Great news has come out in the fight against Ebola. A vaccine against Ebola has been shown to highly effective in trials conducted during the Ebola outbreak in Guinea. In fact, some news sources claim that the vaccine is 100% effective.
Is the vaccine truly 100% effective? Likely no.
In the original study in the Lancet, the researchers identified individuals who were diagnosed with Ebola. Then, clusters of individuals who had close contact with the person diagnosed with Ebola were identified. These were mostly family and close friends. Some individuals received the vaccine and others received the vaccine, but in a delayed manner.
Among the 4,123 people who received immediate vaccination, no one contracted Ebola. This appears to be a 0% infection rate. However, among the 3,528 people who received delayed vaccination, 16 contracted Ebola, which corresponds to a 0.5% infection rate. Because the baseline infection rate is so low, it is highly unlikely that the vaccine is in fact 100% effective. In fact, the authors claim that the true rate of efficacy of the vaccine is more likely to be around 75%. Why is this the case?
If most people out of the 4,123 were likely to get infected, then observing 0 Ebola cases is a strong results. However, based on the results of the delayed vaccination, we would only expect 18 or 19 people to be infected. Decreasing the number infected from an expected 18 or 19 to zero is a significant achievement. However, if infection rates were around 50% and about 2000 people were infected, then this would constitute a much more certain outcome.
In practice, the authors used the following approach to estimate the confidence intervals:
In case of zero cases of Ebola virus disease occurring (ie, vaccine efficacy 100%) a 95% CI was derived by fitting a β-binomial distribution to the cluster-level numerators and denominators and using an inverted likelihood ratio test to identify the lower bound for vaccine efficacy. For comparisons in which events were reported in both groups, a Cox proportional hazards model was fitted using a cluster-level frailty term to adjust for clustering within rings.
Regardless of the statistical methods used, this is certainly a key advance in the fight against Ebola.