192 – Noninferiority Studies
Using Simulation When Only Minimal Information Is Available to Estimate the Design Effect for an Ebola Vaccine Evaluation Study
Charles E. Rose, PhD
Centers for Disease Control and Prevention
Paul Gargiullo, PhD
Centers for Disease Control and Prevention
Benjamin Lopman, PhD
Centers for Disease Control and Prevention
Manoj Gambhir, PhD
Centers for Disease Control and Prevention
Randomized trials (RT) are the benchmark for estimating vaccine efficacy (VE). When individuals cannot be practically randomized to a study arm, a group RT (GRT) may be an attractive alternative despite a loss of statistical efficiency, i.e., an inflated standard error, for estimating VE. This loss of efficiency is often referred to as the design effect (DEFF) or variance inflation factor (VIF). Failure to account for the lack of independence among individuals within a group during the design stage may result in an underpowered GRT with an inadequate sample size. An estimate of the intraclass correlation (ICC) is usually necessary to estimate the DEFF. Estimates of the DEFF or ICC are often challenging to obtain, especially for a novel vaccine or drug. In the study design phase for an evaluation to estimate the efficacy of a candidate Ebola vaccine there is an absence of information on the DEFF or ICC. We use a minimal amount of field data available from Sierra Leone on the Ebola incidence and RT groups (types and sizes) to conduct a simulation to estimate ICC and DEFF.