Evaluating the Burden-of-Disease Endpoint to Assess Vaccine Efficacy

Overview

In this Summary Document, we evaluate, from a clinical perspective, the burden of disease (BOD) endpoint recently proposed by Mehrotra et.al. for its suitability as a primary endpoint for assessing COVID-19 (C-19) vaccine efficacy (VE) in a multiple endpoint approach. This summary document does not intend to recommend a specific BOD endpoint definition. The example 0,1,2,3,4 score in the table is there for illustration only.

The BOD endpoint was developed for vaccines that may affect both disease incidence and disease severity. C-19 vaccines may not be able to prevent infection per se, but may be able to attenuate disease severity in individuals with breakthrough disease. In interventional efficacy trials, vaccine-attenuated disease (VAD) has the effect of lowering VE in preventing C-19, in which VE = 1 – RR and RR represents the rate ratio of vaccine to placebo. If using the BOD endpoint, the VE_BOD in reducing C-19 incidence and disease severity, is a simple, useful, and interpretable statistic when expressed as VE = 1 – RR, in which RR represents the rate ratio of the BOD score vaccine to placebo.

The effect of VAD in interventional trials is that it gradually lowers VE against milder disease manifestations. Typically for vaccines associated with VAD, VE is greatest against death > critical disease > severe disease > moderate disease > mild disease. This lowers the overall VE point estimate, significantly impacting the sample size of C-19 cases required. The lower (lowest) efficacy against mild disease, however, does not necessarily imply that the vaccine has less biological activity against mild disease. The effect of VAD on VE is illustrated with a purely hypothetical example that shows that despite preventing disease in 88% of people who would have had mild disease if unvaccinated, the VE against mild disease is lowest with only 50%, because of the way VE is mathematically calculated. In the example, this is due to accumulation of 23 VAD cases that were classified as mild C-19, out of the total of 30 mild C-19 cases.

The purpose of the example is to illustrate the effect of VAD by “opening the black box” also showing 1 severe VAD case out total of 3 severe C-19 cases, and 6 moderate VAD cases out of 10 moderate C-19 cases, explaining the gradual lowering of VE against milder disease manifestations.
The answer to VAD in interventional trials is to accept a relatively low VE point estimate for success. In the sample size calculations, low VE assumptions significantly increase the number of C-19 cases required for success. This could prove problematic during a pandemic with public health intervention that aim to prevent infection potentially slowing down accrual of C-19 cases. The purpose of this document is to stimulate the discussion around the BOD as an alternative way of addressing VAD that might speed-up demonstration of VE, and to explain that inclusion of moderate disease might be critical to its added value when used as a dual or triple primary endpoint, because of the relatively low incidence of severe C-19.