Mechanisms of Unexplained Neurological Deterioration After Intravenous Thrombolysis
Background and Purpose—Unstable clinical course characterizes the first 24 hours after thrombolysis for anterior circulation stroke, including early neurological deterioration (END), a secondary complication consistently predictive of poor outcome. Apart from straightforward causes, such as intracerebral hemorrhage and malignant edema, the mechanism of END remains unclear in the majority of cases (ENDunexplained). Based on the core/penumbra model, we tested the hypothesis that ENDunexplained is caused by infarct growth beyond the initial penumbra and assessed the associated vascular patterns.
Methods—From our database of consecutive thrombolyzed patients (n=309), we identified 10 ENDunexplained cases who had undergone both admission and 24-hour MRI. Diffusion-weighted imaging lesion growth both within and beyond the acute penumbra (Tmax>6 seconds) was mapped voxelwise. These 10 cases were compared with 30 no-END controls extracted from the database blinded to 24-hour diffusion-weighted imaging to individually match cases (3/case) according to 4 previously identified clinical and imaging variables.
Results—As predicted, lesion growth beyond initial penumbra was present in 9 of 10 ENDunexplained patients (substantial in 8) and its volume was significantly larger in cases than controls (2P=0.047). All ENDunexplained cases had proximal arterial occlusion initially, of which only 2 had recanalized at 24 hours.
Conclusions—In this exploratory study, most instances of ENDunexplained were related to diffusion-weighted imaging growth beyond acute penumbra. Consistent presence of proximal occlusion at admission and lack of recanalization at 24 hours in most cases suggest that hemodynamic factors played a key role, via for instance systemic instability/collateral failure or secondary thromboembolic processes. Preventing END after tissue-type plasminogen activator using, eg, early antithrombotics may therefore be feasible.
The first 24 hours after intravenous recombinant tissue-type plasminogen activator (tPA) are characterized by variable clinical course, most patients improving, others remaining stable, and ≈10% experiencing early neurological deterioration (END).1 Accordingly, at 24 hours, the National Institutes of Health Stroke Scale (NIHSS) becomes highly predictive of final outcome.1 END refers to a significant worsening in neurological score, with the most widely used definition being ≥4 NIHSS points.2 It is an ominous clinical event that consistently predicts poor outcome.3 Although END may have straightforward causes, such as symptomatic intracerebral hemorrhage, malignant edema, early recurrent stroke, and poststroke seizure,2 in the majority of cases no clear mechanism is found.3 For instance, this was true in over two-third of END cases occurring within 24 hours of intravenous recombinant tPA for middle cerebral artery (MCA) stroke, a scenario operationally termed ENDunexplained.4
As no prevention or management guidelines are currently available for unexplained END, understanding its underlying pathophysiology at the tissue and vascular level is an important goal. According to the classic model, both core and penumbra contribute to the clinical deficit and constitute the symptomatic tissue, whereas the surrounding, mildly hypoperfused oligemia is asymptomatic and in principle not at risk of infarction.5 Accordingly, unexplained END would be caused by infarct growth beyond the initial penumbra as a result of secondary hemodynamic or metabolic disruption in the oligemia.6
This straightforward hypothesis was for the first time directly tested in the present study, with reference to the core/penumbra model (Figure 1). Specifically, we compared the volume of extrapenumbral diffusion-weighted imaging (DWI) lesion extension in ENDunexplained patients and in a sample of matched control patients without END. We also assessed the associated vascular patterns, particularly the presence of proximal occlusion on admission imaging and the occurrence of recanalization at follow-up.
From our prospective database of 309 consecutive patients who received only intravenous recombinant tPA within 4.5 hours of onset of MCA stroke between January 2003 and March 2013, patients with ENDunexplained (see below) and MRI obtained both at admission (including both DWI and PWI) and at follow-up (including DWI) were identified. Control patients (3 per case) without END (END−) were then manually extracted from the same population, blinded to follow-up DWI, so as to match as closely as possible individual END cases according to 4 clinical and radiological variables found to be strongly associated with ENDunexplained in our previous study,4 namely initial DWI volume, admission NIHSS, 24-hour recanalization, and site of occlusion, in this specific order.
In accordance with French legislation, Institutional or Ethics Committee approval was not required for this study because it only implied retrospective analysis of anonymized data collected as part of routine clinical care. Likewise, written consent was not necessary for intravenous tPA, which is part of routine care.
