Intensive Care Med (2014) 40:228–234 DOI 10.1007/s00134-013-3149-8
Pedro Kurtz Nicolas Gaspard Anna Sophia Wahl Rebecca Marie Bauer Lawrence J. Hirsch Hannah Wunsch Jan Claassen
Received: 10 July 2013 Accepted: 27 October 2013 Published online: 16 November 2013 Ó Springer-Verlag Berlin Heidelberg and ESICM 2013 Take-home message: Nonconvulsive seizures occur in 16 % of patients admitted to a surgical ICU for post-operative care with altered mental status and are associated with poor outcome at discharge. Risk factors include coma and clinical seizures prior to cEEG monitoring. Electronic supplementary material The online version of this article (doi:10.1007/s00134-013-3149-8) contains supplementary material, which is available to authorized users.
ORIGINAL ARTICLE
Continuous electroencephalography in a surgical intensive care unit
R. M. Bauer Department of Anesthesiology, Feinberg School of Medicine, Northwestern University, Chicago, USA H. Wunsch Division of Critical Care, Department of Anesthesiology, Columbia University, College of Physicians and Surgeons, New York, NY, USA H. Wunsch Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, NY, USA
Abstract Purpose: Our aim was to investigate the prevalence, risk factors, and impact on outcome of nonconvulsive seizures (NCSz), nonconvulsive status epilepticus (NCSE), and periodic epileptiform discharges (PEDs) in surgical intensive care unit (SICU) patients with continuous electroencephalography (cEEG) P. Kurtz A. S. Wahl J. Claassen ()) Division of Critical Care Neurology, monitoring. Methods: This was a Department of Neurology, Columbia retrospective study of SICU patients University, College of Physicians and who underwent cEEG monitoring for Surgeons, New York, NY, USA altered mental status over a 6-year e-mail: [email protected] period. We report the frequency of NCSz (including NCSE) and PEDs P. Kurtz on cEEG. The primary outcome was Department of Intensive Care Medicine, Clı´nica Sa˜o Vicente, Rua Joa˜o Borges 204, death or severe disability at hospital Gavea, Rio de Janeiro, RJ 22451-100, discharge. Multivariable logistic Brazil regression was used to identify whether NCSz (including NCSE) and N. Gaspard L. J. Hirsch Department of Neurology, Comprehensive PEDs were independently associated with poor outcome (death, vegetative Epilepsy Center, Yale University, state or severe disability). Results: New Haven, CT, USA
Of 154 patients, the mean age was 64 ± 14 years old, and 40 % were women. The majority of patients were admitted following abdominal surgery (36 %) and liver transplantation (24 %). Sepsis developed in 100 (65 %) patients. Sixteen percent (n = 24) had NCSz [including 5 % (N = 8) with NCSE], and 29 % (N = 45) had PEDs. All eight patients with NCSE were septic. Clinical seizures prior to cEEG and coma were more common among patients who developed NCSz or NCSE compared to patients without NCSz or NCSE (70 vs. 27 %; p \ 0.01; 75 vs. 52 %; p = 0.046 and 63 vs. 34 %; p = 0.09, respectively). NCSzs (including NCSE) were independently associated with poor outcome (20 vs. 3 %, OR 10.4, 95 % CI 1.0–53.7; p = 0.039). Conclusion: In this retrospective study of SICU patients with cEEG monitoring for altered mental status, NCSz and periodic discharges were frequent and NCSz were independently associated with poor outcome. NCSz were more common when clinical seizures occurred before cEEG. Keywords Nonconvulsive status epilepticus Nonconvulsive seizures Electrographic seizures Periodic discharges Surgical ICU Outcome
229
Introduction
Table 1 Patient characteristics (n = 154)
Acute neurologic dysfunction, including delirium and coma, is common in the acute postoperative period for patients cared for in surgical and trauma intensive care units (ICUs) [1–6] and is associated with worse outcome [7–10]. A subgroup of critically ill patients with impaired mental status, often lumped under the diagnoses of metabolic encephalopathy or delirium, may have ongoing periodic epileptiform discharges (PEDs) and nonconvulsive seizures (NCSz), including nonconvulsive status epilepticus (NCSE), potentially necessitating a change in treatment approach [7, 11]. Seizures and other highly epileptiform patterns are common in ICU patients with acute brain injury [1, 3, 5–7, 11–14], as well as those who receive continuous electroencephalography (cEEG) monitoring without acute brain injury in medical ICUs [7, 9, 11, 15, 16]. These seizures are almost always nonconvulsive and therefore are not detectable without EEG monitoring [7, 11, 12, 16]; they are associated with poor outcome [3, 5, 7, 11–14, 17]. Surgical ICU (SICU) patients, who are primarily admitted from the operating room, are fundamentally different from patients encountered in medical ICUs. Although recommendations for the use of cEEG in critically ill patients exist [18], there is little information available on the utility of cEEG monitoring in SICU patients and on the prognostic significance of EEG findings in this population. Therefore, we studied the prevalence, predictors, and impact on outcome of PEDs, NCSz and NCSE in surgical ICU (SICU) patients without primary acute brain injury.
Age; median years (IQR) Female n (%) Primary admission reason n (%) Transplant Other abdominal Orthopedic/trauma Cardiovascular/thoracic Other Sepsis on admission Past medical history n (%) Epilepsy Stroke Diabetes Chronic renal dysfunction Cardiac disease Chronic liver dysfunction Cancer cEEG monitoring Time from admission to cEEG; days; median (IQR) Length of monitoring; days; median (IQR) Clinical status at time of cEEG n (%) GCS: \8/9–12/13–14/15
Methods Study population Of a total of *5,800 patients admitted to the SICU at Columbia University Medical Center from August 2004 until July 2010, we identified 154 consecutive patients who underwent continuous EEG (cEEG) for a minimum of 12 h and had an indication for cEEG monitoring of unexplained altered mental status (supplementary Table 1 and supplementary information on methods). Patients were identified from the Comprehensive Epilepsy Center electronic log and cEEG database (please refer to the supplementary information for details on the study population, cEEG and clinical variables). Continuous EEG was requested by a neurology consultant in 87 % of cases. The patients’ neurologic status (comatose or not comatose) at the time cEEG was initiated was recorded from documentation by the neurology consultant or by the intensivist. All patients were admitted for postoperative care after a variety of procedures (see Table 1). Patients with a primary acute neurological diagnosis or those that underwent neurosurgical interventions were excluded
Focal neurological signs* Length of stay in ICU/in hospital (days; median [range]) Outcome at hospital discharge: dead/ persistent vegetative/severe disability/ moderate disability/good recovery) n (%)**
65 (54–74) 62 (40) 44 56 15 29 12 58
(29) (36) (9) (19) (8) (37)
18 55 59 61 73 60 55
(12) (35) (38) (39) (47) (39) (36)
6 (4–9) 2 (1–2.5) 86 (55)/35 (23)/30 (19)/3 (2) 17(12) 22 (1–167)/52 (10–201) 73 (47)/7 (5)/37 (24)/ 27 (17)/8 (5)
cEEG continuous EEG, ICU intensive care unit, SD standard deviation Data available in * N = 144 and ** N = 152 patients respectively
from this analysis. Approval for this study was received from the Columbia University Medical Center Hospital Institutional Review Board (IRB).
