Association between brain imaging signs, early and late outcomes, and response to intravenous alteplase after acute ischaemic stroke in the third International Stroke Trial (IST-3): secondary analysis of a randomised controlled trial

1Prompt treatment with intravenous alteplase improves independent survival after acute ischaemic stroke. However, concerns about the risk of symptomatic intracranial haemorrhage and whether patients with early signs of ischaemia on CT should receive alteplase could be deterring use of this treatment.

2Early signs of ischaemia on non-enhanced brain CT include tissue hypoattenuation, lesion swelling, and arterial hyperattenuation from occlusive thrombus. Tissue hypoattenuation might represent irreversible tissue damage and has been associated with increased risk of symptomatic intracranial haemorrhage in some, but not all, studies. These signs are rarely seen alone, but the association between combinations of signs and prognosis after stroke or their interaction with alteplase is unknown. Moreover, in some alteplase trials, patients with specific findings (eg, extensive tissue hypoattenuation) were excluded.

3Other pathological features—such as leukoaraiosis, cerebral atrophy, and old infarcts—are often seen on brain CT in patients with acute stroke. These pre-existing signs might indicate brain frailty (ie, a vulnerability to ischaemia) or increased risk of symptomatic intracranial haemorrhage. The association of pre-existing signs with response to alteplase, with or without early ischaemic signs, is unknown.

4The third International Stroke Trial (IST-3) is the largest (n=3035) randomised controlled trial of alteplase in acute stroke to date and was designed to test whether a wider range of patients could benefit from alteplase up to 6 h after stroke. The primary endpoint of IST-3—an increase in independent survival (defined as an Oxford Handicap Scale [OHS] score of 0–2 at 6 months)—did not differ between alteplase and control groups (odds ratio 1·13, 95% CI 0·95–1·35; p=0·181); however, in prespecified secondary analyses (ordinal shift analysis of the OHS and an OHS score of 0–1), alteplase increased functional outcome in all patients randomised up to 6 h and increased the primary outcome in the patients treated less than 3 h after stroke. The effect of alteplase 18 months after stroke has been published elsewhere. The findings of IST-3 accord with those of previous alteplase trials, at all time windows.

5A secondary objective of IST-3 was to ascertain whether early ischaemic or pre-existing brain CT signs—individually or in combination—were associated with response to alteplase on several clinically relevant early and late outcomes. Here, we present that secondary analysis.

6IST-3 was an international, multicentre, open-label, prospective blinded endpoint, randomised controlled trial done at 156 centres in 12 countries (appendix pp 2–5). Patients were eligible for the study if the treating clinician felt there was no clear indication for or contraindication to alteplase and judged the treatment promising but unproven for that individual. Eligibility criteria were age 18 years or older (no upper age limit) and symptoms of cortical, lacunar, and posterior circulation stroke, of all severities, according to the National Institutes of Health Stroke Scale (NIHSS). All participants had to be able to undergo randomisation and start treatment within 6 h of stroke. We did not exclude patients with early ischaemic or pre-existing imaging signs on brain CT. We excluded individuals younger than 18 years, people with standard contraindications to alteplase, patients with established visible infarction (ie, indicating stroke onset was likely to be more than 6 h previously), and individuals with haemorrhagic stroke or non-stroke lesion as the cause of stroke.

7The IST-3 protocol and statistical analysis plan have been published elsewhere. The study was approved by the Scotland A research ethics committee and by ethics committees and other regulatory bodies of all participating countries, hospitals, and institutions. All patients, or a relative if the patient lacked capacity, provided written informed consent before randomisation into the trial.

8Full details of the IST-3 randomisation procedure, including the minimisation process, are presented elsewhere. Briefly, the randomising clinician recorded baseline data via a central, secure, telephone-based or web-based system. The system randomly allocated patients to either standard best medical care plus immediate thrombolysis with intravenous alteplase (0·9 mg/kg bodyweight; maximum 90 mg; 10% bolus with the remainder over 1 h) or control (standard best medical care alone) within 6 h of ischaemic stroke. Patients allocated to the control group received the same stroke care as did individuals allocated to the alteplase group. The first 276 patients enrolled to the study were treated in a placebo-controlled, double-blind phase (the placebo was an inert product identical in appearance to alteplase); thereafter, patients were randomised to either immediate alteplase or open control.

