Can the Safety of the Yellow Fever Vaccine Be Evaluated by a Retrospective Study of Databases?

Nordin et al. assessed the quality of data in two databases, the Vaccine Safety Datalink (VSD) and the US Department of Defense (DoD), for adverse events that occur after yellow fever vaccine.1 The issues are as follows: (1) Can the adverse events be attributed solely to yellow fever vaccine? No, 85% of the adults and 75% of the pediatric subjects in the VSD database received other vaccines and those vaccines were not stated. The authors state that the DoD database did not list other vaccines given (and military personnel are usually given multiple vaccines). Reports of adverse events after vaccination to pharmacovigilance databases usually involve several vaccines given simultaneously, and identifying whether a specific vaccine was responsible is usually impossible. Thus, separating reports for patients who received only yellow fever vaccine from those who received yellow fever and other vaccines is a key solution. (2) Do the authors describe how the events are reported to the databases? No. We are not told if the databases use passive or active reporting. Nor is there any statement about whether other publications or the authors estimated the percentage of vaccine events reported. This is important because underreporting in both active versus passive databases can be substantial.2,3 (3) Is the authors' choice of ICD‐9 codes to classify events appropriate? No. For “allergic and local reactions” they chose 5 ICD‐9 codes, for “visceral” events they chose 25 broad ICD‐9 codes, and for “neurological” events 7 codes for meningitis and encephalitis, 3 for central nervous system demyelinating disease, 1 for hemiplegia and hemiparesis, 4 for seizures and movement disorders, 8 for peripheral and cranial neuropathy, and 4 for altered mental status. Although the ICD‐9 codes could capture some of the adverse events related to yellow fever vaccine, they are nonspecific and do not incorporate the careful algorithms in the CDC or Brighton Collaboration criteria4–8 for adverse events following yellow fever vaccine and would produce substantially different classifications. (4) Do the databases provide enough data for classification of events? No. The authors correctly identified the problems with their data in these four comments: (i) there were “no detailed clinical data needed to determine whether these events met formal YF‐vaccine‐AVD and YF‐vaccine‐AND case definitions”; (ii) “laboratory and radiology data were not available to confirm whether visceral or neurologic events met formal CDC YFSWG or Brighton case definitions”; (iii) “diagnostic codes for symptoms of neurologic and visceral events have unknown sensitivity and specificity for identifying reactions to YF‐vaccine”; and (iv) “further details were not available to determine whether any of these represented true cases of anaphylaxis.” (5) Did the authors independently check data accuracy? No. The VSD and DoD databases are automated and the authors did not perform any checks on data accuracy. (6) For the case–control procedure the authors chose controls on the basis of nonexposure to yellow fever vaccine: “YF‐vaccine‐exposed subjects were compared with randomly selected YF‐vaccine‐unexposed controls.” But did this random selection produce cases and controls with the same number and type of other vaccines? Moreover, the authors stated that “unexposed subjects were not required to have a visit during the period of observation,” and thus no data are available about ICD‐9 events in the unexposed cohort that did not result in a visit to the specific medical organizations. (7) Do the authors' estimates of rates agree with those of other yellow fever databases? The authors cited three yellow fever pharmacovigilance databases (Khromava 2005, Martin 2001, and Martins 2010) for comparison of rates with their data but did not compare their rates with six other databases9–14 or three systematic reviews of yellow fever adverse events.2,3,15 Some data for US military personnel are also reported in the US VAERS database and the data should have been compared with the databases that reported VAERS data. The VSD and DoD databases could potentially be important sources of information about severe adverse events of vaccine, and the authors have performed a helpful service in assessing these databases. Their results illustrate how the databases could be improved by those designing and maintaining them. The author states that he has no conflicts of interest.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.