Multivariate decay models for longitudinal censored maternal antibody levels against measles, mumps, rubella and varicella

Abstract

The aim is to jointly model maternal antibody decay in infants for measles, mumps, rubella and varicella, and to estimate the duration of passive immunity as well as individual heterogeneity. In a prospective study of pregnant women and their offspring, blood samples were taken during pregnancy (week 36), at birth (cord) and in infants at 1, 3, and 12 months, and randomly at either 6 or 9 months (Leuridan et al., 2010, 2011). Disease-specific antibodies were measured by enzyme linked immunosorbent assay (ELISA). Longitudinal antibody levels are analyzed using non-linear mixed effects models accounting for left-censoring and right-censoring, induced by the lower and upper quantification limits of the ELISAs. We assess the effect of important maternal and infant characteristics and use infant-specific random effects to quantify the heterogeneity between subjects. Non-linear decay models, such as the generalized exponential, Gompertz and logistic model, are used to describe the maternal a