Advertisement

Neurocritical Care

, Volume 29, Issue 3, pp 396–403 | Cite as

Combination of Clinical Exam, MRI and EEG to Predict Outcome Following Cardiac Arrest and Targeted Temperature Management

  • Matthew B. Bevers
  • Benjamin M. Scirica
  • Kathleen Ryan Avery
  • Galen V. Henderson
  • Alexander P. Lin
  • Jong W. Lee
Original Article

Abstract

Background

Despite the widespread adoption of targeted temperature management (TTM), coma after cardiac arrest remains a common problem with a high proportion of patients suffering substantial disability. Prognostication after cardiac arrest, particularly the identification of patients with likely good outcome, remains difficult.

Methods

We performed a retrospective study of 78 patients who underwent TTM after cardiac arrest and were evaluated with both electroencephalography (EEG) and magnetic resonance imaging (MRI). We hypothesized that combining malignant versus non-malignant EEG classification with clinical exam and quantitative analysis of apparent diffusion coefficient (ADC) and fluid-attenuated inversion recovery imaging would improve prognostic ability.

Results

Consistent with prior literature, presence of a malignant EEG pattern was 100% specific for poor outcome. We found that decreased whole brain ADC signal intensity was associated with poor outcome (853 ± 14 vs. 950 ± 17.5 mm2/s, p < 0.0001). Less than 15% total brain volume with ADC signal intensity < 650 mm2/s was predictive of good outcome with 100% sensitivity, 51% specificity and an area under the curve of 0.787. A model combining this ADC marker with non-malignant EEG and flexor-or-better motor response was 100% sensitive and 91.1% specific for good outcome following cardiac arrest and targeted temperature management.

Conclusion

We conclude that in the absence of malignant EEG findings, combination of physical exam and MRI findings can be a useful to identify those patients who have potential for recovery. Variability in timing of imaging and findings in different modalities argue for the need for future prospective studies of multimodal outcome prediction after cardiac arrest.

Keywords

Cardiac arrest Apparent diffusion coefficient (ADC) imaging Fluid attenuation inversion recovery (FLAIR) imaging Continuous EEG 

Introduction

Coma after cardiac arrest is common, with an incidence of 76 per 100,000 person years [1, 2]. Of the 40,000 patients who present each year with cardiac arrest, up to 80% remain in coma [3, 4]. While targeted temperature management (TTM) has revolutionized the management of post-cardiac arrest neurologic injury, a high proportion of patients still suffer substantial disability [5, 6, 7]. Prognostication after cardiac arrest is a difficult problem, particularly in patients who undergo TTM, as the previously established clinical examination findings may no longer be valid [8]. As such, there is greater dependence on neurophysiological and imaging modalities. Electroencephalography (EEG), in particular the presence of a “malignant” pattern is able to identify patients with poor prognosis with high specificity, but many patients with initially favorable EEG still go on to have a poor outcome [9, 10, 11, 12, 13, 14, 15, 16]. Magnetic resonance imaging (MRI) has the potential to aid in prognostication, but existing studies using imaging markers alone do not outperform EEG or clinical exam in determining prognosis [17, 18, 19, 20]. As a result, current guidelines only mention that imaging should be used as an adjunct to other studies, but do not offer specific recommendations on timing or type of imaging [8, 9]. It is likely that improving our ability to prognosticate will require a multimodal approach combining exam, EEG, imaging and other markers [9, 15, 21]. In the current study, we perform a retrospective analysis of patients who underwent targeted temperature management after cardiac arrest and had post-arrest EEG and MRI available for review. We hypothesize that MRI will be able to distinguish good versus poor outcome in those patients with non-malignant EEG following cardiac arrest and TTM.

Methods

Patients

Patients were identified retrospectively from a prospectively collected database of patients admitted to Brigham and Women’s Hospital with cardiac arrest. Patients selected underwent TTM and subsequently had both continuous EEG and MRI performed as part of their clinical care. All patients in the current study had cooling initiated within 6 h of arrest and were cooled to a target temperature of 33 °C, per the standard protocol at our institution [22]. A total of 86 patients were identified. Of these, three were excluded for technical issues with the imaging (two for inability to retrieve images from PACS, one outside hospital scan with incompatible images). Five patients were excluded due to presence of focal intracranial pathology that impacted both image analysis and outcome determination. Two of these had intracranial hemorrhage, two had large focal infarcts and one patient had a pontine infarct. A total of 78 patients were used in the final analysis. Of these, three patients had serial scans. In one, the second scan had artifact in the ADC sequence, so the first scan was used. In the other two the later scan was used. Retrospective review of neurology consult notes was performed to identify key clinical exam findings, namely pupillary light response, corneal reflex and motor response. Motor findings were dichotomized as flexor response or better (including flexion, withdrawal localization or normal function) versus extensor response, triple-flexion or no response. Motor function was scored based on the best response in either upper or lower extremity. Clinical outcome was defined as good in those patients who regained the ability to open eyes and follow commands prior to hospital discharge; any response less than this was considered poor outcome. Cerebral Performance Category score [23] was used as a secondary outcome, with these results reported in Supplementary Table 1.

