Context
Most survivors of cardiac arrest are comatose after resuscitation, and
meaningful neurological recovery occurs in a small proportion of cases. Treatment
can be lengthy, expensive, and often difficult for families and caregivers.
Physical examination is potentially useful in this clinical scenario, and
the information obtained may help physicians and families make accurate decisions
about treatment and/or withdrawal of care.
Objective
To determine the precision and accuracy of the clinical examination
in predicting poor outcome in post–cardiac arrest coma.
Data Sources and Study Selection
We searched MEDLINE for English-language articles (1966-2003) using
the terms coma, cardiac arrest, prognosis, physical examination, sensitivity and specificity, and observer variation. Other sources came from bibliographies of retrieved
articles and physical examination textbooks. Studies were included if they
assessed the precision and accuracy of the clinical examination in prognosis
of post–cardiac arrest coma in adults. Eleven studies, involving 1914
patients, met our inclusion criteria.
Data Extraction
Two authors independently reviewed each study to determine eligibility,
abstract data, and classify methodological quality using predetermined criteria.
Disagreement was resolved by consensus.
Data Synthesis
Summary likelihood ratios (LRs) were calculated from random effects
models. Five clinical signs were found to strongly predict death or poor neurological
outcome: absent corneal reflexes at 24 hours (LR, 12.9; 95% confidence interval
[CI], 2.0-68.7), absent pupillary response at 24 hours (LR, 10.2; 95% CI,
1.8-48.6), absent withdrawal response to pain at 24 hours (LR, 4.7; 95% CI,
2.2-9.8), no motor response at 24 hours (LR, 4.9; 95% CI, 1.6-13.0), and no
motor response at 72 hours (LR, 9.2; 95% CI, 2.1-49.4). The proportion of
individuals' dying or having a poor neurological outcome was calculated by
pooling the outcome data from the 11 studies (n = 1914) and used as an estimate
of the pretest probability of poor outcome. The random effects estimate of
poor outcome was 77% (95% CI, 72%-80%). The highest LR increases the pretest
probability of 77% to a posttest probability of 97% (95% CI, 87%-100%). No
clinical findings were found to have LRs that strongly predicted good neurological
outcome.
Conclusions
Simple physical examination maneuvers strongly predict death or poor
outcome in comatose survivors of cardiac arrest. The most useful signs occur
at 24 hours after cardiac arrest, and earlier prognosis should not be made
by clinical examination alone. These data provide prognostic information,
rather than treatment recommendations, which must be made on an individual
basis incorporating many other variables.