early prognosis in coma after cardiac arrest: a prospective clinical

60ournal of Neurology, Neurosurgery, and Psychiatry 1996;61:610-615

Early prognosis in coma after cardiac arrest: a

prospective clinical, electrophysiological, andbiochemical study of 60 patients

C Bassetti, Fulvio Bomio, Johannes Mathis, Christian W Hess

AbstractBackground-The univariate study ofclinical, electrophysiological, or bio-chemical variables has been shown topredict the outcome in postanoxic comain about 50% of patients for each type ofvariable. Previous studies did not, how-ever, consider the prognostic accuracy ofa multivariate approach.Methods-Sixty patients in coma formore than six hours after cardiac arrestwere prospectively examined by means ofrepeated clinical examinations (includingGlasgow coma score (GCS)), EEG, andmedianus nerve somatosensory evokedpotentials (SEPs). In 16 patients, theearly concentrations of serum neuronspecific enolase and ionised calcium werealso measured.Results-Within the first year after car-diac arrest, 20% of patients made a goodneurological recovery; 80% remained in avegetative state or died. Clinical exami-nation correctly predicted outcome in58% of patients, SEP in 59%, and EEG in41%. The combination of clinical exami-nation, SEP, and EEG raised the percent-age of correct predictions to 82%, withoutfalse pessimistic predictions. Concen-trations of serum neuron specific enolaseand ionised calcium were of no additionalprognostic help. Multivariate regressionanalysis identified the association of GCS< 8 at 48 hours with abnormal or absentearly cortical SEPs as highly predictive ofa bad outcome (risk = 97%, 95% confi-dence interval = 86-99%).Conclusion-The combination of GCS at48 hours, SEP, and if these are non-conclusive, EEG, permits a more reliableprediction ofoutcome in postanoxic comathan clinical examination alone.

Department ofNeurologyC BassettiJ MathisCW HessDepartment ofMedicine, UniversityHospital (Inselspital)Bern, SwitzerlandF BomioCorrespondence to:Dr C Bassetti, Departmentof Neurology, UniversityHospital (Inselspital) 3010Bern, Switzerland.

Received 29 January 1996and in revised form13 August 1996Accepted 19 August 1996

(J Neurol Neurosurg Psychiatry 1996;61:610-615)

Keywords: cardiac arrest; postanoxic coma; Glasgowcoma scale; electroencephalography; somatosensoryevoked potentials

Up to 44% of patients with cardiac arrest can

be resuscitated successfully.' Most remaincomatose initially and only 10%-20% make a

good recovery. '3 Prediction of individual out-come is a difficult task with important med-ical, ethical, and socioeconomic implications.4As the duration and severity of cerebralischaemia are usually unknown or difficult to

estimate, neuronal damage must be assessedindirectly by exploring cerebral functions andevaluating their pattern of recovery over time.In general, the longer the delay in recovery offunction the worse the prognosis. Due to thegreater vulnerability of the cerebral cortex tohypoxia than the brainstem, recovery usuallyoccurs in a caudorostral sequence,5 and themaximal recovery time still compatible with agood outcome depends on the neurologicalfunction being considered. Several studieshave shown that the duration and depth ofcoma are good predictors of outcome.3'5 6 7 Theabsence of brainstem reflexes for more than sixto 24 hours after cardiopulmonary resuscita-tion,3589 absent motor responses to pain formore than one or two days,371011 the persis-tence of a Glasgow coma score (GCS) < 5 formore than two or three days,67 and of a GCS< 8 for more than one week36 all herald a pooroutcome. However, clinical prognosticationwithin the first few days after cardiopulmonaryresuscitation is afflicted with incontestablelimitations. Firstly, there is a risk of false pes-simistic predictions,'2 which may be as high as5% to 1 0%.13 14 Secondly, brainstem functionsnormalise in most patients within a few hoursand accurate prediction of outcome in thesepatients is not possible within the first twodays after cardiopulmonary resuscitation, aseven a GCS of 3 is compatible with a goodoutcome.3 7 15 16 Thirdly, the accuracy of clini-cal parameters is limited in intubated patientswho often need muscle relaxation and seda-tion. For these reasons, an intense search forelectrophysiological and biochemical criteriafor prognosis in postanoxic coma has beenundertaken. The absence of continuous EEGactivity for more than eight hours after car-diopulmonary resuscitation,5 the presence of aperiodic or "burst suppression" EEG,'7-19 andthe absence of early cortical SEPs'6 20-22 in thefirst two to three days have been shown to bereliable predictors of a bad outcome. Morerecently, biochemical parameters such asserum glucose at admission,2 serum ionisedcalcium,23 CSF creatine kinase,22 24 CSF lac-tate,25 and serum neuron specific enolase,26have been suggested as potential "markers" ofdiffuse cerebral ischaemia. Their prognosticaccuracy, however, remains to be confirmed.Despite considerable information about theprognostic accuracy of single parametersalone, the value of a multivariate approach hasnot yet been determined. Clearly, the com-bined analysis of, for example, clinical exami-nation, EEG, and SEP has the potentialadvantage of giving a more complete func-

