Exploring redundant research into the effect of tranexamic acid on surgical bleeding: further analysis of a systematic review of randomised controlled trials

OBJECTIVES: We examined whether apparent redundancy in a cumulative meta-analysis of trials is justified by concern about bias, random error or generalisability of the results. DESIGN: Cumulative meta-analysis, risk of bias assessment, trial sequential analysis, description of study participants over time and a review of rationales for conducting trials. DATA SOURCE: 126 randomised trials included in a systematic review assessing of tranexamic acid on blood transfusion in surgery. RESULTS: The cumulative meta-analysis including all trials shows that the pooled estimate first reached statistical significance after the second trial in 1993. When the analysis was limited to the 38 high-quality trials and adjusted to account for potential systematic and random errors, the uncertainty was resolved after the 22nd trial in 2008. When the analysis was restricted to the two high-quality, prospectively registered trials, the cumulative z-curve crossed p=0.05 but not the monitoring boundary, suggesting an early potentially spurious statistically significant result. As precision of the pooled estimate increased, the number of trials initiated increased, although trial activity appeared to move to other surgery types. Most (62%) reports cited at least one systematic review. Of 118 reports examined, concern about generalisability was the reason for initiating the trial in 60%. Other reasons were to address a question other than the effect on bleeding (26%) and to confirm previously observed results (4%). Unawareness of previous research was apparent in 4% trials, while the rationale was unclear in 3%. CONCLUSIONS: Our results indicate that poor quality is a more important cause of redundant research than the failure to review existing evidence. Concerns about generalisability of results is the main motivation for new trials. Contrary to previous claims, our results suggest that systematic reviews showing treatment effects can stimulate an increase in trial activity rather than reduce it.