cEEG recordings and variables Continuous EEG was recorded using 21 electrodes placed according to the international 10–20 system and interpreted by a board-certified electroencephalographer. We recorded the occurrence of convulsive seizures (CSz), NCSz, convulsive status epilepticus (CSE), NCSE and PEDs [including generalized periodic epileptiform discharges (GPEDs), periodic lateralized epileptiform discharges (PLEDs), triphasic waves and stimulusinduced periodic, rhythmic and ictal appearing discharges (SIRPIDs)] [7, 9, 11, 12, 15, 16, 19]. Clinical variables Data were retrieved via cEEG reports, medical records including discharge summaries and physician sign-out
230
notes. Details of the collected clinical variables are out- Brain imaging lined in the supplementary material. Our primary outcome was functional outcome at hospital discharge as assessed Brain imaging was performed within 24 h of cEEG in 142 patients [all patients received a CT scan and 34 (22 %) using the Glasgow Outcome Scale [7, 19]. also received an MRI]; 23 scans showed acute ischemic lesions (12 territorial infarcts, of which 4 were clinically silent, and 12 multifocal punctate ischemic lesions, all Statistical analysis clinically silent), 6 showed intraparenchymal hemorrhage Data were analyzed with SPSS 16.0 software (Chicago, IL), (of which 3 were clinically silent), and 4 showed radiousing Student’s t test for continuous variables and v2 test for logical signs of posterior reversible encephalopathy categorical variables. Univariate analysis was performed to syndrome (PRES). Signs of chronic or remote injury were identify significant predictors of NCSz (including NCSE) seen in 68 patients. and PEDs. Significant variables at the level of p \ 0.1 were then included in a multivariable logistic regression model to identify independent predictors. The relationship CSF studies between outcome and the presence of NCSz (including NCSE) and the presence of PEDs was analyzed indepen- CSF was analyzed in 24/154 (16 %) patients. An abnordently using a multivariable logistic regression model. A mal protein level ([50 mg/dl) was found in 5/24 (21 %) cases, and an abnormal cell count ([5 cells/ml) was found p value \0.05 was considered significant. in 14/24 (58 %) cases. Glucose levels were in the normal range in all cases, and bacterial cultures were negative.
Results Patient characteristics From August 2004 to July 2010, 154 SICU patients (2.7 %) underwent cEEG monitoring. Median time from SICU admission to initiation of cEEG was 6 days (IQR: 4–9 days), and median duration of cEEG was 2 days (IQR: 1–2.5 days) (Table 1). Most patients (65 %) were admitted after undergoing abdominal surgery, including liver and kidney transplantation (24 and 5 % of total, respectively). Fifty-eight patients (37 %) had sepsis on admission, and 95 % were mechanically ventilated. Half of patients (55 %) were comatose (GCS B8), and 12 % had focal neurological signs prior to cEEG. The etiology of the alteration of consciousness (GCS \15; N = 151), as determined by the clinical team, was NCSE or NCSZ [N = 10 (7 %)], sepsis associated encephalopathy [N = 65 (43 %)], ICU delirium [N = 33 (22 %)], metabolic encephalopathy [N = 28 (19 %)] and acute brain injury [N = 18 (12 %)]. Main complications during the SICU stay involved organ failure, including renal dysfunction and circulatory shock (34 and 63 % of the total cohort, respectively). Clinical seizures occurred in 51/154 patients (33 %) prior to cEEG. Sedation One hundred six (70 %) patients were sedated at the time of cEEG. Propofol [N = 49 (32 %)], midazolam [N = 14 (9 %)], dexmedetomidine [N = 34 (22 %)], and a combination of midazolam and propofol [N = 11 (7 %)] were used.
Continuous EEG findings and predictors of NCSz and PEDs Continuous EEG was performed because of unexplained alteration in mental status in 133/154 (86 %) patients and because of suspected clinical seizure activity in 56/154 patients (36 %). A total of 24 of 154 patients (16 %) had NCSz, including 8 (5 %) with NCSE, and 45 (29 %) developed PEDs during cEEG. Most patients with NCSz [21/24; (88 %)] also had PEDs. After multivariate regression, clinical seizures before cEEG and coma were more common among patients with NCSz (including NCSE) and PEDs (Table 2 and Supplementary Table 1; Figs. 1, 2). Chronic hepatic failure was more common among patients with NCSz in univariate analysis but not after multivariate analysis. Admission diagnosis and the development of sepsis did not differ between patients with NCSz or PEDs and those without these EEG findings. All eight patients who developed NCSE had sepsis. The proportion of patients on sedation during cEEG did not differ significantly between patients with NCSz and those without NCSz [15/24 (63 %) vs. 92/130 (71 %); p = 0.40]. We found no difference in the incidence of NCSz between the type of sedative drug used [3/ 33 (9 %), 6/31 (19 %) and 11/60 (18 %), respectively for dexmedetomidine, midazolam and propofol; p = 0.30]. Of 23 patients with acute ischemic lesions, 5 had seizures [5/23 (22 %) vs. 19/103 (17 %) without ischemic lesions; p = 0.41]; of 6 patients with hemorrhage, 1 had seizures [1/6 (17 %) vs. 23/138 (19 %) without hemorrhagic lesions; p = 0.96]; of 4 patients with PRES, 2 had seizures [2/2 (50 %) vs. 22/142 (15 %) without PRES; p = 0.13]. Overall, the presence of an acute brain lesion
231
Table 2 Predictors of NCSz (including NCSE) and (PEDs) (variables significant in univariate analysis only) NCSz (including NCSE; n = 24) Multivariate analysis* Without Chronic hepatic failure 44 (34) Coma 68 (52) Clinical seizures 35 (27) prior to cEEG
PEDs (n = 45)
With
Univ. p
p
15 (63) 18 (75) 16 (70)
0.007 0.036 \0.001
0.09 – – 38 (35) 0.046 4.1 2.8–7.6 53 (48) \0.001 10.6 4.9–76.5 31 (28)
Data reported as N (%) or median (IQR) NCSz nonconvulsive seizures, NCSE nonconvulsive status epilepticus, PEDs periodic epileptiform discharges, CI confidence interval, Univ. univariate analysis * Statistics of the multivariate model with three independent variables: likelihood ratio test (vs. constant model): G-statistic = 20.97;
OR
95 %CI
Without With
Multivariate analysis** Univ. p p
OR 95 %CI
21 (47) 0.17 33 (73) 0.004 0.0087 5.5 20 (44) 0.007 0.015 5.4
3.1–27.0 3.1–20.7
p \ 0.001 (df = 3) and Hosmer-Lemeshow test: 3.95; p = 0.89 (degrees of freedom = 8) ** Statistics of the multivariate model with two independent variables: likelihood ratio test (vs. constant model): G-statistic = 13.49; p = 0.001 (degrees of freedom = 2) and Hosmer-Lemeshow test: 2.02; p = 0.98 (degrees of freedom = 8)
Fig. 1 a A 70-year-old man admitted to the surgical intensive care unit after the pancreaticoduodenectomy for intraductal papillary mucinous neoplasm. Found to have generalized periodic discharges with superimposed fast frequencies on cEEG. b An 81-year-old man admitted to the surgical intensive care unit after endoscopic retrograde cholangiopancreatography complicated by perforation requiring a Roux-en-Y operation. Connected to cEEG for unresponsiveness demonstrating burst suppression
tended to be associated with seizures, but this was not significant [8/31 (26 %) vs. 16/113 (13 %) p = 0.10]. The presence of an acute brain lesion was not associated with the occurrence of PEDs (Table 2). The presence of chronic lesions was not associated with seizures [12/68 (18 %) vs. 11/76 (14 %); p = 0.65]. Sodium and creatinine levels and PaO2 did not differ between patients with and without NCSz or PEDs (Table 2). The cause of seizures, as determined by the consulting neurologist, was
Fig. 2 (Patient from Fig. 1, b) Evolving seizure with onset depicted on a and offset on d (a–c are consecutive pages of EEG, between c and d 2 min elapsed)
sepsis [N = 11/24 (46 %)], acute brain injury [N = 8/24 (33 %)], renal failure [N = 2/24 (8 %)] and unclear [N = 3/24 (12 %)].