EEG Analysis

Each patient was placed on continuous EEG monitoring as per protocol at our institution. Electrodes were placed as early as possible during the patient’s admission to the cardiac critical care unit. Video EEG was collected using Natus XLTEK system (Pleasanton, CA) according to the international 10–20 system. All patients were recorded for a minimum of 24 h during TTM, throughout rewarming and either until the patient awoke sufficiently to follow commands or the neurology consult team felt it was no longer needed for prognostication or seizure management.

A baseline post-TTM and rewarming was established; visual elements were interpreted using the American Clinical Neurophysiology Society critical care EEG terminology [24] and classified into three categories (highly malignant, malignant, benign), as previously validated by Westhall et al. [25]. All EEG classification was done prospectively, and thus, raters were not aware of patient outcome at the time of EEG analysis. For the purposes of this study, categories were then binarized by combining the highly malignant and malignant patients as the specificity is very high in both groups.

Image Analysis

Whole brain ADC signal intensity was determined using Analyze Pro 1.0 based on our prior methods [26], by a rater (MBB) blinded to EEG characteristics and outcome. Briefly, a semi-automated seed-extraction method was used to generate whole brain maps on diffusion weighted sequences. This map was then transferred to the ADC sequence. Artifact was removed by filtering out signal intensity less than 200 mm2/s, while CSF was removed by filtering signal intensity greater than 2000 mm2/s. The resulting object map was used to determine total brain volume, mean signal intensity, and percent of brain volume < 650 mm2/s.

Regional analysis of ADC and FLAIR images were carried out using FSL (Oxford Center for Functional MRI of the Brain, Oxford, UK). ADC and FLAIR images were first co-registered to the T1 weighted image for each subject. The T1 images were then co-registered to the Montreal Neurologic Institute (MNI) Brain Atlas, and the same transformation was applied to the ADC and FLAIR. Success of automated registration and skull stripping was then manually confirmed. Regional anatomic masks were created using a 50% probability map from the MNI atlas. Mean ADC and FLAIR signal intensity was then calculated for each region. To account for inter-subject and inter-scanner differences in FLAIR signal, a FLAIR ratio (FR) was calculated by dividing the mean signal intensity for each gray matter region with the mean intensity for the same subject’s subcortical white matter [27, 28, 29].

Statistical Analysis

Summary statistics are presented as mean ± standard deviation or median and interquartile range [IQR] as appropriate. Comparisons of continuous data were made using Student’s T test for normally distributed or Wilcoxon rank sum for nonparametric data. Proportional data were compared using Fisher’s exact test. All tests were two-tailed with alpha set at 0.05. Reporting of test characteristics for the various prediction models was done by focusing on good outcome, such that a 100% sensitive test ensures that no patients with potential for recovery are missed. Multimodal models were then constructed to optimize specificity for prediction of good outcome. All analyses were carried out in JMP Pro 13.0 (SAS Institute Inc., Cary, NC).

Results

Description of Study Cohort

The study cohort is summarized in Table 1. Subjects had a mean age of 53 ± 17 years and 63% were male. The most common presenting rhythm was ventricular fibrillation or other shockable rhythm (n = 36, 46%). All patients in this cohort underwent TTM. MRI was performed a median of 4 days after rewarming (IQR 3–5 days). A malignant EEG pattern was observed in 38 patients (49%) and 43 patients died (55%). Of these, 42 had withdrawal of life-sustaining therapy (WLST), and the cited cause was poor neurologic prognosis in 40 cases. WLST occurred a median of 8 days after arrest (IQR 5, 11 days), and in all but one case, death occurred the same day as WLST. In total, 31 patients (40%) had good outcome, defined as regaining the ability to follow commands prior to discharge. A similar distribution was observed for outcome classified by Cerebral Performance Category score (Supplemental Table 1).
Table 1

Characteristics of the study cohort

 

n = 78

Age (years), mean ± SD

53 ± 17

Sex (male), N (%)