610

Early prognosis in coma after cardiac arrest: a prospective clinical, electrophysiological, and biochemical study of 60 patients

tional profile of a patient. To assess whether amultivariate approach provides a better pre-diction of outcome within the first three daysafter cardiopulmonary resuscitation, we pro-spectively studied a consecutive series ofpatients in postanoxic coma with a standardexamination protocol which included repeatedclinical examinations, EEG, and SEP, serumionised calcium, and serum neuron specificenolase.

Patients and methodsWe studied a series of 60 consecutive patientsafter cardiac arrest and cardiopulmonaryresuscitation who remained comatose, asdefined elsewhere,3 for more than six hours.The average age was 62 years, ranging from 19to 86 years, and 49 were men. Patients withcraniocerebral trauma, pre-existing intra-cranial lesions, or drug intoxication wereexcluded. All patients were initially intubated,most received slight hyperventilation, intra-venous nutrition, and slight sedation withmidazolam or morphine. Intensive care treat-ment was guaranteed in all patients for at leastthe first three days after cardiopulmonaryresuscitation.

Repeated clinical examination was per-formed in each patient at regular intervalsbetween six to 12 hours, at 24 hours, at 48hours, and at one week after cardiopulmonaryresuscitation. All clinical assessments wereperformed by two of us (CB, FB). Glasgowcoma score (best 15, worst 3), corneal, pupil-lar, and oculocephalic brainstem reflexes, thepresence of spontaneous flexor ("decortica-tion") or extensor ("decerebration") postur-ing, and myoclonic or epileptic phenomenawere noted.

Electrophysiological studies were carriedout within the first three days after cardiopul-monary resuscitation in all but four patients;EEG and SEP were always recorded 12 hoursor more after cardiopulmonary resuscitation,as even an isoelectric EEG before this intervaldoes not exclude full recovery.27 In 48 patientsboth EEG and SEPs were recorded within 24hours of each other. An eight channel EEGwas recorded in 59 patients for at least 30 min-utes, using the international 10-20 system andincluding testing of the reactivity to acoustic,painful, and photic stimulation. Recordingswere classified by a visual analysis in a fivegrade scale of increasing severity (table 118).SEPs were obtained from 56 patients by elec-trical stimulation of the median nerve at thewrist.28 The central conduction time and theP25/N20 amplitude ratio20 of SEPs weredefined as abnormal when exceeding controldata from healthy subjects in our laboratory bymore than 2-5 SD. Results of the SEP wereclassified in three grades of increasing severity(table 1). Clinical, EEG, and SEPs findingswere classified in three prognostic groups(favourable; uncertain; unfavourable) accord-ing to currently suggested prognostic criteria(see above and table 2). Only very selective cri-teria for unfavourable outcome were chosen toavoid false pessimistic predictions.

Table 1 Grading ofEEG and medianus somatosensoryevoked potentials (SEPs)

EEGGrade I Dominant normally distributed a activity, reactiveGrade II Dominant 0-6 activity, reactiveGrade III 6-0 activity without a activity

Reactive or non-reactiveGrade IV* "a or 0 coma", non-reactiveGrade IV Burst suppression activty

Low voltage 6 activity, non-reactivePeriodic general phenomena with isoelectricintervals

Grade V Very low voltage EEGIsoelelectric EEG

SEPGrade I Normal N20/P25 response bilaterally

(normal amplitude + normal central conductiontime)

Grade II Abnormal N20/P25 response unilaterally orbilaterally (amplitude reduction > 50% orCCT > 7-2 ms, or both)

Grade III Absent N20/P25 bilaterally

Biochemical studies were carried out in 16patients. Serum neuron specific enolase con-centrations were determined in heparinisedblood 12 and 24 hours after cardiopulmonaryresuscitation. Concentrations in serum weremeasured by a commercially availableradioimmunoassay (Pharmacia-PDF-Sweden).The concentration of serum neuron specificenolase measured in healthy subjects in ourlaboratory is < 20 m mol/l. Ionised calcium,total calcium concentrations, and pH weredetermined in an arterial sample ofheparinised blood four and 12 hours after car-diopulmonary resuscitation. Measurementswere performed with a calcium selective elec-trode and corrected for a pH of 7-4. The nor-mal range of ionised calcium concentrations inhealthy subjects in our laboratory is1 20-1 30 mmol/l, and for total calcium it is2-10-2-50 mmol/l.Outcome was defined as the best cerebral

performance achieved at any follow up timeand classified by analogy with the GOS intotwo groups: (1) good outcome: awakeningwith reappearance of signs of cognition(Glasgow outcome scale (GOS) 3-5); (2) bad(poor) outcome: persistent vegetative state ordeath (GOS 1-2). Outcome was evaluated atone, two, and seven days, as well as one monthand one year after cardiopulmonary resuscita-tion. We decided to include patients with aGOS = 3 in the "good outcome" groupbecause some recovery may occur in thispatient group beyond our observation time.