The sequences used for the admission and follow-up MRI are detailed in Methods in online-only Data Supplement. Note that the follow-up MRI, scheduled at ≈24 hours after treatment, did not include PWI. In case of END, additional brain imaging (CT or MRI) was obtained as soon as possible, unless it occurred close to the planned follow-up MR, in which case the latter was brought forward.
END was defined as a ≥4 point increase in NIHSS score (ΔNIHSS=NIHSS24 h−NIHSS0 h) between baseline and 24 hours.2 A neuroradiologist (M.T.) and a neurologist (P.S.) reviewed each END case and adjudicated in consensus for potential causes, taking into account the sudden or progressive onset of the deterioration, the neurological function that worsened, and the findings on imaging. Symptomatic intracerebral hemorrhage was defined as a ΔNIHSS≥4 points presumably caused by a parenchymal hematoma type 2 on follow-up imaging.7,8 Early malignant edema was considered the cause if concomitant imaging showed brain swelling and midline shift associated with worsening of consciousness. Early recurrent ischemic stroke was defined as the occurrence of new neurological symptoms suggesting the involvement of initially unaffected vascular territories and evidence of corresponding ischemic lesions on (follow-up) cranial CT or MRI.9,10 Accordingly, an END associated with extension of the ischemic lesion in the same territory as that already affected on admission MRI was not considered an early recurrent ischemic stroke. Other potential causes of END, eg, early poststroke seizure, were also considered. Thus, ENDunexplained was determined as END without evidence for any of the above mechanisms.
Image processing for DWI and PWI is detailed in the online-only Data Supplement Methods and has been published previously.11 DWI lesions on the admission and follow-up MRI (DWI1 and DWI2, respectively) were segmented blinded to the END category. Any true DWI lesion growth, ie, DWI2 lesion beyond the boundaries of the DWI1 lesion as mapped voxelwise, will be referred to as progressing acute DWI lesion (PAD) below. The severely hypoperfused region was defined as Tmax>6 seconds.12,13 Mismatch was defined as tissue with Tmax>6 seconds outside DWI1. Symptomatic tissue was defined as (DWI1 lesion+mismatch).14 The volumes of PAD within and beyond the boundaries of the acute mismatch (penumbral PAD and extrapenumbral PAD, to be referred to as EP-PAD below) were determined using a voxel-counting algorithm (Matlab). Pretreatment occlusions were categorized on 3-dimensional time-of-flight magnetic resonance angiography (MRA) into proximal (ie, internal carotid artery or proximal [M1] MCA) or not proximal (distal or no visible occlusion). Recanalization was defined on follow-up MRA as Thrombolysis In Cerebral Infarction score ≥2.15
Continuous variables are described as median and interquartile range. Data from cases and controls (clinical and imaging variables including the volume of EP-PAD) were compared using nonparametric Mann-Whitney U tests for quantitative variables and Fisher exact test for qualitative variables, using SPSS 21.0 and SAS 9.3. The quantitative relationships between PAD volumes and clinical recovery were assessed using the Alawneh model, which posits that the larger the volume of eventually noninfarcted penumbra, the better the recovery, and conversely, the larger the volume of eventually infarcted asymptomatic tissue, the worse the recovery, with both effects possibly acting simultaneously and disentangled by multivariate analysis.14 This analysis was performed across the ENDunexplained and END− groups together, using nonparametric tests. Kendall correlations were assessed between (1) relative volume of noninfarcted symptomatic tissue (ie, noninfarcted symptomatic tissue volume divided by symptomatic tissue volume) and %ΔNIHSS (ie, ΔNIHSS 24 hours – 0 hour divided by initial NIHSS); and (2) relative EP-PAD volume and %ΔNIHSS. The latter relationship was finally reassessed using Kendall partial τ correlation controlling for the relative volume of noninfarcted symptomatic tissue. We predicted all 3 correlations to be significant and in the predicted direction.
END unexplained Patients Versus Controls
Twenty-three patients with unexplained END were identified. Of these, 13 were not eligible for the present study because of no initial MRI (n=4), no initial PWI (n=7), initial imaging not available in DICOM format (n=1), and intra-arterial thrombolysis performed immediately after END (n=1). These excluded patients did not significantly differ from the 10 eligible cases for age, initial NIHSS, onset-to-treatment time, or proximal occlusion.