49 (63%)

Presenting rhythm, N (%)

 PEA

18 (23%)

 Asystole

11 (14%)

 VT, VF or AED-shockable

38 (49%)

 Unknown

11 (14%)

Time to MRI (days), median [IQR]

4 [3, 5]

Whole brain ADC, mean ± SD

892 ± 108

% Brain volume < 650 mm/s2

7.6% [3.5–24%]

Malignant EEG pattern, N (%)

38 (49%)

Regained ability to follow commands, N (%)

31 (40%)

Length of stay (days), median [IQR]

10 [7, 17]

Mortality, N (%)

43 (55%)

 WLST due to poor neurologic prognosis

40 (51%)

 WLST due to other medical comorbidities

2 (3%)

 Time to WLST (days), median [IQR]

8 [5, 11]

ADC apparent diffusion coefficient, AED automated external defibrillator, PEA pulseless electrical activity, VF ventricular fibrillation, VT ventricular tachycardia, WLST withdrawal of life-sustaining therapy

Association of Clinical Exam Findings to Outcome

Subjects with good outcome were significantly more likely to have intact pupillary light response (100 vs. 77%, p = 0.003) and corneal reflexes (100 vs. 55%, p < 0.0001). The strongest exam predictor was a motor response of flexor or better, which was present in 100% of those with good outcome versus 21% with poor outcome (p < 0.0001).

Malignant EEG Pattern is Associated with Poor Outcome

Malignant EEG patterns were observed in 38 of 47 patients with poor outcome and no patients with good outcome (p < 0.0001), making it 100% specific and 81% sensitive for poor outcome. Conversely, non-malignant EEG was observed in 9 of 47 patients with poor outcome and all 31 patients with good outcome. Non-malignant EEG was therefore 100% sensitive and 77.5% specific for predicting good outcome.

Quantitative and Qualitative MRI Findings are Associated with Outcome

Whole Brain MRI Analysis

Whole brain ADC signal intensity was lower in those with poor outcome (mean ADC 852 ± 14 vs. 950 ± 18 mm2/s, p < 0.0001, Table 2). Similarly, the percentage of total brain volume with ADC signal less than 650 mm2/s was higher in those with poor outcome (16.2% [IQR 4.6–40%] vs. 4.1% [2.9–6.0%], p < 0.0001). Those with poor outcome were more likely to have a clinical MRI read reporting subjective anoxic brain injury (n = 42 [89%] vs. 5 [16%], p < 0.0001). Of the 42 patients with poor outcome and clinical reads of anoxic injury, this injury was described as diffusely involving cortical and deep structures in 29, cortical in 11, and restricted to basal ganglia in 2. Of the five patients with good outcome, findings were described as diffuse in three and cortical in two.
Table 2

Imaging characteristics based on patient outcome

 

Good outcome

N = 31

Poor outcome

N = 47

p

Whole brain ADC (mm2/s), mean ± SEM

950 ± 17.5

853 ± 14

< 0.0001

% Brain volume < 650 mm2/s, median [IQR]

4.1% [2.9–6.0%]

16.2% [4.6–40%]

< 0.0001

Subjective anoxia on clinical MRI read, n (%)

5 (16%)

42 (89%)

< 0.0001

ADC apparent diffusion coefficient, MRI magnetic resonance imaging

Receiver operating curve (ROC) analysis of percent brain volume < 650 mm2/s found that a cut-off of < 15% was 100% sensitive and 51% specific for predicting good outcome. Area under the curve was 0.787.

Regional Distribution of ADC and FLAIR Changes

The predictive value of regional ADC changes for poor outcome is shown in Fig. 1a. Decreased ADC signal intensity was associated with reduced odds of good outcome in multiple regions including the cortex (OR per 100 mm2/s decrease in ADC 0.43 [95% CI 0.26–0.72], p = 0.001), combined basal ganglia (OR 0.54 [0.25–0.99], p = 0.027) and putamen alone (OR 0.55 [0.31–0.96], p = 0.021). Similar results were observed for FR (Fig. 1b). Increased FR in cortex (OR per 0.1 increase in FR 0.37 [95% CI 0.18–0.76], p = 0.004), combined basal ganglia (OR 0.62 [0.40–0.96], p = 0.025) and putamen alone (OR 0.16 [0.06–0.42], p < 0.0001) was associated with reduced odds of good outcome. ROC analysis of each region was performed, with cutoffs chosen to ensure 100% sensitivity in detecting good outcome (that is, a false positive rate of 0). Results are summarized in Table 3. Only putamen FR outperformed whole brain ADC in terms of area under the curve, but at a threshold for 100% sensitivity had a lower specificity (32 vs. 51% for whole brain ADC).
Fig. 1

