CMStatistics 2017: Start Registration
View Submission - CMStatistics
B0987
Title: Flexible two-stage model proposal for multivariate longitudinal and survival data using spline smoothing Authors:  Ipek Guler - KU Leuven (Belgium) [presenting]
Christel Faes - Hasselt University (Belgium)
Carmen Cadarso Suarez - Universidad de Santiago de Compostela (Spain)
Francisco Gude - Complexo Hospitalario Universitario de Santiago de Compostela (Spain)
Abstract: In many biomedical studies, it is common to follow up subjects repeatedly. These follow-up studies typically produce different types of outcomes including both longitudinal biomarkers and time-to-event outcomes. Often, the interest is on assessing the relationship between the longitudinal and the time-to-event processes. Joint modelling approaches of longitudinal and survival data is an appropriate way to study such relationship. Existent joint models are mostly concentrated on a single longitudinal and survival process. However, many studies collect several longitudinal biomarkers and instead of selecting a single longitudinal biomarker we want to study the relationship between all these biomarkers and the survival outcome. Additionally, flexible regression techniques may be necessary for the non-linear longitudinal and survival trends. For instance, Orthotopic Liver Trasplantation data includes glucose and daily insulin therapy which could predict the risk for death of the patients who underwent transplantation. Both longitudinal trends and the risk for death show non-linear profiles. The joint modelling approaches in a frequentist framework are difficult to implement when the number of longitudinal biomarkers is large or when flexible techniques are needed. For this aim we propose a two-stage based modelling approach for modelling of non-linear multivariate longitudinal and non-linear survival data using spline smoothing.