Skip to main content

Languages

Reporting of interim analyses, stopping rules, and Data Safety and Monitoring Boards in protocols and publications of discontinued randomized trials

Date and Location

Session: 

P4.002

Date

Monday 23 September 2013 - 10:30 - 12:00

Location

Presenting author and contact person

Presenting author

Mihaela Stegert

Contact person

Mihaela Stegert
Abstract text
Background: Pre-planned interim analyses, stopping rules, and the presence of an independent Data Safety and Monitoring Board (DSMB) are means to increase transparency and credibility of the decision-making process in case of randomized controlled trials (RCTs) discontinuation. Objectives: To determine: (1) the prevalence of planned interim analyses, stopping rules and DSMBs in RCTs; (2) the purpose of interim analyses (safety, efficacy, futility, other); (3) the proportion of discontinued trials which report DSMBs; (4) the proportion of discontinued trials with a matching stopping rule; (5) discrepant reporting between protocols and publications; and (6) trial characteristics associated with reporting of planned interim analyses, stopping rules, and DSMBs. Methods: A multi-centre cohort of RCTs was established including protocols approved by six research ethics committees (RECs) in Switzerland, Germany and Canada between 2000 and 2003. Data on trial characteristics were extracted from protocols and corresponding publications. The completion and publication status of RCTs was assessed using information from REC files, literature searches, and by investigators survey. Results: We included 949 RCT protocols. Of these, 302 (31.8%) reported interim analyses, 160 (16.9 %) stopping rules, and 263 (27.7%) the presence of DSMBs. The main reported purposes of interim analyses were: efficacy (26.2%), safety (19.2%), and both combined (21.9%). Of a total of 250 (26.3%) discontinued trials (due to any reason), 63(25.2%) mentioned a DSMB in protocol. We found some evidence for an association between pre-specification of a stopping rule and trial discontinuation (odds ratio 1.56, 95% CI 0.98-2.49, p 0.06). However, only 10/250 discontinued RCTs pre-specified a matching stopping rule in the protocol (benefit (N=3), harm (N=6), futility (N=1)). Full results regarding all outlined objectives will be presented at the Colloquium. Conclusions: Trial discontinuation appears to happen mostly ad-hoc. Systematic reviewers should bear this in mind when assessing the risk of bias of discontinued RCTs.