Regional ADC and FLAIR analysis. ADC and FLAIR sequences were co-registered to the Montreal Neurologic Institute probability atlas, with accurate registration manually confirmed. Masks for each listed brain region were created using the 50% probability area and mean ADC and FLAIR signal measured within each region. FLAIR signal was normalized by dividing each regions FLAIR by the signal intensity of the white matter in the same subject, with results reported as the FLAIR ratio. a Decreased ADC signal intensity was associated with reduced odds of good outcome in the cortex (OR per 100 mm2/s decrease in ADC 0.43 [95% CI 0.26–0.72], p = 0.001), combined basal ganglia (OR 0.54 [0.25–0.99], p = 0.027) and putamen alone (OR 0.55 [0.31–0.96], p = 0.021). b Reduced odds of good outcome were seen with each 10% increase in FLAIR ratio in the cortex (OR 0.37 [95% CI 0.18–0.76], p = 0.004), combined basal ganglia (OR 0.62 [0.40–0.96], p = 0.025) and putamen alone (OR 0.16 [0.06–0.42], p < 0.0001)

Table 3

Test characteristics of imaging parameters to predict good outcome after TTM

Parameter

Cut-off

AUC

Sensitivity (%)

Specificity (%)

Whole brain ADC SI (mm2/s)

> 847

0.754

100

45

% Brain volume < 650 mm2/s

< 15%

0.787

100

51

Cortex ADC SI (mm2/s)

> 923

0.715

100

43

Basal ganglia ADC SI (mm2/s)

> 798

0.654

100

32

Putamen ADC SI (mm2/s)

> 719

0.626

100

30

Cortex FR

< 1.07

0.702

100

27

Basal ganglia FR

< 1.19

0.638

100

18

Putamen FR

< 1.19

0.804

100

32

ADC SI apparent diffusion coefficient signal intensity, AUC area under curve, FR FLAIR ratio

Effect of Time on ADC Signal Intensity

Subjects were grouped according to time of MRI, and then, ADC signal intensity was compared between groups (Fig. 2). ADC varied significantly by time point (p = 0.0481), reaching a nadir at 3 days post-arrest, then rising again. Restricting analysis to only those patients who underwent MRI on days 2 through 4 (n = 44), the ability of ADC to predict good outcome improves. ROC analysis using only these subjects gives an area under the curve of 0.892, with a cut-off of 12.5% brain volume < 650 mm2/s predicting good outcome with sensitivity of 100% and specificity of 73%.
Fig. 2

Mean whole brain ADC signal intensity by date of MRI. Subjects were grouped based on the day post-cardiac arrest they underwent MR imaging. The mean whole brain ADC signal was calculated, and median values were compared across time points. There was a significant difference in median ADC values over time, with a nadir seen at 3 days post-arrest (p = 0.0481)

Relationship of MRI Findings to Outcome in Patients with Non-malignant EEG

In the subgroup of patients with non-malignant EEG (n = 40), mean ADC signal intensity did not differ between outcome groups (913 ± 24 vs. 950 ± 13 mm2/s, poor vs. good outcome, p = 0.17) or median percent of brain volume with ADC < 650 mm2/s (4.2% [IQR 2.0–25%] vs. 4.1% [2.9–6.0%], poor vs. good outcome, p = 0.94). Fewer patients with poor outcome had < 15% of total brain volume with ADC < 650 mm2/s (78 vs. 100%, poor vs. good outcome, p = 0.046). There were no significant differences between outcome groups in regional ADC signal intensity or FR. Patients with non-malignant EEG and poor outcome were more likely to have a clinical read reporting anoxic brain injury (n = 7 [77.8%] vs. 5 [16.1%], poor vs. good outcome, p = 0.0004).

