Article
Pooled-studies publications: an analysis of the relevant characteristics of these articles for systematic reviewers
Search Medline for
Authors
Published: | July 10, 2012 |
---|
Outline
Text
Background: Because of high risk of bias systematic reviewers exclude publications containing pooled data that are not based on a systematic review and perform no critical appraisal.
Objectives: To systematically explore the characteristics of the PSPs, in particular those that might suggest susceptibility to bias.
Methods: We systematically searched the database of exclusions from a recent comparative effectiveness review of second-generation antidepressants conducted for the US Agency for Healthcare Research and Quality. We dually abstracted information from all PSPs, including: number of pooled trials and patients; funding or conflicts of interest; was pooling planned a priori; how trial protocols differed; and was data from unpublished trials included.
Results: We retrieved 65 PSPs containing an average of 6.4 trials and 1489 patients. Sixty-three (97%) of the PSPs were either explicitly funded by a pharmaceutical company or at least one of the authors was an employee of a pharmaceutical company. In total, 3 (4.6%) PSPs stated that pooling of the included trials was planned a prior, 12 (18.5%) pooled trials with identical protocols, and 23 (35.4%) included unpublished data.
Discussion: PSPs are almost exclusively funded by the pharmaceutical industry and regularly ontain unpublished data. Pooled trial often differed in patient population, duration, dosing, and comparator agents.
Implications for guideline developers/users: Because PSPs are almost exclusively industry-funded a high degree of suspicion about the analysis of data presented is warranted; however PSPs contain analyses of unpublished data that might otherwise be unavailable to reviewers.