Sources of variability in the detection of cerebral emboli with transcranial Doppler during cardiac surgery.

摘要

OBJECTIVE: The application of intensity thresholds for embolus detection with transcranial Doppler (TCD) can exclude from analysis an unrecognized proportion of high-intensity transient signals (HITS))whose intensities are below the threshold. The lack of consistent threshold criteria between clinical trials may explain part of the discrepancy in the reported HITS counts. We investigated the effect of choosing different thresholds on the sensitivity and specificity of detecting HITS during cardiopulmonary bypass (CPB). METHODS: Two observers independently analyzed TCD recordings from 8 patients under CPB. Doppler signals were classified as true HITS, equivocal HITS, artifacts, and Doppler speckles according to preestablished criteria. The relative intensity of Doppler signals was measured by two different methods (TCD software vs manual). Receiver Operating Characteristic curves determined the optimal threshold for each of the two intensity methods. RESULTS: Reviewers achieved agreement in 96% of 2190 Doppler signals (kappa = 0.90). Relative intensities calculated with the TCD-software method were 3 dB (95% CI: 3.0-3.4) higher than the manual method. The optimal threshold was found at 10 dB (sensitivity: 99%; specificity: 90.8%) with the software method and at 7 dB with the manual method (sensitivity: 96%; specificity: 83%). The use of an intensity threshold 2 dB higher than the optimal increased the rejection of true HITS by 8% and 14%, respectively. CONCLUSIONS: Using intensity thresholds higher than the optimal for embolus detection decreases HITS counts. Choosing a threshold depends on the type of method used for measuring the signal intensity. Uniform threshold criteria and comparative studies between different Doppler devices are necessary for making clinical trials more comparable.