9To be approved for inclusion in IST-3, all participating centres had to pass minimum image acquisition standards to ensure that brain scans were of diagnostic quality for acute stroke (for CT) and included the minimum correct sequences (for MRI). Before randomisation, all patients had either a CT or MRI brain scan, repeated 24–48 h after stroke and again if any neurological deterioration arose in the first 7 days. The present analysis concerns brain images taken before randomisation, up to 6 h after stroke.

10Participating centres sent all brain images to the IST-3 coordinating centre for adjudication and quality checking; all scans were anonymised. Adjudicators were masked to all information, except whether the scan was prerandomisation or follow-up. They viewed the images and recorded their ratings on a secure web-based system (systematic image review system [SIRS]). All adjudicators were neuroradiologists (JMW, RvK, AvH, LC, NB, ZM, AF, GP) or stroke neurologists (AP, AA) with experience in stroke imaging; they rated 60 scans from the ACCESS study and 25 scans from IST-3 that were selected at random, to ensure satisfactory agreement (defined as κ>0·70) for presence of early ischaemic signs and no more than one category difference for pre-existing signs. Adjudicators completed all analyses before the database was locked and the randomisation code was broken. Every scan was read by one adjudicator. Any discrepancies between the scan rating and clinical data (eg, side of brain affected, presence of haemorrhage) identified during data cleaning before data lock were cross-checked by JMW, who assigned a final rating before unmasking took place and the randomisation code was broken.

11We first checked brain CTs for early ischaemic signs (tissue hypoattenuation, lesion size, swelling, and hyperattenuated artery); we used the term visible infarct to represent any of these signs. We then looked for pre-existing structural signs (old infarcts, leukoaraiosis, and atrophy). We classified images using validated scores. We classed scans as normal only if no early ischaemic changes or pre-existing changes were present.

12We defined the presence and degree of hypoattenuated tissue as either mild (grey matter attenuation equal to normal white matter) or severe (grey and white matter attenuation slightly less than normal white matter but still consistent with onset of stroke within 6 h). We classified the extent of acute ischaemic lesions in three ways: with the one-third middle cerebral artery (MCA) method; with the IST-3 method; and with the Alberta Stroke Program Early CT Stroke (ASPECTS) score. The IST-3 score for infarct extent reflects all arterial territories, whereas the ASPECTS score and one-third MCA method focus only on the MCA territory. Therefore, we used the IST-3 score as the primary measure of infarct size in analyses, condensing the full IST-3 lesion extent score into four groups for analysis: small infarcts, which we defined as lacunar, small cortical, small cerebellar, less than half of brainstem, or less than half of the anterior cerebral artery (ACA) or posterior cerebral artery (PCA) territory; medium infarcts, classed as striatocapsular, the anterior or posterior half of the peripheral MCA territory, or more than half the ACA or PCA territory; large infarcts, defined as the whole of the peripheral MCA territory or all the MCA territory; and very large infarcts, which comprised the whole MCA and PCA territory, all the MCA and ACA territory, or all three territories. Small infarcts on the IST-3 score were equivalent to ASPECTS 8–10, medium infarcts corresponded to ASPECTS 5–7, and large and very large infarcts were similar to ASPECTS 0–4. MCA involvement in small and medium infarcts (according to IST-3 score) was equivalent to less than one-third MCA, and in large and very large infarcts it corresponded to more than one-third MCA. We graded ischaemic lesion swelling on a seven-point scale. We noted the presence or absence and location of any hyperattenuated artery.

13We recorded the location of old infarcts (eg, cortical, lacunar, border zone, and brainstem or cerebellar). We noted the presence and severity of leukoaraiosis on CT; if brain MRI was done, we used the Fazekas scale to grade changes. We classified atrophy as none, moderate, or severe when compared against standard examples. We also classed follow-up scans for any haemorrhage (petechial, haematoma in or remote from infarct, intraventricular, or subarachnoid blood), including whether the haemorrhage was likely to worsen neurological status, as part of the outcome assessment for symptomatic intracranial haemorrhage.

14The primary outcome of IST-3 was the proportion of patients alive and independent, defined by an OHS score of 0–2, at 6 months. The OHS is similar to the modified Rankin scale and has values from 0 to 6, with 0 representing no symptoms or completely independent and 6 signifying the patient has died. OHS score 0–2 indicates good functional outcome. We also analysed two prespecified secondary endpoints: an ordinal analysis of the OHS; and OHS score of 0–1 (ie, favourable outcome). Other secondary outcomes were symptomatic intracranial haemorrhage within 7 days, death within 7 days, and death by 6 months.

