In the absence of studies of breakthrough cases, several teams have attempted to

determine a correlate of protection for COVID-19 vaccines by comparing

published data on vaccine efficacy with those from earlier trials measuring immune
responses to the jabs2,3.

Six months of COVID vaccines: what 1.7 billion doses have taught scientists

These studies suggest that virus-blocking ‘neutralizing’ antibodies are a good

predictor of a vaccine’s success. Those that trigger high levels of these antibodies,
such as the Pfizer–BioNTech and Moderna jabs, are more effective than the
Oxford–AstraZeneca and Johnson & Johnson vaccines, which generated relatively
low levels of neutralizing antibodies.

The Oxford–AstraZeneca study confirms the relationship between higher

neutralizing-antibody levels and protection. The analysis, led by Oxford
biostatistician Merryn Voysey, compared the immune responses in 171
breakthrough cases with those of more than 1,404 people who got the vaccine and
did not develop a symptomatic infection.

Participants who had higher levels of neutralizing antibodies — as well as

‘binding’ antibodies, which recognize the SARS-CoV-2 spike protein — tended to
gain stronger, but not total, protection from a symptomatic infection, the Oxford
team determined. The team used a model to estimate the antibody levels that
corresponded to different levels of vaccine protection against COVID-19 in trials,
ranging from 50% to 90% protection. Other vaccines that trigger similar antibody
responses can be expected to generate similar levels of protection against
symptomatic infections, the Oxford team says.

Miles Davenport, an immunologist at the University of New South Wales in

Sydney, Australia, notes that there was no significant difference in the neutralizing
antibody responses of breakthrough infections and controls. This could occur if
young people at greater risk of infection — because they have more social
contacts, for example — also had higher antibody levels. The Oxford team
accounted for this overlap in their model by estimating participants’ risk of
infection. However, Davenport says that it's a challenge to identify protective
antibody levels based on estimated risk, rather than observed differences in
antibody levels — which would have been possible only if there were clear
differences between breakthroughs and controls.