As Shigella vaccine candidates move toward large Phase 3 trials, a modeling study highlights a challenge for measuring how well they work: people may acquire some immunity through infection during a trial. In places where Shigella is common, that natural protection could narrow the difference between vaccinated and unvaccinated groups over time, making vaccine efficacy appear lower than it would in a population with no prior exposure. [S1, Captured publisher article text]
The study, published in PLOS Medicine, did not test a vaccine in people. It simulated hypothetical trials and compared estimates under different assumptions about infection rates, vaccine effects, and immunity after infection. The authors modeled settings with different Shigella burdens and tested six combinations of infection surveillance and statistical analysis. The aim was to estimate how well each approach recovered vaccine efficacy in a fully susceptible population. [S1, Captured publisher article text]
The logic is that unvaccinated infants could “catch up” immunologically after repeated exposure, while vaccinated infants already have vaccine-conferred protection. If a trial records only diarrhea symptoms, it can miss infections without symptoms that might still confer immunity. Those missed infections matter because the comparison between trial groups changes as immunity accumulates. [S1, Captured publisher article text]
Across the simulated scenarios, only one approach consistently recovered the modeled efficacy in a fully susceptible population: active surveillance that detected infections, including those without symptoms, paired with an analysis that counted only each participant’s first observed outcome. The authors say that approach may be difficult to use in practice because it requires regular testing of people who are not ill. Other approaches generally underestimated efficacy, with larger bias at sites with higher infection burdens. [S1, Captured publisher article text]
One illustrative result shows the scale of the possible difference, but it is not a measurement from a real vaccine trial. In a well-powered simulation where the assumed efficacy against severe Shigella diarrhea was 60%, the active-surveillance, single-outcome approach estimated 59.5% in a high-burden setting. The other five approaches estimated between 47.3% and 51.7% under the same scenario. These figures describe model outputs conditional on its assumptions, not the performance of any vaccine candidate. [S1, Captured publisher article text]
In simulations with more realistic trial sizes, the general pattern remained: as infection burden rose, confidence-interval coverage and the chance of missing a real vaccine effect worsened for most approaches. Active infection surveillance with a single-outcome analysis performed best; symptom-based reporting with a single-outcome analysis was the next-best option in the scenarios described. The results indicate that how a trial finds and counts infections can affect what its efficacy estimate means. [S1, Captured publisher article text]
The distinction is important. The paper separates effectiveness in the trial’s actual setting, where natural immunity is part of the population’s experience, from efficacy in a fully susceptible population, which the authors identify as relevant to regulatory evaluation. A lower estimate in a high-burden setting would not by itself establish that a vaccine has lost biological protection; the study suggests that accumulated natural immunity could contribute to the difference. [S1, Captured publisher article text]
The size of the projected bias is uncertain. The simulations rely on assumptions about Shigella’s natural history and how much protection infection provides, and the authors note that published data on those factors are limited. The model also did not account for differences between Shigella serotypes or variation in participants’ prior immunity. Actual Phase 3 trial results could therefore differ from the numerical examples. [S1, Captured publisher article text]
The work offers a framework for anticipating how surveillance choices and natural exposure may shape future trial estimates, not a forecast of a specific vaccine’s efficacy. The captured PLOS Medicine page identifies the article as an uncorrected proof. [S1, Captured publisher article text]