15Local staff at every centre followed up all patients for 7 days and recorded early outcomes (symptomatic intracranial haemorrhage, death, and likely cause of death before 7 days), which were sent to the central trials office. At 6-month follow-up, central researchers who were masked to group allocations sent questionnaires by post to patients or their carers, to record independence (using the OHS); if we knew the patient had died, we sent the questionnaire to their family doctor, to obtain the date of death and likely cause.

16Analyses of early ischaemic and pre-existing structural brain signs, their effect on functional outcome, death, and symptomatic intracranial haemorrhage, and interactions with alteplase, were prespecified in the trial protocol and analysis plan. We analysed imaging data for patients whose prerandomisation scans were received at the central trials office. We did statistical analyses with SAS version 9.3.

17We used logistic regression to ascertain associations between imaging signs and: (a) age, NIHSS score, and time to randomisation; and (b) outcomes at 7 days (symptomatic intracranial haemorrhage) and 6 months (OHS score of 0–2 and OHS score of 0–1), which were judged representative of the most clinically relevant early hazard and late benefit. We did multivariate logistic regression, adjusting for the linear effects of age, NIHSS score, and time to randomisation, to identify whether any combinations of imaging variables were associated with symptomatic intracranial haemorrhage or OHS score of 0–2. We then tested for interactions of imaging signs (presence, absence, or severity) and response to alteplase (adjusting for age, NIHSS score, and time to randomisation) with symptomatic intracranial haemorrhage, death within 7 days, and OHS score of 0–2 at 6 months; as a secondary analysis, we assessed interactions of imaging signs and response to alteplase with an OHS score of 0–1 and by ordinal OHS at 6 months (a more statistically sensitive ordinal model than analysis of dichotomous OHS categories). We tested whether associations between imaging signs and outcomes and imaging signs and alteplase differed in patients randomised within 3 h, 3–4·5 h, or 4·5–6 h of stroke. We tested whether the response to alteplase differed by the extent of early ischaemic signs (IST-3 score and ASPECTS score 0–7 vs 8–10), and we did a prespecified meta-analysis of data from IST-3 with imaging data from other thrombolysis trials. We judged p values less than 0·05 significant, except for analyses of interactions between imaging signs and response to alteplase, for which we used a significance level less than 0·01, to minimise false-positive results.

18This trial is registered at ISRCTN.com, number ISRCTN25765518.

19The funding sources had no role in study design, data collection, data analysis, data interpretation, or writing of the report. All authors had full access to all data in the study and the corresponding author had final responsibility for the decision to submit for publication.

20Between May 1, 2000, and July 31, 2011, 3035 patients were recruited to IST-3 and underwent randomisation. Brain scans were available for 1507 patients assigned to alteplase and for 1510 controls; scans for 18 patients were not received at the central trials office and, therefore, were excluded from analyses (figure 1). Table 1 shows baseline clinical and imaging characteristics, which were well balanced between arms and no variables were missing.

21Of 3017 prerandomisation brain scans, 2962 (98%) were obtained by CT and 55 (2%) by MRI. The expert panel judged 269 (9%) scans normal (ie, no early ischaemic or pre-existing signs); 1224 (41%) patients had early ischaemic signs (ie, visible infarct, whether or not pre-existing signs were also present) and 1524 (51%) had pre-existing signs (but no early ischaemic signs). The commonest early ischaemic sign was tissue hypoattenuation, seen in 1203 (40%) patients, and the least frequent sign was hyperattenuated artery, recorded in 735 (24%) patients (table 1). Pre-existing signs were common: 1336 (44%) patients had signs of an old infarct, 1547 (51%) had leukoaraiosis, and 2327 (77%) had evidence of atrophy.

22Some strong associations were noted between individual imaging variables and age, NIHSS score, and time to randomisation (table 2; appendix pp 20–22). Every point increase in NIHSS score was associated with a roughly 10% increase in the odds of tissue hypoattenuation, swelling, hyperattenuated artery, or large lesion. Every delay of 1 h increased the odds of tissue hypoattenuation, but not other early or pre-existing structural signs. Every increase in age by 1 year decreased the odds of tissue hypoattenuation, large lesion, swelling, or hyperattenuated artery by about 2% (table 2), but the odds of old infarct, atrophy, and leukoaraiosis were increased.

