Predicting neurologic outcome after targeted temperature management for cardiac arrest: systematic review and meta-analysis.

OBJECTIVES: Targeted temperature management improves survival and neurologic outcomes for adult out-of-hospital cardiac arrest survivors but may alter the accuracy of tests for predicting neurologic outcome after cardiac arrest.

DATA SOURCES: We systematically searched Medline, Embase, CINAHL, and CENTRAL from database inception to September 2012.

STUDY SELECTION: Citations were screened for studies that examined diagnostic tests to predict poor neurologic outcome or death following targeted temperature management in adult cardiac arrest survivors.

DATA EXTRACTION: Data on study outcomes and quality were abstracted in duplicate. We constructed contingency tables for each diagnostic test and calculated sensitivity, specificity, and positive and negative likelihood ratios.

DATA SYNTHESIS: Of 2,737 citations, 20 studies (n = 1,845) met inclusion criteria. Meta-analysis showed that three tests accurately predicted poor neurologic outcome with low false-positive rates: bilateral absence of pupillary reflexes more than 24 hours after a return of spontaneous circulation (false-positive rate, 0.02; 95% CI, 0.01-0.06; summary positive likelihood ratio, 10.45; 95% CI, 3.37-32.43), bilateral absence of corneal reflexes more than 24 hours (false-positive rate, 0.04; 95% CI, 0.01-0.09; positive likelihood ratio, 6.8; 95% CI, 2.52-18.38), and bilateral absence of somatosensory-evoked potentials between days 1 and 7 (false-positive rate, 0.03; 95% CI, 0.01-0.07; positive likelihood ratio, 12.79; 95% CI, 5.35-30.62). False-positive rates were higher for a Glasgow Coma Scale motor score showing extensor posturing or worse (false-positive rate, 0.09; 95% CI, 0.06-0.13; positive likelihood ratio, 7.11; 95% CI, 5.01-10.08), unfavorable electroencephalogram patterns (false-positive rate, 0.07; 95% CI, 0.04-0.12; positive likelihood ratio, 8.85; 95% CI, 4.87-16.08), myoclonic status epilepticus (false-positive rate, 0.05; 95% CI, 0.02-0.11; positive likelihood ratio, 5.58; 95% CI, 2.56-12.16), and elevated neuron-specific enolase (false-positive rate, 0.12; 95% CI, 0.06-0.23; positive likelihood ratio, 4.14; 95% CI, 1.82-9.42). The specificity of available tests improved when these were performed beyond 72 hours. Data on neuroimaging, biomarkers, or combination testing were limited and inconclusive.

CONCLUSION: Simple bedside tests and somatosensory-evoked potentials predict poor neurologic outcome for survivors of cardiac arrest treated with targeted temperature management, and specificity improves when performed beyond 72 hours. Clinicians should use caution with these predictors as they carry the inherent risk of becoming self-fulfilling.