Neurocrit Care (2014) 20:60–68 DOI 10.1007/s12028-013-9898-y
ORIGINAL ARTICLE
Comparison of Non-invasive and Invasive Arterial Blood Pressure Measurement for Assessment of Dynamic Cerebral Autoregulation Nils H. Petersen • Santiago Ortega-Gutierrez • Andres Reccius Arjun Masurkar • Amy Huang • Randolph S. Marshall
•
Published online: 14 November 2013 Ó Springer Science+Business Media New York 2013
Abstract Background There is a growing interest in measuring cerebral autoregulation in patients with acute brain injury. Non-invasive finger photo-plethysmography (Finapres) is the method of choice to relate arterial blood pressure to changes in cerebral blood flow. Among acutely ill patients, however, peripheral vasoconstriction often limits the use of Finapres requiring direct intravascular blood pressure measurement. We evaluated how these two different forms of blood pressure monitoring affect the parameters of dynamic cerebral autoregulation (DCA). Methods We performed 37 simultaneous recordings of BP and cerebral blood flow velocity in 15 patients with acute brain injury. DCA was estimated in the frequency domain using transfer function analysis to calculate phase shift, gain, and coherence. In addition the mean velocity index (Mx) was calculated for assessment of DCA in the time domain. Results The mean patient age was 58.1 ± 15.9 years, 80 % (n = 12) were women. We found good inter-method agreement between Finapres and direct intravascular measurement using Bland–Altman and correlation analyses. Finapres gives higher values for the efficiency of N. H. Petersen S. Ortega-Gutierrez A. Masurkar A. Huang R. S. Marshall Stroke Division, Department of Neurology, Columbia University, New York, NY, USA N. H. Petersen (&) Neuroscience Intensive Care Unit, Massachusetts General Hospital, 55 Fruit Street, Lunder 650, Boston, MA 02114, USA e-mail: [email protected] A. Reccius Department of Critical Care, Clinica Alemana, Universidad del Desarrollocation, Santiago, Chile
123
dynamic CA compared with values derived from radial artery catheter, as indicated by biases in the phase (26.3 ± 11.6° vs. 21.7 ± 10.5°, p = 0.001) and Mx (0.571 ± 0.137 vs. 0.649 ± 0.128, p < 0.001). Gain in the low frequency range did not significantly differ between the two arterial blood pressure methods. The average coherence between CBFV and ABP was higher when BP was measured with arterial catheter for frequencies above 0.05 Hz (0.8 vs. 0.73, p < 0.001). Conclusion Overall, both methods yield similar results and can be used for the assessment of DCA. However, there was a small but significant difference for both mean Mx and phase shift, which would need to be adjusted for during monitoring of patients when using both methods. When available, invasive arterial blood pressure monitoring may improve accuracy and thus should be the preferred method for DCA assessment in the ICU. Keywords Dynamic cerebral autoregulation Cerebral blood flow Transcranial Doppler ultrasound Arterial blood pressure Finapres Transfer function analysis
Introduction Cerebral autoregulation is the homeostatic mechanism whereby the brain maintains a relatively constant cerebral blood flow despite changes in systemic blood pressure and hence cerebral perfusion pressure (CPP). [1, 2] Short-term fluctuations in CPP prompt adjustments in cerebrovascular resistance via complex neurogenic, myogenic, and metabolic mechanisms to preserve a stable cerebral blood flow. This process is known as dynamic cerebral autoregulation (DCA). DCA insures maintenance of an adequate supply of
Neurocrit Care (2014) 20:60–68
glucose and oxygen to the brain to meet its high metabolic demands. Impaired autoregulation has been described after traumatic brain injury (TBI), [3–5] ischemic stroke [6–9], and intracerebral hemorrhage (ICH) [10, 11], and has been associated with worse outcomes [5, 10]. To improve outcomes unregulated cerebral blood flow may require active blood pressure management in the NICU setting within very specific parameters. This concept has led to a growing interest in measuring autoregulation to guide clinical management. However, there is currently no ‘‘gold standard’’ of measuring autoregulation; various methods have been described. Transcranial Doppler ultrasonography (TCD) combined with servo-controlled finger photoplethysmography (Finapres) has emerged as a non-invasive means of assessing DCA. This technique gives excellent temporal resolution by measuring the blood flow velocity response to induced or spontaneous changes in blood pressure [12, 13]. The two most commonly used methods to assess DCA are the time correlation method and the transfer function analysis. The time correlation method measures cerebral blood flow velocity (CBFV) and systemic blood pressure simultaneously for a given time period, and yields the correlation coefficient between time-averaged CBFV and MAP, or mean velocity index (Mx) [4]. Alternatively, DCA can be assessed from spontaneous oscillations in blood pressure and middle cerebral artery flow velocity at particular frequencies using transfer function analysis [14, 15]. The first method uses the data in a time domain; the second uses the same data in a frequency domain. The two methods have been cross-validated non-invasively using Finapres [16]. Finapres is not always possible in the NICU population, however, due to peripheral vasoconstriction [17]: such patients may be treated with vasoactive medications or have high sympathetic tone. In these cases blood pressure is commonly measured invasively via intra-arterial catheter. Although non-invasive measurement of arterial blood pressure (ABP) with the Finapres device has been validated against intravascular ABP for resting measurements [17], very little is known about whether they are equivalent in assessing DCA. We therefore tested the hypothesis that invasive blood pressure monitoring using arterial catheter can achieve comparable results to assessment with Finapres.