Multimodal Models to Predict Good Outcome

In an effort to optimize sensitivity and specificity for identifying good outcome, we constructed multimodal models using EEG, MRI and exam markers. Non-malignant EEG alone was 100% sensitive but only 77.5% specific for the prediction of good outcome. Adding the finding of < 15% total brain volume with ADC < 650 mm2/s improved specificity to 81.6%. Specificity was improved with the use of qualitative MRI results; however, sensitivity decreased to 83.9%. Addition of a motor exam of flexor or better to EEG improved specificity to 88.6%, while all three (EEG + motor exam + ADC) was 100% sensitive and 91.1% specific for good outcome (Table 4).
Table 4

Multimodal prediction models for good outcome

Model

Sensitivity (%)

Specificity (%)

EEG

100

77.5

EEG + ADC

100

81.6

EEG + clinical read of anoxic injury

83.9

92.9

EEG + motor

100

88.6

EEG + motor + ADC

100

91.1

EEG: scored as positive if no malignant features were observed

ADC: scored as positive if < 15% of total brain volume had ADC SI of < 650 mm2/s

Motor: scored as positive if response to noxious stimuli was flexor or better

ADC apparent diffusion coefficient, EEG electroencephalography

Only three patients had poor outcome despite non-malignant EEG, low percentage of brain volume with ADC < 650 mm2/s and better than flexor motor response. All three underwent WLST according to family wishes. Two patients had pre-existing malignancy and care withdrawal was related to their overall prognosis from non-neurologic disease. The third patient had qualitative evidence of anoxic brain injury restricted to basal ganglia, and post-cardiac arrest course was complicated by severe autonomic instability, which drove the family’s decision to withdraw care.

Conclusion

We found that a combination of EEG, physical exam and MRI analysis was superior to EEG alone for predicting good outcome following cardiac arrest and TTM. Furthermore, our results confirm findings in other studies regarding the ability of malignant EEG to predict poor outcome. Similarly, our findings on the association of whole brain ADC and proportion of brain volume with ADC < 650 mm2/s are consistent with prior reports in the literature [17, 18, 19, 20]. Quantitative regional ADC and FLAIR analysis found that both restricted diffusion and FLAIR hyperintensity were associated with poor outcome in structures known to be vulnerable to global ischemic injury, including cortex and basal ganglia [18, 30]. It is somewhat surprising that we did not see significant ADC or FLAIR changes in the hippocampus, given its known vulnerability to hypoxic/ischemic injury [31]. This is likely explained by the fact that its location and size make it vulnerable to both artifact and errors in accurate identification when using a co-registered probability atlas.

In our cohort, 22% of subjects with non-malignant EEG still had poor outcome. While the only significant imaging finding in this subset of patients was the proportion of patients with < 15% brain volume with ADC < 650 mm2/s, we are likely underpowered to detect quantitative changes given the relatively small sample size of patients with non-malignant EEG. Interestingly, qualitative reads identified evidence of anoxic injury in the majority of these patients; all were noted to be subtle findings or strictly regional. MR changes in these patients likely represent small but significant injury to brain regions in variable locations; thus, the lack of detectable quantitative changes across this particular population. This may also reflect the fact that the qualitative reads could better account for changes across multiple MR sequences and may have been particularly useful in our cohort where timing of imaging was quite variable. We found that ADC signal intensity varied over time, reaching a nadir at 3 days out and then increasing again. Some of this may be due to pseudo-normalization rather than recovery, a finding that would likely be reflected in increased FLAIR signal at later time points [32, 33].

It is also possible that qualitative reads were more predictive of outcome because these reads were available to the treating teams. They would therefore contribute to a “self-fulfilling-prophecy” such that those with a clinical read reporting anoxia were more likely to undergo withdrawal of care. A similar effect may have been present for the EEG reads. This is a limitation in any retrospective study of outcome after cardiac arrest or other severe neurologic injury [34, 35]. Given that the vast majority of patients in our cohort with poor outcome had WLST based on prognostic information provided by the neurology team, it is likely that the “self-fulfilling-prophecy” was a confounding factor. This emphasizes the importance of future prospective trials of multimodal outcome prediction including clinical exam, quantitative MRI and EEG. Such trials should further include blinding of all participants to whatever degree is ethically possible and the use of systematic prognostication algorithms to minimize any potential bias.

We chose to report test characteristics in this study with a focus on good outcome, which stands in contrast to current literature which instead focuses on optimizing specificity for poor outcome [9, 10, 11, 12, 13, 14, 15, 16]. Given that presence of malignant EEG findings has been established to be 100% specific for poor outcome, we instead felt that the key existing question is to develop a model for predicting good outcome. This led us to choose a baseline of 100% sensitivity for good outcome and then identify imaging parameters and multimodal models that optimize the specificity of that prediction. The end result of focusing on 100% specificity for poor outcome or 100% sensitivity for good outcome is the same, in that no patient with potential for neurology recovery is “missed” by the prediction model.