STATISTICAL ANALYSISPredictive values were determined for favour-able and unfavourable factors. Favourable pre-dictive values (percentage of patients with thetest result whose good outcome was correctlypredicted), and unfavourable predictive values(percentage of patients with the test resultwhose bad outcome was correctly predicted)and their approximate 95% confidence inter-vals (95% CIs) were calculated for clinical,EEG, and SEP parameters. The number (andpercentage) of correct and wrong predictionsof both good and bad outcome were calcu-lated for clinical examination, EEG, and SEPseparately, according to the prognostic criteriashown in table 2. We decided to use these cri-

611

Bassetti, Bomio, Mathis, Hess

Table 2 Published prognostic criteria of clinical examination, EEG and SEPs in theliterature

GCS at 48-72 hoursBR at 6-12 hours EEG grade SEPs

Favourable > 8 I-II Normaland normal BR

Uncertain 5-7 III, IV* Abnormaland normal BR N20/P25

Unfavourable < 5 IV-V Absentor > 1 BR absent N20/P25 bilaterally

*"a or 6 coma".GCS = Glasgow coma score; BR = brainstem reflexes (corneal, pupillary, oculocephalic); SEP- medianus somatosensory evoked potentials.

teria because their accuracy has already beenproved in several large series of patients (seeintroduction). The prognostic value of thecombination of the three modalities wasassessed as follows. A good outcome was pre-dicted if there were (1) no unfavourable find-ings, and (2) at least one favourable finding onclinical examination, EEG, or SEP. Con-versely, a bad outcome was predicted if therewere one or more unfavourable findings onclinical examination, EEG, or SEP. The num-ber (and percentage) of correct and wrongpredictions of both good and bad outcomewere calculated for this multivariate approach.A multivariate logistic regression analysis wasused to identify from a series of different para-meters (brainstem reflexes at 6-12 hours,GCS at 24 hours, GCS at 48 hours, motorresponses at 24 and at 48 hours, EEG, andSEP) the combination which best predicts theoutcome. The comparison of patients withgood and bad outcome was performed forparametric variables with the unpaired t test,and for non-parametric variables with theMann-Whitney U test, X2 test, and compari-son of proportions test.

ResultsOf 60 patients in coma for more than six hoursafter cardiopulmonary resuscitation, 12 (20%)had a good outcome and 48 (80%) had a badoutcome. Five of 12 patients with good out-come remained in coma (GCS < 8) for two ormore days after cardiopulmonary resuscita-tion. However, all but one of the 12 patients

Table 3 Prognostic value of clinicalfindings, EEG, and SEPs in 60 patients withpostanoxic coma

Time after cardiac arrest (h)

6-12 24 48

Predictors ofgood outcome(percentage of patients with the test result whose good outcome was correctly prer3 BRpresent 29 (28) 33 (21) 33 (27)GCS > 8 50 (4) 50 (8) 78 (9)Spontaneous eye opening 17 (6) 23 (13) 53 (15)Favourable EEG (grade I-II)Favourable SEP (N20/P25normal bilaterally)

Predictors of bad outcome(percentage of patients with the test result whose bad outcome was correctly predi> 1 BR absent 100 (10) 100 (5) 100 (3)GCS < 5 81 (27) 87 (31) 100 (20)Poor motor response* 79 (38) 85 (48) 92 (37)Seizures or myoclonusUnfavourable EEG (Grade IV-V)Unfavourable SEP (N20/P25absent bilaterally)

Values are % with total number of patients in parentheses (95% CIs are giveBR = brainstem reflexes (comeal, pupillary, oculocephalic); GCS = GlasgovSEP = medianus nerve somatosensory evoked potential.*No or only stereotyped (extensor, flexor) motor response to pain.

awakened and reached their best level of func-tion within the first week, and nine patientseventually became independent in daily activi-ties. Thirteen of 60 patients (22%) died withinthe first three days, six of whom were certifiedbrain dead. Within the next month a further31 patients died, mostly (29/3 1) withoutawakening from coma. Four patients remainedin a vegetative state for more than one monthand died 70, 72, 89, and 280 days after car-

diopulmonary resuscitation. Table 3 sum-

marises the predictive values of clinicalfindings, EEG, and SEP.

