As you may recall from George Hripcsak’s @hripcsa presentation at the OHDSI symposium about “CRIUS” (comparative-effectiveness research about immunization used for SARS-COV-2), we’ve conducted analyses of the absolute effectiveness of COVID vaccines in the Columbia data. We learned at a lot, for example that:
-
Observational data can indeed be used to assert absolute vaccine effectiveness, which was proven by matching the results to the RCT’s results for Pfizer, Moderna and Janssen;
-
Doing so requires avoiding or correcting for a number of biases that we uncovered;
-
We observed a difference in baseline patient characteristics in patients vaccinated with mRNA vaccines and Janssen vaccine (the latter group on average having more co-morbidities and drug exposures) as well as the difference in temporal distribution of vaccines (for example, Moderna was administered early on in 2021 when COVID-19 infectious rates were high).
Based on our findings, we think there is a a good opportunity for an OHDSI network study to conduct a comparative benefit-risk analysis of COVID-19 vaccines. The study would focus on direct head-to-head comparisons between available COVID-19 vaccines in each population (e.g. in US: Pfizer, Moderna, and Janssen) and will involve population-level effect estimation of the key ‘benefit’ outcomes (COVID-19 infection and COVID-19-related hospitalization) as well as the known ‘safety’ outcomes that are listed on the vaccine product labels (e.g. myocarditis, anaphylaxis, Guillian-Barre Syndrome, angioedema, thrombosis with thrombocytopenia). We will examine the comparative benefit-risk within each database, and also produce composite effect estimates for each target-comparator-outcome via random effects meta-analysis. Following OHDSI best practices, we propose CohortDiagnostics to evaluate the exposure and outcome phenotypes for participating databases, and study diagnostics for the comparative cohort design (including attrition, PS distribution, covariate balance, and negative control calibration). Details of the entire analysis will be formalized in a protocol and codified in an R analysis package that we’d post on github.com/ohdsi-studies.
At this point, we are looking to gauge interest from members of the OHDSI community to participate in this COVID-19 vaccine comparative benefit-risk analysis. Please share your interest, and what data source(s) you may be able to bring to the collaboration.