Imaging and treatment response after ischaemic stroke.

838 www.thelancet.com/neurology Vol 11 October 2012 Use of brain imaging to predict treatment response after ischaemic stroke, a test of which is reported by Maarten Lansberg and colleagues in this issue of The Lancet Neurology, has a long history. In 1993, Gilles Marchal and colleagues used information about both blood fl ow and metabolic activity obtained with PET to predict diff erential clinical outcomes. They identifi ed three outcome groups: fi rst, when perfusion was normal or increased and metabolism was normal, clinical outcome was uniformly favourable, and spontaneous recanalisation was assumed to have occurred; second, when blood fl ow was severely reduced and brain metabolism in the ischaemic region was very low, clinical outcomes were uniformly poor: irreversible damage had already occurred; and third, when perfusion was reduced but oxygen consumption was reduced although still present, clinical outcomes were variable. Patients in the last group had a substantial volume of tissue with features consistent with ischaemic penumbra, a state of critical hypoperfusion associated with impaired neuronal activity that is amenable to salvage by restoration of perfusion, but which is otherwise likely to progress to infarction. A penumbral pattern on brain imaging might be used to defi ne the subset of patients for whom intervention might aff ect clinical outcome, and who therefore could be the best target for clinical trials. Investigation of penumbral imaging with the clinically practical approaches of CT and MRI has been a focus of acute stroke imaging. The MRI signature of irreversible tissue injury is a lesion on diff usion-weighted imaging. When combined with information about tissue perfusion from dynamic susceptibility weighted contrast MRI (perfusion-weighted imaging), the region of mismatch on the MRI scans can be used to defi ne the penumbra. However, optimum parameters for image processing and patient selection have emerged from small datasets. The combined data from two well-conducted small studies, Diff usion and Perfusion Imaging Evaluation for Understanding Stroke Evolution (DEFUSE), in which all participants received intravenous alteplase, and the Echoplanar Imaging Thrombolytic Evaluation Trial (EPITHET), a randomised controlled trial of intravenous alteplase compared with placebo, have yielded tissue viability thresholds—ie, values of brain perfusion that discriminate viable from non-viable tissue, and viable tissue that is at risk of infarction if hypoperfusion persists from tissue that will survive despite being hypoperfused—and pragmatic clinical trial selection criteria that also incorporate lesion volume. To minimise observer variability, these criteria have been included in a scanner-independent software package (RAPID) that forms the basis for ongoing trials of late intravenous alteplase (Extending the Time for Thrombolysis in Emergency Neurological Defi cits [EXTEND] and European Cooperative Acute Stroke Study [ECASS] 4) and the DEFUSE 2 study. The DEFUSE 2 investigators postulated that diff erent treatment responses would occur depending on pretreatment MRI characteristics in patients undergoing an intra-arterial revascularisation procedure within 12 h after onset of stroke. Selection of such population enabled the investigators to accurately characterise early recanalisation and reperfusion status, the major determinants of both radiological and clinical outcomes, in 99 patients. The study confi rmed a diff erential response to successful reperfusion: patients with target mismatch had a more favourable clinical outcome (increase in National Institutes of Health Stroke Scale score of 8 points or more) than those who did not have target mismatch (odds ratio 8·8, 95% CI 2·7–29·0 vs 0·2, 0·0–1·6; p=0·003). DEFUSE 2 adds support for the use of MRI mismatch as an entry criterion for phase 2 clinical trials—it could minimise harm among those unlikely to benefi t, and also substantially reduce sample sizes needed—and more specifi cally for adoption of the RAPID defi nitions in multicentre trials, as opposed to several alternative approaches to mismatch defi nition and patient selection that are in use. Limitations of use of MRI target mismatch approaches for large trials include restriction to sites with access to imaging equipment and exclusion of up to 20% of patients, for example because of ferromagnetic implants or acute illness, evident in very slow recruitment rates in previous studies. The additional time needed for mismatch imaging, including acquisition, processing, and interpretation of images is a consideration when the benefi t of treatment decreases rapidly with time—for example, with intravenous alteplase, which has a 4·5 h window. In such situations, non-contrast CT remains the standard of care. CT perfusion imaging—potentially more widely available than MRI—has yet to be standardised suffi ciently for investigators to be wholly confi dent in Imaging and treatment response after ischaemic stroke