CLINICAL EXAMINATION

Only eight of 28 patients with normal brain-stem reflexes at 6-12 hours had a good out-come (favourable predictive value = 29%,95% CI 13-49%). On the other hand, all 10patients in whom two or more brainstemreflexes were absent at 6-12 hours died incoma (unfavourable predictive value = 100%,69-100%). The GCS in the first 24 hoursranged between 3 and 14 in patients with goodoutcome, and between 3 and 9 in patients withbad outcome. However, seven of nine patientswith a GCS > 8 at 48 hours had a good out-come (favourable predictive values = 78%,40-97%), whereas all 20 patients with a GCS <

5 at 48 hours had a bad outcome(unfavourable predictive value = 100%,83-100%). Spontaneous eye opening was

often seen within the first 48 hours in patientswith both good and bad outcome. Hence, thefavourable predictive values of spontaneouseye opening was only 23% (5%-54%) at 24hours, and 53% (27-79%) at 48 hours. Poormotor responses to pain (absent or stereotypedresponses) heralded a poor outcome, particu-larly at 48 hours (favourable predictive value= 92%, 78-98%). Myoclonic or other epilepticphenomena were seen within the first threedays after cardiopulmonary resuscitation in 25patients, and were limited in eight of them tothe periocular region. Twenty four patientshad a bad outcome (unfavourable predictivevalue = 96%, 80-100%). Spontaneous"decerebration" and "decortication" spasmswere found in 11 patients including fourpatients with a good outcome. One patientwith good outcome had these up to 34 hoursafter cardiopulmonary resuscitation.

12-72 ELECTROPHYSIOLOGYFour of 10 patients with favourable EEG

dicted) (grade I-II) had a good recovery (favourablepredictive value = 40%, 12-74%). Anunfavourable EEG (grade IV-V) was found in

40 (10) 20 (34%) of 59 patients. All 20 patients had a

50 (20) bad outcome (unfavourable predictive value= 100%, 83-100%). Fourteen of 59 patients

icted) (24%) had an "a or 0 coma" EEG (grade IV).In 11 of them cortical SEPs were absent bilat-erally and none recovered. In two patients "a

190 (20) coma was "incomplete"'8 as the EEG activitywas either diffusely distributed or intermit-

100 (23) tently reactive. Both patients had normaln in the text). SEPs, and one of them had a good recovery.w coma score; SEPs were obtained in 56 patients. Normal

SEPs were recorded in 20 patients, and 10 of

612


Table 4 Comparison ofpatients with good and bad outcome after postanoxic coma

Good outcome Bad outcome(n = 12) (n = 48) P value

Age (y (SD)) 64.8 (12-4) 60-8 (16-5) NSBrainstem reflexes(at 6-12 hours, n (%)):

All three present 8 (67) 20 (48) NSTwo or more absent 0 10 (24) < 0-05

Poor motor responses (no or onlystereotyped responses, nAt 6-12 hours 8 (80) 30 (97) NSAt 24 hours 7 (70) 39 (87) NSAt 48 hours 3 (25) 34 (92) < 0-05

Spontaneous eye opening (n (%)):At6-12hours 1 (11) 5 (17) NSAt24hours 3 (33) 10 (22) NSAt48 hours 8 (67) 7 (18) < 0 05

Glasgow coma scale (score (SD)):At 6-12hours 5 (3) 4 (2) NSAt 24 hours 6 (3) 5 (2) NSAt48 hours 10 (4) 5 (2) < 0 05

Seizure or myoclonic activity(within first 3 days, n (%)) 1 (8) 24 (52) < 0-05EEG findings (59 patients, n (%)):

Favourable or uncertain 11 (100) 28 (58)Unfavourable 0 20 (42) < 0 05

SEP findings (56 patients, n (%)):Favourable or uncertain 11 (100) 22 (49)Unfavourable 0 23 (51) < 0 05

SEP = medianus nerve somatosensory evoked potentials.

them had a good outcome (favourable predic-tive value = 50%, 30-74%). Delayed or lowamplitude cortical (N20/P25) evoked poten-tials were found in 12 patients, and only one ofthem had a good outcome. Cortical SEPs wereabsent in 23 patients. All of them had a badoutcome (unfavourable predictive value =100%, 90-100%). Table 4 shows the compar-ison of patients with good and bad outcome.The absence of two or more brainstem reflexesat 6-12 hours, poor motor responses at 48hours, absent spontaneous eye opening at 48hours, seizure or myoclonic activity, andunfavourable EEG and SEP findings wereassociated with a bad outcome. There was alsoa significant difference between patients withgood and bad outcome in the GCS at 48hours, but not in the GCS at 6-12 hours and24 hours.

