Influence of Penumbral Reperfusion on Clinical Outcome Depends on Baseline Ischemic Core Volume
Background and Purpose—In alteplase-treated patients with acute ischemic stroke, we investigated the relationship between penumbral reperfusion at 24 hours and clinical outcomes, with and without adjustment for baseline ischemic core volume.
Methods—Data were collected from consecutive acute ischemic stroke patients with baseline and follow-up perfusion imaging presenting to hospital within 4.5 hours of symptom onset at 7 hospitals. Logistic regression models were used for predicting the effect of the reperfused penumbral volume on the dichotomized modified Rankin Scale (mRS) at 90 days and improvement of National Institutes of Health Stroke Scale at 24 hours, both adjusted for baseline ischemic core volume.
Results—This study included 1507 patients. Reperfused penumbral volume had moderate ability to predict 90-day mRS 0 to 1 (area under the curve, 0.77; R2, 0.28; P<0.0001). However, after adjusting for baseline ischemic core volume, the reperfused penumbral volume was a strong predictor of good functional outcome (area under the curve, 0.946; R2, 0.55; P<0.0001). For every 1% increase in penumbral reperfusion, the odds of achieving mRS 0 to 1 at day 90 increased by 7.4%. Improvement in acute 24-hour National Institutes of Health Stroke Scale was also significantly related to the degree of reperfused penumbra (R2, 0.31; P<0.0001). This association was again stronger after adjustment for baseline ischemic core volume (R2, 0.41; P<0.0001). For each 1% of penumbra that was reperfused, the 24-hour National Institutes of Health Stroke Scale decreased by 0.069 compared with baseline.
Conclusions—In patients treated with alteplase, the extent of the penumbra that is reperfused is a powerful predictor of early and late clinical outcomes, particularly when baseline ischemic core is taken into account.
Salvaging the ischemic penumbra from infarction has been the principal treatment goal for intravenous therapy in acute ischemic stroke for several decades. Predicting the clinical outcome in response to this treatment is complex. However, reperfusion of the ischemic region has been reported to be associated with increased odds of good clinical outcome.1–3 Computed tomographic (CT) perfusion (CTP) offers a rapid and well-validated assessment of acute tissue pathophysiology. Perfusion imaging has the potential to identify patients who would benefit from reperfusion treatment4 and also those who may be harmed.5 Post-treatment perfusion imaging is also clinically useful because it allows quantitative assessment of the degree of early reperfusion and penumbral salvage.6 Conventionally, studies have assessed the relationship between reperfusion and clinical outcome by dichotomizing reperfusion status2,7,8 and shown that penumbral salvage predicts better clinical outcomes.9 It is, therefore, intuitive that reperfusion of penumbra must underlie this association, there are limited studies directly assessing whether the extent of penumbral reperfusion is a major driver in altering clinical outcome for the better. Furthermore, the relationship between reperfusion, penumbral salvage, and baseline ischemic core volume (a known strong predictor of outcome) on clinical outcome has not been previously investigated.3–10 Thus, the aim of the current study in patients with hyperacute ischemic stroke treated with alteplase was to assess the relationship between reperfusion, penumbral salvage, baseline ischemic core, and clinical outcome. We hypothesized that the reperfused penumbral volume would be more strongly associated with clinical outcome when taking into account the baseline ischemic core volume.
Consecutive acute ischemic stroke patients presenting within 4.5 hours of symptom onset who were eligible for treatment with intravenous alteplase in 7 centers across Australia, China, and Canada between 2012 and 2016 were prospectively recruited into the International Stroke Perfusion Imaging Registry (INSPIRE). From the INSPIRE registry, patients were included in this study if they underwent baseline multimodal CT imaging, including noncontrast CT, CTP, CT angiography, and 24-hour follow-up magnetic resonance imaging, including diffusion-weighted imaging and perfusion-weighted imaging or multimodal CT, which included CTP. Stroke severity was assessed at baseline and at 24 hours using the National Institutes of Health Stroke Scale (NIHSS). Early clinical response was assessed by the change of NIHSS (difference between 24-hour and baseline NIHSS). Patients were treated with intravenous thrombolysis (alteplase) if they were eligible based on institutional guidelines. Functional outcome was assessed at day 90 after stroke using the modified Rankin Scale (mRS). Endovascular treated patients were not included in the INSPIRE database during the study recruitment. All patients gave written informed consent for their clinical and imaging information to be collected for the registry, and the INSPIRE study was approved by local ethics committees in accordance with the Australian National Health and Medical Research Council guidelines.