23Individually, all early ischaemic signs predicted worse outcomes (table 3) with the exception of severe hypoattenuation, although relatively few patients had this sign. Individually, of all pre-existing signs, only old infarct predicted symptomatic haemorrhage (table 2); no pre-existing signs predicted death within 7 days. Leukoaraiosis and severe atrophy predicted death by 6 months, all pre-existing structural signs predicted reduced chance of being alive and independent (OHS score of 0–2) at 6 months, and leukoaraiosis and atrophy predicted diminished chance of a favourable outcome (OHS score of 0–1) at 6 months. Similar associations between imaging signs and outcomes were seen for patients randomised within 3 h, 3–4·5 h after stroke, and 4·5–6 h after stroke (appendix pp 6–7).

24The multivariate logistic regression model for symptomatic intracranial haemorrhage included age, NIHSS score, time to randomisation, individual imaging signs, alteplase, and use of antiplatelet drugs immediately before stroke. Increasing NIHSS score, alteplase, antiplatelet treatment, and pre-existing old infarcts predicted a significantly higher risk of symptomatic intracranial haemorrhage (table 4). With standard stepwise model selection methods, retaining age, NIHSS score, time to randomisation, and treatment group, both old infarcts and hyperattenuated arteries were potentially significant predictors of symptomatic intracranial haemorrhage (table 5). The multivariate logistic regression model for good functional outcome (OHS score of 0–2) at 6 months showed that increasing age and NIHSS score were the strongest adverse predictors, but hyperattenuated arteries, large lesion, and leukoaraiosis each individually predicted less chance of a good outcome by about 25–30% (table 4). With standard stepwise regression methods, the selected model retained large lesion, hyperattenuated arteries, and leukoaraiosis, with coefficients very similar to those in the full model (table 5). Similar effects were seen in patients randomised 0–3 h, 3–4·5 h, and 4·5–6 h after stroke (appendix pp 8–9). However, potential differences—eg, that tissue hypoattenuation (rather than old infarcts or hyperattenuated arteries) predicts symptomatic intracranial haemorrhage in patients randomised 4·5–6 h after stroke—should be interpreted with caution because of the weak significance level and smaller sample.

25No interaction was recorded between any individual or combined imaging variable and alteplase, for either functional outcome (OHS score of 0–2, figure 2) or symptomatic intracranial haemorrhage (figure 3). Furthermore, no interaction was noted between any individual or combined image variable and alteplase in the ordinal shift analysis or in analyses restricted to patients randomised within 3 h of stroke or 3–4·5 h or 4·5–6 h after stroke (appendix pp 10–17).

26Despite the absence of definite interactions between imaging signs and alteplase, the absolute increase in symptomatic intracranial haemorrhage after alteplase with combined imaging signs was substantial. The combination of old infarcts and hyperattenuated arteries (adjusting for age, NIHSS score, and time to randomisation) predicted nearly three-fold increased odds of symptomatic intracranial haemorrhage (odds ratio 2·98, 95% CI 1·71–5·16) versus patients with neither sign (both signs present, absolute excess of events with alteplase 13·8%, 95% CI 6·9–20·7; both signs absent, absolute excess with alteplase 3·2%, 1·4–5·1; appendix p 18). Similar absolute effects on symptomatic intracranial haemorrhage were seen for patients randomised within 3 h of stroke (appendix pp 12–14). A difference in absolute effects was not seen for functional outcome at 6 months (OHS score of 0–2) in the whole study group (appendix p 19) and by separate time windows (data not shown).

27To date, IST-3 is the largest randomised controlled trial of thrombolysis versus control after ischaemic stroke in patients for whom alteplase was judged promising but unproven. Although IST-3 was neutral on the primary endpoint (OHS score of 0–2), this finding does not preclude the presence of clinically relevant interactions with treatment. Therefore, we planned a priori a detailed secondary analysis of the association of imaging signs with thrombolysis effects. Imaging signs are powerful prognostic markers; however, we recorded no unequivocal evidence that any individual imaging sign modified response to alteplase in patients presenting up to 6 h after ischaemic stroke. In this population, no interaction was noted between extensive early ischaemia—commonly cited as an exclusion for thrombolysis treatment—and alteplase, which accords with previous findings (panel, figure 4). Some combinations of imaging features (eg, old infarct and hyperattenuated artery) were associated with an increased absolute excess of symptomatic intracranial haemorrhage after alteplase.