BIOCHEMICAL PARAMETERSSerum neuron specific enolase concentrationswere above normal in seven patients after 12hours, and in eight patients after 24 hours.The unfavourable predictive values of highserum neuron specific enolase were 86% at 12hours and 100% at 24 hours. Ionised calciumconcentrations were below normal in ninepatients after four hours, and in 12 patientsafter 12 hours despite normal concentrationsof total calcium. The unfavourable predictivevalues of low ionised calcium were 89% atfour hours and 83% at 12 hours.

UNVARIATE AND MULTIVARIATE APPROACHThe percentage of total correct predictions ofoutcome obtained applying a univariateapproach (criteria shown in table 2) was 58%

Table S Risk ofbad outcome in coma after cardiac arrest (multivariate logistic regressionanalysis of 60 patients)

GCS > 8 at 48 hours GCS < 8 at 48 hours GCS < 8 at 48 hoursand or andSEP normal SEP abnormal or absent SEP abnormal or absent

12% (95% CI, 2%-46%) 68% (95% CI, 47%-84%) 97% (95% CI, 86%-99%)

GCS = Glasgow coma score; SEP = medianus nerve somatosensory evoked potential.

for the clinical examination, 41% for the EEG,and 59% for the SEPs. The combination ofthree modalities raised the percentage of cor-rect predictions of outcome to 82%. Therewere no false negative predictions. A correctprediction of bad outcome was possible in 36of 48 (75%) patients combining clinical exam-ination and SEP alone, a total which wasraised by one to 37 (77%) by also consideringthe EEG. Prediction of a bad outcome waseasier in patients who died within the firstthree days (92%) than in those who survivedlonger (71%). Electrophysiological tests wereof significant prognostic help in nine (36%) of25 patients in whom prediction based on clini-cal signs was considered to be uncertain. Inaddition, unfavourable EEG or SEP findingswere found in 23 (66%) of 35 patients withunfavourable clinical signs. Conversely, nor-mal or abnormal biochemical tests were of noadditional prognostic help. A correct predic-tion of good outcome was possible applyingthe multivariate approach in 11 of 12 patients.There were 39% false optimistic predictions.A multivariate analysis identified the combina-tion of GCS after 48 hours and SEP as thebest predictive combination. The risk of a badoutcome was calculated to be 97% (86%-99%) for patients with a GCS < 8 and abnor-mal (or absent) SEP, but only 12% (2-46%)for patients with a GCS > 8 and normal SEPs(table 5).

DiscussionComa following resuscitation after cardiacarrest is a serious condition, and a good out-come can be expected in less than 20% ofpatients, as shown in this and previous largerseries."3Although awakening from coma usu-ally occurs within the first three days after car-diopulmonary resuscitation'° 25 we confirm thatrecovery after this interval can occur.36 Becausedecisions about limitation or maintenance ofintensive care often need to be made within thefirst few days after cardiopulmonary resuscita-tion, prognostic criteria in postanoxic coma areneeded. In this study we tried to avoid some ofthe methodological problems of previous stud-ies of prognosis in coma."3 Maximal treatmentwas guaranteed for at least the first three daysafter cardiopulmonary resuscitation, with theexception of patients in whom the clinical crite-ria of brain death were satisfied; data were col-lected prospectively by two observers only;electrophysiological studies were performedwithin a narrow interval of time from the onsetof coma; and 95% CIs for the suggested predic-tive model were provided. There are threemajor findings from our data. Firstly, normal oronly mildly abnormal EEG and SEP findingsare unreliable predictors of a good outcome.Secondly, currently suggested univariate clini-cal, EEG, and SEP criteria for prognosis of badoutcome (table 2) are reliable, but identify onlyabout 50% of the patients who die or remain ina vegetative state. Thirdly, the combination ofGCS at 48 hours, SEP, and EEG raises the per-centage of accurate prediction of bad outcometo 77% of the patients.