Methods
61
monitor ABP for clinical purposes. No further selection criteria were applied as each patient served as his/her own control; our aim was to achieve broad generalizability of the results. Consent was obtained from the patient or surrogate decision maker prior to enrollment. All patients were provided with detailed written information regarding the intent and methods of the study. Approval for the study was obtained from the Institutional Review Board at Columbia University Medical Center. DCA measurements were performed in supine position with the subjects breathing spontaneously. Cerebral blood flow velocities were assessed using transcranial Doppler (DWL-Multidop-X, Sipplingen, Germany). The proximal middle cerebral artery was insonated through the temporal window with a 2 MHz probe attached to a head frame. Depth of insonation was 45–60 mm. Blood pressure was recorded simultaneously from the intravascular catheter inserted into the radial artery and via servo-controlled finger plethysmography (Finometer Pro, Amsterdam, the Netherlands). The appropriate finger cuff (Size: small, medium or large) was placed on the middle phalanx of the left or right middle finger. After establishing a stable recording, the Physiocal procedure, an intermittently occurring calibration routine, was turned off. Data were recorded for 10 min. All analog signals were digitized and stored for editing and offline analysis. Data sampling frequency was 100 Hz. Data Analysis An initial editing process was performed to provide temporal synchronization of the blood pressure and blood flow velocity waveforms using ICUpilot software (Dipylon Medical, Solna, Sweden). This was followed by visual inspection and removal of all major artifacts (due to brief loss of TCD or ABP signal or other noise). The relationship between changes in arterial pressure and changes in CBFV was assessed with transfer function analysis and time correlation method with an in-house program using Matlab (MathWorks, Natick, USA), described below. Transfer Function Analysis Estimation of spectra and transfer function was based on the method described by Welch [18]. In brief, ABP and CBFV were normalized with respect to their mean prior to frequency spectrum analysis. The transfer function H(f) relating each CBFV to ABP was approximated by assuming linearity and time invariance and calculated as follows:
Subjects and Measurements Hð f Þ ¼ Patients admitted to the Neurological ICU at Columbia University Medical Center were eligible for the study. All patients underwent placement of a radial artery catheter to
Sxy ðf Þ Sxx ðf Þ
where Sxx(f) represented the autospectrum of ABP and Sxy(f) represented the cross spectrum of ABP and each
123
62
CBF. The gain |H(f)| and phase U(f) of the system were calculated from H(f) as: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi jHðf Þj ¼ jHr ðf Þj2 þjHi ðf Þj2 Hi ðf Þ Uðf Þ ¼ arctan Hr ðf Þ where Hr(f) represented the real component and Hi(f) represented the imaginary component of the transfer function H(f). The squared coherence c2(f) was approximated as follows: Sxy ðf Þ2 2 c ðf Þ ¼ Sxx ðf ÞSyy ðf Þ where Sxy(f) and Sxx(f) were as described above and Syy(f) represented the autospectrum of each CBF. Spectra were calculated using individual segments 1/6th the total length of the time series with 50 % overlap. Given the 10-min recordings at 100 Hz, this resulted in segments of 10000 points with 5000 points of overlap. Smoothing was accomplished by employing a Hanning window w(n) of length L equivalent to the individual segment: n wðnÞ ¼ 0:5 1 cos 2p ;0nN N where L = N + 1. Coherence significance criterion (cmin), above which coherence differs significantly from 0, was derived from the degrees of freedom m of the spectral estimate at a significance level a of 0.05 [15, 19]. pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 cmin ¼ 1 av2 Degrees of freedom are dependent on the details of the spectral estimation method. Estimates of the degrees of freedom of the Welch method with smoothing nearly identical to the Hanning window have been determined [18]. Employing this estimate, with 50 % overlap between segments, m equals 19.6. Using the above formula for coherence significance level, cmin equals 0.5372. Average phase shift and gain were therefore calculated in the low frequency range (0.06–0.12 Hz) with a minimum coherence of 0.53. This frequency range was chosen because as it has been shown to be the most meaningful measure of autoregulation in previous studies [8, 15, 16, 20]. Less (lower) phase shift indicated poorer autoregulation. Time Correlation Method The Mx was first described in patients with TBI [4] and later adapted to patients with other types of neurologic disease, such as stroke or carotid occlusion [7, 16]. The Mx was calculated according to a previously described method [16]. In brief, CBFV and ABP were averaged over 3-s
123
Neurocrit Care (2014) 20:60–68
