Amita Manatunga
Department of Biostatistics
Emory University
Modeling the Relationship Between Two Correlated Survival Outcomes
Often, disease occurrence time of the same subject can be measured by
using two different methods. The primary interest is to determine whether
both methods agree in measuring the disease freedom. In this presentation,
we provide a statistical framework of this problem by distinguishing from
approaches of comparing two marginal distributions or measuring the
association between event times. We propose several indices to quantify the
agreement between the event times and discuss their statistical properties
and modeling procedures with covariates. We illustrate our results with a
prostate cancer epidemiology study. Patients with prostate cancer are
treated with simultaneous radiotherapy and their prostate specific antigen
(PSA) is measured every six months after the treatment. The disease
recurrence times are measured by two different methods based on the
post-treatment PSA levels.
Back to Colloquium Series