Change-point analysis of spontaneous adverse event reporting following human papillomavirus vaccination

초록

Background: Detecting signals of sharp increase in temporal trends of adverse event reporting is important to notice public safety concerns and to take prompt actions, especially for the vaccines under national immunization program (NIP). Objectives: To assess applicability of change-point analysis for signal detection of vaccine safety surveillance using spontaneous reporting database in Korea. Methods: Among the individual case safety reports (ICSRs) reported to Korea Adverse Event Reporting System (KAERS) between 2008 and 2017, ICSRs containing vaccines were extracted. Vaccine of interest was human papillomavirus (HPV) (WHO-ATC, J07BM01-J07BM03) vaccine, which was recently included in the NIP in Korea. We assessed demographic characteristics of reported cases and characteristics of ICSRs. We defined syncope and dizziness as the target adverse events (AEs). We analyzed the number of reports based on the quarterly recognition date AEs. To derive safety signals, we used the change-point analysis approach in which cumulative sum (CUSUM) and bootstrapping techniques were iteratively used to detect the changes and to assign a confidence level to each change. The p-value of the detected change point using CUSUM was cut off at the significance level of 0.05 and it was defined as a signal. Results: Among total of 26,293 vaccine ICSRs in KAERS database, the number of ICSRs for HPV was 2,842. Reports for cases aged 11 - 64 years old were 56.6% and the unknown 42% in HPV. Among report type, 56.8% were reported as the spontaneous report and were higher in HPV vaccine than other vaccines (P<0.01). The change-point analysis detected four significant change points in the number of reports: Q2 2013 (P=0.02), Q4 2014 (P=0.01), Q3 2016 (P<0.01), Q2 2017 (P<0.01). A significant increase in the number of reports from the first, third and fourth change point (Q2 2013, Q3 2016, Q2 2017), and a significant decrease in reports from the second change point (Q4 2014) was observed. Conclusions: The detected change points were consistent with increased public concerns about adverse reactions of HPV through mass media, and inclusion into NIP in Korea. The change-point analysis, which focuses on detecting changes within time-series data could be used as a useful tool for signal detection within spontaneous adverse event reporting system.