4F-02 - Screening for Pediatric Safety Alerts Following Antipsychotic Treatment Initiation using a Tree-Based Scan Statistics Approach

Background: Antipsychotics (APs) are commonly prescribed to children. While pediatric studies have provided evidence on adverse events for widely used APs (i.e., aripiprazole, olanzapine, quetiapine, risperidone, and ziprasidone), safety information on newer APs (i.e., asenapine, brexpiprazole, cariprazine, lurasidone, and paliperidone) remains sparse.

Objectives: Using tree-based scan statistics (TBSS) to systematically screen across a broad range of outcomes, while tightly controlling the type 1 error, we evaluated whether known associations for common APs could be identified, and whether previously unrecognized adverse effects associated with newer APs would be detected.

Methods: We identified children initiating APs (with quetiapine as the reference) within two large US health insurance databases 2000-2021. Leveraging a pruned version of the Multi-Level Clinical Classification tree (after removing codes irrelevant for the pediatric population under study), which grouped 8,128 diagnostic codes into increasingly aggregated higher-level clinical concepts, we scanned for incident outcomes recorded within 6 months of AP initiation. Incident outcomes were defined using a washout of 15 months prior to AP initiation. We used propensity score fine stratification to adjust for a broad range of potential confounders and an unconditional Poisson scan statistic to calculate relative risks (RR) and p-values for each outcome node. A p-value threshold of 0.05 was used to define statistical alerts for potential safety signals.

Results: The number of children exposed to common APs was 376,717 for aripiprazole, 92,701 for olanzapine, 664,967 for risperidone, and 41,575 for ziprasidone. For these, compared to quetiapine (N=286,192), multiple elevated risks, consistent with previous evidence, were identified, including endocrine, cardiovascular, hematologic, metabolic, neurologic, and immunologic conditions. Among newer APs, no signals were detected for asenapine (N=3,262) and brexpiprazole (N=780). For cariprazine (N=534), one alert for viral disease (RR=2.6; 26 exposed cases) was found. Lurasidone (N=9,664) was associated with metabolic, urinary, pituitary and platelet disorders (RR range: 1.2-24.9; range of N exposed cases: < 11-810), and paliperidone (N=5,893) with extrapyramidal and pituitary disorders and allergies (RR range: 1.8-18.0; range of N exposed cases: < 11-75).

Conclusions: The detection of known associations supports TBSS’ feasibility for pediatric drug safety surveillance. Continued monitoring of the newer drugs as more data accrue, in-depth assessment of clinical plausibility and replication in independent data sources remains essential.