Letter to the Editor: A Discussion About Pharmacovigilance Study Methods

To the Editor:

We read with interest the article by Kim DH, et al.1 (March 4, 2024 issue, JKMS) about a pharmacovigilance study to explore potential risks of coronavirus disease 2019 (COVID-19) vaccines in comparison to other vaccines with a disproportionality analysis using the World Health Organization pharmacovigilance database, VigiBase. The study findings were consistent with the known safety profile of COVID-19 vaccines,2, 3 highlighting myocarditis/pericarditis as a fingerprint of mRNA vaccine risks. Although the study result was meaningful, a discussion about study methods would still be required.

Medical Dictionary for Regulatory Activities (MedDRA) is a medical terminology developed and maintained by the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use.4 As MedDRA has a hierarchical structure and is updated twice a year with version control, MedDRA provides several guidance documents to enable the MedDRA users to make the best use of the terminology.4 Some descriptions in the article raise a concern as to whether MedDRA were appropriately used for data management following the guidance.

The point that drew our attention was the version of MedDRA. In Supplementary Table 1 of the index article,1 MedDRA version was designated as 24.1,1 which was released in September 2021.5 As the study data consisted of adverse events collected during the period from January 2020 to July 2023, it is expected that the MedDRA versions from 22.1 to 26.0 had been used for individual case safety report (ICSR) processing,5 and we could guess that the extracted data were coded in version 26.0. Inconsistency may exist between the version that was used for data extraction, 24.1 and the version that was probably used for data coding, 26.0. Besides, as the Standardised MedDRA Query (SMQ) of Noninfectious myocarditis/pericarditis was developed in version 25.0,5 version 24.1 referenced in Supplementary Table 1 of the index article1 would be an incorrect match for the terms extracting myocarditis/pericarditis events.

There is another discussion point about adverse events which were collected before the creation of current MedDRA terms and thus were coded in different MedDRA terms. The Preferred Term (PT) of multisystem inflammatory syndrome in children was newly created in version 23.1, and the PT of multisystem inflammatory syndrome (MIS) was first introduced in version 24.1.5 If there had been suspected MIS in children cases collected before the release of version 23.1, would those cases still have been categorized into the case group?6 We are curious about in what way this issue was addressed. Inappropriate MedDRA use could cause unexpected biases like a misclassification bias.

In addition, we would like to discuss other method-related points. Fig. 1 of the index article1 shows that a disproportionality analysis was performed by using numbers of cases and non-cases under the different exposures which were counted at the adverse event level. This approach could provide an opportunity that a single report can present in multiple cells on two-by-two contingency table for reporting odds ratio (ROR) calculation; hence, the independence assumption for the statistical method can be jeopardized.7, 8 The analysis based on numbers counted at the report level would provide more valid ROR values. Also, the inconsistent use of adverse event and ICSR number as a denominator is suspected when the proportion of adverse events of special interest (AESI) cases was calculated. For example, the proportion of myocarditis/pericarditis cases for COVID-19 vaccines was calculated out of the adverse event number, whereas the proportion for other vaccines was calculated out of the ICSR number.1 Probably, we can also discuss the appropriateness of using reports from study for the analysis, as those reports may be different in many aspects compared to spontaneous reports.

These factors would not affect the overall study conclusion, but it would be worth discussing them because they are closely related to fundamental principles in pharmacovigilance.