Use of composite outcomes in cluster randomized control trials of influenza vaccines

Despite being widely used in other fields, composite outcomes are seldom used in influenza vaccine randomized trials. Vaccine trialists may benefit from the increased event rates and power of composite outcomes. We explored the use of a composite outcome on a series of randomized control trials conducted within Canadian Hutterite communities.

Using data from three blinded, cluster randomized control trials of pediatric influenza vaccination, we assessed the use of a composite outcome of antimicrobial prescription, school or work-related absenteeism, and medically attended visits for respiratory illness. Key comparisons between component outcomes and the composite included odds reduction, frequency, and confidence intervals.

In Trial 1, 65 Hutterite colonies were randomly assigned to receive either the influenza vaccine or the hepatitis A vaccine. Among all season 1 and 3 participants in Trial 1, the composite outcome yielded a greater event rate, significance, precision, and a greater protective effect than its components. Within participants across all seasons in Trial 2, the composite yielded a greater event rate, lower significance, a null protective effect, and greater precision than two out of three of its components. Among all participants across all seasons in Trial 3, the composite outcome yielded a greater event rate and a weaker protective effect than its components, with no notable differences in precision or significance.

Composite outcomes can be a viable approach to raise event rates and increase the precision of estimates in influenza vaccine trials and should be explored further within this field.