The demographics and relevant clinical and imaging data of the 10 eligible patients are shown in Table 1. No patient but one (#4) was on antiplatelet agents before their stroke. The follow-up MRI took place 29.4 hours [20.6–54.4] (median [interquartile range]) after the start of tPA in the END cases, which was not different from the controls (25.4 hours [22.5–30.8]; P=0.288). The END occurred 0.5 to 23 hours after the start of tPA. All 10 patients had proximal occlusion on admission MRA. At follow-up, 2 patients only (1 and 7) had recanalized, while in the remaining the initial occlusion was still present. Finally, functional outcome was poor (3-month mRS>2) in 8 of 10 patients.
EP-PAD was present in 9 of the 10 eligible patients, with volumes ranging from 7 to 137 mL (>10 mL in 8; Table 1). Figure 2 illustrates the acute and follow-up DWI and MRA images and binarized acute Tmax map and Figure 3, the topography of EP-PAD relative to the initial Tmax>6 seconds lesion, in the same 3 patients.
There was no significant difference for any of the demographics or admission clinical or imaging data between cases and controls, including age, initial NIHSS, onset-to-treatment time, presence of proximal occlusion, volume of DWI1, or volume of symptomatic tissue (Table 2). Also, there was no significant difference in occurrence of recanalization, but as expected both the 24-hour NIHSS and DWI2 volume were significantly greater in the END group. Importantly, all 3 volumes for total PAD, penumbral PAD, and EP-PAD were significantly larger in ENDunexplained compared with END− patients (P=0.01, 0.01, and 0.047, respectively) (Table 2). The plot of the individual EP-PAD volumes in the cases and controls is shown in the Figure in online-only Data Supplement.
Topography of Extrapenumbral PAD
In all 9 patients with EP-PAD, the topography of the latter was contiguous with the penumbra area, as illustrated in Figures 2 and 3, although in patient 1 there was an additional, small area of EP-PAD located remote from the initial penumbra, involving the anterior cerebral artery (ACA) territory.
Relationships Between EP-PAD Volume and Clinical Course
Across the entire sample (n=40, ie, cases and controls together), there was a strong correlation between relative volume of noninfarcted symptomatic tissue and %ΔNIHSS (τ=−0.43; P<0.001) in the expected direction, ie, the larger this volume, the better the recovery (Figure 4A). The predicted inverse correlation between %EP-PAD and %Δ NIHSS was also present (τ=0.32; P=0.004) (Figure 4B), though did not survive after adjustment for the volume of noninfarcted symptomatic tissue; of note, the correlation between EP-PAD and ΔNIHSS was also present using the absolute data (τ=0.25; P=0.026). Finally, the larger the %EP-PAD, the smaller the relative volume of noninfarcted symptomatic tissue (τ=−0.65; P<0.001) (Figure 4C).
The main findings from this study are as follows: (1) consistent with our hypothesis, EP-PAD was present in the vast majority of ENDunexplained cases (9/10 patients; >10 mL in 8); and (2) its volume was significantly larger in ENDunexplained cases than in END− controls. In addition, the volume of EP-PAD negatively influenced clinical course and negatively correlated with the volume of surviving symptomatic tissue.
Although the difference in volume of EP-PAD between cases and controls was marginally significant according to the usual 2-tailed cutoff, this comparison tested the strong a priori hypothesis that EP-PAD would be larger, not smaller, in ENDunexplained patients, which could be tested using 1-tailed P.16 As detailed in the Introduction section, this hypothesis was underpinned by robust mechanistic reasoning and previous evidence.6,14 Furthermore, the absolute volume of EP-PAD was markedly larger in cases compared with controls (median: 16 mL versus 5 mL, respectively).
Also supporting our findings, the topography of EP-PAD roughly matched the neurological items that deteriorated, as illustrated for 3 patients in Figure 3. Across the sample, EP-PAD involving the motor system or primary somatosensory area was associated with motor or sensory deterioration in 8 of 9 and 5 of 5 patients, respectively, and EP-PAD involving the Broca area or the left insula with dysphasia in 2 of 2 patients (data not shown). More detailed analysis was precluded by current lack of voxel-based atlases of the relationship of lesion topography to NIHSS items.
