The Analysis of Biased Time-to-Event Data from Pregnancy Registries

• Activity Number: 479
• Type: Invited
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Ronghui Xu*+ and Walter Faig
• Companies: University of California, San Diego and University of California, San Diego
• Keywords: left truncation ; cure rate ; Cox regression ; weighted likelihood ; semiparametric estimate ; nonparametric maximum likelihood

Abstract

We consider pregnancy outcomes such as spontaneous abortion or preterm delivery, in the context of observational studies of drug exposure. These are essentially binary endpoints. However, due to accrual through pregnancy registries, women can enter a study any time during their pregnancy. Not counting for such left truncation leads to bias in the estimated rates. In addition, a substantial portion of the women will not have the events of interest, a portion termed 'cured' in survival analysis. While left truncation is relatively easily dealt with in the Cox proportional hazards regression, with a cured proportion new methodology is needed. We investigate approaches using the exact semiparametric likelihood, an approximate likelihood, and a weighted (complete data) likelihood. Variance estimates are derived with closed-form expressions. Time permitting efficiency consideration will be discussed.