613

Bassetti, Bomio, Mathis, Hess

PROGNOSIS OF GOOD OUTCOME

Prognosis of good outcome in postanoxiccoma is difficult. We confirm that early recov-

ery of brainstem functions and spontaneouseye opening do not necessarily herald a goodoutcome, in that the cortex may have beendamaged to a degree that prevents cognitiverecovery. Also, the percentage of false opti-mistic predictions is high because of a signifi-cant extracerebral mortality.7 29 Nevertheless,78% of our patients with a GCS > 8 after 48hours had a good outcome, a percentage simi-lar to the 82% reported by Mullie et al in a

series of 360 patients.6 Conversely, favourableEEG (grade I-II) and normal SEPs were unre-

liable predictors of good recovery. Our resultsagree with those of previous series in that50%-60% of patients with favourable EEGand SEP findings still had a poor out-

come. 821222530 Hence, as long as the patientremains clinically in coma, the absence of rele-vant EEG and SEP abnormalities should beconsidered an indispensible condition ratherthan a "guarantee" for a good outcome. Apossible exception is the preservation of longlatency (N70) cortical SEPs.'6

UNIVARIATE PROGNOSIS OF BAD OUTCOME

We confirm that the absence of more than one

brainstem reflex for more than 6-12 hours,339 a

GCS < 5 at 48 hours or later, 67 an EEG gradeIV-V,'7 19 and the bilateral absence of corticalSEPs1620-22 at 12-72 hours after cardiopul-monary resuscitation, all herald a bad out-come without false pessimistic predictions. Wealso corroborate previous studies that foundthe following criteria to be unreliable predic-tors of poor outcome: prolonged absence of a

single brainstem reflex,7 a low GCS, and poor

motor responses within the first 24 hours aftercardiopulmonary resuscitation,7 16 spontaneousflexor and extensor posturing,3' an "a coma"EEG,'82 532 and abnormal but still detectablecortical SEPs. Poor motor responses to pain at48-72 hours3 7 "and epileptic and myoclonicphenomena in the first three days33 after car-

diopulmonary resuscitation were confirmed tobe the single most sensitive clinical predictorsof bad outcome. However, a few patients withthese findings enjoyed a good outcome, bothin our series and previous ones. 123436 For thisreason we think that the use of a combinationof clinical signs (for example, GCS), should bepreferred to using a single clinical parameterfor prognostication of postanoxic coma.

Prognostically unfavourable SEPs and EEGwere found both in 40%-50% of patients withbad outcome. The lower than expected EEGsensitivity for severe brain damage may be inpart related to the fact that we included a

coma patterns in the uncertain rather thanunfavourable EEG signs. This decision was

prompted by a few reports (confirmed in thisstudy) of patients with a coma and goodoutcome.'x 25 32 37 Although SEP and EEGhave a similar sensitivity in detecting severe

postanoxic brain damage, and the EEG mayhelp in the recognition of subclinical seizureactivity, we think that SEP should be preferredfor prognostication in coma after cardiac

arrest; SEPs are faster to perform and easier tointerpret, and also they are not influenced bysedating medications. Also, we found thatEEG only rarely added prognostic informationwhen clinical examination and SEP resultswere considered together. Although we con-firmed previous reports that over 80% ofpatients with raised serum neuron specificenolase and decreased ionised calcium have abad outcome,2326 the limited number ofpatients evaluated in this and previous studiesand the non-negligible percentage of false pes-simistic predictions call for caution. Thedetermination of CSF enzymes, including cre-atine kinase, may yield a higher sensitivity andspecificity in predicting outcome after cardiacarrest. None of these biochemical tests can,however, be suggested for routine use at thepresent time.38

MULTIVARIATE PROGNOSIS OF BAD OUTCOMEIn this, as well as previous reports the univariateapplication of selective criteria allowed theidentification of only about 50% of thepatients with bad outcome.367 Therefore, themost relevant result of this study is theincrease in the percentage of correct predic-tions of bad outcome from 58% with clinicalexamination alone to 77% with the combina-tion of clinical examination and electrophysi-ology. In other words, the recording of SEPand EEG allowed the prediction of outcome inone third of those patients, in whom clinicalsigns alone were not (yet) conclusive. In addi-tion, unfavourable electrophysiological para-meters supported the clinical assessment inpredicting a bad outcome in two thirds ofpatients with unfavourable clinical signs. It isof clinical relevance to point out that the utilityof electrophysiological recordings applied notonly to patients who died within the first threedays but also to those who survived longer.Conversely, in the 16 patients tested, bio-chemical tests were of no additional prognos-tic help. A multivariate logistic regressionanalysis of our data identified the associationof a GCS < 8 at 48 hours with an abnormal orabsent SEP as highly predictive of a bad out-come (risk 97%, 95% CI 86%-99%). Thismeans that the combination of clinical signsand SEPs may allow the application of lessselective prognostic criteria for prognostica-tion of bad outcome than those suggestedwhen these variables are used in isolation(table 2). There have been only a few studiesassessing the prognostic value of a multivariateapproach and none including clinical examina-tion, EEG, SEP, and biochemical parameters.Edgren et a125 correctly predicted the outcomein about 75% of 32 patients combining a mod-ified GCS at 48 hours with the EEG at 48hours, and the CSF lactate at 24 hours. In thatstudy, however, the percentage of false pes-simistic predictions was as high as 29%. In astudy of 40 patients, Rothstein et al found, aswe did, that SEPs were more reliable thanEEG in predicting outcome. These authorsdid not consider, however, any clinical para-meter, and the exact timing of electrophysio-logical recordings was not specified. Becker et