The current study has other limitations related to its retrospective nature. Ancillary tests, such as somatosensory evoked potentials (SSEPs), may also contribute prognostic information. Due to resource limitation during the time period, these patients were admitted, SSEPs were typically only performed if the neurology consult team felt additional information was needed. As noted above, timing of imaging was quite variable. While we saw increased sensitivity of ADC signal intensity when MR was performed between days 2 and 4, we do not have serial imaging available to better describe the evolution of ADC over time and identify the ideal imaging time point.

This dataset is also limited as it lacks long-term functional outcomes. Other studies of imaging prognostication after cardiac arrest have used both early markers such as eye opening by 1 week [18], pupillary and motor responses at 72 h [17] or death prior to hospital discharge [20]. Longer-term outcomes have included 6 months modified Rankin Scale [18] and Glasgow Outcome Scale at 3 [19] or 6 months [17]. Here, we opted to use the ability to follow commands prior to discharge as the differentiator between good and poor outcome, which reflects outcome at a median of 10 days after arrest (IQR 7, 17). While it would be ideal to validate this with longer-term outcomes, the fact that only four patients in our poor outcome group survived the hospital stay limits any potential bias introduced by looking only at early outcome.

Despite these limitations, the current study provides additional evidence for the importance of using multiple testing modalities when attempting to prognosticate after cardiac arrest and TTM. In absence of other major concurrent central nervous system lesions and serious medical comorbidities prompting withdrawal of care, a combination of non-malignant EEG, at least flexor motor response and MRI with < 15% of total brain volume with low ADC was able to accurately identify patients with neurologic recovery to at least following commands. As such, it is reasonable that every patient with a non-malignant EEG but without improvement to at least following commands by day 2 following cardiac arrest obtain an MRI. In the absence of imaging evidence of anoxic brain injury, decisions regarding goals of care should be made with the understanding that neurologic prognosis would be considered indeterminant.

Where there is likely a role for imaging in addition to EEG, future prospective study is needed. Future investigations of serial imaging, including other imaging modalities such as magnetic resonance spectroscopy, may help improve our ability to identify those patients with the potential to recover. Integrating these findings with consistent clinical exams and other techniques such as quantitative EEG is likely the key to building an effective model for predicting outcome after cardiac arrest.

Notes

Author contributions

MBB analyzed the MRI data and was involved in statistical analysis, study conception and design, and primary drafting of manuscript. BMS, KRA, GVH were involved in acquisition of clinical data and critical revisions of the manuscript. APL interpreted the imaging data and critically revised the manuscript. JWL analyzed the EEG and MRI data, critically revised the manuscript, and was involved in study conception and design.

Sources of Support

None.

Compliance with Ethical Standards

Conflict of interest

Matthew B. Bevers reports a grant from the American Academy of Neurology, personal fees from Oakstone publishing, personal fees from DynaMed LLC, all outside the submitted work. Benjamin M. Scirica reports Grants and personal fees from AstraZeneca, grants from Daiichi Sankyo, grants and personal fees from Eisai, grants and personal fees from Gilead, grants from Novartis, grants and personal fees from Merck, grants from Poxel, personal fees from Biogen, personal fees from Boehringer Ingelheim, personal fees from Boston Clinical Research Institute, personal fees from Covance, personal fees from Elsevier Practice Update Cardiology, personal fees from GlaxoSmithKline, personal fees from Lexicon, personal fees from NovoNordisk, personal fees from Sanofi, personal fees from St. Jude’s Medical, other support from Health at Scale, all outside the submitted work. Kathleen Ryan Avery, Galen V. Henderson, Alexander P. Lin have Nothing to disclose. Jong W. Lee reports Contract work with SleepMed/DigiTrace and Advance Medical, grants from NIH, all outside the submitted work.

Ethical Approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The current study was approved by the Partners Healthcare Institutional Review Board. As a retrospective study, it was approved with waiver of informed consent.

Supplementary material

12028_2018_559_MOESM1_ESM.docx (13 kb)
Supplementary material 1 (DOCX 12 kb)

