Multivariate extension of meta-analysis (Doctoral thesis)
Standard methods for meta-analysis are limited to the case of comparing two interventions. In real life clinical practice, however, there are usually many alternative competing interventions that can be used to treat the same disease, while studies may contrast different sets of these interventions, thus forming a network of evidence. In such complicated cases of data availability pairwise meta-analyses cannot give a definite answer as to which intervention works best for the target condition. Network meta-analysis (NMA) is an extension of the standard, pairwise meta-analysis, and can be used to jointly analyze evidence regarding multiple interventions in order to produce clinically relevant estimates. In Chapter 2 of this dissertation we described an updated review of methods for NMA, which we performed in order to summarize the state-of-the-art in the field. Our scope was to provide a comprehensive account of the currently available methods, which can be used by researchers interested in assessing the quality of published NMAs, in applying NMA to answer new clinical questions, or in conducting further methodological research. The second aim of this dissertation was to advance the statistical methodology for jointly analyzing multiple correlated outcomes in NMA. In Chapter 3 we introduced a multiple outcomes network meta-analysis (MONMA) model which focused on the case of analyzing multiple dichotomous outcomes while accounting for the correlations between them. The model synthesizes information from randomized controlled trials augmented by external evidence, which can be obtained from expert clinicians. In Chapter 4 we presented two additional MONMA models. Both models can be used to synthesize multiple dichotomous, continuous, or time-to-event outcomes. In order to illustrate our methods, we applied all our MONMA models to a network of antimanic drugs, where 15 drugs and placebo were compared in terms of efficacy and acceptability. We found that our models provided more precise estimates for most treatment comparisons, for both outcomes. Βased on our findings we recommend researchers to consider both univariate and multivariate approaches when possible, to ascertain if clinical conclusions about the ranking of treatments for each outcome remain consistent under different model assumptions. As a final, concluding remark, we believe that the research presented in this dissertation is an important advancement in the field of NMA. We also think that our models constitute the best available method for the network meta-analysis of multiple correlated outcomes, and that their implementation is in practice straightforward.
|Institution and School/Department of submitter:||Πανεπιστήμιο Ιωαννίνων. Σχολή Επιστημών Υγείας. Τμήμα Ιατρικής|
|Keywords:||Μετα-ανάλυση,Βιοστατιστική,Μετα-ανάλυση δικτύου,Meta-analysis,Biostatistics,Network meta-analysis|
|Appears in Collections:||Διδακτορικές Διατριβές|
Files in This Item:
|Δ.Δ. ΕΥΘΥΜΙΟΥ ΟΡΕΣΤΗΣ 2017.pdf||2.86 MB||Adobe PDF||View/Open|
Please use this identifier to cite or link to this item:This item is a favorite for 0 people.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.