Current issues of ACP Journal Club are published in Annals of Internal Medicine


Review: 4 clinical tests most accurately predict poor outcome in patients with anoxic-ischemic coma

ACP J Club. 1999 July-Aug;131:22. doi:10.7326/ACPJC-1999-131-1-022

Source Citation

Zandbergen EG, de Haan RJ, Stoutenbeek CP, Koelman JH, Hijdra A. Systematic review of early prediction of poor outcome in anoxic-ischaemic coma. Lancet. 1998 Dec 5;352:1808-12.



Which early neurologic and neurophysiologic tests predict clinical outcomes in patients with anoxic-ischemic coma?

Data sources

Studies in English, German, and French were identified by using MEDLINE (from 1966) and EMBASE/Excerpta Medica (from 1982) with the search terms anoxia (cerebral), ischemia (cerebral), heart arrest, hypotension, shock, postoperative complications, respiratory insufficiency, resuscitation, drowning, coma, and Glasgow Coma Scale. Bibliographies of papers were also scanned.

Study selection

Studies that assessed the association between neurologic or neurophysiologic tests and outcome were selected if they had data for patients with anoxic-ischemic coma; reported unselected, consecutive cases; recorded the description, classification, and timing of clinical and neurophysiologic features; studied patients ≥ 10 years of age; and had a combined outcome of death or vegetative state that could be compared with other outcome states. Studies of patients whose comas resulted from other medical conditions or trauma were excluded.

Data extraction

Data were extracted on the following prognostic tests: Glasgow Coma Scale, pupillary light reflex, corneal reflexes, eye movements, epilepsy or myoclonus, somatosensory-evoked potentials (SSEPs) from median nerve stimulation, and electroencephalography (EEG). The outcome of death or vegetative state was compared with any other outcome state.

Main results

1667 studies were identified, and 33 met the inclusion criteria. For the outcome of death or vegetative state, 3 prognostic tests—absence of pupillary reactions to light on day 3 of coma (3 studies), absence of motor response to pain on day 3 of coma (3 studies), and bilateral absence of cortical response to median nerve SSEP within week 1 of coma (11 studies)—had 100% specificity in all studies. Burst suppression of isoelectric EEG within week 1 had 100% specificity in 5 of 6 studies. The accuracy of these prognostic tests was determined by using meta-analytic techniques (Table).


Among patients with anoxic-ischemic coma, absence of pupillary reactions to light on day 3 of coma, absence of motor response to pain on day 3 of coma, bilateral absence of cortical response to median nerve somatosensory-evoked potentials within week 1 of coma, and burst-suppression of isoelectric electroencephalography within week 1 of coma more accurately predict death or vegetative state than other neurologic and neurophysiologic tests.

Source of funding: University of Amsterdam.

For correspondence: Dr. E.G. Zandbergen, Departments of Neurology and Clinical Neurophysiology, Academic Medical Centre, University of Amsterdam, P.O. Box 22660, 1100 DD Amsterdam, The Netherlands. FAX 31-20-697-1438.

Table. Pooled +LRs of prognostic tests for death or vegetative state*

Type and timing of test Reaction Pooled +LR (95% CI)
Pupillary reactions to light, day 3 None 10.5 (2.1 to 52.4)
Motor response to pain, day 3 None 16.8 (3.4 to 84.1)
Response to median nerve SSEP, week 1 None 12.0 (5.3 to 27.6)
Isoelectric EEG, week 1 Burst suppression 9.0 (2.5 to 33.1)

*EEG = electroencephalography; SSEP = somatosensory-evoked potential. LRs defined in Glossary. +LRs for most studies were generated by using a correction factor where specificity was 100%.


The prognosis for patients who have cardiac arrest and remain unconscious for > 12 hours is bleak. Many of these patients die, those who enter a vegetative state have poor outcomes (1), and those who regain consciousness are likely to have severe and irreversible dementia.

Nevertheless, as Zandbergen and colleagues suggest, if prognostic markers of poor outcome are used to make decisions about treatment cessation, they must have 100% specificity (i.e., no patient whose test result predicts a poor prognosis should subsequently recover consciousness).

3 prognostic tests met this standard. All 187 patients who had no cortical responses to median nerve SSEP within the first week had poor outcomes. However, this is only an estimate made on the basis of the patients studied—the CI for a false-positive result shows that as many as 1 in 50 patients without cortical responses could regain consciousness. The other 2 tests were used on fewer patients, making the CIs even larger.

SSEPs are least susceptible to metabolic changes and drugs and are therefore the best single predictor of poor outcome. The authors could not test for additive effects of multiple predictors of poor outcome. Nevertheless, their proposed clinical guidelines recommend that withdrawal of active treatment should depend on more than just a poor test result using SSEP alone. In all patients, assessment should be postponed until ≥ 72 hours after coma onset. Other causes of coma need to be excluded.

Simon Fleminger, MB, PhD
Maudsley HospitalLondon, England, UK


1. Kennard C, Illingworth R. Persistent vegetative state. J Neurol Neurosurg Psychiatry. 1995;59:347-8.