232
Table 3 Clinical and EEG predictors of functional outcome at hospital discharge Good (n = 34)
Clinical variables Age [65 years Gender (female) Sepsis Coma Acute renal failure Acute liver failure Respiratory failure Acute brain injury Thrombocytopenia Circulatory shock Cardiac dysfunction Clinical seizures before cEEG EEG variables NCSz (incl. NCSE) PEDs Non reactive EEG
6 18 15 12 14 17 8 5 17 10 3 10
(18) (53) (44) (35) (41) (50) (24) (15) (50) (29) (9) (29)
1 (3) 5 (15) 3 (9)
Poor (n = 118)
67 73 83 72 85 71 29 28 89 85 48 41
(57) (62) (70) (61) (72) (60) (25) (25) (75) (72) (41) (35)
23 (20) 39 (33) 25 (24)
Univ.
Multivariate analysis*
p
OR
p
95 % CI
\0.001 0.35 0.005 0.008 \0.001 0.29 0.90 0.26 0.004 \0.001 0.001 0.23
10.9 – – – 2.8 –
0.001 – 0.43 0.45 0.045 –
3.4-35.0 – – – 1.1–7.9 –
3.8 3.4 – –
0.014 0.026 0.19 –
1.3–11.1 1.2–9.9 – –
0.02 0.04 0.13
10.4 –
0.039 0.47
1.0–53.7 –
Poor outcome was defined as death, vegetative state or severe disability at hospital discharge. Data reported as N (%) SOFA sequential organ failure assessment, ICU intensive care unit, NCSz nonconvulsive seizures, NCSE nonconvulsive status epilepticus, PEDs periodic epileptiform discharges, CI confidence interval, Univ. univariate analysis
* Statistics of the multivariate model with nine independent variables: likelihood ratio test (vs. constant model): G-statistic = 53.58; p \ 0.001 (degrees of freedom = 3) and Hosmer-Lemeshow test: 9.97; p = 0.27 (degrees of freedom = 8)
Organ failure, NCSz and outcome
included post-anoxic brain injury [N = 2/7 (29 %)], massive or multifocal ischemic stroke [N = 2/7 (29 %)], severe sepsis [N = 2/7 (29 %)] and fulminant liver failure [N = 1/7 (14 %)]. Of the 71 patients with a GOS 3–5 (i.e., better than vegetative state), detailed examination was available in 52 (73 %). Twenty-four patients (34 %) had signs of critical illness neuromyopathy. Seven had NCSz during cEEG, and among those seven patients, five (70 %) had signs of encephalopathy at discharge [vs. 10/55 (18 %) without NCSz; p = 0.007].
Overall, 77 % of patients had poor outcomes (Glasgow Outcome Scale score 1–3; death, vegetative state or severe disability) at hospital discharge. Clinical variables independently associated with poor outcome were age [65, acute renal failure, circulatory shock and thrombocytopenia (Table 3). In the same multivariate model, we found that NCSzs (including NCSE) were also independently associated with poor outcome [OR: 10.4 (95 % CI: 1.0–53.7)]; Table 3). Twenty-three out of 24 (96 %) patients with NCSz or NCSE had a poor outcome as opposed to 95/130 (73 %) of patients without these findings. PEDs were not associated with outcome in this analysis. The absence of EEG reactivity tended to be associated with a worse outcome, but this was not significant (25/118 vs. 3/34; p = 0.13). The cause of death could be identified in 67/73 (93 %) patients and was as follows: cardiac arrest [N = 11 (15 %)], refractory shock, [N = 5 (7 %)], acute respiratory distress syndrome [N = 5 (7 %)] and withdrawal of care, usually in the setting of refractory multiple organ failure [N = 46 (63 %)]. Patients with PEDs persisting for more than 24 consecutive hours during monitoring had a worse outcome than patients with transient PEDs [32/33 (97 %) vs. 7/12 (58 %); p = 0.01]. All patients with seizures except for one had poor outcome [23/24 (96 %)]. Seven patients were in a vegetative state when discharged from our institution. Causes of vegetative state
Treatment of clinical seizures and NCSz Fifty-four out of 154 (36 %) patients received antiepileptic drugs during their stay in the SICU. The following anticonvulsant drugs were used: levetiracetam [N = 37/154 (24 %)], phenytoin [N = 33/154 (20 %)], valproate [N = 5/154 (3 %)] and others [N = 27/154 (18 %)]. Forty-six out of 154 (30 %) were treated prior to cEEG for suspected clinical seizures, and all 24 patients with NCSz were treated. NCSzs were controlled in 14 of 24 (58 %) cases; 7 of 14 (50 %) patients died after seizures were controlled, while all 10 patients with uncontrolled seizures died. In those with uncontrolled seizures, the cause of death was withdrawal of care in the vast majority of cases (80 %) and multiple organ failure in the remaining 20 % of patients.
233
Discussion We studied a cohort of surgical ICU patients who underwent cEEG recordings for altered mental status to rule out subclinical seizures. We found that 29 % of patients developed periodic epileptiform discharges (PEDs), 16 % had electrographic seizures (NCSz), and 5 % had nonconvulsive status epilepticus (NCSE). Comatose patients were more likely to have NCSzs and PEDs, and NCSZs were also more common in patients with chronic hepatic failure. Sepsis and acute organ dysfunction were not predictors of electrographic findings. However, NCSzs were independently associated with poor outcome. Our cohort had a high level of severity of illness as evidenced by the prevalence of organ dysfunction. Although we do not have admission APACHE II scores, organ dysfunction status during cEEG recordings may reflect severity better than data from the first 24 h of the postoperative period; high severity of illness translated into poor outcome (death, vegetative state or severe disability) in 77 % of our cohort. Even in such a critically ill group, the development of NCSz (including NCSE) was an independent predictor of poor outcome. Periodic discharges were also associated with poor outcome, but only in univariate analysis. This is probably explained by the co-occurrence of NCSz in patients with PEDs. Other studies have analyzed the occurrence of NCSz and PEDs in comatose patients in the ICU. Our study is in agreement with the increasing number of reports that showed independent associations between NCSz and poor outcome. Twenty-three out of 24 (96 %) of patients with NCSz were severely disabled or worse on hospital discharge as compared to 95/130 (73 %) of patients without those findings (Table 3). NCSz have been described in critically ill patients without acute brain injury, but only a limited number of studies analyzed the occurrence of NCSz in patients admitted to non-neurological ICUs [7, 9]. One study that included patients with and without primary brain injuries found an 8 % occurrence rate of NCSz [7, 12]. A more recent study including only medical ICU patients without known acute brain injury found an incidence of electrographic seizures of 10 % [7]. Another recent study found an incidence of 11 % in a population of patients admitted to a medical or surgical ICU [9]. In this mixed cohort, only 12/105 patients were admitted to a SICU, and the overall severity of critical illness was lower than in our study, as indicated by a lower prevalence of sepsis, organ dysfunction and altered mental status. Also, the median duration of cEEG monitoring was 13 h, whereas it was more than 24 h for most patients in our cohort. This probably explains why we found a greater incidence (16 %) of NCSz. Previous data indeed suggested that sepsis was a significant risk factor for NCSz or PEDs [7]. In our cohort, all patients with NCSE had sepsis, but the association was not significant.