References

  1. 1.
    Rea TD, Eisenberg MS, Sinibaldi G, White RD. Incidence of EMS-treated out-of-hospital cardiac arrest in the United States. Resuscitation. 2004;63(1):17–24.CrossRefGoogle Scholar
  2. 2.
    Mozaffarian D, Benjamin EJ, Go AS, et al. Heart disease and stroke statistics-2016 update a report from the American Heart Association. 2016.Google Scholar
  3. 3.
    Holzer M. Targeted temperature management for comatose survivors of cardiac arrest. N Engl J Med. 2010;363(13):1256–64.CrossRefGoogle Scholar
  4. 4.
    Silverman MG, Scirica BM. Cardiac arrest and therapeutic hypothermia. Trends Cardiovasc Med. 2016;26(4):337–44.CrossRefGoogle Scholar
  5. 5.
    Bernard S, Gray T, Buist M, et al. Treatment of comatose survivors of out-of-hospital cardiac arrest with induced hypothermia. N Engl J Med. 2002;346:557–63.CrossRefGoogle Scholar
  6. 6.
    Nielsen N, Wetterslev J, Cronberg T, et al. Targeted temperature management at 33°C versus 36°C after cardiac arrest. N Engl J Med. 2013;369(23):2197–206.CrossRefGoogle Scholar
  7. 7.
    HACA Study Group. Mild therapeutic hypothermia to improve the neurologic outcome after cardiac arrest. N Engl J Med. 2002;346(8):549–56.CrossRefGoogle Scholar
  8. 8.
    Wijdicks EFM, Hijdra A, Young GB, Bassetti CL, Wiebe S. Practice parameter: prediction of outcome in comatose survivors after cardiopulmonary resuscitation (an evidence-based review). Report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology. 2006;67(2):203–10.CrossRefGoogle Scholar
  9. 9.
    Sandroni C, Cariou A, Cavallaro F, et al. Prognostication in comatose survivors of cardiac arrest: an advisory statement from the european resuscitation council and the european society of intensive care medicine. Resuscitation. 2014;85(12):1779–89.CrossRefGoogle Scholar
  10. 10.
    Legriel S, Hilly-Ginoux J, Resche-Rigon M, et al. Prognostic value of electrographic postanoxic status epilepticus in comatose cardiac-arrest survivors in the therapeutic hypothermia era. Resuscitation. 2013;84(3):343–50.CrossRefGoogle Scholar
  11. 11.
    Crepeau AZ, Rabinstein AA, Fugate JE, et al. Continuous EEG in therapeutic hypothermia after cardiac arrest: prognostic and clinical value. Neurology. 2013;80(4):339–44.CrossRefGoogle Scholar
  12. 12.
    Rossetti AO, Carrera E, Oddo M. Early EEG correlates of neuronal injury after brain anoxia. Neurology. 2012;78(11):796–802.CrossRefGoogle Scholar
  13. 13.
    Rossetti AO, Urbano LA, Delodder F, Kaplan PW, Oddo M. Prognostic value of continuous EEG monitoring during therapeutic hypothermia after cardiac arrest. Crit Care. 2010;14(5):R173.CrossRefGoogle Scholar
  14. 14.
    Rossetti AO, Oddo M, Logroscino G, Kaplan PW. Prognostication after cardiac arrest and hypothermia: a prospective study. Ann Neurol. 2010;67(3):301–7.Google Scholar
  15. 15.
    Oddo M, Rossetti AO. Early multimodal outcome prediction after cardiac arrest in patients treated with hypothermia. Crit Care Med. 2014;42(6):1340–7.CrossRefGoogle Scholar
  16. 16.
    Fantaneanu TA, Sarkis R, Avery K, et al. Delayed deterioration of EEG background rhythm post-cardiac arrest. Neurocrit Care. 2017;26(3):411–9.CrossRefGoogle Scholar
  17. 17.
    Wijman CAC, Mlynash M, Caulfield AF, et al. Prognostic value of brain diffusion-weighted imaging after cardiac arrest. Ann Neurol. 2009;65(4):394–402.CrossRefGoogle Scholar
  18. 18.
    Wu O, Sorensen AG, Benner T, Singhal AB, Furie KL, Greer DM. Comatose patients with cardiac arrest: predicting clinical outcome with diffusion-weighted MR imaging. Radiology. 2009;252(1):173–81.CrossRefGoogle Scholar
  19. 19.
    Choi SP, Park KN, Park HK, et al. Diffusion-weighted magnetic resonance imaging for predicting the clinical outcome of comatose survivors after cardiac arrest: a cohort study. Crit Care. 2010;14(1):R17.CrossRefGoogle Scholar
  20. 20.
    Järnum H, Knutsson L, Rundgren M, et al. Diffusion and perfusion MRI of the brain in comatose patients treated with mild hypothermia after cardiac arrest: a prospective observational study. Resuscitation. 2009;80(4):425–30.CrossRefGoogle Scholar
