New time-frequency method for cerebral autoregulation in newborns: predictive capacity for clinical outcomes.

Abstract

To describe an alternative analysis in the frequency-domain of the temporal relationship between 2 biological signals and evaluate the method's predictive capacity for classifying infants at risk for an adverse outcome. We studied 54 infants (mean gestational age 27 weeks) with invasive mean arterial blood pressure monitoring. The bivariate autoregressive spectral coherence (BiAR-COH) method and the spectral coherence methods were used to analyze the relationship between spontaneous changes in mean arterial blood pressure and the near-infrared tissue oxygenation index. The mean postnatal age at the beginning and end of the autoregulation study was 6.0 (3.0) and 29.0 (7.5) hours, respectively. The BiAR-COH was superior to the spectral coherence in predicting low superior vena cava (SVC) flow (≤ 41 mL/kg per minute), with an area under the receiver operating characteristic curve of 0.84 (95% CI, 0.77-0.90; P < .001). The BiAR-COH threshold for identifying low SVC flow was 0.577, with 0.8 sensitivity and 0.76 specificity. After adjusting for the repeated measures effect (multiple epochs) in a given patient, the averaged BiAR-COH per patient and averaged COH per patient were calculated as the average value per patient. The pBiAR-COH (but not the pCOH) was associated with intraventricular hemorrhage grades 3 and 4 and predicted mortality. The BiAR-COH classifier identifies low SVC flow infants who are at risk for brain hypoperfusion. The BiAR-COH is superior to frequency domain methods in predicting adverse outcomes in infants.