Characterization of gastrointestinal adverse effects reported in clinical studies of corticosteroid therapy.

OBJECTIVES: To examine whether 159 studies included in a previous meta-analysis reported on gastrointestinal bleeding or perforation in accordance with the CONSORT extension for reporting harms outcomes (CONSORT Harms recommendations checklist); whether differences were associated with funding source, journal, or publication year; and whether the CONSORT Harms checklist is a suitable tool for evaluation of adverse effects reporting.

STUDY DESIGN AND SETTING: Articles were assessed for fulfillment of the CONSORT Harms recommendations, funding source, publication type, and year. Agreement between reviewers was assessed by comparing scores for each study.

RESULTS: The mean CONSORT Harms score was 5.25 out of 10 (standard deviation ± 2.09). Most studies included information on participant withdrawals (133 studies, 83.6%), absolute risk of gastrointestinal bleeding or perforation (130 studies, 81.8%), and how harms-related information was collected (118 studies, 74.2%). Reporting of gastrointestinal bleeding or perforation increased with higher scores (odds ratio 1.173, P = 0.042). There was no significant association between CONSORT Harms score achieved and publication year or funding source, but there was a trend toward higher scores in studies published in the major medical journals (score difference 0.78, P = 0.052). Definitions of gastrointestinal bleeding differed between studies. Reviewer agreement was fair to moderate with large variations.

CONCLUSION: Few studies in the systematic review received high scores using the CONSORT Harms criteria. Most studies reported on the most important criteria regarding risk of gastrointestinal bleeding or perforation. Reviewer agreement showed large variations due to imprecise texts and ambiguous criteria. Routine scoring according to fulfillment of the CONSORT Harms recommendations would be inadvisable without qualified judgment.