“Independent” reanalysis of landmark starch solutions trial was published by original authors

The commentary by Doshi raises apposite questions about the analysis of the CHEST data. Doshi points out that the ongoing refusal of the CHEST authors to make their data available for open analysis remains a cause for concern, as the recently presented re-analysis is open to question.

The adverse events analysis (and re-analysis) was performed on a so-called “safety set” of all patients who received HES regardless of randomisation. However, this analysis could only have been performed post hoc. The original statistical plan stipulated that all data would be analysed by intention to treat, but the adverse events data using intention to treat analysis have not been presented (as they ought to have been) prior to the safety set analysis. As far as can be ascertained from available data, the intention to treat analysis of the original data suggested no significant between-group differences for the adverse events.

This raises a further question. In terms of the original protocol, renal replacement therapy (RRT) was not a specified endpoint, but could be regarded as an adverse event. It is surely reasonable to analyse the RRT data using the “safety set” grouping. Such an analysis would show if all patients exposed to starch at any stage were at risk of increased need for RRT.

We echo the call by Doshi and others for the full data set from this study to be available to the scientific community as a whole.