LUT VAN DEN BROECK3DOOR LAUWAERT4MARK WEEGHMANS5LEO BOSSAERT7PAUL LEWI8
Department of Critical Care Medicine, Akademisch Ziekenhuis StJan, Brugge, Belgium;1 Departments of Emergency Medicine,
Universitair Ziekenhuis, Gent;2 Universitair Ziekenhuis St Rafaël-Gasthuisberg, Leuven;3 Akademisch Ziekenhuis, Vrije Universiteit,Brussel;4 Imeldaziekenhuis, Bonheiden;5 Jan Palfijnziekenhuis,
Merksem;6 Department of Information Science, JanssenPharmaceutica, Beerse,8 and Department of Intensive Care
Medicine, University Ziekenhuis-UIA, Antwerpen7
Summary The Glasgow coma score (GCS) duringdays 1-6 after cardiac arrest was used to
predict neurological outcome in 360 resuscitated victims ofout-of-hospital cardiac arrest. A predictive rule based on thebest GCS of 216 patients resuscitated in 1983-84
(prediction group) was constructed, and its predictive powerwas tested on 133 patients treated in 1985 (test group).Neurological outcome was correctly predicted 2 days aftercardiac arrest in 80% of the prediction group, with a bestGCS of 10 or above and 4 or below as cutoff points. Forpatients with a best GCS of 5-9, prediction of outcome was
possible 6 days after cardiac arrest, with a best GCS of 8during the first 6 days as the single cutoff point. The rule wasthen validated in the test group: the sensitivity was 96%; thespecificity 86%; the negative predictive value 97%; and thepositive predictive value 77%. These data suggest that thissimple GCS-based rule can be helpful in predictingoutcome in patients resuscitated after out-of-hospitalcardiac arrest, but confirmation of these data is required in aprospective study in a larger number of patients.
Introduction
MOBILE intensive-care units (MICU) have improved theprognosis for victims of out-of-hospital cardiac arrest.1-3Longstreth et al4 proposed a clinical rule, based on motorresponse, reaction by pupils to light, spontaneous eyemovements, and blood glucose, to predict awakening.However, this rule is exclusively constructed on
retrospectively collected data and has been criticised5because of its high error rate in identifying patients with apoor prognosis.
Since January, 1983, the Cerebral Resuscitation StudyGroup of the Belgian Society for Intensive Care has
prospectively collected information on victims of out-of-hospital cardiac arrest, in an attempt to identify factorsrelated to neurological outcome.6,7 A preliminary analysis ofthe data showed that the Glasgow coma score (GCS) gavethe best correlation with neurological outcome.8 Othershave suggested that the GCS may be useful in the predictionof outcome.9We report here our data on the GCS in a large number of
resuscitated victims of out-of-hospital cardiac arrest andpropose a simple rule that might help the clinician in
predicting neurological outcome.
138
Patients and Methods
All cases of out-of-hospital cardiac arrest (1832) attended
between Jan 1, 1983, and Dec 31,1985, by the MICU teams of theparticipating centres were registered. Each MICU team consists ofa physician and a nurse, experienced in prehospital advanced lifesupport, and is integrated in the nationwide emergency medicalsystem. Of the 1832 patients, 360 were successfully resuscitated.Only these patients, in whom initial restoration of spontaneouscirculation was achieved, were admitted to a hospital and areincluded in this analysis.
Patient management, although not strictly standardised in awritten treatment protocol, was based in each centre and during thewhole study on the American Heart Association standards andguidelines of advanced cardiac life supportlO and of cardio-pulmonary cerebral resuscitation (CPCR)."The outcome for the 360 patients was evaluated after 14 days and
expressed in cerebral performance outcome groups as definedpreviously.6 235 (65%) were CPCR failures and 125 (35%) CPCRsuccesses. Of the 235 CPCR failures, 132 died from cerebral failure,82 from other causes (eg, cardiac failure, sepsis), and 21 were still ina vegetative state at 14 days (18 eventually dying without regainingconsciousness, 2 being still alive without regaining consciousness,and 1 patient being wrongly classified as vegetative, because ofheavy sedation given for tetanus, and eventually awakening).Of the 125 CPCR successes, 61 had recovered to their prearrest
status, 28 to moderate disabilities and 36 to severe disability.For 349 of the 360 patients, data on the GCS at 12 h, 24 h, and 2,
3,4,5, and 6 days after cardiopulmonary resuscitation (CPR) and onconcomitant sedative drug use were available. The GCS wasrecorded in the clinical notes, except in one centre, where the scorewas estimated from written or personal information from theintensive-care personnel. For the whole study period thisestimation was done by the same two research nurses, who were notinvolved in the further analysis of the data.A predictive rule of cerebral performance outcome was
constructed based on the analysis of the GCS of the patients studiedin 1983 and 1984 (prediction group; n = 216). The predictive powerof this rule was tested in the cohort of patients seen in 1985 (testgroup; n = 133). The characteristics of the prediction and testgroups were comparable as regards age, sex, underlying disease,clinical findings, and treatment protocol, as well as final outcome.
