Khoury · Nature medicine 2021 · Predictive mathematical modeling and meta-regression · n=?

Neutralizing antibody levels are highly predictive of immune protection from symptomatic SARS-CoV-2 infection.

Cited 4249 times in the scientific literature.

Level 3 - non-randomized controlled study

Predictive mathematical modeling and meta-regression of aggregate clinical trial and cohort data

PubMed 34002089 · doi:10.1038/s41591-021-01377-8 · record verified 2026-08-31

What was done

Authors analyzed the relationship between in vitro neutralizing antibody titers and clinical protection from SARS-CoV-2 infection using published data from seven COVID-19 vaccines and convalescent cohorts. They developed a mathematical model to calculate the neutralization threshold for 50% protective efficacy, modeled antibody decay over 250 days post-immunization, and projected efficacy against viral variants of concern.

What was found

The neutralization titer required for 50% protection against detectable infection was estimated at 20.2% of the mean convalescent level (95% CI = 14.4–28.4%). The level needed for 50% protection from severe infection was significantly lower at 3% of the mean convalescent level (95% CI = 0.7–13%, P = 0.0004). Decay modeling over 250 days predicted significant loss of protection from infection over time, whereas protection from severe disease was predicted to be largely retained.

Why it matters

This study provided an early quantitative benchmark establishing neutralizing antibodies as a predictive correlate of vaccine efficacy, facilitating estimates of waning immunity and protection against emerging variants.

Limits

The analysis relies on aggregate study-level data from seven vaccine trials rather than individual participant data, and total sample size is not stated in the abstract. Predictions regarding long-term antibody decay over 250 days and cross-variant efficacy are model-based extrapolations rather than direct long-term observational measurements, and non-neutralizing mechanisms such as T-cell immunity were not modeled.

Cited by