Defining superiority, equivalence and non-inferiority in clinical trials of second-generation malaria vaccines. Summary of the possible trial types, outcomes and considerations when testing second-generation malaria vaccines compared to a partially efficacious first-generation licensed vaccine. Error bars correspond to possible trial outcomes and indicate two-sided 95% confidence intervals (CI). Δ (the superiority margin (+Δ), non-inferiority margin (-Δ) and equivalence margin (-Δ to +Δ) can be defined by an absolute or a relative difference in actual malaria outcomes. Interpretation of trials depends on where the CI for the true difference in outcome falls relative to Δ and the null effect (0). For superiority trials, to conclude superiority, the trial effect may be bigger or smaller than Δ but the 95% CI must be above 0 (scenarios A and B). For equivalence trials, equivalence requires the CI to lie wholly within a bidirectional symmetrical equivalence margin (-Δ and Δ, scenarios C and D). If the effect estimates lie outside the bidirectional symmetrical equivalence margins, the second-generation malaria vaccine is either better or worse than the first-generation vaccine . In a non-inferiority trial, the prime interest is determining whether the new malaria vaccine is no worse than the non-inferiority margin (-Δ) which, if exceeded, defines the new treatment as being inferior to RTS,S. For non-inferiority trials, if the CI lies completely to the right of the prespecified margin (-Δ) a conclusion of non-inferiority of the second-generation malaria vaccine is reached (scenarios E and F). If the CI includes -Δ it is concluded that the new malaria vaccine is inferior to the first-generation malaria vaccine.
Fowkes et al. BMC Medicine 2013 11:232 doi:10.1186/1741-7015-11-232