Article type
Year
Abstract
Background: Cochrane methods systematic reviews typically adapt methods for systematic literature searches, 'Risk of bias' assessment, data extraction, and results synthesis developed for intervention reviews. We originally used such methods to develop 'Full publication of results initially presented in abstracts', a methods review, but approaches for assessing risk of bias and identifying heterogeneity appeared inadequate for our ongoing update of the review.
Objectives: To develop:
1. a unique 'Risk of bias' tool for studies examining full publication outcomes of conference abstracts; and
2. robust algorithms to classify subgroups to facilitate applicability of synthesized results.
Methods: We identified methods used in included studies to assess risks of selection, information, and ascertainment bias. We compared these biases with those described by Schmucker et al in a methods review on full publication of studies approved by ethics committees or included in trial registries. We arrived at consensus on items to include. We also grouped all included studies by abstract source (i.e. how investigators identified the cohort of abstracts assessed for full publication).
Results: We included five components in the 'Risk of bias' tool (see Table):
1. sampling frame;
2. length of follow-up;
3. identification of full publications;
4. match of abstract to full publications; and
5. adjustment for factors possibly associated with full publication.
From 428 included studies, we identified unique subgroups based on abstract source (specific conference, specialized register, unique set of authors); location (international, national, or regional conferences); study design (randomized trial, systematic review, etc.); and medical specialty. Work concerning how to organize results synthesis using these subgroups is ongoing.
Conclusions: While some of the standard methods used to conduct Cochrane Intervention Reviews can be adapted for conducting a methods review, distinctive features of included methods studies require unique adaptations of the Cochrane Review methodology for assessing risk of bias and synthesizing results.
Objectives: To develop:
1. a unique 'Risk of bias' tool for studies examining full publication outcomes of conference abstracts; and
2. robust algorithms to classify subgroups to facilitate applicability of synthesized results.
Methods: We identified methods used in included studies to assess risks of selection, information, and ascertainment bias. We compared these biases with those described by Schmucker et al in a methods review on full publication of studies approved by ethics committees or included in trial registries. We arrived at consensus on items to include. We also grouped all included studies by abstract source (i.e. how investigators identified the cohort of abstracts assessed for full publication).
Results: We included five components in the 'Risk of bias' tool (see Table):
1. sampling frame;
2. length of follow-up;
3. identification of full publications;
4. match of abstract to full publications; and
5. adjustment for factors possibly associated with full publication.
From 428 included studies, we identified unique subgroups based on abstract source (specific conference, specialized register, unique set of authors); location (international, national, or regional conferences); study design (randomized trial, systematic review, etc.); and medical specialty. Work concerning how to organize results synthesis using these subgroups is ongoing.
Conclusions: While some of the standard methods used to conduct Cochrane Intervention Reviews can be adapted for conducting a methods review, distinctive features of included methods studies require unique adaptations of the Cochrane Review methodology for assessing risk of bias and synthesizing results.