Procedures

Baseline demographic and clinical characteristics were collected upon hospital presentation and randomisation. Neurological severity was measured on the National Institutes of Health Stroke Scale (NIHSS; range 0–42, with higher scores indicating greater severity)⁶ at baseline, and 24 and 72 h. The functional outcome of participants was assessed on the modified Rankin scale (range 0 [no symptoms] to 6 [death]), administered through telephone or in-person interview by trained researchers blind to treatment allocation, at 7 (or at hospital discharge, if earlier), 28 and 90 days. All computerised tomography (CT) and magnetic resonance imaging (MRI) on participants, undertaken at baseline (i.e. confirmation of diagnosis) and on follow-up at 24–36 h (routine assessment of infarct size and reperfusion injury), and at other time points as clinically indicated (i.e. neurological deterioration), were uploaded to a secure purpose-built web-based system for central analysis at The George Institute for Global Health. Training and instruction of clinician assessors in the interpretation of abnormalities on brain imaging included the structured interpretation of acute (presence/absence of acute ischaemic lesion, its size, swelling, any haemorrhagic transformation, arterial thrombus or angiography findings) and pre-stroke (old infarcts, leukoaraiosis, brain volume loss) imaging findings derived from the University of Edinburgh, and previously used in third International Stroke Trial (IST-3).⁷ The brain imaging assessor group included a radiologist (GBZ) and three experienced neurologists (CC, LYZ, and SO), who undertook online training on the software and data forms, which included the rating of 21 CT scans from participants of the ENCHANTED alteplase-dose arm study.^8,9 Full details of their performance are provided in the Appendix (p 4).

The primary outcome was the size of post-thrombolysis cerebral infarction measured on follow-up brain imaging. The primary analysis was on CT scans undertaken at 24–36 h post-randomisation, as undertaken according to usual standard of care. Three sets of sensitivity analysis were conducted on other approaches to assessing the brain imaging: follow-up MRI alone at 24–36 h, combined measures from either a CT or MRI at 24–36 h, and measures from all follow-up brain imaging. In each analysis set, for patients with multiple scans at different time points, the earliest follow-up scan was selected for analysis. Secondary outcomes were intracranial haemorrhage and large parenchymal haematoma (PH2), defined according to the Heidelberg bleeding classification system,¹⁰ and the type and volume of intracranial haemorrhage, location of cerebral infarction, and functional recovery according to scores on the modified Rankin scale at 90 days.

Brain imaging in Digital Imaging and Communications in Medicine (DICOM) format, identified only with a patient's unique study number, were used to measure the size of cerebral infarction (ml) by computer-assisted multi-slice morphometric and voxel threshold techniques using software MiStar version 3.2 (Apollo Medical Imaging Inc, Melbourne, VIC, Australia).¹¹ This software was automatically applied to each brain imaging (CT or MRI) slice to calculate the size of identifiable cerebral infarction (Hounsfield units for CT with hypodensity, imaging sequence for diffusion weighted imaging of MRI), which was then summed manually to calculate total size. Reliability of this method has been validated elsewhere.^12,13

All follow-up scans were reviewed by at least one assessor; a second assessor confirmed any and parenchymal haemorrhage. If there was any disagreement, a third assessor was required to finalise the diagnosis and classification. Other brain imaging abnormalities, such as cerebral atrophy and old infarcts, were also collected. All assessors were blinded to treatment group, outcomes, and time of scan. Finally, 10% of all scans were double reviewed to assess inter-rater quality¹⁴ (Appendix p 4).