Tree-based data mining for safety assessment of first COVID-19 booster doses in the Vaccine Safety Datalink

• Published in: Vaccine Vol. 41; no. 2; pp. 460 - 466
• Main Authors: Katherine Yih, W., Daley, Matthew F., Duffy, Jonathan, Fireman, Bruce, McClure, David, Nelson, Jennifer, Qian, Lei, Smith, Ning, Vazquez-Benitez, Gabriela, Weintraub, Eric, Williams, Joshua T.B., Xu, Stanley, Maro, Judith C.
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier Ltd 09.01.2023; Elsevier Limited
• Subjects: Ad26COVS1; Allergies; Allergy and Immunology; BNT162 Vaccine; Clustering; Codes; Coronaviruses; COVID-19; COVID-19 - prevention & control; COVID-19 infection; COVID-19 vaccines; COVID-19 Vaccines - adverse effects; Data Mining; Dermatitis, Atopic; Disease control; Drug-Related Side Effects and Adverse Reactions - epidemiology; Emergency medical care; Emergency medical services; Epidemiologic research design; Exanthema; FDA approval; fever; Humans; Hypersensitivity (delayed); immunity; Immunization; Mining accidents & safety; monitoring; Monte Carlo method; Monte Carlo simulation; mRNA; mRNA vaccines; Myalgia; Population; Pre-existing conditions; risk; Safety; safety assessment; secondary immunization; Side effects; Statistical analysis; Surveillance; Urticaria; Vaccination; Vaccines; United States > US; COVID-19; FDA; CDC; Vaccination; VSD; ACIP; Data-mining; Epidemiologic research design; ICD-10-CM; Vaccines; Advisory Committee on Immunization Practices; Vaccine Safety Datalink; Centers for Disease Control and Prevention; International Classification of Diseases, Tenth Revision, Clinical Modification; Food and Drug Administration
• ISSN: 0264-410X; 1873-2518; 1873-2518
• DOI: 10.1016/j.vaccine.2022.11.053

The Centers for Disease Control and Prevention’s Vaccine Safety Datalink (VSD) has been performing safety surveillance for COVID-19 vaccines since their earliest authorization in the United States. Complementing its real-time surveillance for pre-specified health outcomes using pre-specified risk intervals, the VSD conducts tree-based data-mining to look for clustering of a broad range of health outcomes after COVID-19 vaccination. This study’s objective was to use this untargeted, hypothesis-generating approach to assess the safety of first booster doses of Pfizer-BioNTech (BNT162b2), Moderna (mRNA-1273), and Janssen (Ad26.COV2.S) COVID-19 vaccines. VSD enrollees receiving a first booster of COVID-19 vaccine through April 2, 2022 were followed for 56 days. Incident diagnoses in inpatient or emergency department settings were analyzed for clustering within both the hierarchical ICD-10-CM code structure and the follow-up period. The self-controlled tree-temporal scan statistic was used, conditioning on the total number of cases for each diagnosis. P-values were estimated by Monte Carlo simulation; p = 0.01 was pre-specified as the cut-off for statistical significance of clusters. More than 2.4 and 1.8 million subjects received Pfizer-BioNTech and Moderna boosters after an mRNA primary series, respectively. Clusters of urticaria/allergy/rash were found during Days 10–15 after the Moderna booster (p = 0.0001). Other outcomes that clustered after mRNA boosters, mostly with p = 0.0001, included unspecified adverse effects, common vaccine-associated reactions like fever and myalgia, and COVID-19. COVID-19 clusters were in Days 1–10 after booster receipt, before boosters would have become effective. There were no noteworthy clusters after boosters following primary Janssen vaccination. In this untargeted data-mining study of COVID-19 booster vaccination, a cluster of delayed-onset urticaria/allergy/rash was detected after the Moderna booster, as has been reported after Moderna vaccination previously. Other clusters after mRNA boosters were of unspecified or common adverse effects and COVID-19, the latter evidently reflecting immunity to COVID-19 after 10 days.