Meta-analytical approaches for pooling time to event individual data : an application to non Hodgkin lymphoma survival studies

Carreras, Giulia and Pizzo, Anna Maria (2010) Meta-analytical approaches for pooling time to event individual data : an application to non Hodgkin lymphoma survival studies. Biomedical statistics and clinical epidemiology, 4 (1). pp. 15-25. ISSN 1973-2430

PDF portiere.pdf - Published Version Restricted to Repository staff only Download (639kB) | Request a copy

Abstract

Introduction. In combining data from several studies, meta-analytical methods must be adopted. Meta-analysis involves combining summary information or pooling individual data from different studies to integrate findings. Meta-analyses are performed by using fixed or random effects models, the latter adopted for analysing heterogeneity. Objectives. This work aims to examine and compare meta-analytical techniques applied to survival analysis for individual data. Methods. In survival data, analyzed by Cox regression models, heterogeneity may be attributed to differences in baseline functions or to the investigated effect. Three different random effect models were identified depending on variables that model heterogeneity: the baseline hazard, the covariate effect or both. A fixed effect model was also analyzed. Data from four studies on survival from Non-Hodgkin lymphoma were gathered. The pooled effect of smoking on overall survival was evaluated by applying the four models. Results. Results showed a better performance of random effect models respect to the fixed effect one. The baseline hazard appeared to better capture heterogeneity, suggesting that heterogeneity was mainly due to differences among study-populations. Conclusions. Random effect models resulted easy to be implemented in meta-analysis of individual survival data, necessary for studying heterogeneity and in our application better performing respect to fixed ones.

Item Type:	Article
Uncontrolled Keywords:	Cox model regression, individual patient data, meta-analysis, random effects, shared frailty model
Subjects:	600 Tecnologia - Scienze applicate > 610 Medicina e salute (Classificare qui la tecnologia dei servizi medici) 600 Tecnologia - Scienze applicate > 610 Medicina e salute (Classificare qui la tecnologia dei servizi medici) > 610.7 Medicina - Educazione e Insegnamento; Medicina - Ricerca > 610.72 Statistica medica
Depositing User:	Emanuele Zinevrakis
Date Deposited:	07 Nov 2011 08:14
Last Modified:	29 Mar 2017 16:11
URI:	http://eprints.bice.rm.cnr.it/id/eprint/3495

Actions (login required)

View Item