the most commonly initiated statins. We identified 160 significant statin-marker class dyad signals, of which 35.6% (n = 57) were classified as potential prescribing cascades. Of the top 25 strongest signals (lowest NNTH), 12 were classified as potential prescribing cascades, including osmotically acting laxatives (NNTH, 44, 95% CI 43-46), opioids + non-opioid combination analgesics (81, 95% CI 74-91), and first-generation cephalosporins (204, 95% CI 175-246). CONCLUSIONS: Using high-throughput sequence symmetry analysis screening, we identified previously known prescribing cascades as well as potentially new prescribing cascades based on known and unknown statin-related adverse events.