Baseline Multimodal CT Protocol
Baseline CT imaging included brain noncontrast CT, CTP, and CT angiography using either 64-, 128-, or 320-detector scanners (GE Lightspeed, Siemens Definition Flash dual source, Philips Brilliance iCT, and Toshiba Aquilion One). Axial slice coverage ranged from 41 to 160 mm. CT angiography was performed after perfusion CT with acquisition from the aortic arch to vertex11 (scanner details are summarized in Table I in the online-only Data Supplement).
24-Hour Follow-Up Imaging
All included patients underwent a stroke magnetic resonance imaging, or multimodal CT, as close as possible to 24 hours after stroke onset. The stroke magnetic resonance imaging protocol was performed on a 1.5-T or 3-T scanner and included an axial isotropic diffusion-weighted imaging, time-of-flight MR angiography, and bolus-tracking perfusion-weighted imaging and fluid-attenuated inversion recovery imaging.
All perfusion images were postprocessed with MIStar (Apollo Medical Imaging Technology, Melbourne, Australia) with single value deconvolution with delay and dispersion correction.12 The perfusion lesion was defined by threshold-relative delay time >3 seconds, and ischemic core was defined by relative cerebral blood flow <30%.13 Penumbral volume was calculated from total perfusion lesion minus the volume of ischemic core. The reperfused penumbral volume was determined as the difference between the follow-up infarct core volume and the baseline penumbra lesion volume. Degree of reperfused penumbra defined as the reperfusion percentage.
Statistical analyses were performed with SAS v9.4 (SAS Institute, Cary, NC), Stata v13.0 (StataCorp, Ltd, College Station, TX), and R (R Foundation for Statistical Computing, Vienna, Austria). Descriptive results and quantitative baseline patients’ characteristics were presented as mean and SD or median and interquartile range. The patient outcome variables for predictive modelling were (1) 90-day mRS between 0 to 1 (good) versus 5 to 6 (poor) and (2) early clinical response: assessed by the change in NIHSS between baseline and 24 hours. For each clinical response outcome (mRS and change in NIHSS), we assessed the predictive ability of the reperfused penumbra volume and the degree of reperfused penumbra (reperfusin percentage), both with and without baseline core volume.14 Logistic regression models were used to model the probability of good functional outcome, and linear regression was used to model early clinical response. For the logistic regression, linearity was assessed by inspecting plots of the empirical logit for the outcome variable versus quintiles of the predictor variable, where there was evidence of nonlinearity; we used the quintile (labeled Q0, Q1, Q2, Q3, and Q4) of the variable as a predictor. Odds ratios (ORs) with 95% confidence intervals (CIs) and P values are presented, model discrimination was assessed by the C statistic, the area under the curve (AUC), and the Nagelkerke pseudo R-squared statistic; calibration was assessed by the Hosmer–Lemeshow test. For the linear regression models, the R-squared values are presented together with regression coefficients, 95% CIs, and P values.
During the study period, 1507 patients were recruited in the INSPIRE registry, for this study: 247 were excluded because they were outside 4.5 hours from symptom onset, 687 were excluded because they did not have follow-up perfusion imaging, 82 because they contained incomplete clinical data, and 57 because the imaging was unable to be processed either because of excessive motion, incomplete scan, or other technical errors. Thus, a total of 434 patients treated with IV alteplase within 4.5 hours of stroke onset were included in this study. Of the included study patients, the median patient age was 74 (65–80) years, and the median baseline NIHSS was 14 (11–17). The median 24-hour NIHSS was 8 (4–15). The mean baseline ischemic core volume was 27.14±32.27 mL, mean penumbra volume was 64.61±50.07 mL, mean reperfused penumbral volume was 40.43±46.08 mL, and mean degree of reperfused penumbra was 66.43±39.01%.
Reperfused Penumbral Volume Predicting Clinical Outcomes
Quintiles of the reperfused penumbral volume were used as the predictor variable for the logistic regression model because of evidence of a nonlinear effect for the logistic regression model (Figure 1). The quintiles of the reperfused penumbral volume were the reperfused penumbral volume divided into 5 categories: Q0 (0–5.6 mL), Q1 (5.6–14.7 mL), Q2 (14.7–34 mL), Q3 (34.1–73 mL), and Q4 (73.3–351 mL) shown in Table 1.
