Multi-level modelling of conception in artificial insemination by donor

Area of Science:

• Reproductive medicine
• Biostatistics
• Statistical modeling

Background:

• Donor sperm insemination data exhibit complex hierarchical structures.
• Both female (e.g., ovulatory cycles) and male (e.g., sperm donation quality) factors influence success rates.
• Existing models may not fully capture these intertwined hierarchical influences.

Purpose of the Study:

• To develop and apply a crossed random multi-level logistic model for analyzing donor sperm insemination data.
• To accurately estimate fixed effects and understand influences at each hierarchical level.
• To investigate the impact of sperm donation quality on donor basal fecundability and specific donations.

Main Methods:

• Utilized a crossed random multi-level logistic model to address the data's hierarchical nature.
• Implemented an efficient algorithm employing alternating Expectation-Maximization (EM) steps for model fitting.
• Incorporated compositional covariates to assess donation quality information.

Main Results:

• The proposed model provides improved estimation of fixed effects compared to simpler models.
• The analysis revealed significant influences at both female and male hierarchical levels.
• Compositional covariates offered insights into donor fecundability and donation-specific characteristics.

Conclusions:

• Crossed random multi-level logistic models are effective for analyzing complex hierarchical data in assisted reproduction.
• This approach enhances the understanding of factors influencing donor sperm insemination success.
• The methodology allows for a nuanced evaluation of donor and donation-specific contributions to fertility outcomes.