614


al performed repeated clinical examination,EEG, SEP, and proton magnetic resonancespectroscopy (MRS) in a series of 25 patientswith postanoxic coma.29 Demonstration ofbrain lactate in MRS and absent cortical SEPswere found to be reliable predictors of badoutcome. The limitations of this study werethe high cardiac mortality, which reduced thenumber of patients available for analysis ofcerebral prognosis, and the lack of analysis ofthe prognostic value of clinical examinationand EEG.The accuracy of the multivariate approach

used in this study needs to be verified in alarger group of patients before it can beadopted in clinical practice. Nevertheless, thefollowing approach to prognostication aftercardiac arrest can be suggested integrating ourdata with the information available in the liter-ature. In the first 48 hours after cardiopul-monary resuscitation serial clinicalexamination including assessment of brain-stem reflexes and GCS should be performed.Electrophysiological testing need not be con-sidered in the first 24 hours, unless statusepilepticus or status myoclonicus are sus-pected. If by 24-48 hours the patient does notawaken, SEPs should be recorded. If clinicalexamination and SEPs are prognostically non-conclusive an EEG could be performed. Wethink that the combination of GCS at 48hours, SEPs, and, when necessary, EEG canassist the clinician in predicting the outcomeof coma after cardiac arrest with reasonableaccuracy. As any available prognostic model inpostanoxic coma carries some risk of error,"the limitation or withdrawal of medical treat-ment should always be made on an individualbasis, involving the patient's family in the deci-sion and considering medical as well as psy-chosocial issues.

Dr G van Melle, Department of Statistics, University Hospital,Lausanne, Switzerland, helped with the statistical analysis. DrsT Herren and T Weiss, Department of Internal Medicine,University Hospital, Bem, Switzerland, collaborated in therecruitment and care of the patients. The measurement of theserum neuron specific enolase was performed by Dr P Huber,Gynecological Laboratory Medicine, University Hospital,Basel, Switzerland, and that of the serum ionised calcium byDr E Peheim, Chemisches Zentrallabor, University Hospital,Bem, Switzerland. Dr F Glocker and Mrs J Schmid assisted inthe recording of SEPs. Drs I Drury and D Gelb, Department ofNeurology, University of Michigan Hospitals, Ann Arbor, MI,USA, made helpful comments.

1 Bedell SE, Delbanco TL, Cook EF, Epstein FH. Survivalafter cardiopulmonary resuscitation in the hospital. NEnglJMed 1983;309:569-76.

2 Longstreth WT, Diehr P, Inui TS. Prediction of awakeningafter out-of-hospital cardiac arrest. N Engl Jf Med1983;308: 1378-82.

3 Levy DE, Caronna JJ, Singer BH, et al. Predicting the out-come from hypoxic-ischemic coma. JAMA 1985;254:1171.

4 Gray WA, Capone RJ, Most AS. Unsuccessful emergencymedical resuscitation-Are continued efforts in the emer-gency department justified? N Engl J7 Med 1991;325:1393-8.

5 Jorgensen EO. Natural history of global and critical brainischemia. Resuscitation 1981;9:131-88.

6 Mullie A, Buylaert W, Michem N, et al. Predictive value ofGlasgow coma score for awakening after out-of-hospitalcardiac arrest. Lancet 1988;137-40.

7 Edgren E, Hedstrand U, Kelsey S, Sutton-Tyrell K, SafarP and BRCT I Study Group. Assessment of neurologicalprognosis in comatose survivors of cardiac arrest. Lancet1994;343: 1055-9.

8 Earnest MP, Breckenridge JC, Yarnell PR, Oliva PB.Quality of survival after out-of hospital cardiac arrest:predictive value of early neurologic evaluation. Neurology1979;29:56-60.

9 Snyder BD, Gumnit RJ, Leppik IE. Neurologic prognosisafter cardiopulmonary arrest: IV. Brainstem reflexes.Neurology 1981;31:1092-7.

10 Longstreth WT, Inui TS, Cobb LA, Copass MK.Neurologic recovery after out-of-hospital cardiac arrest.Ann Intern Med 1983;98:588-92.

11 Hamel MB, Goldman L, Teno J, et al. Identification ofcomatose patients at high risk for death or severe disabil-ity. JAAMA 1995;273:1842-8.