For statistical analysis the Mann-Whitney U test and the
chi-square test were used.12
Results
GCS in Prediction Group
Considering all 199 patients still alive 12 h after CPR,there was a large overlap in the 90% confidence intervals ofthe median GCS between CPCR failures and successes (fig1). However, when the CPCR failures who had died fromcauses other than cerebral failure (55 patients) were
excluded, CPCR successes and failures could be separated;after 12 h the median GCS was higher in the CPCRsuccesses but there was still an overlap in the 90%confidence intervals of cases at that time (median GCS[90% confidence interval] of failures 3 [3-6], of successes 11[3-15]). By day 2, the 90% confidence intervals no longeroverlapped (failures 3 [3-6], successes 14 [8-15]).
Furthermore, 63 of the 65 patients (97%) in whom thebest GCS during the first 2 days after resuscitation was 10 orabove could be classified as CPCR successes 14 days afterCPR. In contrast, of the 60 patients with a best GCS of 4 orless in the first 2 days, not a single patient could be classifiedas a CPCR success. In the remaining group of 32 patients(best GCS of 5-9 in the first 2 days), 9 patients (28%)eventually recovered and were classified as CPCR successes.If the best GCS during the first 6 days of these 32 patients isconsidered, of the 12 patients with a best GCS of 8 or above,8 (67%) were CPCR successes, whereas of the 20 patients
Fig I-Time course of GCS (median and 90% confidence interval[shaded]) in prediction group (1983 and 1984).
Figures in parentheses = numbers of patients alive at each time.
with a best GCS below 8, only 1 (5%) was a CPCR success.5 of these 32 patients received at least intermittent heavysedation up to day 6; 4 of the 5 had a best GCS below 8 anddid not recover, and 1 had a best GCS above 8 andrecovered.
Predictive Value of GCS
These data suggest that, at least in the subset of patientswho did not recover from cerebral failure, correct predictionof neurological outcome was possible in 80% of patients,based on the best GCS after 2 days, with cutoff points atGCS 10 or above and 4 or below. For patients with a score of5-9, a prediction based on the best GCS during the first 6days, with 8 as the single cutoff point, seems possible.The validity of this predictive method was tested on all
133 cardiac-arrest patients seen in 1985, irrespective ofwhether they died from cerebral failure or other causes (fig2).The outcome was predicted correctly in 46 of the 48
patients considered CPCR successes (sensitivity 96%; 95%CI 89-100%), and in 71 of the 85 patients who did notrecover (CPCR failures; specificity 84%; 95 % CI 75-91 %).Of the 60 patients predicted to recover, 46 patients did
actually recover; the positive predictive value was therefore77%. Of the 73 patients predicted not to recover, 71 did not;the negative predictive value was 97%.
139
DiscussionThe aim of our study was to test the value of the GCS
measured during the first 6 days after resuscitation inpredicting outcome for victims of out-of-hospital cardiacarrest admitted to intensive-care units. The final outcomewas evaluated at day 14; it could be argued that this
evaluation point was too early. However, of the 21 patientsin a vegetative state at 14 days, only 1 later recovered and to astate of persisting severe disability as described in anotherreport.13 Moreover, published data show that in patientsresuscitated after cardiac arrest, the cumulative mortalityreaches a plateau after 14 days.l’ The mortality in ourpatients at 14 days was 59%, which is similar to that in otherstudies in which final outcome after cardiac arrest wasevaluated at the end of the hospital stay.2,14 6% of ourpatients remained in a vegetative state, and about half of theCPCR successes recovered to their initial state. These ratesare also similar to those in other studies in resuscitated
out-of-hospital cardiac-arrest victims.2We found that in our patient population the GCS after
cardiac arrest improved more rapidly in the recovering thanin the non-recovering patients; however, a clear separationbetween the two classes was not possible even after 6 days.We postulated that this overlap was due mainly to
non-recovering patients who had high GCS but died fromcauses other than cerebral failure, and when these patientswere excluded from the analysis there was a clear separationbetween failure and success after 2 days. Since our primaryaim was to help clinicians to predict potential neurologicalrecovery after cardiac arrest, patients dying from causesother than cerebral failure, thus "contaminating" the valueof the median GCS in the CPCR failure group, wereexcluded for the construction of the predictive rule.
