背景:传统的活性疫苗安全性监测涉及预先指定健康结果和生物学上合理的结果特定的关注时间窗口,限制可以评估的不良事件。在这项研究中,我们使用基于树的扫描统计量来广泛寻找二价COVID-19疫苗接种后>6万起可能的不良事件.
方法:在2022年11月之前接受Moderna或Pfizer-BioNTech二价COVID-19疫苗的年龄≥5岁的疫苗安全数据链参与者在疫苗接种后随访56天。在ICD-10-CM的分级诊断代码“树”中以及在疫苗接种后随访中对住院或急诊科的事件诊断进行了聚类分析。使用条件自控树时间扫描统计量,在整个树的扫描风险窗口期间,每个诊断的病例总数和任何诊断发生的病例总数。P=0.01是用于统计显著性的预先指定的截止值。
结果:分析包括352,509剂Moderna和979,189剂Pfizer-BioNTech二价疫苗。Moderna疫苗接种后,没有发现有统计学意义的聚类.在辉瑞生物技术公司之后,有一组未指定的不良事件(第1-3天,p=0.0001-0.0007),流感(第35-56天,p=0.0001),咳嗽(第44-55天,p=0.0002),和COVID-19(第52-56天,p=0.0004)。
结论:仅适用于辉瑞生物技术公司,我们检测到:(1)未指明的不良反应,正如在使用这种方法的其他疫苗研究中观察到的那样,(2)随访结束时的呼吸道疾病。呼吸道集群可能是由于随访与呼吸道合胞病毒的传播重叠,流感,和COVID-19,即,季节性混杂。该方法的非目标性质及其对许多诊断和评估的风险区间的固有调整是独特的优势。局限性包括对时变混杂的敏感性,与针对较少结果的传统研究相比,评估特定结果风险的统计能力较低,以及错过不良事件的可能性在时间上或在“树”内没有强烈聚集。\"
Traditional active vaccine safety monitoring involves pre-specifying health outcomes and biologically plausible outcome-specific time windows of concern, limiting the adverse events that can be evaluated. In this study, we used tree-based scan statistics to look broadly for >60,000 possible adverse events after bivalent COVID-19 vaccination.
Vaccine Safety Datalink enrollees aged ≥5 years receiving Moderna or Pfizer-BioNTech bivalent COVID-19 vaccine through November 2022 were followed for 56 days post-vaccination. Incident diagnoses in inpatient or emergency department settings were analyzed for clustering within the hierarchical ICD-10-CM diagnosis code \"tree\" and temporally within post-vaccination follow-up. The conditional self-controlled tree-temporal scan statistic was used, conditioning on total number of cases of each diagnosis and total number of cases of any diagnosis occurring during the scanning risk window across the entire tree. P = 0.01 was the pre-specified cut-off for statistical significance.
Analysis included 352,509 doses of Moderna and 979,189 doses of Pfizer-BioNTech bivalent vaccines. After Moderna vaccination, no statistically significant clusters were found. After Pfizer-BioNTech, there were clusters of unspecified adverse events (Days 1-3, p = 0.0001-0.0007), influenza (Days 35-56, p = 0.0001), cough (Days 44-55, p = 0.0002), and COVID-19 (Days 52-56, p = 0.0004).
For Pfizer-BioNTech only, we detected clusters of: (1) unspecified adverse effects, as have been observed in other vaccine studies using this method, and (2) respiratory disease toward the end of follow-up. The respiratory clusters were likely due to overlap of follow-up with the spread of respiratory syncytial virus, influenza, and COVID-19, i.e., confounding by seasonality. The untargeted nature of the method and its inherent adjustment for the many diagnoses and risk intervals evaluated are unique advantages. Limitations include susceptibility to time-varying confounding, lower statistical power for assessing risks of specific outcomes than in traditional studies targeting fewer outcomes, and the possibility of missing adverse events not strongly clustered in time or within the \"tree.\"