intervals with the use of waveform time integration. For every 20 consecutive values of averaged ABP and CBFV a Pearson’s correlation coefficient was calculated for each method of ABP recording. The resulting set of 1-min correlation coefficients was then averaged, yielding the Mx. A positive coefficient signified a positive association between FV and perfusion pressure, which may be interpreted as impaired autoregulation. A very low correlation coefficient signified a lack of association between the two signals, implying intact autoregulation. Some authors regard a correlation coefficient of >0.3 as a threshold for disturbed autoregulation [21]. It is important to note that there is considerable heterogeneity in the calculation of the correlation coefficient and other lengths of the primary data segments (5 or 10 s instead of 3 s), or the use of a moving window for calculation of the correlation coefficients have been used. In addition, longer periods of monitoring have been suggested to filter out random fluctuations of ABP and CBFV [22]. For reasons of simplicity, since the aim of our study was to evaluate how different forms of BP monitoring affect parameters of DCA and therefore length of data assessment is not such a relevant factor, we decided on 10-min recordings. Statistical Analysis Continuous variables were tested for normality using the Shapiro–Wilk test. Normally distributed data were reported as means and standard deviation (SD); non-normally distributed as median, 25th and 75th percentile. The agreement between the estimates of DCA (phase, gain, and Mx) obtained via Finapres and invasive intra-arterial catheter was assessed with Bland–Altman analysis for repeated measurements [23, 24]. For each subject the difference between two parameters (e.g., phase from Finapres and arterial catheter) was plotted against the mean of these two parameters. Given approximate normality in the distribution of the differences, the 95 % confidence interval for inter-method agreement is the mean difference ±1.96 standard deviations. A small mean difference indicates small inter-method bias, while a small variance indicates good inter-method agreement. For additional assessment of agreement, Pearson’s correlation coefficient between subject means was calculated. The dynamic relationship between the two ABP methods was also quantified by the coherence function between them. For analysis of intermethod differences, a paired t test was used to compare means, Wilcoxon signed-rank test for median comparison. To account for the correlation among repeated measures in the same person, mean scores for each subject were calculated [25]. A p value of less than 0.05 was considered statistically significant.
Neurocrit Care (2014) 20:60–68
63
Results Twenty patients were recruited for the study; 5 were excluded due to inability to obtain a good Finapres signal. The final analysis contained 15 patients in whom we measured CBFV with TCD and blood pressure with arterial catheter (ABPArt) and Finapres (ABPFin). In 10 patients we performed repeated measurements for a total of 37 observations for each hemisphere. Admission diagnoses included aneurysmal subarachnoid hemorrhage (n = 7), acute ischemic stroke (n = 4), ICH (n = 3), and reversible cerebral vasoconstriction syndrome (n = 1). The mean age was 58.1 ± 15.9 years, 80 % (n = 12) were women. Basic demographics, clinical characteristics, and medications assessed at each measurement are given in Table 1. ABP was similar for Finapres and arterial catheter (mean difference 1.92 mmHg, 95 % CI -5.82 to 1.99), p = 0.31). Figure 1 illustrates a high coherence between mean ABPFin and mean ABPArt for frequencies above 0.05 Hz Table 2.
Table 1 Demographics and Baseline Characteristics Parameters Age, years (mean, SD)
58.1 (15.9)
Gender (male/female)
12/3
Ethnicity (n, %) Hispanic
9 (60)
White
5 (33.3)
Black
1 (6.7)
Diagnosis (n, %) SAH
7 (46.67)
Ischemic Stroke
4 (26.7)
ICH
3 (20)
RCVS
1 (6.7)
Our study patients demonstrated significant impairment in DCA with abnormal values for phase shift and Mx, which is likely due to their underlying neurological injury. The mean phase shift in the low frequency range (0.06–0.12 Hz) was 26.3 ± 11.6° with Finapres, and 21.7 ± 10.5° using arterial catheter (p = 0.001). Gain in the LF range did not significantly differ between the two ABP methods. The mean difference in the autoregulatory index Mx between Finapres and arterial catheter was 0.08 (95 % CI 0.01–0.05, p < 0.0001). In summary, the autoregulatory parameters measured with invasive ABP assessment resulted in lower absolute values for both the transfer analysis and time correlation methods compared with the Finapres. Figure 2a–c illustrates this finding, showing the phase shift, gain, and Mx group averages for both Finapres and arterial catheter. The average coherence between CBFV and ABP shows consistently higher values for arterial catheter when compared with Finapres for frequencies above 0.05 Hz (Fig. 2a). Median coherence was excellent for both methods—on the left side median coherence was 0.79 (0.65, 0.85) for Finapres and 0.87 (0.75, 0.92) for arterial line (p < 0.0073). For the right side values were 0.72 (0.52, 0.81) for Finapres and 0.82 (0.71, 0.86) when measured with arterial catheter (p < 0.0001). The correlation between the autoregulatory parameters phase shift and Mx was r = 0.54, p < 0.01 for Finapres and r = 0.55, p < 0.01 for arterial catheter. Figure 3a–c shows the correlation between estimates of DCA obtained via Finapres and corresponding values derived from arterial catheter. Correlation coefficients and mean biases are given in Table 3. For the calculation of phase shift and gain 1.1 1
