Assessing the effect of vaccine on spontaneous abortion using time‐dependent covariates Cox models

Abstract

In studying the safety of vaccine for influenza A (H1N1) given during pregnancy, spontaneous abortion (SAB) is one of the important points to consider. Because women may receive the vaccine any time during their pregnancy, evaluation of the effect of the vaccine on SAB should only take place after vaccination and in the risk window for SAB, that is, the first 20 weeks of gestation. In addition, when such studies are conducted through pregnancy registries where recruitment occurs after pregnancy recognition, the accrued subjects are left truncated in the sense that they are not followed from the start of pregnancy. As previously reported, left truncation needs to be properly handled using survival analysis methods to avoid bias. In the context of time-dependent vaccine exposure, a time-dependent covariate Cox model can be used to simultaneously take into account the left truncation and the vaccine exposure timing. In this communication, we illustrate the approach using the Vaccine and Medication in Pregnancy Surveillance System data. We explain in details how the model is fitted using different software. We recommend survival analysis methods together with collection of necessary data to study the effects of vaccine on SAB.