One blatant exception to the above was patient 3, who had no EP-PAD despite an 8-point END, which occurred 8 hours after tPA. The follow-up MRI, performed 8 hours later, showed persistent proximal M1 occlusion, without any new DWI lesion. Perhaps, in some situations, the relationship between symptomatic tissue and clinical deficit is disrupted because of unclear factors, such as heterogeneity within the DWI lesion,17 with clinical worsening occurring without detectable macroscopic DWI growth. Conversely, as shown in the Figure in online-only Data Supplement, a few END− controls had relatively large EP-PAD. For instance, the largest 2 EP-PAD controls had right hemisphere stroke and no deterioration at all. One possibility in such cases is that the EP-PAD occurred within noneloquent tissue and that there was concomitant improvement in other NIHSS items because of simultaneous salvage of symptomatic tissue.14
Here, we document the presence of significant DWI growth beyond the initial penumbra in patients with unexplained END. Sizeable infarction of acutely silent tissue has been previously reported,14 but its timing could not be assessed as follow-up structural imaging was performed at 1 month, while its relationship to occurrence of END was not addressed. More recently, new DWI lesions outside the initial perfusion deficit (not further defined) were reported in 23% patients18; however, the clinical course after admission was not reported. Finally, Bang et al19 reported the association of both new lesions and infarct growth on 7-day MRI follow-up with larger volumes of, and apparently often located within, initially mild perfusion deficits, but this relationship was not analyzed voxelwise and END was not mentioned. Despite these limitations, the imaging findings of these 3 studies seem consistent with ours.
That ENDunexplained after thrombolysis is largely due to DWI lesion growth beyond the initial penumbra has clinical implications. Thus, preventing or treating this complication positively impacts outcome, which is almost consistently poor after END3,4 as also found here. To this end, however, the precise pathological process(es) underlying extrapenumbral DWI growth would need to be identified. One key finding from the present study is that all ENDunexplained cases had proximal occlusion on admission time-of-flight MRA, consistent with the idea that ENDunexplained tends to occur in the context of hemodynamic compromise.4,6 Extrapenumbral DWI growth within the same vascular territory may be caused by secondary events worsening perfusion or neuronal status in oligemia, such as extension of the original thrombus, new embolic events, blood pressure drops, or hyper/hypoglycemia.6 Significant blood pressure drops, long-lasting hyperglycemia, or hypoglycemic episodes were not recorded in any case here (data not shown). Of note, 2 patients (Nos 4 and 10) had atrial fibrillation, a risk factor for drops of cardiac output and hence intracranial hemodynamic failure. Recurrent embolism in the affected MCA territory is unlikely, given the presence of proximal occlusion although cannot be formally excluded. Also against recurrent embolism and in favor of hemodynamic failure and potential collateral flow instability, the EP-PAD was consistently in continuity with the penumbral area, as illustrated in Figure 2. Recurrent embolism in a different vascular territory was in principle excluded, given the criteria for END used. However, the notion of vascular territory in the definition of early recurrent ischemic stroke9,10 is potentially confusing, for instance, both the MCA and the ACA belong to the carotid territory. Consistent with the clinical deterioration pattern (Table 1), the EP-PAD involved not only the originally affected MCA territory but also the ipsilateral ACA territory in patients 1 and 2, slightly and substantially so, respectively (Figure 3). Recurrent embolism in the ACA could have caused the DWI lesion to grow inside the MCA territory because of the shift of the ACA-MCA borderzone after the MCA occlusion. Finally, prospective longitudinal studies including thrombus imaging are required to assess potential extension of the original thrombus.
Proximal occlusion was still present in 8 of 10 patients at follow-up, a further putative risk factor for infarct extension within and also beyond the penumbra.6 In support of this scenario, DWI growth within the initial penumbra was significantly larger in ENDunexplained cases (Table 2), and furthermore, the volume of EP-PAD was inversely proportional to the amount of salvaged symptomatic tissue (Figure 4C), suggesting that the same process that caused the penumbra to proceed to infarction also facilitated the occurrence of EP-PAD, and in turn of END. Finally, in both patients (Nos. 1 and 7) with complete or subtotal recanalization, the follow-up MRA was performed several hours after the END (data not shown), raising the possibility of futile recanalization.20
The clinical significance of extrapenumbral infarct growth was further strengthened by its clinical correlation in the expected direction, ie, the larger the EP-PAD, the worse the recovery (Figure 4B). Previously, Alawneh et al14 also reported that infarction of asymptomatic tissue negatively influenced clinical recovery at 1 month in 2 samples that included only 1 deteriorating case each. Although in this previous report this relationship remained significant after adjustment for the positive influence of salvaged penumbra, this was not true in the present study. Note, however, that the present analysis regarded 24-hour data and was run on a selected sample that included END.