Medications taken at time of measurement (n, %)
0.9
31 (83.8)
b-Adrenoreceptor Antagonist
9 (24.3)
0.8
ACE/ARB
3 (8.1)
0.7
Propofol
4 (10.8)
Dexmedetomidine
1 (2.7)
Phenylephrine
4 (10.8)
Coherence
Calcium channel blocker
0.6 0.5 0.4
Dopamine ABPFin, mmHg (mean, SD)
2 (5.4) 99.35 (17.2)
0.3
ABPArt, mmHg (mean, SD)
101.3 (17.3)
0.2
CBFV Left, cm s-1 (mean, SD)
75.4 (38.5)
CBFV Right, cm s-1 (mean, SD)
75.4 (23.1)
Study population n = 15, measurements n = 37. Medications were assessed with each measurement SAH subarachnoid hemorrhage, ICH intracerebral hemorrhage, RCVS reversible cerebral vasoconstriction syndrome, ACE angiotensin converting enzyme inhibitor, ARB angiotensin II receptor antagonist, ABP arterial blood pressure, CBFV cerebral blood flow velocity
0.1 0 0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Frequency (Hz) Fig. 1 Average coherence function between mean arterial blood pressure obtained from intravascular catheter and Finapres (solid line) and standard deviation (dashed line)
123
64
Neurocrit Care (2014) 20:60–68
Table 2 Descriptive statistics of autoregulatory parameters for transfer function and time correlation method Side Left
N
Phase shift Finapres
Phase shift arterial
Gain Finapres
Gain arterial
Coherence Finapres
Coherence arterial
Mx Finapres
Mx arterial
Valid
35
35
35
35
37
37
37
37
Missing
2
2
2
2
0
0
0
0
Mean
24.67
20.50
0.67
0.81
0.73
0.81
0.58
0.66
Median
27.12
20.83
0.46
0.54
0.78
0.86
0.60
0.67
8.87
9.61
0.57
0.72
0.14
0.12
0.14
0.12
10.68
5.55
0.11
0.09
0.41
0.47
0.35
0.46
25
36.10 14.56
34.11 13.72
1.87 0.22
2.37 0.28
0.89 0.65
0.91 0.76
0.77 0.45
0.84 0.56
50
27.12
20.83
0.46
0.54
0.78
0.86
0.60
0.67
75
32.55
30.31
0.88
1.37
0.82
0.88
0.70
0.75
Valid
35
35
35
35
37
37
37
37
Missing
SD Minimum Maximum Percentiles
Right
N
2
2
2
2
0
0
0
0
Mean
27.92
22.82
0.78
0.83
0.72
0.79
0.56
0.64
Median
27.41
21.62
0.62
0.57
0.72
0.81
0.58
0.65
SD
13.99
11.54
0.52
0.50
0.11
0.10
0.14
0.14
Minimum
-6.53
-2.93
0.27
0.25
0.47
0.60
0.33
0.39
Maximum
57.44
50.55
1.90
1.94
0.90
0.93
0.78
0.88
Percentiles
25
22.60
19.26
0.40
0.37
0.66
0.74
0.41
0.54
50
27.41
21.62
0.62
0.57
0.72
0.81
0.58
0.65
75
33.33
26.62
1.15
1.19
0.78
0.86
0.68
0.73
Mx mean velocity index
two measurements had to be removed because of coherence less than 0.53 in the LF range. Figure 4a–c shows the Bland–Altman plots comparing phase shift, gain, and Mx measured by Finapres and arterial catheter. The differences did not vary in a systematic way over the range of measurement.
Discussion There is considerable methodological diversity among studies measuring DCA. TCD combined with Finapres has been the method of choice for the non-invasive assessment of DCA. Among acutely ill patients in the intensive care setting, however, vasoconstriction often limits the use of non-invasive finger plethysmography, and direct intravascular blood pressure measurement is used for blood pressure monitoring. Thus it is important that arterial approaches also be available for the assessment of cerebral autoregulation in this population [5]. Precision and accuracy comparing Finapres to invasive arterial assessment has been demonstrated in measuring blood pressure [17]; however, it has been less certain whether these different techniques may be used interchangeably to measure cerebral autoregulation. Lavinio et al. compared the autoregulatory index calculated as the correlation coefficient between fluctuations of CPP and
123
CBFV with a completely non-invasive approach using Finapres instead of CPP. For the calculation of CPP they measured intravascular radial artery blood pressure and intracranial pressure [26]. This implied two assumptions: (1) negligibility of ICP changes on DCA calculation and (2) accuracy of Finapres blood pressure slow-wave estimation. Although the non-invasive Mx satisfactorily described DCA in comparison to the invasive approach (R = 0.755), limits of agreement calculated according to Bland–Altman were fairly wide (±0.36). It remained unclear, whether this was due differences in blood pressure measurement or the result of ICP changes. In our study, we found good agreement between the two methods as blood pressure sources. There was no significant difference in mean ABP assessed with Finapres and arterial catheter with a mean bias of only 1.9 mmHg. Coherence between the two signals was excellent for frequencies above 0.05 Hz. The drop in coherence for very low frequencies has been described before and might be due to greater variability in mean ABP with Finapres in this frequency range. Our study thus confirms the results of others that showed similar beat-to-beat blood pressure measurements comparing Finapres to intra-arterial catheter [27, 28]. Importantly, we found a good correlation between the estimates of DCA obtained with Finapres and arterial
