Table 2. Key features of individually randomized and cluster-randomized trials.

Feature	Individually randomized trial	Cluster-randomized trial
Unit of analysis	Individuals randomized to the investigational vaccine or control arm. Ratio of participants in the investigational vaccine to control arm is typically 1:1 (most statistically efficient with fixed number of participants) but can be 2:1 or 3:1 (for increasing the safety database for the vaccine).	Clusters or groups of participants randomized to the investigational vaccine or control arm. Often, clusters are defined geographically (e.g., villages), but they can also be defined based on social contacts (e.g., Ebola ring vaccination trial) [19].
Differential in risk if vaccine proves effective	Between individuals who receive investigational vaccine and those who receive comparator	Between those in clusters randomized to receive vaccine and those in control clusters.*
Statistical efficiency	Greater statistical efficiency compared to cluster randomization [22]. All else equal, this leads to faster time to completion and earlier opportunity to administer a vaccine to control participants once it is judged effective [21].	Analysis incurs a statistical “penalty,” known as the design effect, to account for correlations in outcomes among members of the same cluster, leading to larger sample size requirements [23,24]. The design effect can be especially large when incidence is highly clustered in space and time [18,25–27].
Effects measured	Only measures direct protective effects [20].	Measures the combination of direct and indirect effects. Only in special circumstances can be designed and analyzed to elucidate the relative contributions of each effect.

*Note: If a “control” vaccine or placebo is used in the control clusters, then vaccine efficacy assessment can be focused on the comparison of disease incidence between those who actually received the vaccine in the vaccine clusters and those who actually received the control intervention in the control clusters. However, if no control vaccine or placebo is used in the control clusters (e.g., delayed vaccination), then the only unbiased comparison is between all those eligible to receive the vaccine in both types of cluster, including those who refused vaccination in the vaccine clusters (as this group cannot be separated out in the control clusters).