Secondly, in constructing a predictive rule we were anxiousto predict correctly all recovering patients-ie, to produceprimarily a sensitive rule. Whereas an incorrect positiveprediction has no adverse consequences for the patient, anincorrect negative prediction is unacceptable. Cutoff pointsfor the GCS of 4 and 10 at day 2 seemed acceptable andallowed correct classification in 80% of the patients at thattime. At day 6, the last day of our recording the GCS, a
single cutoff point (8) was set to classify all remaining (20%)patients, again with a minimum negative predictive error.The use of the test population seemed justifiable; the
distribution of the outcome groups in the test and predictionpopulation was comparable, as regards age, sex, clinicalfindings, treatment protocol, and outcome. The low
specificity (84%) of the proposed rule results from ourdetermination to avoid any false prediction of non-recovery.The sensitivity (96%) and specificity of this rule comparesfavourably with those of Longstreth et al (sensitivity 92%,specificity 65%).4
In the 133 tested patients, there were 2 incorrect negativepredictions and 14 incorrect positive predictions. The firstincorrect negative prediction was on an 83-year-old man,who remained in deep coma for 3 days after the arrest (GCSbelow 4 on day 2). He was therefore predicted a CPCRfailure. He eventually improved to a level of awakening butremained severely disabled, for reasons other than
neurological disease, and died 35 days after the cardiacarrest. The second incorrect negative prediction was on a31-year-old man (GCS 6 on day 2). He required heavysedation, and the GCS remained at 6 up to day 6; he wastherefore predicted a CPCR failure. However, he graduallyrecovered but substantial neurological damage remained.Although sedation of the patients did not seem to interferewith correct estimation in the prediction group, the effect ofsedative drugs may still have affected the GCS in this patientand contributed to this error.Of the 14 incorrect positive predictions, 8 patients died
from causes other than cerebral failure (secondary cardiacand septic causes). We expected that in this subset ofpatients the GCS would correlate poorly with the outcome,as occurred in the prediction group. 4 patients with incorrectpositive predictions died within 14 days in cerebral failure. 2other patients survived but recovered only to a vegetativestate. This amount of "unrewarding" intensive care may beunavoidable if one wishes to err on the conservative side in
decision-making. isTo exclude the possibility that secular changes during the
study could have affected the data, and to show the timelesscharacter of the predictive rule, we also analysed the patient
Fig 2-Prediction of outcome based on GCS in resuscitated out-of-hospital cardiac-arrest victims in test group (1985).
140
data in randomly selected subsets; we constructed thepredictive rule on a randomly selected subset of two thirds ofthe whole data set and tested it on the remaining third. Theresults obtained were identical.Our findings suggest that the overall GCS 2 days and 6
days after cardiac arrest have a high predictive value forrecovery. However, several facts must be taken into accountbefore this rule can be recommended for use in daily clinicalpractice.
First, patients with the same total GCS may differ inclinical status.16 Second, in the patients from one centreincluded in the study, the GCS was estimated rather thanaccurately measured. Since this was true for both predictionand test groups, the data from this centre could be includedfor testing the predictive rule. Moreover, separate analysis ofthe data for this centre did not reveal any difference from thewhole study population. However, we cannot exclude thepossibility that more accurate measurement of the GCS inall patients would have led to another choice of cutoff points.Third, although the sensitivity of the proposed predictiverule is high, this estimate is based on a small sample. Fourth,some subsets of patients should be viewed with great care inpredicting outcome after cardiac arrest: patients still underheavy sedation and/or in metabolic encephalopathy, forwhom the time course of the GCS is obscured by eventsother than the natural course of the postischaemicencephalopathy, patients with recurrent episodes of
hypotension or rearrests, in whom the GCS is greatlyinfluenced by the variability of the cardiovascular system;children, in whom the time for decision-making shouldcertainly be increased. In our study, the number of patientsin these special categories is obviously too small for us todraw any conclusions. We do not believe that simplepredictive rules for neurological outcome can be applied forsuch patients..
A prospective study in a larger number of patients isneeded before the simple predictive rule proposed here canbe safely used in daily clinical practice.
Correspondence should be addressed to A. M., Department of CriticalCare Medicine, AZ-St Jan, B-8000 Brugge, Belgium.
REFERENCES
1 Bedell SE, Delbanco IL, Cook EF, Epstein EH Survival after cardiopulmonaryresuscitation in the hospital N Eng J Med 1983, 309: 569-76
2 Earnest MP, Breckinridge JD, Yamell PR, Oliva PB Quality of survival after
out-of-hospital cardiac arrest predictive value of early neurologic evaluation.Neurology 1979, 29: 56-60.
3. Willoughby JO, Leach BG Relation of neurological findings after cardiac arrest tooutcome. Br Med J 1974: 3: 437-39
4. Longstreth WT, Diehr P, Inui TS. Prediction of awakening after out-of-hospitalcardiac arrest. N Eng J Med 1983: 308: 1378-82
5. Levy DE, Caronna JJ, Singer EH, Lapinski RH, Frydman H, Plum F. Predictingoutcome from hypoxic ischemic coma. JAMA 1985; 253: 1420-26.
