Looking in the medicine cabinet: methods for using real-world data to assess the impact of measles, mumps and rubella (MMR) and recombinant adjuvanted varicella-zoster vaccines on coronavirus disease 2019 (COVID-19) prevention and case fatality

Abstract

Background: Analysis of real-world data can be used to identify promising leads and dead ends among products being repurposed for clinical practice for coronavirus disease 2019 (COVID-19).  This paper uses real-world data from Cerner Labs collected from 90 source institutions in the United States to assess the potential impact of live viral vaccines on COVID-19 case fatality rates. Methods: We identified 373,032 polymerase chase reaction (PCR)-positive COVID-19 cases in the Cerner Labs database between 01-MAR-2020 and 31-DEC-2020 and identified patients that had received measles, mumps and rubella (MMR) or a recombinant adjuvanted varicella-zoster vaccine within the previous 5 years. We calculated heterogeneity scores to support interpretation of results across institutions, and used stepwise forward variable selection to construct covariable-based propensity scores. These scores were used to match cases and control for biasing and confounding issues inherent in observational data. Results: Neither the recombinant adjuvanted varicella-zoster vaccine nor MMR showed significant efficacy in prevention of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. We could not derive clinically significant results on the impact of MMR for case fatality rates due to persistently high rates of heterogeneity between institutions. However, we were able to achieve acceptable levels of heterogeneity for the analysis of the recombinant adjuvanted varicella-zoster vaccine, and found a clinically meaningful benefit of reduced case fatality rate, with an odds ratio of 0.43 (95% confidence interval [CI]: 0.38 – 0.48). Conclusions: Using propensity score matching and heterogeneity statistics can help guide our interpretation of real-world data, and rigorous statistical methods are needed to reduce bias or disparities in data interpretation. Applying these methods to the impact of live viral vaccines on COVID-19 case fatalities yields actionable findings for further analysis.