Neurocrit Care (2014) 20:60–68
A
C
65
B
D
Fig. 2 Average coherence (a), phase shift (b), gain (c), and Mx (d) between mean ABP and CBFV for Finapres (solid line) and arterial catheter (dashed line). Gray areas represent LF range (0.06–0.12 Hz)
catheter. However, there was a significant difference in absolute values for mean phase shift and Mx. Since both were calculated using the same CBFV signal, any difference measured is attributable to differences in the ABP signals. As measured with the arterial catheter, the Mx was on average 0.08 points higher (suggesting poorer autoregulation) when compared to Finapres. It is possible that slightly increased noise in the Finapres signal led to the lower correlation coefficient. It becomes more difficult to explain the differences in phase shift, which was on average 4.6° lower (suggesting poorer autoregulation) with arterial catheter when compared to Finapres. Our findings are in agreement with the only other study testing the effects of invasive ABP measurement on parameters of DCA. Sammons et al. [28] analyzed phase shifts obtained via Finapres and intra-aortic catheter in 27 patients undergoing coronary catheterization. The absolute values obtained in their study using arterial catheter were 5.7° lower than values obtained with Finapres. They suggested that this might be due to the difference in vascular compliance between aorta and peripheral circulation, leading to the bias in autoregulation parameters. We found a similar bias and all our measurements were obtained from the radial artery making it less likely that amplification of the pulse pressure waveform from aorta to the peripheral
circulation is an important factor. However, it is still possible that the slightly more distal location of the Finapres may have contributed to the difference in phase shift, as this site has been shown to have a greater effect on rhythmic changes in arteriolar tone, thus resulting in increased phase shift [29]. Sammons et al. further hypothesized that medication effect might have played a role, as 76 % of their patients were taking beta-adrenergic antagonists. Our results do not support this hypothesis as we found similar discrepancies, but only 24 % of the patients in our study were taking beta-adrenergic antagonists. In addition, the phase shift bias was not significantly different among those taking medication compared to those without (5.6° vs. 3.7°, p = 0.819). An alternative explanation is that Finapres is not able to precisely reproduce the dynamic changes in the low frequency range as well as invasive monitoring. Omboni and colleagues found that Finapres overestimated systolic blood pressure powers also in the very low-frequency range, which may have resulted in the increased phase shift. They hypothesized that these discrepancies between finger and intra-arterial data are due to an intrinsic defect of the Finapres method. Although absolute differences in measurements of autoregulation may not matter in patients being followed with one method or the other as individual
123
66
Neurocrit Care (2014) 20:60–68
A
B
C
Fig. 3 Scatterplot with regression line of phase shift (a) gain (b), and Mx (c) measured with Finapres and arterial catheter. Triangles represent values from the left MCA, circles those from the right MCA Table 3 Bias and correlation coefficients of parameters of autoregulation using Finapres and Arterial catheter Parameter
Left phase shift (°) Right phase shift (°)
Bias ABPFin-Art
4.17
SD
7.21
95 % CI of the Difference Lower
Upper
0.18
8.17
p value
Correlation coefficient
0.042
0.70
p value
0.004
5.10
6.77
1.36
8.85
0.011
0.88
ORIGINAL ARTICLE
Comparison of Non-invasive and Invasive Arterial Blood Pressure Measurement for Assessment of Dynamic Cerebral Autoregulation Nils H. Petersen • Santiago Ortega-Gutierrez • Andres Reccius Arjun Masurkar • Amy Huang • Randolph S. Marshall
•
Published online: 14 November 2013 Ó Springer Science+Business Media New York 2013
