WP 14: Statistical analysis

Objectives

Develop the final statistical analysis plan (SAP). Appropriate use of statistical methodology and analysis is essential for evaluating benefits and harms of cooling in the EuroHYP-1 trial and for providing reliable and valid interpretation of the results. The EuroHYP-1 statistical analysis plan (SAP) for the final analysis will specify the hypotheses that are to be tested and the treatment effects to be estimated in order to satisfy the primary and secondary objectives of the EuroHYP-1 trial. In all elements, the statistical methodology employed will minimise the risk of systematic errors (bias) and random errors (type I and type II). The SAP will be finalised before the EuroHYP-1 trial provides any data for analyses. The statistician performing the final analysis will be fully independent from the Data Safety and Monitoring Committee (DSMC) and the Executive Committee and will remain blinded to the intervention allocation throughout the analysis and reporting of the results.

Work package Description

In order to maximise efficiency of the procedures relating to the final statistical analyses we will pilot three processes: the data transfer system; the statistical analyses; and the data validation systems. This pilot is performed, not as an analysis per se, but to ensure the feasibility, efficiency, validity, and adequate responsiveness of the systems.

To meet the primary objective of the EuroHYP-1 trial, statistic analyses will assess whether systemic cooling to a target temperature of 34-35°C improves the functional outcome at three months in patients with acute ischaemic stroke. The primary outcome measure is the modified Rankin scale (mRS) at 90 days using multi-level, ordinal logistic regression. The mRS is an ordinal score indicating functional outcome in patients with stroke, fittingly, we will employ the use of ordinal distribution or ‘shift analysis’ to analyse this outcome. Pre-defined covariates for the adjusted primary analysis will be: age (≥ 70), sex, stroke severity (NIHSS ≥12), intention to give alteplase, and centre. Where more than 5% of data are missing the risk of bias caused by missingness will be minimised. This will involve using three analytical methods: a complete case analysis (which is usually biased); analysis using multiple imputations; and sensitivity analysis of the missing data tests. The SAP and analyses will be developed following The Copenhagen Trial Unit’s standard operating procedure (SOP) ‘Statistical analysis of clinical trial data’.

To meet the secondary objective of the EuroHYP-1 trial, statistical analyses will assess whether one or more particular patient groups are more or less likely to benefit from cooling. Primary and secondary outcome measures and all subgroup analyses and sensitivity analyses will be pre-defined in the protocol and the SAP in order to limit systematic bias, spurious findings and prevent ‘data-driven’ analysis. To limit systematic bias in making inferences from the data the statistician will remain blinded to intervention allocation during the analysis and reporting to the Steering Committee. The statistical analysis report will contain the results of the data analysed according to the methodological features described in the SAP.