6. Studygroup "Cerebral Resuscitation" of the Belgian Society of Intensive CareMedicine (SIZ). A multicentre registration project of CPR in Belgium ActaAnaesth Belg 1984, 35 S1: 29.
7 Come L, Buylaert W, Mennes J, De Cock R, Delooz H, Lewi P Prognostic factors inpatients with out-of-hospital cardiac arrest Arch Emergency Med 1986; 3: 65
8. Mullie A. Studygroup "Cerebral Resuscitation", Belgian Society of Intensive Care(SIZ). Clinical Study of cerebral and cardiopulmonary resuscitation by Belgianmobile intensive care units. Reanimazione 1985; 2: 297-310
9. Urban P, Cereda JM. Glasgow coma score 1 hour after cardiac arrest. Lancet 1985, ii1012.
10 American Heart Association Textbook of advanced cardiac life support Dallas AHA,1982
11. Safar P. Cardiopulmonary cerebral resuscitation. Philadelphia W. B. Saunders, 1981.12. Siegel S Nonparametric statistics for the behavioral sciences. New York. McGraw-
Hill, 198013. Jennett B, Plum F Persistent vegetative state after brain damage. A syndrome in
search of a name Lancet 1972, i. 734-37.
14 Snyder BD, Loewenson RB, Gummit RJ. Neurologic prognosis after
cardiopulmonary arrest II. Level of consciousness. Neurology 1980, 30: 52-5815. Falk RH. The death of death with dignity. Am J Med 1984, 77: 775-7616. Teasdale G, Jennett B, Murray L. Murray G Glasgow coma scale to sum or not to
sum ? Lancet 1983; ii: 678.
EFFECT OF VITAMIN AND MINERAL
SUPPLEMENTATION ON INTELLIGENCE OF ASAMPLE OF SCHOOLCHILDREN
DAVID BENTON1 GWILYM ROBERTS
Department of Psychology, University College, Swansea SA28PP;1 and Darland High School, Rossett, Wrexham LL12 0EN
Summary 90 schoolchildren aged twelve and thirteenyears kept a dietary diary for three days. In
most cases the average intake of vitamins was close to therecommended daily allowance, although for a minority theintake was low; with minerals the recommended dailyallowance was less commonly achieved. To examine thepossibility that deficiency of dietary minerals and vitaminswas preventing optimum psychological function, a multi-vitamin/mineral supplement or a placebo was administereddouble-blind for eight months to 60 of the children. Thesupplement group, but not the placebo group or the
remaining 30 who took no tablets, showed a significantincrease in non-verbal intelligence.
Introduction
THE adequacy of the diet in industrialised societies is atopic that generates great controversy. On one side are thosewho argue that a diet that contains too much refined food canbe so poor in minerals and vitamins as to hamperbiochemical functions. On the other are those who maintainthat in most cases the diet is adequate, often supplyingminerals and vitamins in excess of our needs. Althoughspecific disorders are associated with low intake of particularvitamins, these occur only after serious, and prolongeddietary deficiency of a kind that is rare in western societies.What of subclinical deficiencies? With some vitamins andminerals there are no suitable biochemical indices of
deficiency, and the recommended daily allowances (RDAs)are based on scant information. In this context there is
mounting evidence that certain subclinical dietaryinadequacies have measurable psychological effects. Aspectsof diet have been found to be associated with personality;1-4intellectual functioning" motivation,9,10 and abnormalitiesof attention and perception.5,6,9-12 In the study reported herewe tested the hypothesis that the diet is broadly adequate byexamining the influence of supplementary minerals andvitamins on the intelligence of schoolchildren. In view of thewide range of vitamins that have been linked with
psychological functioning, and the finding in the
preliminary dietary survey that for each of the vitamins andminerals a sizeable minority showed a low intake, we used amulti-vitamin/mineral supplement rather than a singlenutrient. An additional reason for adopting this approachwas that combinations of micronutrients may be synergic.
Methods
In September, 1986, 90 schoolchildren (45 boys, 45 girls) filled indietary diaries for three days and completed two intelligence tests.Aged twelve and thirteen years, they were almost a complete year’sintake at Darland High School, a comprehensive school in
Wrexham, Wales (1 child did not participate, all the others, andtheir parents, gave full consent). Three groups of 30 children werematched by their form teacher for sex, school performance, andhome background. They were then randomly assigned 3 at a time,by an individual who did not know them and who was not directlyinvolved in the trial, to one of the three options. The first groupreceived no tablets. The other two took either the mineral/vitaminsupplement or placebo, administered double-blind-ie, no parent,teacher, or child could know if a child was taking active or placebo
