Sequential predictions of menstrual cycle lengths.

Area of Science:

• Reproductive endocrinology and statistical modeling.
• Biostatistics and computational biology.
• Population health and family planning.

Background:

• Accurate forecasting of menstrual cycle length and phases is crucial for infertility management and natural family planning.
• Existing methods may not fully capture the temporal dynamics of individual menstrual cycles.
• Large datasets of repeated cycle measurements are available for analysis.

Purpose of the Study:

• To develop a novel Bayesian hierarchical dynamic approach for forecasting menstrual cycle length and its phases.
• To explicitly account for the temporal nature of menstrual cycle data.
• To enable prediction of conception probability in future cycles when combined with a fecundability model.

Main Methods:

• Utilized a large English database with repeated measurements of cycle length and preovular phase.
• Employed a state-space process to model the temporal behavior of cycle lengths for individual women.
• Embedded individual processes into a multivariate Bayesian hierarchy allowing for subject-specific parameter variations.

Main Results:

• The proposed Bayesian hierarchical dynamic model effectively captures the temporal dynamics of menstrual cycle length.
• The model allows for individual variability in cycle parameters.
• Integration with a fecundability model enables forecasting of conception probability based on intercourse patterns.

Conclusions:

• The developed Bayesian dynamic approach offers a robust method for menstrual cycle forecasting.
• This method enhances the potential for personalized infertility management and natural family planning strategies.
• The model provides a framework for predicting conception risk by incorporating temporal cycle data and intercourse behavior.