Protracted PEDs (lasting more than 24 h) were associated with a worse outcome than transient PEDs. This result might suggest that PEDs exert a negative effect on the brain and outcome. It is also very plausible that longer duration of PEDs merely reflects the severity of the underlying condition. Although it fails to reach significance, the absence of reactivity tended to be associated with a poor outcome in our study, similarly to recent studies in patients with anoxic brain injury [20]. Our population however differs from the post-anoxic patients because of the presence of multiple organ complications that also concur to a poor outcome. This may explain why the lack of EEG reactivity had a weaker predictive value in our study. Also, stimulation was often performed while patients received sedation, a likely confounder. We observed that coma and clinical seizures prior to cEEG, as previously shown [7, 11, 21], were associated with a higher risk of NCSz. We also found that chronic hepatic disease tended to be associated with NCSz. Although our results do not give insights into the mechanisms underlying this observation, patients with chronic hepatic dysfunction (many of whom had undergone liver transplants) were at high risk of experiencing acute liver and/or renal failure, infection and sepsis, electrolyte imbalances or adverse effects from drugs, all of which might cause seizures. Interestingly, patients receiving sedation were not at a lower risk of developing seizures, regardless of the type of sedation. This may be explained by the fact that some sedative drugs, such as dexmedetomidine, do not possess anticonvulsant effects and that the dose of the other drugs (midazolam, propofol) used for sedation was insufficient to prevent seizure activity. Assessment of sedative doses was outside of the scope of this study and should be studied in a prospective fashion. We did not find a significant association between acute brain injury or metabolic imbalances and NCSzs or PEDs, although some trends were present (e.g., seizures tended to be more frequent in patients with acute brain injury and creatinin levels, temperature and PaO2 tended to be higher in patients with PEDs). This is in contrast to a recent study that found that malignant EEG patterns were associated with acute ischemia in patients with sepsis [22]. It is likely that our study was underpowered to identify such associations, which should be further investigated, ideally in a prospective matched cohort study. Our study has a number of limitations. Similar to other studies on NCSz, ours has a major selection bias since only patients that had cEEG ordered by their clinicians were included. This limits the conclusions we can draw regarding the frequency of NCSz among surgical ICU patients and also compromises generalizability to other surgical populations. The retrospective nature of our study did not allow us to investigate in detail the influence
234
of progressive changes in organ function on the risk of seizures. Finally, the study group is very heterogeneous, with admission diagnosis ranging from orthopedic surgery to liver transplantation.
Conclusions
seizures and periodic epileptiform discharges occurred in 16 and 29 % of patients, respectively. They were more frequent in comatose patients and after clinical seizures. Nonconvulsive seizures showed strong independent association with death and severe disability at hospital discharge. Prospective studies are needed to determine the exact prevalence and clinical impact of seizures and to investigate whether treatment will improve patients’ outcomes.
In this retrospective study of critically ill surgical patients without primary acute brain injury but with altered mental Conflicts of interest None of the authors has any competing status who underwent cEEG monitoring, nonconvulsive interest to declare.
References 1. Claassen J, Mayer SA, Hirsch LJ (2005) Continuous EEG monitoring in patients with subarachnoid hemorrhage. J Clin Neurophysiol 22:92–98 2. Pandharipande P, Cotton BA, Shintani A et al (2007) Motoric subtypes of delirium in mechanically ventilated surgical and trauma intensive care unit patients. Intensive Care Med 33:1726–1731. doi: 10.1007/s00134-007-0687-y 3. Vespa PM, Miller C, McArthur D et al (2007) Nonconvulsive electrographic seizures after traumatic brain injury result in a delayed, prolonged increase in intracranial pressure and metabolic crisis. Crit Care Med 35:2830–2836. doi:10.1097/01.CCM.0000295667. 66853.BC 4. Pandharipande P, Cotton BA, Shintani A et al (2008) Prevalence and risk factors for development of delirium in surgical and trauma intensive care unit patients. J Trauma 65:34–41. doi: 10.1097/TA.0b013e31814b2c4d 5. Claassen J, Jette´ N, Chum F et al (2007) Electrographic seizures and periodic discharges after intracerebral haemorrhage. Neurology 69:1356–1365. doi: 10.1212/01.wnl.0000281664.02615.6c 6. Carrera E, Claassen J, Oddo M et al (2008) Continuous electroencephalographic monitoring in critically ill patients with central nervous system infections. Arch Neurol 65:1612–1618. doi: 10.1001/archneur.65.12.1612 7. Oddo M, Carrera E, Claassen J et al (2009) Continuous electroencephalography in the medical intensive care unit. Crit Care Med 37:2051–2056. doi:10.1097/CCM. 0b013e3181a00604
8. Lat I, McMillian W, Taylor S et al (2009) The impact of delirium on clinical outcomes in mechanically ventilated surgical and trauma patients. Crit Care Med 37:1898–1905. doi: 10.1097/CCM.0b013e31819ffe38 9. Kamel H, Betjemann JP, Navi BB et al (2012) Diagnostic yield of electroencephalography in the medical and surgical intensive care unit. Neurocrit Care. doi: 10.1007/s12028-012-9736-7 10. Brown CVRC, Daigle JBJ, Foulkrod KHK et al (2011) Risk factors associated with early reintubation in trauma patients: a prospective observational study. J Trauma 71:32–37. doi: 10.1097/TA.0b013e31821e0c6e 11. Claassen J, Mayer SA, Kowalski RG et al (2004) Detection of electrographic seizures with continuous EEG monitoring in critically ill patients. Neurology 62:1743–1748 12. Young GB, Jordan KG, Doig GS (1996) An assessment of nonconvulsive seizures in the intensive care unit using continuous EEG monitoring: an investigation of variables associated with mortality. Neurology 47:83–89 13. Towne AR, Waterhouse EJ, Boggs JG et al (2000) Prevalence of nonconvulsive status epilepticus in comatose patients. Neurology 54:340–345 14. Vespa PM, O’Phelan K, Shah M et al (2003) Acute seizures after intracerebral haemorrhage: a factor in progressive midline shift and outcome. Neurology 60:1441–1446 15. Hirsch LJ, Claassen J, Mayer SA, Emerson RG (2004) Stimulus-induced rhythmic, periodic, or ictal discharges (SIRPIDs): a common EEG phenomenon in the critically ill. Epilepsia 45:109–123
16. Chong DJ, Hirsch LJ (2005) Which EEG patterns warrant treatment in the critically ill? Reviewing the evidence for treatment of periodic epileptiform discharges and related patterns. J Clin Neurophysiol 22:79–91 17. Vincent JLJ, Moreno RR, Takala JJ et al (1996) The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the working group on sepsisrelated problems of the European society of intensive care medicine. Intensive Care Med 22:707–710 18. Claassen J, Taccone FS, Horn P et al (2013) Recommendations on the use of EEG monitoring in critically ill patients: consensus statement from the neurointensive care section of the ESICM. Intensive Care Med. doi: 10.1007/s00134-013-2938-4 19. Jennett B, Bond M (1975) Assessment of outcome after severe brain damage. Lancet 1:480–484. doi: 10.1016/S0140-6736(75)92830-5 20. Rossetti AO, Oddo M, Logroscino G, Kaplan PW (2010) Prognostication after cardiac arrest and hypothermia: a prospective study. Ann Neurol 67:301–307. doi:10.1002/ana.21984 21. Jette N, Claassen J, Emerson RG, Hirsch LJ (2006) Frequency and predictors of nonconvulsive seizures during continuous electroencephalographic monitoring in critically ill children. Arch Neurol 63:1750–1755. doi: 10.1001/archneur.63.12.1750 22. Polito A, Eischwald F, Le Maho A-L et al (2013) Pattern of brain injury in the acute setting of human septic shock. Crit Care 17:R204. doi:10.1186/ cc12899
Pedro Kurtz Nicolas Gaspard Anna Sophia Wahl Rebecca Marie Bauer Lawrence J. Hirsch Hannah Wunsch Jan Claassen
Received: 10 July 2013 Accepted: 27 October 2013 Published online: 16 November 2013 Ó Springer-Verlag Berlin Heidelberg and ESICM 2013 Take-home message: Nonconvulsive seizures occur in 16 % of patients admitted to a surgical ICU for post-operative care with altered mental status and are associated with poor outcome at discharge. Risk factors include coma and clinical seizures prior to cEEG monitoring. Electronic supplementary material The online version of this article (doi:10.1007/s00134-013-3149-8) contains supplementary material, which is available to authorized users.