Abstract Background There is a growing interest in measuring cerebral autoregulation in patients with acute brain injury. Non-invasive finger photo-plethysmography (Finapres) is the method of choice to relate arterial blood pressure to changes in cerebral blood flow. Among acutely ill patients, however, peripheral vasoconstriction often limits the use of Finapres requiring direct intravascular blood pressure measurement. We evaluated how these two different forms of blood pressure monitoring affect the parameters of dynamic cerebral autoregulation (DCA). Methods We performed 37 simultaneous recordings of BP and cerebral blood flow velocity in 15 patients with acute brain injury. DCA was estimated in the frequency domain using transfer function analysis to calculate phase shift, gain, and coherence. In addition the mean velocity index (Mx) was calculated for assessment of DCA in the time domain. Results The mean patient age was 58.1 ± 15.9 years, 80 % (n = 12) were women. We found good inter-method agreement between Finapres and direct intravascular measurement using Bland–Altman and correlation analyses. Finapres gives higher values for the efficiency of N. H. Petersen S. Ortega-Gutierrez A. Masurkar A. Huang R. S. Marshall Stroke Division, Department of Neurology, Columbia University, New York, NY, USA N. H. Petersen (&) Neuroscience Intensive Care Unit, Massachusetts General Hospital, 55 Fruit Street, Lunder 650, Boston, MA 02114, USA e-mail: [email protected] A. Reccius Department of Critical Care, Clinica Alemana, Universidad del Desarrollocation, Santiago, Chile
123
dynamic CA compared with values derived from radial artery catheter, as indicated by biases in the phase (26.3 ± 11.6° vs. 21.7 ± 10.5°, p = 0.001) and Mx (0.571 ± 0.137 vs. 0.649 ± 0.128, p < 0.001). Gain in the low frequency range did not significantly differ between the two arterial blood pressure methods. The average coherence between CBFV and ABP was higher when BP was measured with arterial catheter for frequencies above 0.05 Hz (0.8 vs. 0.73, p < 0.001). Conclusion Overall, both methods yield similar results and can be used for the assessment of DCA. However, there was a small but significant difference for both mean Mx and phase shift, which would need to be adjusted for during monitoring of patients when using both methods. When available, invasive arterial blood pressure monitoring may improve accuracy and thus should be the preferred method for DCA assessment in the ICU. Keywords Dynamic cerebral autoregulation Cerebral blood flow Transcranial Doppler ultrasound Arterial blood pressure Finapres Transfer function analysis
Introduction Cerebral autoregulation is the homeostatic mechanism whereby the brain maintains a relatively constant cerebral blood flow despite changes in systemic blood pressure and hence cerebral perfusion pressure (CPP). [1, 2] Short-term fluctuations in CPP prompt adjustments in cerebrovascular resistance via complex neurogenic, myogenic, and metabolic mechanisms to preserve a stable cerebral blood flow. This process is known as dynamic cerebral autoregulation (DCA). DCA insures maintenance of an adequate supply of
Neurocrit Care (2014) 20:60–68
glucose and oxygen to the brain to meet its high metabolic demands. Impaired autoregulation has been described after traumatic brain injury (TBI), [3–5] ischemic stroke [6–9], and intracerebral hemorrhage (ICH) [10, 11], and has been associated with worse outcomes [5, 10]. To improve outcomes unregulated cerebral blood flow may require active blood pressure management in the NICU setting within very specific parameters. This concept has led to a growing interest in measuring autoregulation to guide clinical management. However, there is currently no ‘‘gold standard’’ of measuring autoregulation; various methods have been described. Transcranial Doppler ultrasonography (TCD) combined with servo-controlled finger photoplethysmography (Finapres) has emerged as a non-invasive means of assessing DCA. This technique gives excellent temporal resolution by measuring the blood flow velocity response to induced or spontaneous changes in blood pressure [12, 13]. The two most commonly used methods to assess DCA are the time correlation method and the transfer function analysis. The time correlation method measures cerebral blood flow velocity (CBFV) and systemic blood pressure simultaneously for a given time period, and yields the correlation coefficient between time-averaged CBFV and MAP, or mean velocity index (Mx) [4]. Alternatively, DCA can be assessed from spontaneous oscillations in blood pressure and middle cerebral artery flow velocity at particular frequencies using transfer function analysis [14, 15]. The first method uses the data in a time domain; the second uses the same data in a frequency domain. The two methods have been cross-validated non-invasively using Finapres [16]. Finapres is not always possible in the NICU population, however, due to peripheral vasoconstriction [17]: such patients may be treated with vasoactive medications or have high sympathetic tone. In these cases blood pressure is commonly measured invasively via intra-arterial catheter. Although non-invasive measurement of arterial blood pressure (ABP) with the Finapres device has been validated against intravascular ABP for resting measurements [17], very little is known about whether they are equivalent in assessing DCA. We therefore tested the hypothesis that invasive blood pressure monitoring using arterial catheter can achieve comparable results to assessment with Finapres.
