Loading...
1 results
Search Results
Now showing 1 - 1 of 1
- Métodos bayesianos aplicados à modelagem conjunta de dados longitudinais e de sobrevivênciaPublication . Martins, R.; Silva, Giovani Loiola da; Andreozzi, Valeska Lima, 1974-Joint modeling of longitudinal and survival data has received increasing attention in the recent years, especially related to AIDS and cancer studies. Generally, these data have been analyzed considering survival (time-to-event) outcome and longitudinal (repeated measures) outcome separately. The existence of individual speci c processes which vary over time and contribute both to the longitudinal and survival result justify the joint modeling of that information. This approach allows to consider simultaneously the correlation between the repeated measures for the individual and its survival time. The motivation of the work was to provide a survival analysis considering the in uence of a longitudinal biomarker in the survival time of HIV/AIDS patients who are residents in Brazil. We employed a bayesian methodology for jointly modelling the two types of data, making inferences for the parameters of interest via Markov chain Monte Carlo (MCMC) methods. We proposed several joint models, namely models with spatial random e ects (to account for the unobserved heterogeneity amongst individuals from the same region), cure fraction models (to deal with possible long term survivors) and penalized B-spline models (to allow a more exible non-linear longitudinal trajectory). The results show that the joint Bayesian models present considerable improvements in the distribution of the median survival time in comparison with those ones obtained by separate modeling. The introduction of spatial random e ects showed the absence of regional extravariation of patients from di erent Brazilian states. The assumption of a cure fraction in the population revealed no improvement. As expected, the inclusion of splines provides better exibility in modeling the longitudinal biomarker trajectory.