Amendment on the article “Missing outcome data management in acute stroke trials testing iv thrombolytics: Is there risk of bias?”

Dear editor

It has come to our attention that by the time our article was published, Mr Young-Saver and colleagues had published a similar article on the same topic, dated December 2018. ¹ Our work forms part of a larger project registered within VISTA-acute database back in 2016, aimed at exploring missing outcome data features. Our results were presented in May 2019 within the European Stroke Organization Conference 2019, ² and then prepared and sent for publication in ESJ by October 2019.³ We performed a thorough scientific search on the topic around the fall of 2018 but regret not having updated it before sending our work to ESJ, so missing the work by Mr Young-Saver and colleagues.

Nevertheless, it is gratifying to note how both groups developed such similar approaches for assessing the way different replacing methods for missing outcome data may alter trial results, and reassuring to verify that two independently conducted analysis of distinct datasets reached similar conclusions. A similar two-step analysis was conducted, first roughly reanalysing complete cases to compare real data with that coming from each individual imputation method; and finally simulating missingness by censoring a growing proportion of cases to observe the effect of those methods on the final OR and p values in a full trial scenario. In both cases taking into consideration too, the two accepted methods for mRS analysis. And we gladly conclude that, faced with missing data, multiple imputation seems to be the most appropriate way to estimate outcomes, preserving study power to detect a treatment benefit, in the particular scenario of the stroke recovery process.