Methods
61
monitor ABP for clinical purposes. No further selection criteria were applied as each patient served as his/her own control; our aim was to achieve broad generalizability of the results. Consent was obtained from the patient or surrogate decision maker prior to enrollment. All patients were provided with detailed written information regarding the intent and methods of the study. Approval for the study was obtained from the Institutional Review Board at Columbia University Medical Center. DCA measurements were performed in supine position with the subjects breathing spontaneously. Cerebral blood flow velocities were assessed using transcranial Doppler (DWL-Multidop-X, Sipplingen, Germany). The proximal middle cerebral artery was insonated through the temporal window with a 2 MHz probe attached to a head frame. Depth of insonation was 45–60 mm. Blood pressure was recorded simultaneously from the intravascular catheter inserted into the radial artery and via servo-controlled finger plethysmography (Finometer Pro, Amsterdam, the Netherlands). The appropriate finger cuff (Size: small, medium or large) was placed on the middle phalanx of the left or right middle finger. After establishing a stable recording, the Physiocal procedure, an intermittently occurring calibration routine, was turned off. Data were recorded for 10 min. All analog signals were digitized and stored for editing and offline analysis. Data sampling frequency was 100 Hz. Data Analysis An initial editing process was performed to provide temporal synchronization of the blood pressure and blood flow velocity waveforms using ICUpilot software (Dipylon Medical, Solna, Sweden). This was followed by visual inspection and removal of all major artifacts (due to brief loss of TCD or ABP signal or other noise). The relationship between changes in arterial pressure and changes in CBFV was assessed with transfer function analysis and time correlation method with an in-house program using Matlab (MathWorks, Natick, USA), described below. Transfer Function Analysis Estimation of spectra and transfer function was based on the method described by Welch [18]. In brief, ABP and CBFV were normalized with respect to their mean prior to frequency spectrum analysis. The transfer function H(f) relating each CBFV to ABP was approximated by assuming linearity and time invariance and calculated as follows:
Subjects and Measurements Hð f Þ ¼ Patients admitted to the Neurological ICU at Columbia University Medical Center were eligible for the study. All patients underwent placement of a radial artery catheter to
Sxy ðf Þ Sxx ðf Þ
where Sxx(f) represented the autospectrum of ABP and Sxy(f) represented the cross spectrum of ABP and each
123
62
CBF. The gain |H(f)| and phase U(f) of the system were calculated from H(f) as: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi jHðf Þj ¼ jHr ðf Þj2 þjHi ðf Þj2 Hi ðf Þ Uðf Þ ¼ arctan Hr ðf Þ where Hr(f) represented the real component and Hi(f) represented the imaginary component of the transfer function H(f). The squared coherence c2(f) was approximated as follows: Sxy ðf Þ2 2 c ðf Þ ¼ Sxx ðf ÞSyy ðf Þ where Sxy(f) and Sxx(f) were as described above and Syy(f) represented the autospectrum of each CBF. Spectra were calculated using individual segments 1/6th the total length of the time series with 50 % overlap. Given the 10-min recordings at 100 Hz, this resulted in segments of 10000 points with 5000 points of overlap. Smoothing was accomplished by employing a Hanning window w(n) of length L equivalent to the individual segment: n wðnÞ ¼ 0:5 1 cos 2p ;0nN N where L = N + 1. Coherence significance criterion (cmin), above which coherence differs significantly from 0, was derived from the degrees of freedom m of the spectral estimate at a significance level a of 0.05 [15, 19]. pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 cmin ¼ 1 av2 Degrees of freedom are dependent on the details of the spectral estimation method. Estimates of the degrees of freedom of the Welch method with smoothing nearly identical to the Hanning window have been determined [18]. Employing this estimate, with 50 % overlap between segments, m equals 19.6. Using the above formula for coherence significance level, cmin equals 0.5372. Average phase shift and gain were therefore calculated in the low frequency range (0.06–0.12 Hz) with a minimum coherence of 0.53. This frequency range was chosen because as it has been shown to be the most meaningful measure of autoregulation in previous studies [8, 15, 16, 20]. Less (lower) phase shift indicated poorer autoregulation. Time Correlation Method The Mx was first described in patients with TBI [4] and later adapted to patients with other types of neurologic disease, such as stroke or carotid occlusion [7, 16]. The Mx was calculated according to a previously described method [16]. In brief, CBFV and ABP were averaged over 3-s
123
Neurocrit Care (2014) 20:60–68
intervals with the use of waveform time integration. For every 20 consecutive values of averaged ABP and CBFV a Pearson’s correlation coefficient was calculated for each method of ABP recording. The resulting set of 1-min correlation coefficients was then averaged, yielding the Mx. A positive coefficient signified a positive association between FV and perfusion pressure, which may be interpreted as impaired autoregulation. A very low correlation coefficient signified a lack of association between the two signals, implying intact autoregulation. Some authors regard a correlation coefficient of >0.3 as a threshold for disturbed autoregulation [21]. It is important to note that there is considerable heterogeneity in the calculation of the correlation coefficient and other lengths of the primary data segments (5 or 10 s instead of 3 s), or the use of a moving window for calculation of the correlation coefficients have been used. In addition, longer periods of monitoring have been suggested to filter out random fluctuations of ABP and CBFV [22]. For reasons of simplicity, since the aim of our study was to evaluate how different forms of BP monitoring affect parameters of DCA and therefore length of data assessment is not such a relevant factor, we decided on 10-min recordings. Statistical Analysis Continuous variables were tested for normality using the Shapiro–Wilk test. Normally distributed data were reported as means and standard deviation (SD); non-normally distributed as median, 25th and 75th percentile. The agreement between the estimates of DCA (phase, gain, and Mx) obtained via Finapres and invasive intra-arterial catheter was assessed with Bland–Altman analysis for repeated measurements [23, 24]. For each subject the difference between two parameters (e.g., phase from Finapres and arterial catheter) was plotted against the mean of these two parameters. Given approximate normality in the distribution of the differences, the 95 % confidence interval for inter-method agreement is the mean difference ±1.96 standard deviations. A small mean difference indicates small inter-method bias, while a small variance indicates good inter-method agreement. For additional assessment of agreement, Pearson’s correlation coefficient between subject means was calculated. The dynamic relationship between the two ABP methods was also quantified by the coherence function between them. For analysis of intermethod differences, a paired t test was used to compare means, Wilcoxon signed-rank test for median comparison. To account for the correlation among repeated measures in the same person, mean scores for each subject were calculated [25]. A p value of less than 0.05 was considered statistically significant.