The model containing the variables reperfused penumbral volume quintile categories (Q0, Q1, Q2, Q3, and Q4) had a modest AUC at predicting 90-day mRS 0 to 1 (AUC, 0.77; 95% confidence interval [CI], 0.72–0.84; R2, 0.28; P<0.0001; Hosmer–Lemeshow test P=1.000; Table 2). Patients located on the first quintile (Q0) with no reperfused penumbral volume (<5.6 mL) had a low probability of achieving mRS 0 to 1 (6%), and patients located in subsequent quintiles (Q1–Q4) with greater reperfused penumbral volumes had a sharp increase in the probability of achieving mRS 0 to 1 (>38%). Relative to the first quintile of patients with no reperfused penumbra (Q0), the ORs of patients achieving mRS 0 to 1 were high for Q1 (OR, 57.50; 95% CI, 17.55–188.43; P=0.0001), Q2 (OR, 20.14; 95% CI, 6.99–61.20; P=0.386), Q3 (OR, 27.60; 95% CI, 9.04–84.27; P=0.058), and Q4 (OR, 30.05; 95% CI, 10.113–89.313; P=0.021; Table 2). There was a significant increase in the rate of hemorrhagic transfomation between quintiles, with the higher quintiles experience a greater rate of PH2 as more penumbra was reperfused (PH2 rate Q0, 0%; Q1, 3.41%; Q2, 4.65%; Q3, 5.57%; Q4, 6.74%; P=0.007; Table 1).
Following adjustment for baseline ischemic core volume, the AUC of the model predicting mRS 0 to 1 increased substantially (AUC, 0.95; 95% CI, 0.92–0.98; R2, 0.55; P<0.0001; Hosmer–Lemeshow test P=0.190; Figure 1). The model with adjustment for baseline ischemic core volume was significantly more predictive of good patient outcome (AUC 0.77 versus AUC 0.95; P<0.001). Relative to the first quintile (Q0), the odds of patients achieving mRS 0 to 1 were high in the model adjusting for baseline ischemic core volume for Q1 (OR, 25.64; 95% CI, 5.83–112.59; P=0.165), Q2 (OR, 13.22; 95% CI, 3.27–53.18; P=0.929), Q3 (OR, 21.86; 95% CI, 4.87–98.04; P=0.305), and Q4 (OR, 65.15; 95% CI, 12.623–336.06; P=0.002; Table 2). In the model where baseline ischemic core was not adjusted for, patients with >20 mL of reperfused penumbra had a 60% probability of achieving mRS 0 to 1 (Figure 1), whereas in the model adjusting for baseline ischemic core, patients with >20 mL of reperfused penumbra had a 80% probability of achieving mRS 0 to 1 (Figure 1).
The degree of reperfused penumbra was a significant predictor of a good functional outcome (mRS 0–1), for every 1% of the degree of reperfused penumbra increased, the odds of achieving mRS 0 to 1 at day 90 increased 7.4% (Figure 2; Table 2).
With respect to early clinical response, change in NIHSS was significantly related to the degree of reperfused penumbra (R2, 0.39; P<0.0001; Figure 3; Table 3). The degree of reperfused penumbra remained strongly associated with NIHSS decrease after adjustment for the baseline ischemic core volume (R2, 0.41; P<0.0001). For each 1% of penumbra that was reperfused, the 24-hour NIHSS decreased by 0.069 U compared with baseline NIHSS (Table 3). For each additional 1 mL of reperfused penumbra, the 24-hour NIHSS was expected to drop 0.05 U compared with baseline NIHSS (Table 3).
We have demonstrated that the degree of penumbra salvaged had a significant influence on early and late patient outcome, particularly after adjustment for baseline ischemic core volume. This indicates that penumbra salvaged is an important metric to predict individual patient outcome which can be used to measure benefit from therapy. This analysis has demonstrated that for every 1% of penumbra salvaged, the odds of a patient having a good 3-month clinical outcome increased by 7.4%. Additionally, the baseline core volume had a major influence on the relationship between penumbral reperfusion and outcome in patients with a salvaged penumbra volume of >20 mL having an 80% chance of an excellent recovery (mRS 0–1). This indicates that penumbra salvaged is an important metric that influences individual patient outcome and can be used to assess potential benefit from therapy. However, without adjustment for baseline ischemic core, the statistical model showed that with 20 mL of penumbral salvage, patients only had a 60% chance of an excellent recovery (mRS 0–1). Therefore, adjusting for baseline core is a vital step in ensuring that our models are accurate and that assessing reperfusion or penumbra alone has limited prognostic power until the infarct core volume is accounted for. Lastly, there was a significant stepwise increase in the rate of PH2 with increasing reperfused penumbra volume, indicating that while the rate of good clinical outcome increased with reperfused penumbra volume, so too did the rate of hemorrhage, although the overall rate of PH2 was still low for this study.