ORIGINAL ARTICLE
Continuous electroencephalography in a surgical intensive care unit
R. M. Bauer Department of Anesthesiology, Feinberg School of Medicine, Northwestern University, Chicago, USA H. Wunsch Division of Critical Care, Department of Anesthesiology, Columbia University, College of Physicians and Surgeons, New York, NY, USA H. Wunsch Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, NY, USA
Abstract Purpose: Our aim was to investigate the prevalence, risk factors, and impact on outcome of nonconvulsive seizures (NCSz), nonconvulsive status epilepticus (NCSE), and periodic epileptiform discharges (PEDs) in surgical intensive care unit (SICU) patients with continuous electroencephalography (cEEG) P. Kurtz A. S. Wahl J. Claassen ()) Division of Critical Care Neurology, monitoring. Methods: This was a Department of Neurology, Columbia retrospective study of SICU patients University, College of Physicians and who underwent cEEG monitoring for Surgeons, New York, NY, USA altered mental status over a 6-year e-mail: [email protected] period. We report the frequency of NCSz (including NCSE) and PEDs P. Kurtz on cEEG. The primary outcome was Department of Intensive Care Medicine, Clı´nica Sa˜o Vicente, Rua Joa˜o Borges 204, death or severe disability at hospital Gavea, Rio de Janeiro, RJ 22451-100, discharge. Multivariable logistic Brazil regression was used to identify whether NCSz (including NCSE) and N. Gaspard L. J. Hirsch Department of Neurology, Comprehensive PEDs were independently associated with poor outcome (death, vegetative Epilepsy Center, Yale University, state or severe disability). Results: New Haven, CT, USA
Of 154 patients, the mean age was 64 ± 14 years old, and 40 % were women. The majority of patients were admitted following abdominal surgery (36 %) and liver transplantation (24 %). Sepsis developed in 100 (65 %) patients. Sixteen percent (n = 24) had NCSz [including 5 % (N = 8) with NCSE], and 29 % (N = 45) had PEDs. All eight patients with NCSE were septic. Clinical seizures prior to cEEG and coma were more common among patients who developed NCSz or NCSE compared to patients without NCSz or NCSE (70 vs. 27 %; p \ 0.01; 75 vs. 52 %; p = 0.046 and 63 vs. 34 %; p = 0.09, respectively). NCSzs (including NCSE) were independently associated with poor outcome (20 vs. 3 %, OR 10.4, 95 % CI 1.0–53.7; p = 0.039). Conclusion: In this retrospective study of SICU patients with cEEG monitoring for altered mental status, NCSz and periodic discharges were frequent and NCSz were independently associated with poor outcome. NCSz were more common when clinical seizures occurred before cEEG. Keywords Nonconvulsive status epilepticus Nonconvulsive seizures Electrographic seizures Periodic discharges Surgical ICU Outcome
229
Introduction
Table 1 Patient characteristics (n = 154)
Acute neurologic dysfunction, including delirium and coma, is common in the acute postoperative period for patients cared for in surgical and trauma intensive care units (ICUs) [1–6] and is associated with worse outcome [7–10]. A subgroup of critically ill patients with impaired mental status, often lumped under the diagnoses of metabolic encephalopathy or delirium, may have ongoing periodic epileptiform discharges (PEDs) and nonconvulsive seizures (NCSz), including nonconvulsive status epilepticus (NCSE), potentially necessitating a change in treatment approach [7, 11]. Seizures and other highly epileptiform patterns are common in ICU patients with acute brain injury [1, 3, 5–7, 11–14], as well as those who receive continuous electroencephalography (cEEG) monitoring without acute brain injury in medical ICUs [7, 9, 11, 15, 16]. These seizures are almost always nonconvulsive and therefore are not detectable without EEG monitoring [7, 11, 12, 16]; they are associated with poor outcome [3, 5, 7, 11–14, 17]. Surgical ICU (SICU) patients, who are primarily admitted from the operating room, are fundamentally different from patients encountered in medical ICUs. Although recommendations for the use of cEEG in critically ill patients exist [18], there is little information available on the utility of cEEG monitoring in SICU patients and on the prognostic significance of EEG findings in this population. Therefore, we studied the prevalence, predictors, and impact on outcome of PEDs, NCSz and NCSE in surgical ICU (SICU) patients without primary acute brain injury.
Age; median years (IQR) Female n (%) Primary admission reason n (%) Transplant Other abdominal Orthopedic/trauma Cardiovascular/thoracic Other Sepsis on admission Past medical history n (%) Epilepsy Stroke Diabetes Chronic renal dysfunction Cardiac disease Chronic liver dysfunction Cancer cEEG monitoring Time from admission to cEEG; days; median (IQR) Length of monitoring; days; median (IQR) Clinical status at time of cEEG n (%) GCS: \8/9–12/13–14/15
Methods Study population Of a total of *5,800 patients admitted to the SICU at Columbia University Medical Center from August 2004 until July 2010, we identified 154 consecutive patients who underwent continuous EEG (cEEG) for a minimum of 12 h and had an indication for cEEG monitoring of unexplained altered mental status (supplementary Table 1 and supplementary information on methods). Patients were identified from the Comprehensive Epilepsy Center electronic log and cEEG database (please refer to the supplementary information for details on the study population, cEEG and clinical variables). Continuous EEG was requested by a neurology consultant in 87 % of cases. The patients’ neurologic status (comatose or not comatose) at the time cEEG was initiated was recorded from documentation by the neurology consultant or by the intensivist. All patients were admitted for postoperative care after a variety of procedures (see Table 1). Patients with a primary acute neurological diagnosis or those that underwent neurosurgical interventions were excluded
Focal neurological signs* Length of stay in ICU/in hospital (days; median [range]) Outcome at hospital discharge: dead/ persistent vegetative/severe disability/ moderate disability/good recovery) n (%)**
65 (54–74) 62 (40) 44 56 15 29 12 58
(29) (36) (9) (19) (8) (37)
18 55 59 61 73 60 55
(12) (35) (38) (39) (47) (39) (36)
6 (4–9) 2 (1–2.5) 86 (55)/35 (23)/30 (19)/3 (2) 17(12) 22 (1–167)/52 (10–201) 73 (47)/7 (5)/37 (24)/ 27 (17)/8 (5)
cEEG continuous EEG, ICU intensive care unit, SD standard deviation Data available in * N = 144 and ** N = 152 patients respectively
from this analysis. Approval for this study was received from the Columbia University Medical Center Hospital Institutional Review Board (IRB).
