Objective: To investigate whether accurate prognostic rules can be derived from the combined results of studies concerning prediction of poor prognosis in anoxic-ischaemic coma with biochemical markers of brain damage in cerebrospinal fluid (CSF) or serum.
Design: A meta-analysis of prognostic studies in anoxic-ischaemic coma, selected from Medline and EMBASE databases, according to predefined criteria.
Subjects: Twenty-eight studies, with a total of 802 unselected, consecutive patients, in which tests, sampling time and outcome measures were described unequivocally and results were described using clear cut-off values or raw data.
Main outcome measures: Poor outcome, defined as death or vegetative state, versus good outcome, defined as any other outcome state.
Analyses: The overall prognostic accuracy of these variables was expressed as the 95% CIs of the pooled false-positive test rate and the pooled positive-likelihood ratios.
Results: Only markers in CSF (creatine kinase isoenzyme (CKBB) >204 U/l, neuron specific enolase (NSE) >33 ng/ml, lactate dehydrogenase (LDH) >82 U/l and glutamate oxaloacetate (GOT) >62 U/l) reached a 0% false-positive rate. However, due to small sample sizes, the confidence limits were wide. The accuracy of prediction of poor outcome seemed acceptably high for CSF-CKBB (pooled false-positive rate 0% [95% CI 0-2.3%]; pooled positive-likelihood ratio 33.2 [95% CI 4.8-230.2]), but this result was based on two retrospective studies without blinding of the treating physicians for the test result.
Conclusions: Because of small numbers of patients studied and methodological limitations the combined results are not sufficiently accurate to provide a solid basis for non-treatment decisions.