Baseline ischemic core volume has been reported to be an independent predictor of clinical outcome in previous studies, with a larger core volume being associated with poorer clinical outcomes.14 In the current study, all analyses were performed with and without adjustment for baseline ischemic core volume and demonstrated a significant improvement in the capacity of penumbral reperfusion to predict outcome. Patients with minimal penumbral salvage (<5.6 mL) had a low probability achieving mRS 0 to 1 at day 90, likely because their initial perfusion lesion was either completely infarcted or reperfused before acute imaging, and there was minimal benefit to be gained from intravenous therapy. Relative to the patients with no penumbral salvage, the patients with larger reperfused penumbral volumes had a significantly better chance of achieving mRS 0 to 1 at day 90. This is clearly demonstrated in Figure 1 where there is a sharp increase in good functional outcome with increasing penumbral reperfusion. A similar effect was seen on early clinical response. For example, 20 mL of penumbra that was reperfused translated into a 1-point 24-hour NIHSS decrease compared with baseline NIHSS. This indicates that patients with a substantial volume of penumbra in the acute phase have the most to gain from effective reperfusion therapy but also, that they have the most to lose if the reperfusion is ineffective. The most important clinical point of this article is that using acute CTP, we can accurately predict patient response to interventions based on whether or not reperfusion occurs, in advance of acute therapy. Of course, we used perfusion imaging at 24 hours to quantify reperfusion as part of the model, but a major point of clinical relevance is that using this model, a clinician can accurately predict potential response to reperfusion before therapy. Thus, our model does not actually require the clinician to wait until 24 hours to assess reperfusion to predict outcome. Essentially, the clinician can predict with high accuracy, probability of good outcome with full penumbral salvage (ie, if complete reperfusion occurs) versus probability of good outcome if no reperfusion occurs. For example, if there is >20 mL of penumbra to salvage, then the probability of good outcome with complete reperfusion is >80%. If there is no penumbral reperfusion, then the probability of good outcome is 6%. This is extremely powerful information for a clinician to help make a treatment decision and inform the patients and their family.
Previous studies assessing the relationship between reperfusion and clinical outcomes, which have dichotomized the reperfusion status into reperfusion versus no reperfusion, have demonstrated that patients with reperfusion have better clinical outcomes.1,15 Importantly, by adjusting for baseline core volume, we took into account individual patient characteristics and hence saw a high AUC of 0.95 for reperfused penumbral volume in predicting good functional outcome. Based on our predictive model, a patient with a baseline ischemic core of 30 mL and a penumbra of 70 mL who is treated with alteplase would expect to have an mRS of 1 or 2 with reperfusion and an mRS of 4 or 5 without reperfusion. However, a patient with no ischemic core on baseline imaging but with a penumbra of 30 mL could expect to be mRS 0 to 1 with reperfusion, and mRS 3 without it.14 This information should be generalizable because we included the entire spectrum of acute ischemic stroke patients in the INSPIRE database presenting within 4.5 hours of acute ischemic stroke who were treated with alteplase, which includes a wide range of acute ischemic core and perfusion lesion volumes.
A limitation of this analysis is that we only assessed patients treated with intravenous therapy because endovascular treatment was not readily available at the study sites during patient recruitment. Validation of our findings in endovascular patients would be valuable because endovascular therapy has been demonstrated to result in earlier and more complete reperfusion.16,17
In conclusion, we have developed a highly accurate model of patient outcome by investigating the relationship between the baseline ischemic core, reperfused penumbra volume, and clinical outcome in patients receiving intravenous alteplase. By correcting for baseline ischemic core volume, we were able to account for a major cause of individual patient variability in clinical response to reperfusion therapy. This model shows that for each percentage increase in penumbral reperfusion, the odds of achieving an excellent clinical outcome were increased by 7.4%. Such a powerful observation was possible because of the large multicenter database, including a wide range of patients treated with alteplase, and should make these findings generalizable.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.018587/-/DC1.
- Received June 29, 2017.
- Revision received July 30, 2017.
- Accepted August 3, 2017.
- © 2017 American Heart Association, Inc.
- Soares BP,
- Tong E,
- Hom J,
- Cheng SC,
- Bredno J,
- Boussel L,
- et al
- De Silva DA,
- Fink JN,
- Christensen S,
- Ebinger M,
- Bladin C,
- Levi CR,
- et al
- Yassi N,
- Parsons MW,
- Christensen S,
- Sharma G,
- Bivard A,
- Donnan GA,
- et al
- Lin L,
- Cheng X,
- Bivard A,
- Levi CR,
- Dong Q,
- Parsons MW
- Davis SM,
- Donnan GA,
- Parsons MW,
- Levi C,
- Butcher KS,
- Peeters A,
- et al
- Inoue M,
- Mlynash M,
- Straka M,
- Kemp S,
- Jovin TG,
- Tipirneni A,
- et al
- Bivard A,
- Levi C,
- Lin L,
- Cheng X,
- Aviv R,
- Spratt NJ,
- et al