28The plain CT scans obtained in IST-3 reflect imaging done in practice, with speed and accessibility being important factors in view of the steep time dependency of alteplase benefit. The scans reflected common problems, such as movement of the patient or oblique positioning. To date, the brain imaging dataset in IST-3 is the largest in an acute stroke trial (more than 7000 scans) to be read centrally. Scan reading was structured and masked, and follow-up was complete. We pretested the imaging rating system thoroughly and the assessors were skilled in acute stroke imaging. Although the expert panel might have detected subtle changes in tissue attenuation, we showed previously that those who were less experienced could still identify hyperattenuated arteries, established tissue ischaemia, and pre-existing structural changes with good reliability. These results should inform clinical practice.

29Although more patients were randomised within 3 h of stroke in IST-3 (n=846) than in any other previous trial, we might have missed an interaction between an imaging sign and alteplase in the subgroup of patients treated within 3 h, because very few symptomatic intracranial haemorrhage events arose in controls, and the estimate of effect on OHS score had wide CIs in each subgroup. Associations between clinical and imaging signs reflect the specific IST-3 population; inference to the wider population with ischaemic stroke—including individuals not judged eligible for the trial—remains speculative. Few patients with obvious tissue hypoattenuation (ie, in whom the acute ischaemic tissue was of lower attenuation than normal grey and white matter) were included in the trial; severe tissue hypoattenuation, including appearances suggesting a lesion older than 6 h, was an exclusion criterion. Additionally, patients with extensive ischaemic lesions (ASPECTS 0–7) comprised only a quarter of the study population, although this proportion is higher than in any previous trial. Hence, although IST-3 was larger than previous trials, it might not have had enough statistical power to ascertain whether alteplase treatment in this subgroup of patients adds risk or benefit. We might have identified a spurious association, but we have been cautious in the interpretation, including 99% CIs for tests of interaction. The use of stepwise methods to identify parsimonious predictive models should not be interpreted as ruling out effects of those imaging signs that were excluded from the models. Different subsets of the full set of predictors could produce almost equally good results in this dataset; therefore, further research to assess the contribution of different signs to the risk of adverse outcomes would be valuable.

30Previous analyses of imaging predictors of the response to alteplase focused on individual acute ischaemic signs. Here, we have shown that pre-existing signs are important. The association between symptomatic intracranial haemorrhage and old infarcts could account for why the risk of symptomatic intracranial haemorrhage does not change with increasing time to alteplase up to 6 h; another factor present before the stroke, or that is independent of time after stroke, might increase risk. Worsening of functional outcome with old infarcts, atrophy, and leukoaraiosis might indicate increased susceptibility to acute ischaemia or alteplase hazards and could provide useful markers of so-called brain frailty. Further research is warranted to ascertain the strength of association between pre-existing signs on brain CT and clinical markers of brain frailty. Our findings also emphasise the value to future trials of minimising easy-to-detect imaging prognostic signs to avoid baseline imbalances.

31The presence of combinations of imaging signs in patients after stroke might provide additional information to decision making when clinical uncertainty exists about the likely benefit of alteplase—eg, in a patient presenting close to the latest time window or for whom the likelihood of benefit was marginal. Additional factors affecting risk of symptomatic intracranial haemorrhage (eg, taking antiplatelet drugs) might contribute to decision making in such patients. A hint, in stepwise modelling, that early ischaemic tissue hypoattenuation might be associated with increased risk of symptomatic intracranial haemorrhage with alteplase 4·5–6 h after stroke (appendix pp 8–9) should be treated with caution, because this finding was not confirmed in the formal test of interaction (appendix pp 12–17). Helpfully for routine practice, the key prognostic CT imaging variables identified here are easy to detect: hyperattenuated artery has the best observer reliability of all early ischaemic signs across a wide range of observers, and old infarcts, atrophy, and leukoaraiosis are also easy to detect. Perceived difficulties in detecting early tissue hypoattenuation might have reduced confidence in use of CT scanning before alteplase. However, our findings confirm that neither early tissue hypoattenuation nor large infarct extent on ASPECTS score should exclude patients from alteplase, hopefully improving confidence in use of CT scanning in acute stroke.

32Correspondence to: Prof J M Wardlaw, Brain Research Imaging Centre, Centre for Clinical Brain Sciences, University of Edinburgh, Western General Hospital, Edinburgh EH4 2XU, UK joanna.wardlaw@ed.ac.uk

33Correspondence to: Prof J M Wardlaw, Brain Research Imaging Centre, Centre for Clinical Brain Sciences, University of Edinburgh, Western General Hospital, Edinburgh EH4 2XU, UK joanna.wardlaw@ed.ac.uk