Understanding the spectrum of vaccine efficacy measures

Phase III covid-19 vaccine efficacy trials have returned encouraging results, exceeding the 50% efficacy threshold specified by the World Health Organization (WHO) and the Food and Drug Administration (FDA). Multiple vaccines are now available for use. These phase III trials address the central question of a vaccine’s effect on a meaningful clinical outcome. In nearly all of the trials, the primary aim is to measure efficacy against laboratory confirmed symptomatic disease, including mild symptoms. But this is not the only endpoint that policy makers and individuals care about when making decisions. In fact, we can think about it as one measure of vaccine efficacy that lies alongside others on a spectrum.

Efficacy against severe disease is perhaps of greatest clinical relevance. Hospitalization and death due to covid-19 are the priority public health burdens that we aim to prevent. But we have comparatively fewer data on vaccine efficacy against severe disease because it is a rarer outcome, particularly in younger populations. A vaccine trial would need to be larger and/or require longer follow-up to have enough statistical power to measure efficacy against severe disease only. For example, Pfizer-BioNTech’s Phase III vaccine trial reported only 10 cases of severe disease, many of which occurred before both doses of vaccine or placebo were received and would not be counted in a standard analysis. Companies largely opted for an endpoint including mild and moderate symptoms to balance clinical relevance and feasibility.

In addition, we leverage knowledge that vaccines usually have higher efficacy against severe disease than against an endpoint that includes milder symptoms. For example, efficacy of an inactivated influenza vaccine against flu of any severity was 55.4% (95% confidence interval [CI] 39.1 to 67.3%), but efficacy against moderate-to-severe disease was 73.1% (95% CI 47.1 to 86.3%). Efficacy of a pentavalent rotavirus vaccine against severe rotavirus gastroenteritis was 36.0% (95% CI 11.7 to 53.6%), but efficacy against the most severe cases was 60.5% (95% CI 17.7 to 81.0%). Similar relationships have been observed for dengue, pertussis, malaria, varicella, and cholera vaccines.

These relationships between measures of efficacy are relevant for covid-19 vaccines, including how public health officials communicate trial results to the public. Though we have considerably less data on the prevention of severe disease for covid-19, the basic trend of higher efficacy against severe disease is holding up across vaccine platforms. For example, researchers at the Butantan Institute reported that an inactivated vaccine from Sinovac had 50.4% efficacy against mild to severe disease, but 100% efficacy against moderate to severe cases. While we see considerable variability in how well vaccines protect against mild disease, it is possible that different vaccines could end up working similarly well against severe disease. 

On the other end of the spectrum, the public is very interested in understanding vaccine efficacy against infection with or without symptoms, as this drives the vaccine’s ability to slow transmission. Preventing infection clearly also prevents transmission and symptoms, yet vaccines can offer clinical benefit by priming the immune system to prevent symptoms even despite infection. Thus, vaccine efficacy against infection cannot be higher than efficacy against disease. For an example from another disease, efficacy of a herpes simplex vaccine against herpes simplex virus 1 genital disease was 58% (95% CI 12 to 80%), but efficacy against herpes simplex virus 1 infection was 35% (95% CI 13 to 52%). 

Yet vaccine efficacy for preventing infection is harder to measure reliably. For SARS-CoV-2, it requires either frequent PCR screening, which is logistically complex for trials with tens of thousands of participants, or measuring a non-spike protein antibody response. Recently, limited antibody data from Johnson & Johnson’s trial indicate a reduction in asymptomatic infection. More detailed data on vaccine efficacy against infection is expected for the Moderna and Pfizer-BioNTech trials. Because some infection endpoints will be misclassified due to imperfect sensitivity and specificity, the results could make the vaccine appear to perform worse.

Yet there are several promising signals of how much protection from infection may occur. The Oxford-AstraZeneca trials reported combined efficacy against symptomatic disease of 70.4% (95% CI 54.8 to 80.6%, based on 131 events), and efficacy against infection of 55.7% confirmed by RT-PCR testing (95% CI 41.1 to 66.7%, based on 221 events). Moderna trial participants were swabbed before receiving their second dose, and there was an approximately 2/3 reduction in positivity in vaccinated versus unvaccinated participants (14 versus 38 infections), also confirmed by RT-PCR. This could reflect either fewer infections or a shortened period of viral shedding. Encouraging results from Israel show a 41% drop in confirmed covid-19 infections in adults aged 60 and older, where the Pfizer-BioNTech vaccine is in wide use.

Importantly, there is another measure of vaccine efficacy that lies along our spectrum, because a vaccine that does not prevent infection entirely may still reduce infectiousness and thus transmission. Vaccines can reduce the duration of the infectious period, as well as the pathogen load. More early data out of Israel suggests that infections in people who had previously been vaccinated had lower viral loads than infections in unvaccinated patients. In addition, there is accruing evidence from studies before vaccines were available that individuals who never develop symptoms are less infectious than presymptomatic or symptomatic individuals. Thus, one could posit that a vaccine that prevents symptoms could also reduce infectiousness. Efficacy against infection can therefore serve as a lower bound for the vaccine’s impact on transmission, and there are reasons to be optimistic that the real benefit is much higher.

In understanding the population-level impact of the covid-19 vaccines, it is useful to think about the various measures of vaccine efficacy as lying along a spectrum. Though it will take longer to accrue reliable data on efficacy against severe disease or against infectiousness, we can use estimates of efficacy against disease or against infection to guide our expectations. We can also communicate to the public the expected benefits of these vaccines and how they will be confirmed. While examples do exist of vaccines that reduce disease while failing to reduce shedding and transmission of the pathogen, it is worth pointing out that these are few and far between. The majority of vaccines that provide an individual-level benefit also provide a population-level benefit, which is why it is critical that we increase vaccine coverage worldwide.

Natalie Dean, Assistant Professor, Department of Biostatistics, University of Florida, USA.

Zachary Madewell, Postdoctoral Associate, Department of Biostatistics, University of Florida, USA.

Competing interests: We declare no competing interests.