OUTCOMES IN CLINICAL INTERVENTION RESEARCH: SELECTION, SPECIFICATION, AND REPORTING

outcomes;systematic reviews;clinical trials;social network analysis;not listed

Saldanha, Ian JudeKatz, Joanne ;

Johns Hopkins University

关键词: outcomes;    systematic reviews;    clinical trials;    social network analysis;    not listed;   

Others  :  https://jscholarship.library.jhu.edu/bitstream/handle/1774.2/44618/SALDANHA-DISSERTATION-2015.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: JOHNS HOPKINS DSpace Repository

【 摘 要 】

IntroductionSound clinical intervention research relies on the use of the right outcomes dependably and without bias. This dissertation addresses important research gaps related to selection, specification, data collection and analysis, and reporting of outcomes in research.Methods We examined systematic reviews (SRs) and randomized controlled trials (RCTs). For Aim 1, related to outcome selection, we conducted a case study of outcomes in all Cochrane SRs addressing HIV/AIDS (June 2013), to evaluate whether social network analysis methods could be used to identify central outcomes for core outcome sets. For Aim 2, related to outcome specification and data collection and analysis, we examined all Cochrane SR protocols (June 2013) addressing four major eye conditions for completeness of pre-specification and comparability of outcomes. For Aim 3, related to outcome reporting, we examined all conference abstracts of RCTs presented at the Association for Research in Vision and Ophthalmology (ARVO) 2001-2004 conferences to: (1) evaluate agreement in main outcome results comparing abstracts and corresponding full publications, and (2) evaluate the association between conflicts of interest and full publication. ResultsFor Aim 1, we applied social network analysis methods to identify central outcomes, and found that the most central outcomes often differed from the most frequent outcomes. For Aim 2, outcome pre-specification in SRs was largely incomplete. For Aim 3, only 44.8% of abstracts describing RCTs were published in full, and more than half (54.7%) of the 86 conference abstract/full publication pairs had some form of discordance in reported results for the main outcome. First author conflicts of interest were associated with full publication, irrespective of whether the main outcome results in the abstract were statistically significant, not statistically significant, or the statistical significance was not reported. ConclusionsThis dissertation identifies causes for concern, such as incomplete outcome pre-specification in SR protocols, and discrepancies in results comparing conference abstract/full publication pairs of the same RCT. We suggest ways forward, such as use of social network analysis methods to identify central outcomes for core outcome sets, and incorporation of a five-element framework for outcome pre-specification in SR and RCT protocols.