A method for longitudinal prospective evaluation of markers for a subsequent event

Area of Science:

• Biostatistics
• Epidemiology
• Longitudinal Studies

Background:

• Longitudinal comparisons of event markers are crucial for understanding disease progression and treatment efficacy.
• Existing methods may not adequately capture the predictive value and population impact of potential markers.
• The menopausal transition provides a relevant clinical context for evaluating markers of a subsequent event (final menstrual period).

Purpose of the Study:

• To present a novel statistical method for longitudinal comparison of alternative markers of a subsequent event.
• To decompose the marker evaluation into discriminatory ability and prevalence factor.
• To apply and compare the method using markers for the menopausal transition.

Main Methods:

• The proposed method assesses aggregate prediction gain from marker occurrence over time.
• It decomposes the prediction gain into discriminatory ability (time difference to event) and prevalence factor (marker proportion).
• The method was applied to four proposed markers of the menopausal transition.

Main Results:

• The method provides an exact decomposition of marker performance into two key components.
• Application to menopausal transition markers allowed for comparison with previous findings.
• The discriminatory ability and prevalence factor offer distinct insights into marker utility.

Conclusions:

• The developed method offers a robust framework for evaluating longitudinal markers.
• It enhances understanding of marker utility by separating predictive accuracy from population prevalence.
• This approach is valuable for identifying effective markers in various health-related longitudinal studies.