cEEG recordings and variables Continuous EEG was recorded using 21 electrodes placed according to the international 10–20 system and interpreted by a board-certified electroencephalographer. We recorded the occurrence of convulsive seizures (CSz), NCSz, convulsive status epilepticus (CSE), NCSE and PEDs [including generalized periodic epileptiform discharges (GPEDs), periodic lateralized epileptiform discharges (PLEDs), triphasic waves and stimulusinduced periodic, rhythmic and ictal appearing discharges (SIRPIDs)] [7, 9, 11, 12, 15, 16, 19]. Clinical variables Data were retrieved via cEEG reports, medical records including discharge summaries and physician sign-out
230
notes. Details of the collected clinical variables are out- Brain imaging lined in the supplementary material. Our primary outcome was functional outcome at hospital discharge as assessed Brain imaging was performed within 24 h of cEEG in 142 patients [all patients received a CT scan and 34 (22 %) using the Glasgow Outcome Scale [7, 19]. also received an MRI]; 23 scans showed acute ischemic lesions (12 territorial infarcts, of which 4 were clinically silent, and 12 multifocal punctate ischemic lesions, all Statistical analysis clinically silent), 6 showed intraparenchymal hemorrhage Data were analyzed with SPSS 16.0 software (Chicago, IL), (of which 3 were clinically silent), and 4 showed radiousing Student’s t test for continuous variables and v2 test for logical signs of posterior reversible encephalopathy categorical variables. Univariate analysis was performed to syndrome (PRES). Signs of chronic or remote injury were identify significant predictors of NCSz (including NCSE) seen in 68 patients. and PEDs. Significant variables at the level of p \ 0.1 were then included in a multivariable logistic regression model to identify independent predictors. The relationship CSF studies between outcome and the presence of NCSz (including NCSE) and the presence of PEDs was analyzed indepen- CSF was analyzed in 24/154 (16 %) patients. An abnordently using a multivariable logistic regression model. A mal protein level ([50 mg/dl) was found in 5/24 (21 %) cases, and an abnormal cell count ([5 cells/ml) was found p value \0.05 was considered significant. in 14/24 (58 %) cases. Glucose levels were in the normal range in all cases, and bacterial cultures were negative.
Results Patient characteristics From August 2004 to July 2010, 154 SICU patients (2.7 %) underwent cEEG monitoring. Median time from SICU admission to initiation of cEEG was 6 days (IQR: 4–9 days), and median duration of cEEG was 2 days (IQR: 1–2.5 days) (Table 1). Most patients (65 %) were admitted after undergoing abdominal surgery, including liver and kidney transplantation (24 and 5 % of total, respectively). Fifty-eight patients (37 %) had sepsis on admission, and 95 % were mechanically ventilated. Half of patients (55 %) were comatose (GCS B8), and 12 % had focal neurological signs prior to cEEG. The etiology of the alteration of consciousness (GCS \15; N = 151), as determined by the clinical team, was NCSE or NCSZ [N = 10 (7 %)], sepsis associated encephalopathy [N = 65 (43 %)], ICU delirium [N = 33 (22 %)], metabolic encephalopathy [N = 28 (19 %)] and acute brain injury [N = 18 (12 %)]. Main complications during the SICU stay involved organ failure, including renal dysfunction and circulatory shock (34 and 63 % of the total cohort, respectively). Clinical seizures occurred in 51/154 patients (33 %) prior to cEEG. Sedation One hundred six (70 %) patients were sedated at the time of cEEG. Propofol [N = 49 (32 %)], midazolam [N = 14 (9 %)], dexmedetomidine [N = 34 (22 %)], and a combination of midazolam and propofol [N = 11 (7 %)] were used.
Continuous EEG findings and predictors of NCSz and PEDs Continuous EEG was performed because of unexplained alteration in mental status in 133/154 (86 %) patients and because of suspected clinical seizure activity in 56/154 patients (36 %). A total of 24 of 154 patients (16 %) had NCSz, including 8 (5 %) with NCSE, and 45 (29 %) developed PEDs during cEEG. Most patients with NCSz [21/24; (88 %)] also had PEDs. After multivariate regression, clinical seizures before cEEG and coma were more common among patients with NCSz (including NCSE) and PEDs (Table 2 and Supplementary Table 1; Figs. 1, 2). Chronic hepatic failure was more common among patients with NCSz in univariate analysis but not after multivariate analysis. Admission diagnosis and the development of sepsis did not differ between patients with NCSz or PEDs and those without these EEG findings. All eight patients who developed NCSE had sepsis. The proportion of patients on sedation during cEEG did not differ significantly between patients with NCSz and those without NCSz [15/24 (63 %) vs. 92/130 (71 %); p = 0.40]. We found no difference in the incidence of NCSz between the type of sedative drug used [3/ 33 (9 %), 6/31 (19 %) and 11/60 (18 %), respectively for dexmedetomidine, midazolam and propofol; p = 0.30]. Of 23 patients with acute ischemic lesions, 5 had seizures [5/23 (22 %) vs. 19/103 (17 %) without ischemic lesions; p = 0.41]; of 6 patients with hemorrhage, 1 had seizures [1/6 (17 %) vs. 23/138 (19 %) without hemorrhagic lesions; p = 0.96]; of 4 patients with PRES, 2 had seizures [2/2 (50 %) vs. 22/142 (15 %) without PRES; p = 0.13]. Overall, the presence of an acute brain lesion
231
Table 2 Predictors of NCSz (including NCSE) and (PEDs) (variables significant in univariate analysis only) NCSz (including NCSE; n = 24) Multivariate analysis* Without Chronic hepatic failure 44 (34) Coma 68 (52) Clinical seizures 35 (27) prior to cEEG
PEDs (n = 45)
With
Univ. p
p
15 (63) 18 (75) 16 (70)
0.007 0.036 \0.001
0.09 – – 38 (35) 0.046 4.1 2.8–7.6 53 (48) \0.001 10.6 4.9–76.5 31 (28)
Data reported as N (%) or median (IQR) NCSz nonconvulsive seizures, NCSE nonconvulsive status epilepticus, PEDs periodic epileptiform discharges, CI confidence interval, Univ. univariate analysis * Statistics of the multivariate model with three independent variables: likelihood ratio test (vs. constant model): G-statistic = 20.97;
OR
95 %CI
Without With
Multivariate analysis** Univ. p p
OR 95 %CI
21 (47) 0.17 33 (73) 0.004 0.0087 5.5 20 (44) 0.007 0.015 5.4
3.1–27.0 3.1–20.7
p \ 0.001 (df = 3) and Hosmer-Lemeshow test: 3.95; p = 0.89 (degrees of freedom = 8) ** Statistics of the multivariate model with two independent variables: likelihood ratio test (vs. constant model): G-statistic = 13.49; p = 0.001 (degrees of freedom = 2) and Hosmer-Lemeshow test: 2.02; p = 0.98 (degrees of freedom = 8)
Fig. 1 a A 70-year-old man admitted to the surgical intensive care unit after the pancreaticoduodenectomy for intraductal papillary mucinous neoplasm. Found to have generalized periodic discharges with superimposed fast frequencies on cEEG. b An 81-year-old man admitted to the surgical intensive care unit after endoscopic retrograde cholangiopancreatography complicated by perforation requiring a Roux-en-Y operation. Connected to cEEG for unresponsiveness demonstrating burst suppression
tended to be associated with seizures, but this was not significant [8/31 (26 %) vs. 16/113 (13 %) p = 0.10]. The presence of an acute brain lesion was not associated with the occurrence of PEDs (Table 2). The presence of chronic lesions was not associated with seizures [12/68 (18 %) vs. 11/76 (14 %); p = 0.65]. Sodium and creatinine levels and PaO2 did not differ between patients with and without NCSz or PEDs (Table 2). The cause of seizures, as determined by the consulting neurologist, was
Fig. 2 (Patient from Fig. 1, b) Evolving seizure with onset depicted on a and offset on d (a–c are consecutive pages of EEG, between c and d 2 min elapsed)
sepsis [N = 11/24 (46 %)], acute brain injury [N = 8/24 (33 %)], renal failure [N = 2/24 (8 %)] and unclear [N = 3/24 (12 %)].
