Statistical analysis plan for the Crystalloid Versus Hydroxyethyl Starch Trial (CHEST)

Abstract

Background: The Crystalloid Versus Hydroxyethyl Starch Trial (CHEST) is a 7000-patient, multicentre, randomised controlled trial comparing the effects of 6% hydroxyethyl starch (130/0.4) to normal saline for fluid resuscitation in intensive care patients. The trial design is based on the Saline Versus Albumin Fluid Evaluation (SAFE) study and will be the largest fluid resuscitation trial conducted to date.

Objective: In accordance with two other trials conducted by the investigators, a predetermined statistical analysis plan (SAP) has been described and made 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.

Methods: The SAP was designed by the chief investigators and statisticians and approved by the CHEST 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 Board. The data shells were produced from a previously published protocol. Statistical analyses are described in broad detail. Specifically, information relevant to baseline characteristics and processes of care were defined, and statistically relevant descriptive elements described, with appropriate comparisons between groups. Trial outcomes were selected, categorised into primary, secondary and tertiary outcomes, and appropriate statistical comparisons between groups were planned and described.

Results: A standard SAP for CHEST was developed. A trial profile outline and list of mock tables were produced. Descriptions of analyses of baseline characteristics, processes of care, measures of efficacy and outcomes were described. Six prespecified subgroups were defined and statistical comparisons between groups in these subgroups were described. In addition, analyses of tertiary outcomes, including health economic and functional outcome assessment, were described.

Conclusion: We have developed a predetermined SAP for CHEST. This plan accords with high-quality standards of internal validity to minimise analysis bias.