Ch 28 (s28.6, s28.18); Ch 16
The signal is the asymmetry between hemispheres, not the absolute saturation that clinicians actually watch. A single-sensor installation discards the information. 98 percent specificity with 39 percent sensitivity is the profile of a rule-in test: sustained asymmetry over 8 percent should trigger imaging, and a symmetrical normal trace excludes almost nothing — it missed 61 percent of injuries in this series.
Cerebral desaturation alone (consecutive hours of regional saturation under 40 percent)
DOI 10.1097/MAT.0000000000002032
Mean left-right regional saturation asymmetry across the run, and asymmetric desaturation (saturation under 40 percent with asymmetry over 10 percent for over 10 consecutive hours)
Single centre, 69 patients, 18 events. The confidence interval of 4.1 to 381.4 is a direct signal of sparse data and the point estimate of 39.4 should not be quoted as a stable effect size. Both thresholds (8 percent; 10 percent over 10 hours) were derived in this dataset and have not been prospectively validated. Cannot be reconciled with the sensitivity 86 percent and specificity 55 percent figures attributed in the ELSO Red Book to Hunt and colleagues.
Mean asymmetry over 8 percent: adjusted odds ratio 39.4 (95 percent CI 4.1 to 381.4) for acute brain injury, sensitivity 39 percent, specificity 98 percent, AUC 0.86. Asymmetric desaturation: adjusted odds ratio 5.2 (1.2 to 22.2), sensitivity 33 percent, specificity 88 percent, AUC 0.72. Neither criterion alone was significantly associated with injury.
69 adult venoarterial ECMO patients (mean age 56 years, 65 percent male) undergoing cerebral near-infrared spectroscopy with left and right sensors, values collected hourly from cannulation. 18 patients (26 percent) had acute brain injury, defined as intracranial haemorrhage, ischaemic stroke, hypoxic-ischaemic brain injury or brain death during ECMO.