12 Golby A, McGuire D, Bayne L. Unexpected recovery fromanoxic-ischemic coma. Neurology 1995;45:1629-30.

13 Shewmon DA, De Giorgio CM. Early prognosis in anoxiccoma. Neurol Clinl989;7:823-43.

14 Bates D. Defining prognosis in medical coma. Jf NeurolNeurosurg Psychiatry 1991;54:569-71.

15 Sacco RL, Van Gool R, Mohr JP, et al. Non traumaticcoma. Glasgow coma score and coma etiology as predic-tors of 2-week outcome. Arch Neurol 1990;47: 1181-4.

16 Madl C, Grimm G, Kramer L, et al. Early prediction ofindividual outcome after cardiopulmonary resuscitation.Lancet 1993;341:855-60.

17 Prior. The EEG in acute cerebral anoxia. Amsterdam:Excerpta Medica, 1973.

18 Bassetti C, Karbowski K. Prognostischer Wert derElektroenzephalographie bei nicht-traumatischenKomata. Schweiz med Wschr 1990;120:1425-34.

19 Synek VM. Value of a revised EEG coma scale for progno-sis after cerebral anoxia and diffuse head injury. ClinElectroencephalogr 1990;21:25-30.

20 Walser H, Mattle H, Keller HM, Janzer R. Early corticalmedian nerve somatosensory evoked potentials.Prognostic value in anoxic coma. Arch Neurol 1986;42:32-8.

21 Brunko E, Zegers de Beyl D. Prognostic value of early cor-tical somatosensory evoked potentials after resuscitationfrom cardiac arrest. Electroencephalogr Clin Neurophysiol1987;66: 15-24.

22 Rothstein TL, Thomas EM, Sumi SM. Predicting outcomein hypoxic-ischemic coma. A prospective clinical andelectrophysiological study. Electroencephalogr ClinNeurophysiol 1991;79:101-7.

23 Urban P, Scheidegger D, Buchmann B, Barth D. Cardiacarrest and blood ionized calcium levels. Ann Intern Med1988;109:110-3.

24 Longstreth WT, Clayson KJ, Sumi SM. Cerebrospinalfluid and serum creatine kinase BB activity after out-of-hospital cardiac arrest. Neurology 1981;31:455-8.

25 Edgren E, Hedstrand U, Nordin M, et al. Prediction of out-come after cardiac arrest. Crit Care Med 1987;15:820-5.

26 Roine RO, Somer H, Kaste M, Viinikka L, Karonen S.Neurological outcome after out of hospital cardiac arrest:prediction by cerebrospinal fluid enzyme analysis. ArchNeurol 1989;46:753-6.

27 Jorgensen EO. EEG without detectable cortical activity andcranial nerve areflexia as parameters of brain death.Electroencephalogr Clin Neurophysiol 1974;36:70-5.

28 Bassetti C, Mathis J, Hess CW. Multimodal electrophysio-logical studies including motor evoked potentials inpatients with locked-in syndrome: report of six patients. J7Neurol Neurosurg Psychiatry 1994;57: 1403-6.

29 Berek K, Lechleimer P, Luef G, et al. Early determination ofneurologic outcome after prehospital cardiopulmonaryresuscitation. Stroke 1995;26:543-9.

30 Scollo-Lavizzari G, Bassetti C. Prognostic value of theEEG in post-anoxic coma after cardiac arrest. EuropNeurol 1987;26: 161-70.

31 Brendler SJ, Selverstone B. Recovery from decerebration.Brain 1970;93:381-92.

32 Austin E, Wilkus RJ, Longstreth WT. Etiology and progno-sis of alpha coma. Neurology 1988;33:773-7.

33 Wijdicks EFM, Parisi JE, Sharbrough FW. Prognosticvalue of myoclonus status in comatose survivors of car-diac arrest. Ann Neurol 1994;35:239-43.

34 Celesia GG, Grigg MM, Ross E. Generalized statusmyoclonicus in acute anoxic and toxic-metabolicencephalopathies. Arch Neurol 1988;45:781-4.

35 Krumholz A, Stern BJ, Weiss HD. Outcome from comaafter CPR: relation to seizures and myoclonus. Neurology1988;38:401-5.

36 Arnoldus EPJ, Lammers GJ. Potsanoxic coma: good recov-ery despite myoclonic status. Ann Neurol 1995;38:697-8.

37 Soerensen K, Thomassen A, Wernberg M. Prognostic sig-nificance of alpha frequency EEG rhythm in coma aftercardiac arrest. J Neurol Neurosurg Psychiatry 1978;41:840-2.

38 Longstreth WT. Neurologic complications of cardiac arrest.In: Aminoff A, ed. Neurology and general medicine, 2nd ed.Edinburgh: Churchill-Livingstone, 1994:159-82.

615

early prognosis in coma after cardiac arrest: a prospective clinical

Documents