• Laurent Billot, Balasubramanian Venkatesh, John Myburgh, Simon Finfer, Jeremy Cohen, Steve Webb, Colin McArthur, Christopher Joyce, Rinaldo Bellomo, Andrew Rhodes, Anders Perner, Yaseen Arabi, Dorrilyn Rajbhandari, Parisa Glass, Kelly Thompson, Maryam Correa, and Meg Harward.
• The George Institute for Global Health, Sydney, NSW, Australia. bmvenkat@bigpond.net.au.
• Crit Care Resusc. 2017 Jun 1; 19 (2): 183-191.

BackgroundThe Adjunctive Corticosteroid Treatment in Critically Ill Patients with Septic Shock (ADRENAL) trial, a 3800-patient, multicentre, randomised controlled trial, will be the largest study to date of corticosteroid therapy in patients with septic shock.ObjectiveTo describe a statistical analysis plan (SAP) and make it public before completion of patient recruitment and data collection. The SAP will be adhered to for the final data analysis of this trial, to avoid analysis bias arising from knowledge of study findings.MethodsThe SAP was designed by the chief investigators and statisticians and approved by the ADRENAL management committee. All authors were blind to treatment allocation and to the unblinded data produced during two interim analyses conducted by the Data Safety and Monitoring Committee. The data shells were produced from a previously published protocol. Statistical analyses are described in broad detail. Trial outcomes were selected and categorised into primary, secondary and tertiary outcomes, and appropriate statistical comparisons between groups are planned and described in a way that is transparent, available to the public, verifiable and determined before completion of data collection.ResultsWe developed a standard SAP for the ADRENAL trial, and have produced a trial profile outline and list of mock tables. We describe analyses of baseline characteristics, processes of care, measures of efficacy and outcomes. Six pre-specified subgroups were defined, and statistical comparisons between groups in these subgroups are described.ConclusionWe have developed an SAP for the ADRENAL trial. This plan accords with high-quality standards of internal validity to minimise analysis bias.

41 keywords...

hide…