Estimating causal effects from multiple cycle data in studies of in vitro fertilization

Area of Science:

• Reproductive Epidemiology
• Biostatistics
• Clinical Trials

Background:

• Prospective reproductive studies often collect data over multiple cycles, such as in vitro fertilization and embryo transfer (IVF-ET).
• Time-varying covariates (e.g., age, diagnostic markers) are common.
• Observational studies face challenges with exposure probability depending on subject characteristics and significant attrition, which can be influenced by prior outcomes.

Purpose of the Study:

• To illustrate the application of causal modeling techniques for analyzing multiple-cycle reproductive data.
• To address complications arising from time-varying exposures and subject characteristics in observational studies.
• To provide a detailed description of inference using weighted estimating equations.

Main Methods:

• Causal modeling framework applied to longitudinal reproductive data.
• Weighted estimating equations for robust inference.
• Detailed review of key assumptions for causal inference in this context.

Main Results:

• Demonstration of causal modeling's utility in handling complex reproductive datasets.
• Application to a specific study on the effects of hydrosalpinx in women undergoing IVF-ET.
• Improved methods for measuring exposure effects in the presence of attrition and time-varying factors.

Conclusions:

• Causal modeling offers a powerful approach for analyzing multiple-cycle reproductive outcomes.
• The proposed methods, including weighted estimating equations, provide a robust framework for inference.
• This methodology is particularly valuable for observational studies with potential confounding and attrition, such as IVF-ET research.