646 – Missing Data Methods 2
Statistical Techniques for Comparing Immune Response of Quadrivalent Vaccines with Trivalent Vaccines
Ayca Ozol-Godfrey
Sanofi Pasteur
Robert D. Small
Sanofi Pasteur
A multivalent vaccine is one that has antigens for more than a single virus strain or species. An important question is whether the different antigens can interfere with the human immune response to the other antigens. A trivalent vaccine has three of the antigens of a quadrivalent vaccine. For a flu vaccine the various antigens are chosen from among a number of antigens. A common design then is to compare a quadrivalent vaccine with two trivalent vaccines whose antigens are different subsets of three of the antigens in the quadrivalent vaccine. Usually the various comparisons are made in a univariate manner. This does not use all of the information in the data since the correlations between immune response is significant. We propose multivariate methods that view the trivalent vaccines as a quadrivalent vaccine with missing responses. We use both direct maximum likelihood (ML) methods and missing data methods (multiple imputation-MI) to use all of the information in the data. The result is greater precision with shorter CIs for the comparisons. The methods can be extended to other vaccines than flu and to multivalent vaccines that have occasional missing titer values.