Interpreting vaccine efficacy trial results for infection and transmission

Abstract

Randomized controlled trials (RCTs) have shown high efficacy of multiple vaccines against SARS-CoV-2 disease (COVID-19), and recent studies have shown the vaccines are also effective against infection. Evidence for the effect of each of these vaccines on ability to transmit the virus is also beginning to emerge. We describe an approach to estimate these vaccines' effects on viral positivity, a prevalence measure which under the reasonable assumption that vaccinated individuals who become infected are no more infectious than unvaccinated individuals forms a lower bound on efficacy against transmission. Specifically, we recommend separate analysis of positive tests triggered by symptoms (usually the primary RCT outcome) and cross-sectional prevalence of positive tests obtained regardless of symptoms. The odds ratio of carriage for vaccine vs. placebo provides an unbiased estimate of vaccine effectiveness against viral positivity, under certain assumptions, and we show through simulations that likely departures from these assumptions will only modestly bias this estimate. Applying this approach to published data from the RCT of the Moderna vaccine, we estimate that one dose of vaccine reduces the potential for transmission by at least 61%, possibly considerably more. We describe how these approaches can be translated into observational studies of vaccine effectiveness.

Keywords: COVID-19; SARS-CoV-2; Trials; Vaccine efficacy.

Fig. 1

Vaccine efficacy for viral positivity Results are shown of a 300-day simulation of a trial of 100,000 participants randomized 1:1 to vaccine or placebo on day 100 and exposed to a constant force of infection of 0.001 throughout the simulation. The different panels represent (left to right) simulations with ${\mathit{VE}}_S=0,0.3,0.6.0.9$ and (top to bottom) ${\mathit{VE}}_D=0,0.3,0.6.0.9$. We simulate a 2-dose regimen, 28 days apart with the first dose giving half the full efficacy and the effect of each dose starting one week after it is given, that is, on days 107 and 135 of the simulation. The solid black lines give the dose-1 and dose-2 predicted values for ${\mathit{VE}}_V$based on eq. (7), while the curves show the estimates obtained from the simulated data using eq. (6). Panel A shows the situation under the assumption that individuals naturally infected who recover (clear infection) become once again susceptible to reinfection (SEIS). Panel B makes the opposite assumption, that individuals naturally infected (whatever their vaccine status) are completely protected against reinfection for the duration of the simulation (SEIR).

Fig. 2

Vaccine efficacy for viral positivity and a combination of symptoms and testing This figure shows the same simulations as Fig. 1B with different analyses of the simulated data, comparing scenarios in which 1% and 80% of unvaccinated infections are symptomatic. The solid black lines give the dose-1 and dose-2 predicted values for ${\mathit{VE}}_V$based on eq. (7), while the solid curves show the estimates obtained from the simulated data using eq. (6) (the solid red line is the same as Fig. 1B). The dashed lines give the dose-1 and dose-2 predicted values for ${\mathit{VE}}_{\mathit{SP}}$, based on equation (2), while the dashed curves show the estimates of ${\mathit{VE}}_{\mathit{combined}}$obtained from the simulated data. When only 1% of infected individuals are symptomatic, the solid red and dashed red lines are nearly identical. However, when 80% are symptomatic , the dashed blue line increases over time but falls below the expected ${\mathit{VE}}_{\mathit{SP}}$.