  21. 21.
    Lee BK, Jeung KW, Lee HY, Jung YH, Lee DH. Combining brain computed tomography and serum neuron specific enolase improves the prognostic performance compared to either alone in comatose cardiac arrest survivors treated with therapeutic hypothermia. Resuscitation. 2013;84(10):1387–92.CrossRefGoogle Scholar
  22. 22.
    Szumita PM, Baroletti S, Avery KR, et al. Implementation of a hospital-wide protocol for induced hypothermia following successfully resuscitated cardiac arrest. Crit Pathw Cardiol. 2010;9(4):216–20.CrossRefGoogle Scholar
  23. 23.
    Jennett B, Bond M. Assessment of outcome after severe brain damage. Lancet. 1975;1(7905):480–4.CrossRefGoogle Scholar
  24. 24.
    Hirsch LJ, LaRoche SM, Gaspard N, et al. American clinical neurophysiology society’s standardized critical care EEG terminology. J Clin Neurophysiol. 2013;30(1):1–27.CrossRefGoogle Scholar
  25. 25.
    Westhall E, Rossetti AO, Van Rootselaar AF, et al. Standardized EEG interpretation accurately predicts prognosis after cardiac arrest. Neurology. 2016;86(16):1482–90.CrossRefGoogle Scholar
  26. 26.
    Bevers MB, Vaishnav NH, Pham L, Battey TWK, Kimberly WT. Hyperglycemia is associated with more severe cytotoxic injury after stroke. J Cereb Blood Flow Metab. 2017;37(7):2577–83.CrossRefGoogle Scholar
  27. 27.
    Focke NK, Bonelli SB, Yogarajah M, Scott C, Symms MR, Duncan JS. Automated normalized FLAIR imaging in MRI-negative patients with refractory focal epilepsy. Epilepsia. 2009;50(6):1484–90.CrossRefGoogle Scholar
  28. 28.
    Huppertz HJ, Wagner J, Weber B, House P, Urbach H. Automated quantitative FLAIR analysis in hippocampal sclerosis. Epilepsy Res. 2011;97(1–2):146–56.CrossRefGoogle Scholar
  29. 29.
    Jha R, Battey TWK, Pham L, et al. Fluid-attenuated inversion recovery hyperintensity correlates with matrix metalloproteinase-9 level and hemorrhagic transformation in acute ischemic stroke. Stroke. 2015;45(4):1040–5CrossRefGoogle Scholar
  30. 30.
    Mlynash M, Campbell DM, Leproust EM, et al. Temporal and spatial profile of brain diffusion-weighted MRI after cardiac arrest. Stroke. 2010;41(8):1665–72.CrossRefGoogle Scholar
  31. 31.
    Greer DM, Scripko PD, Wu O, et al. Hippocampal magnetic resonance imaging abnormalities in cardiac arrest are associated with poor outcome. J Stroke Cerebrovasc Dis. 2013;22(7):899–905.CrossRefGoogle Scholar
  32. 32.
    Schlaug G, Siewert B, Benfield A, Edelman RR, Warach S. Time course of the apparent diffusion coefficient (ADC) abnormality in human stroke. Neurology. 1997;49(1):113–9.CrossRefGoogle Scholar
  33. 33.
    Ahlhelm F, Schneider G, Backens M, Reith W, Hagen T. Time course of the apparent diffusion coefficient after cerebral infarction. Eur Radiol. 2002;12(9):2322–9.CrossRefGoogle Scholar
  34. 34.
    Zahuranec DB, Brown DL, Lisabeth LD, et al. Early care limitations independently predict mortality after intracerebral hemorrhage. Neurology. 2007;68(20):1651–7.CrossRefGoogle Scholar
  35. 35.
    Greer DM, Rosenthal ES, Wu O. Neuroprognostication of hypoxic–ischaemic coma in the therapeutic hypothermia era. Nat Rev Neurol. 2014;10(4):190–203.CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media, LLC, part of Springer Nature and Neurocritical Care Society 2018

Authors and Affiliations

  • Matthew B. Bevers
    • 1
  • Benjamin M. Scirica
    • 2
  • Kathleen Ryan Avery
    • 3
  • Galen V. Henderson
    • 1
  • Alexander P. Lin
    • 4
  • Jong W. Lee
    • 5
  1. 1.Divisions of Stroke, Cerebrovascular and Critical Care NeurologyBrigham and Women’s HospitalBostonUSA
  2. 2.Cardiovascular DivisionBrigham and Women’s HospitalBostonUSA
  3. 3.Department of NursingBrigham and Women’s HospitalBostonUSA
  4. 4.Department of RadiologyBrigham and Women’s HospitalBostonUSA
  5. 5.Division of Epilepsy and Seizure DisordersBrigham and Women’s HospitalBostonUSA

Personalised recommendations