232
Table 3 Clinical and EEG predictors of functional outcome at hospital discharge Good (n = 34)
Clinical variables Age [65 years Gender (female) Sepsis Coma Acute renal failure Acute liver failure Respiratory failure Acute brain injury Thrombocytopenia Circulatory shock Cardiac dysfunction Clinical seizures before cEEG EEG variables NCSz (incl. NCSE) PEDs Non reactive EEG
6 18 15 12 14 17 8 5 17 10 3 10
(18) (53) (44) (35) (41) (50) (24) (15) (50) (29) (9) (29)
1 (3) 5 (15) 3 (9)
Poor (n = 118)
67 73 83 72 85 71 29 28 89 85 48 41
(57) (62) (70) (61) (72) (60) (25) (25) (75) (72) (41) (35)
23 (20) 39 (33) 25 (24)
Univ.
Multivariate analysis*
p
OR
p
95 % CI
\0.001 0.35 0.005 0.008 \0.001 0.29 0.90 0.26 0.004 \0.001 0.001 0.23
10.9 – – – 2.8 –
0.001 – 0.43 0.45 0.045 –
3.4-35.0 – – – 1.1–7.9 –
3.8 3.4 – –
0.014 0.026 0.19 –
1.3–11.1 1.2–9.9 – –
0.02 0.04 0.13
10.4 –
0.039 0.47
1.0–53.7 –
Poor outcome was defined as death, vegetative state or severe disability at hospital discharge. Data reported as N (%) SOFA sequential organ failure assessment, ICU intensive care unit, NCSz nonconvulsive seizures, NCSE nonconvulsive status epilepticus, PEDs periodic epileptiform discharges, CI confidence interval, Univ. univariate analysis
* Statistics of the multivariate model with nine independent variables: likelihood ratio test (vs. constant model): G-statistic = 53.58; p \ 0.001 (degrees of freedom = 3) and Hosmer-Lemeshow test: 9.97; p = 0.27 (degrees of freedom = 8)
Organ failure, NCSz and outcome
included post-anoxic brain injury [N = 2/7 (29 %)], massive or multifocal ischemic stroke [N = 2/7 (29 %)], severe sepsis [N = 2/7 (29 %)] and fulminant liver failure [N = 1/7 (14 %)]. Of the 71 patients with a GOS 3–5 (i.e., better than vegetative state), detailed examination was available in 52 (73 %). Twenty-four patients (34 %) had signs of critical illness neuromyopathy. Seven had NCSz during cEEG, and among those seven patients, five (70 %) had signs of encephalopathy at discharge [vs. 10/55 (18 %) without NCSz; p = 0.007].
Overall, 77 % of patients had poor outcomes (Glasgow Outcome Scale score 1–3; death, vegetative state or severe disability) at hospital discharge. Clinical variables independently associated with poor outcome were age [65, acute renal failure, circulatory shock and thrombocytopenia (Table 3). In the same multivariate model, we found that NCSzs (including NCSE) were also independently associated with poor outcome [OR: 10.4 (95 % CI: 1.0–53.7)]; Table 3). Twenty-three out of 24 (96 %) patients with NCSz or NCSE had a poor outcome as opposed to 95/130 (73 %) of patients without these findings. PEDs were not associated with outcome in this analysis. The absence of EEG reactivity tended to be associated with a worse outcome, but this was not significant (25/118 vs. 3/34; p = 0.13). The cause of death could be identified in 67/73 (93 %) patients and was as follows: cardiac arrest [N = 11 (15 %)], refractory shock, [N = 5 (7 %)], acute respiratory distress syndrome [N = 5 (7 %)] and withdrawal of care, usually in the setting of refractory multiple organ failure [N = 46 (63 %)]. Patients with PEDs persisting for more than 24 consecutive hours during monitoring had a worse outcome than patients with transient PEDs [32/33 (97 %) vs. 7/12 (58 %); p = 0.01]. All patients with seizures except for one had poor outcome [23/24 (96 %)]. Seven patients were in a vegetative state when discharged from our institution. Causes of vegetative state
Treatment of clinical seizures and NCSz Fifty-four out of 154 (36 %) patients received antiepileptic drugs during their stay in the SICU. The following anticonvulsant drugs were used: levetiracetam [N = 37/154 (24 %)], phenytoin [N = 33/154 (20 %)], valproate [N = 5/154 (3 %)] and others [N = 27/154 (18 %)]. Forty-six out of 154 (30 %) were treated prior to cEEG for suspected clinical seizures, and all 24 patients with NCSz were treated. NCSzs were controlled in 14 of 24 (58 %) cases; 7 of 14 (50 %) patients died after seizures were controlled, while all 10 patients with uncontrolled seizures died. In those with uncontrolled seizures, the cause of death was withdrawal of care in the vast majority of cases (80 %) and multiple organ failure in the remaining 20 % of patients.
233
Discussion We studied a cohort of surgical ICU patients who underwent cEEG recordings for altered mental status to rule out subclinical seizures. We found that 29 % of patients developed periodic epileptiform discharges (PEDs), 16 % had electrographic seizures (NCSz), and 5 % had nonconvulsive status epilepticus (NCSE). Comatose patients were more likely to have NCSzs and PEDs, and NCSZs were also more common in patients with chronic hepatic failure. Sepsis and acute organ dysfunction were not predictors of electrographic findings. However, NCSzs were independently associated with poor outcome. Our cohort had a high level of severity of illness as evidenced by the prevalence of organ dysfunction. Although we do not have admission APACHE II scores, organ dysfunction status during cEEG recordings may reflect severity better than data from the first 24 h of the postoperative period; high severity of illness translated into poor outcome (death, vegetative state or severe disability) in 77 % of our cohort. Even in such a critically ill group, the development of NCSz (including NCSE) was an independent predictor of poor outcome. Periodic discharges were also associated with poor outcome, but only in univariate analysis. This is probably explained by the co-occurrence of NCSz in patients with PEDs. Other studies have analyzed the occurrence of NCSz and PEDs in comatose patients in the ICU. Our study is in agreement with the increasing number of reports that showed independent associations between NCSz and poor outcome. Twenty-three out of 24 (96 %) of patients with NCSz were severely disabled or worse on hospital discharge as compared to 95/130 (73 %) of patients without those findings (Table 3). NCSz have been described in critically ill patients without acute brain injury, but only a limited number of studies analyzed the occurrence of NCSz in patients admitted to non-neurological ICUs [7, 9]. One study that included patients with and without primary brain injuries found an 8 % occurrence rate of NCSz [7, 12]. A more recent study including only medical ICU patients without known acute brain injury found an incidence of electrographic seizures of 10 % [7]. Another recent study found an incidence of 11 % in a population of patients admitted to a medical or surgical ICU [9]. In this mixed cohort, only 12/105 patients were admitted to a SICU, and the overall severity of critical illness was lower than in our study, as indicated by a lower prevalence of sepsis, organ dysfunction and altered mental status. Also, the median duration of cEEG monitoring was 13 h, whereas it was more than 24 h for most patients in our cohort. This probably explains why we found a greater incidence (16 %) of NCSz. Previous data indeed suggested that sepsis was a significant risk factor for NCSz or PEDs [7]. In our cohort, all patients with NCSE had sepsis, but the association was not significant.
