Biostatistics Weekly Seminar

Time-to-event analysis is a widely used tool for modeling disease progression data. Measuring the true survival outcome, such as the time to abnormality of cerebrospinal fluid biomarkers, can be costly and invasive. Consequently, the true survival outcome is only accessible for a small subset of participants, leading to limitations in sample size, loss in estimation efficiency, and bias if the missingness mechanism is not properly accounted for. An inexpensive/less invasive auxiliary outcome that is correlated with the true outcome may be collected. We propose a likelihood-based method and an E-M algorithm for Cox regression models which incorporate the error-prone auxiliary outcomes and improve estimation performance. Our proposed method also accommodates truncation and censoring in the true and auxiliary event times. We assess the performance of the method in finite sample scenarios through simulation studies and illustrate the proposed method using the Alzheimer's Disease Neuroimaging Initiative (ADNI) study.

Virtual: Zoom Link to Follow
18 January 2024
2:30pm