Our study has some limitations. First, only 10 cases were available; yet, this sample was extracted from a large prospective database, including all tPA-treated MCA stroke patients admitted in our center over 10 years, and both full admission and follow-up clinical and imaging data sets were required for eligibility. Nevertheless, confirmation in a larger sample is needed. Second, our findings may not be generalizable to other vascular territories, nor to nonthrombolysed patients or those seen beyond 4.5 hours. Third, the case-control matching process used has intrinsic limitations. Blinded to the occurrence of PAD, 3 controls per case were manually extracted from our database so as to match the cases as far as possible using 4 clinical and radiological criteria previously found to be strongly associated with ENDunexplained.4 As a result, the matching was not perfect, with, eg, higher age, longer OTT, greater proximal occlusions, and more no recanalization in the cases, although neither these nor any other major clinical or radiological features significantly differed between the 2 groups. Although mismatch volume was not selected as criterion, it did not significantly differ between cases and controls, which was expected given the matching for admission NIHSS, DWI lesion volume, and presence of proximal occlusion. Fourth, we used a fixed Tmax>6 seconds as penumbra threshold, which may not be perfectly accurate in all patients and across tissue characteristics such as gray versus white matter and does not consider the possibility of heterogeneity within the penumbra.17 However, this threshold has been validated against PET12,13 and is widely used in randomized trials, eg, DEFUSE221 and ECASS4. It would, however, be of interest to assess whether EP-PAD occurs within oligemic areas (eg, Tmax 4–6 seconds) or even near-normally perfused areas (eg, Tmax<4 seconds). Fifth, we speculate but did not prove that ENDunexplained is subtended by extension of the penumbra secondary to hemodynamic deterioration of the oligemia. To address this point would require to repeat perfusion imaging soon after the occurrence of otherwise unexplained END.
In conclusion, in agreement with our working hypothesis, this study documented significant extrapenumbral infarct growth in patients with unexplained END after intravenous thrombolysis. Furthermore, the universal presence of proximal arterial occlusion on admission imaging and the absence of early recanalization in most patients would be consistent with some hemodynamic compromise occurring at the time of END. Systemic factors, such as blood pressure drops, did not seem to underlie such hemodynamic impairment. Although secondary thromboembolic processes were not directly assessed here, if present they would raise the issue of how early and best to prevent them after intravenous thrombolysis over and beyond the currently recommended low-dose aspirin therapy started 24 hours after treatment.22 To address these issues, future studies should implement comprehensive longitudinal vascular, thrombus, and perfusion imaging, as well as intensive/continuous monitoring of heart rate, blood pressure, oxygen saturation and blood glucose, to detect potentially untoward events and collaterals mapping as a potential risk factor for END.
Sources of Funding
Dr Tisserand and P. Seners are supported by the Fondation pour la Recherche Médicale.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.006745/-/DC1.
- Received July 12, 2014.
- Revision received September 23, 2014.
- Accepted September 25, 2014.
- © 2014 American Heart Association, Inc.
- Saver JL,
- Altman H
- Seners P,
- Turc G,
- Oppenheim C,
- Baron JC
- Seners P,
- Turc G,
- Tisserand M,
- Legrand L,
- Labeyrie MA,
- Calvet D,
- et al
- Alawneh JA,
- Moustafa RR,
- Baron JC
- Awadh M,
- MacDougall N,
- Santosh C,
- Teasdale E,
- Baird T,
- Muir KW
- Georgiadis D,
- Engelter S,
- Tettenborn B,
- Hungerbühler H,
- Luethy R,
- Müller F,
- et al
- Labeyrie MA,
- Turc G,
- Hess A,
- Hervo P,
- Mas JL,
- Meder JF,
- et al
- Zaro-Weber O,
- Moeller-Hartmann W,
- Heiss WD,
- Sobesky J
- Takasawa M,
- Jones PS,
- Guadagno JV,
- Christensen S,
- Fryer TD,
- Harding S,
- et al
- Alawneh JA,
- Jones PS,
- Mikkelsen IK,
- Cho TH,
- Siemonsen S,
- Mouridsen K,
- et al
- Tomsick T,
- Broderick J,
- Carrozella J,
- Khatri P,
- Hill M,
- Palesch Y,
- et al
- Usnich T,
- Albach FN,
- Brunecker P,
- Fiebach JB,
- Nolte CH
- Molina CA