Protracted PEDs (lasting more than 24 h) were associated with a worse outcome than transient PEDs. This result might suggest that PEDs exert a negative effect on the brain and outcome. It is also very plausible that longer duration of PEDs merely reflects the severity of the underlying condition. Although it fails to reach significance, the absence of reactivity tended to be associated with a poor outcome in our study, similarly to recent studies in patients with anoxic brain injury [20]. Our population however differs from the post-anoxic patients because of the presence of multiple organ complications that also concur to a poor outcome. This may explain why the lack of EEG reactivity had a weaker predictive value in our study. Also, stimulation was often performed while patients received sedation, a likely confounder. We observed that coma and clinical seizures prior to cEEG, as previously shown [7, 11, 21], were associated with a higher risk of NCSz. We also found that chronic hepatic disease tended to be associated with NCSz. Although our results do not give insights into the mechanisms underlying this observation, patients with chronic hepatic dysfunction (many of whom had undergone liver transplants) were at high risk of experiencing acute liver and/or renal failure, infection and sepsis, electrolyte imbalances or adverse effects from drugs, all of which might cause seizures. Interestingly, patients receiving sedation were not at a lower risk of developing seizures, regardless of the type of sedation. This may be explained by the fact that some sedative drugs, such as dexmedetomidine, do not possess anticonvulsant effects and that the dose of the other drugs (midazolam, propofol) used for sedation was insufficient to prevent seizure activity. Assessment of sedative doses was outside of the scope of this study and should be studied in a prospective fashion. We did not find a significant association between acute brain injury or metabolic imbalances and NCSzs or PEDs, although some trends were present (e.g., seizures tended to be more frequent in patients with acute brain injury and creatinin levels, temperature and PaO2 tended to be higher in patients with PEDs). This is in contrast to a recent study that found that malignant EEG patterns were associated with acute ischemia in patients with sepsis [22]. It is likely that our study was underpowered to identify such associations, which should be further investigated, ideally in a prospective matched cohort study. Our study has a number of limitations. Similar to other studies on NCSz, ours has a major selection bias since only patients that had cEEG ordered by their clinicians were included. This limits the conclusions we can draw regarding the frequency of NCSz among surgical ICU patients and also compromises generalizability to other surgical populations. The retrospective nature of our study did not allow us to investigate in detail the influence
234
of progressive changes in organ function on the risk of seizures. Finally, the study group is very heterogeneous, with admission diagnosis ranging from orthopedic surgery to liver transplantation.
Conclusions
seizures and periodic epileptiform discharges occurred in 16 and 29 % of patients, respectively. They were more frequent in comatose patients and after clinical seizures. Nonconvulsive seizures showed strong independent association with death and severe disability at hospital discharge. Prospective studies are needed to determine the exact prevalence and clinical impact of seizures and to investigate whether treatment will improve patients’ outcomes.
In this retrospective study of critically ill surgical patients without primary acute brain injury but with altered mental Conflicts of interest None of the authors has any competing status who underwent cEEG monitoring, nonconvulsive interest to declare.
References 1. Claassen J, Mayer SA, Hirsch LJ (2005) Continuous EEG monitoring in patients with subarachnoid hemorrhage. J Clin Neurophysiol 22:92–98 2. Pandharipande P, Cotton BA, Shintani A et al (2007) Motoric subtypes of delirium in mechanically ventilated surgical and trauma intensive care unit patients. Intensive Care Med 33:1726–1731. doi: 10.1007/s00134-007-0687-y 3. Vespa PM, Miller C, McArthur D et al (2007) Nonconvulsive electrographic seizures after traumatic brain injury result in a delayed, prolonged increase in intracranial pressure and metabolic crisis. Crit Care Med 35:2830–2836. doi:10.1097/01.CCM.0000295667. 66853.BC 4. Pandharipande P, Cotton BA, Shintani A et al (2008) Prevalence and risk factors for development of delirium in surgical and trauma intensive care unit patients. J Trauma 65:34–41. doi: 10.1097/TA.0b013e31814b2c4d 5. Claassen J, Jette´ N, Chum F et al (2007) Electrographic seizures and periodic discharges after intracerebral haemorrhage. Neurology 69:1356–1365. doi: 10.1212/01.wnl.0000281664.02615.6c 6. Carrera E, Claassen J, Oddo M et al (2008) Continuous electroencephalographic monitoring in critically ill patients with central nervous system infections. Arch Neurol 65:1612–1618. doi: 10.1001/archneur.65.12.1612 7. Oddo M, Carrera E, Claassen J et al (2009) Continuous electroencephalography in the medical intensive care unit. Crit Care Med 37:2051–2056. doi:10.1097/CCM. 0b013e3181a00604
8. Lat I, McMillian W, Taylor S et al (2009) The impact of delirium on clinical outcomes in mechanically ventilated surgical and trauma patients. Crit Care Med 37:1898–1905. doi: 10.1097/CCM.0b013e31819ffe38 9. Kamel H, Betjemann JP, Navi BB et al (2012) Diagnostic yield of electroencephalography in the medical and surgical intensive care unit. Neurocrit Care. doi: 10.1007/s12028-012-9736-7 10. Brown CVRC, Daigle JBJ, Foulkrod KHK et al (2011) Risk factors associated with early reintubation in trauma patients: a prospective observational study. J Trauma 71:32–37. doi: 10.1097/TA.0b013e31821e0c6e 11. Claassen J, Mayer SA, Kowalski RG et al (2004) Detection of electrographic seizures with continuous EEG monitoring in critically ill patients. Neurology 62:1743–1748 12. Young GB, Jordan KG, Doig GS (1996) An assessment of nonconvulsive seizures in the intensive care unit using continuous EEG monitoring: an investigation of variables associated with mortality. Neurology 47:83–89 13. Towne AR, Waterhouse EJ, Boggs JG et al (2000) Prevalence of nonconvulsive status epilepticus in comatose patients. Neurology 54:340–345 14. Vespa PM, O’Phelan K, Shah M et al (2003) Acute seizures after intracerebral haemorrhage: a factor in progressive midline shift and outcome. Neurology 60:1441–1446 15. Hirsch LJ, Claassen J, Mayer SA, Emerson RG (2004) Stimulus-induced rhythmic, periodic, or ictal discharges (SIRPIDs): a common EEG phenomenon in the critically ill. Epilepsia 45:109–123
16. Chong DJ, Hirsch LJ (2005) Which EEG patterns warrant treatment in the critically ill? Reviewing the evidence for treatment of periodic epileptiform discharges and related patterns. J Clin Neurophysiol 22:79–91 17. Vincent JLJ, Moreno RR, Takala JJ et al (1996) The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the working group on sepsisrelated problems of the European society of intensive care medicine. Intensive Care Med 22:707–710 18. Claassen J, Taccone FS, Horn P et al (2013) Recommendations on the use of EEG monitoring in critically ill patients: consensus statement from the neurointensive care section of the ESICM. Intensive Care Med. doi: 10.1007/s00134-013-2938-4 19. Jennett B, Bond M (1975) Assessment of outcome after severe brain damage. Lancet 1:480–484. doi: 10.1016/S0140-6736(75)92830-5 20. Rossetti AO, Oddo M, Logroscino G, Kaplan PW (2010) Prognostication after cardiac arrest and hypothermia: a prospective study. Ann Neurol 67:301–307. doi:10.1002/ana.21984 21. Jette N, Claassen J, Emerson RG, Hirsch LJ (2006) Frequency and predictors of nonconvulsive seizures during continuous electroencephalographic monitoring in critically ill children. Arch Neurol 63:1750–1755. doi: 10.1001/archneur.63.12.1750 22. Polito A, Eischwald F, Le Maho A-L et al (2013) Pattern of brain injury in the acute setting of human septic shock. Crit Care 17:R204. doi:10.1186/ cc12899
