Do systematic reviewers and clinical trialists in the same field consider similar outcomes to be important? A case study in HIV/AIDS




Poster session 2


Monday 24 October 2016 - 15:30 to 16:00


All authors in correct order:

Saldanha I1, Li T1, Williamson P2, Dickersin K1
1 Cochrane United States, and Cochrane Eyes and Vision US Satellite, USA
2 University of Liverpool Institute of Translational Medicine, UK
Presenting author and contact person

Presenting author:

Ian Saldanha

Contact person:

Abstract text
Background: Systematic reviewers select outcomes they perceive as relevant; yet trialists addressing the same research question may not report similar outcomes. Understanding the amount and type of overlap in outcomes between reviews and trials could inform whether core outcome sets should incorporate outcomes examined in trials, reviews, or both.

Objective: To examine overlap between outcomes examined in reviews addressing HIV/AIDS and trials included in them.

Methods: Eligible reviews were completed, published Cochrane Reviews of HIV/AIDS examining at least one trial as of June 2013. We identified all outcomes (domains) examined in the reviews and the trials. We calculated the percent positive agreement (PPA) as the proportion of all outcomes that occurred in both trials and reviews (Box). We predefined four intervention subgroups: clinical management, biomedical prevention, behavioral prevention, and health services.

Results: Of 140 published Cochrane reviews of HIV/AIDS, 99 were completed and 84 included at least one trial. Most reviews (72/84; 86%) were published from 2008-2012. The 84 reviews included 524 trials; most (78%) published from 1993-2007. The 84 reviews examined 218 unique outcomes (median 7.5 outcomes each, interquartile range (IQR) 4-11). The trials examined 779 unique outcomes (median 8, IQR 5-12), 3.6 times the number of unique outcomes as the reviews (779 vs 218). PPA ranged from 20% for health services to 33% for clinical management. When comparing the most frequent outcomes within intervention subgroups (Table), trials more frequently examined interim, short-term, and safety outcomes (e.g. adherence, viral load, and adverse events (specified)); reviews more frequently examined long-term and perhaps more patient-important outcomes (e.g. quality-of-life, intervention acceptability).

Conclusions: Although numbers of outcomes per review and per trial were similar, the outcomes were not. Differences in perspectives and goals between these two sets of researchers may explain the differences in outcomes they examine. Developers of core outcome sets should note that reviews and trials often provide complementary types of outcomes.