Back to search

University of Ioannina

Studies with missing values in network meta-analysis

Abstract

dc:description

According to systematic reviews in psychiatry, the dropout rate isconsiderable high in the studies of this field. Despite the high dropoutout, the reporting on the extent of missing outcome data, the analyticalstrategies undertaken and the acknowledgment of the implication ofmissing outcome data on the results have been particularly insufficient.190 systematic reviews in the mental health field published in theCochrane Database of Systematic Reviews after 1/1/2009 by threeCochrane Review Groups were considered in order to provide empiricalevidence on addressing missing outcome data. The findings of thisreview indicated that the description of the amount and the extend ofmissing outcome data and the inclusion of their implications in thefindings of the systematic reviews seem to be particularly suboptimal.Another importance consideration in the missing outcome data issueis the minimization of their occurrence. In order to prevent as much aspossible the missing outcome data, it is important to investigate and tounderstand the reasons that the patients drop out early from trials.Therefore, an extensive research was initiated to identify trialcharacteristics that have an impact on premature discontinuation inantipsychotic trials for schizophrenia. The findings of this researchrevealed that high dropout rates in antipsychotic trials can be associatedwith various characteristics, particularly with the use of placebo ascontrol and the size of the study.A conventional meta-analysis model has been suggested to betterquantify uncertainty in the imputations of the missing outcome data byincorporating in the model an unknown parameter that refers to binarymissing outcome data and it has been called the ‘informativemissingness odds ratio’ (IMOR). This parameter describes therelationship between the unknown occurrence of the outcome amongmissing participants and the known outcome in observed participants.The present research extended the idea of IMOR into a network metaanalysissetting. Two dataset comparing anti-manic treatments andantidepressants were used. In the original meta-analyses, missingparticipants were assumed to have failed regardless the treatment theywere randomized to. This assumption was evaluated by consideringseveral IMOR models that reflected different assumption on the missingoutcome data. In both datasets, the relative effectiveness of thetreatments was affected only by the worst- and best-case analyses.Moreover, heterogeneity increased in both datasets under these twoextreme scenarios. Overall, there were small changes on the ranking ofthe anti-manic and antidepressant treatments. The posterior and priorIMOR distributions were very similar implying that there was littleinformation about the true outcome in missing participants.

Degree

thesis:*
Grantor dc:publisher
University of Ioannina
Year dc:date
2014

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Spineli, Loukia
  • Σπινέλη, Λουκία

Subjects

dc:subject × 20

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/39557

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Spineli, Loukia; Σπινέλη, Λουκία. Studies with missing values in network meta-analysis. University of Ioannina, 2014. http://hdl.handle.net/10442/hedi/39557