Objectives: The accuracy of noninvasive measurements of arterial blood pressure (BP), in comparison to intra-arterial recordings, is normally quantified by the bias, which is usually assumed to be a constant parameter. We tested the hypothesis that continuous beat-to-beat differences are not random and aimed to describe the temporal pattern of any transient drifts.
Methods: Forty participants were studied after undergoing elective cardiac catheterization. Continuous recordings of noninvasive finger BP (Finapres), intra-aortic BP (Millar catheter-tip transducer) and electrocardiogram were carried out for two periods of 10 min each. The mean of continuous beat-to-beat differences between the two BP sources was removed for the entire recording (null bias) and any linear trends were also removed. Stochastic properties of the time series of differences were described by the autocorrelation function. The temporal pattern of transient drifts was characterized by the coherent averaging of positive and negative transients, synchronized by their peak/trough values.
Results: On average, autocorrelation functions for differences between mean, systolic and diastolic values were significantly greater than zero for at least 24 beats, implying nonrandom differences. For mean BP, the peak value of positive and negative transient drifts were 4.98+/-3.61% and -6.75+/-7.48%, respectively. The average duration of these transients was approximately 20 s. Patients receiving (N=26) and not receiving beta-blockers (N=14) had similar autocorrelation functions indicative of nonrandom differences. The former showed lower peak/trough values in comparison with the latter, but differences were not statistically significant.
Conclusion: The presence of transient drifts in beat-to-beat differences between finger and aortic BP can lead to erroneous results if biases are calculated using very short segments of data. Clinical and/or research applications using the Finapres, such as dynamic cerebral autoregulation modelling or sequence analysis for assessment of baroreceptor sensitivity can also be corrupted by the presence of nonrandom short-term fluctuations of the difference signal unless results are averaged for multiple segments of data.