Reliability of the mean flow index (Mx) for assessing cerebral autoregulation in healthy volunteers

BACKGROUND Mean flow index (Mxa) for evaluating dynamic cerebral autoregulation is derived using varying approaches for calculation, which may explain that the reliability ranges from poor to excellent. The comparability, repeatability, stability, and internal consistency of approaches have not previously been assessed. METHODS We included 60 recordings from resting healthy volunteers and calculated Mxa using four different approaches: three without overlapping calculations, using intervals for averaging wave-form data (blocks) of 3, 6, and 10 s, and correlation periods (epochs) of 60, 240, and 300 s (3-60-F, 6-240-F, and 10-300-F); and one using 10-second blocks, 300 s epochs, and overlaps of 60 s (10-300-60). The comparability between the approaches was assessed using Student's t test, intraclass correlation coefficients (ICC), and Bland-Altman plot. RESULTS Overall, 3-60-F resulted in a higher Mxa than the other indices (p < 0.001, for all). The reliability when comparing all the approaches ranged from moderate to good (ICC: 0.68; 95%CI: 0.59-0.84), which was primarily due to similarities between 10-300-F and 10-300-60 (ICC: 0.94; 95%CI: 0.86-0.98). The reliability when comparing the first and last half was poor for 10-300-F and ranged from poor to moderate for the other approaches. Additional random artifacts resulted in poor reliability for 10-300-F, while the other approaches were more stable. CONCLUSIONS Mxa in general has a low sensitivity to artifacts, but otherwise seems highly dependent on the approach, with a repeatability that is moderate at best. The varying accuracy and precision renders Mxa unreliable for classifying impaired cerebral autoregulation when using healthy adults for comparison.