The simplest design involves randomization of equal numbers of subjects to the candidate vaccine and control groups (that is, 1:1). In trials that employ a control group that is not vaccinated against the disease to be prevented, but some clinical data are available to support the likely efficacy of the candidate vaccine, it may be appropriate (subject to statistical considerations and an assessment of the impact on the total trial sample size) to use unbalanced randomization (for example, 2:1 or 3:1) to reduce the chance that individual subjects will be randomized to the control group, thus ensuring that the majority of trial subjects receive the candidate vaccine.
Trials may be planned to follow trial subjects for a fixed period after the last dose of the primary series. The time at which the primary analysis is conducted should take into account the anticipated rates of the disease under study in each treatment group, including the unvaccinated control group if applicable. Other considerations regarding the timing of the primary analysis may include the possible importance of having some information on the duration of protection before licensure occurs, the feasibility of following up subjects for prolonged periods, and whether or not the vaccine could address a pressing unmet need (for example, in an outbreak situation where there is no approved vaccine to prevent the disease).
Alternatively, a case-driven approach may be taken based on the anticipated rates of the primary efficacy end-point in the control group and the expected or minimum desirable level of efficacy of the candidate vaccine. In this design the primary analysis is conducted once a pre-specified total number of cases has been detected – based, in a double-blind setting, on the anticipated numbers in test and control groups required to demonstrate the projected vaccine effect.
Alternative designs that allow for comparison with a control group that is not vaccinated against the disease to be prevented may, at least in the short term, include the following:
■ In a randomized stepped wedge trial, the candidate vaccine is administered to predefined groups in a sequential fashion. Each predefined group is a unit of randomization. These may be geographical groups or groups defined by host factors (for example, age) or other factors (for example, attendance at a specific school or residence within a specific health-care facility catchment area). Such a design may be chosen when there is good evidence to indicate that the vaccine will do more good than harm (affecting the equipoise associated with randomization to a control group that is not vaccinated against the disease to be prevented) and/or when it is impossible to deliver the intervention to all trial participants within a short time frame.
■ In a ring vaccination trial, the direct contacts (and sometimes secondary contacts) of a case may be randomized to vaccine or control or may be randomized to receive immediate vaccination or vaccination after a period of delay (21). This type of post-exposure cohort trial usually requires smaller sample sizes than prospective randomized controlled trials.
Ring vaccination trials may be particularly applicable when the infectious disease to be prevented is associated with a relatively high incidence of secondary cases in susceptible populations. Therefore the use of this trial design requires prior knowledge of the infectivity of the infectious agent and of the proportion of infections that are clinically apparent, as well as of the general susceptibility of the trial population.
Ring vaccination trials may not be appropriate if the vaccination regimen requires multiple doses over an extended period to induce a protective immune response.
The follow-up period for subjects after contact with the index case should extend to the upper limit of the incubation period, taking into account both the period during which the index cases were infectious and the contact period. The inclusion period for new cases and controls and their contacts following the detection of the first case should be stated in the protocol. The duration of the inclusion period should take into account the potential for introducing bias if the disease incidence changes over time.