Neurocrit Care (2014) 20:60–68
63
Results Twenty patients were recruited for the study; 5 were excluded due to inability to obtain a good Finapres signal. The final analysis contained 15 patients in whom we measured CBFV with TCD and blood pressure with arterial catheter (ABPArt) and Finapres (ABPFin). In 10 patients we performed repeated measurements for a total of 37 observations for each hemisphere. Admission diagnoses included aneurysmal subarachnoid hemorrhage (n = 7), acute ischemic stroke (n = 4), ICH (n = 3), and reversible cerebral vasoconstriction syndrome (n = 1). The mean age was 58.1 ± 15.9 years, 80 % (n = 12) were women. Basic demographics, clinical characteristics, and medications assessed at each measurement are given in Table 1. ABP was similar for Finapres and arterial catheter (mean difference 1.92 mmHg, 95 % CI -5.82 to 1.99), p = 0.31). Figure 1 illustrates a high coherence between mean ABPFin and mean ABPArt for frequencies above 0.05 Hz Table 2.
Table 1 Demographics and Baseline Characteristics Parameters Age, years (mean, SD)
58.1 (15.9)
Gender (male/female)
12/3
Ethnicity (n, %) Hispanic
9 (60)
White
5 (33.3)
Black
1 (6.7)
Diagnosis (n, %) SAH
7 (46.67)
Ischemic Stroke
4 (26.7)
ICH
3 (20)
RCVS
1 (6.7)
Our study patients demonstrated significant impairment in DCA with abnormal values for phase shift and Mx, which is likely due to their underlying neurological injury. The mean phase shift in the low frequency range (0.06–0.12 Hz) was 26.3 ± 11.6° with Finapres, and 21.7 ± 10.5° using arterial catheter (p = 0.001). Gain in the LF range did not significantly differ between the two ABP methods. The mean difference in the autoregulatory index Mx between Finapres and arterial catheter was 0.08 (95 % CI 0.01–0.05, p < 0.0001). In summary, the autoregulatory parameters measured with invasive ABP assessment resulted in lower absolute values for both the transfer analysis and time correlation methods compared with the Finapres. Figure 2a–c illustrates this finding, showing the phase shift, gain, and Mx group averages for both Finapres and arterial catheter. The average coherence between CBFV and ABP shows consistently higher values for arterial catheter when compared with Finapres for frequencies above 0.05 Hz (Fig. 2a). Median coherence was excellent for both methods—on the left side median coherence was 0.79 (0.65, 0.85) for Finapres and 0.87 (0.75, 0.92) for arterial line (p < 0.0073). For the right side values were 0.72 (0.52, 0.81) for Finapres and 0.82 (0.71, 0.86) when measured with arterial catheter (p < 0.0001). The correlation between the autoregulatory parameters phase shift and Mx was r = 0.54, p < 0.01 for Finapres and r = 0.55, p < 0.01 for arterial catheter. Figure 3a–c shows the correlation between estimates of DCA obtained via Finapres and corresponding values derived from arterial catheter. Correlation coefficients and mean biases are given in Table 3. For the calculation of phase shift and gain 1.1 1
Medications taken at time of measurement (n, %)
0.9
31 (83.8)
b-Adrenoreceptor Antagonist
9 (24.3)
0.8
ACE/ARB
3 (8.1)
0.7
Propofol
4 (10.8)
Dexmedetomidine
1 (2.7)
Phenylephrine
4 (10.8)
Coherence
Calcium channel blocker
0.6 0.5 0.4
Dopamine ABPFin, mmHg (mean, SD)
2 (5.4) 99.35 (17.2)
0.3
ABPArt, mmHg (mean, SD)
101.3 (17.3)
0.2
CBFV Left, cm s-1 (mean, SD)
75.4 (38.5)
CBFV Right, cm s-1 (mean, SD)
75.4 (23.1)
Study population n = 15, measurements n = 37. Medications were assessed with each measurement SAH subarachnoid hemorrhage, ICH intracerebral hemorrhage, RCVS reversible cerebral vasoconstriction syndrome, ACE angiotensin converting enzyme inhibitor, ARB angiotensin II receptor antagonist, ABP arterial blood pressure, CBFV cerebral blood flow velocity
0.1 0 0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Frequency (Hz) Fig. 1 Average coherence function between mean arterial blood pressure obtained from intravascular catheter and Finapres (solid line) and standard deviation (dashed line)
123
64
Neurocrit Care (2014) 20:60–68
Table 2 Descriptive statistics of autoregulatory parameters for transfer function and time correlation method Side Left
N
Phase shift Finapres
Phase shift arterial
Gain Finapres
Gain arterial
Coherence Finapres
Coherence arterial
Mx Finapres
Mx arterial
Valid
35
35
35
35
37
37
37
37
Missing
2
2
2
2
0
0
0
0
Mean
24.67
20.50
0.67
0.81
0.73
0.81
0.58
0.66
Median
27.12
20.83
0.46
0.54
0.78
0.86
0.60
0.67
8.87
9.61
0.57
0.72
0.14
0.12
0.14
0.12
10.68
5.55
0.11
0.09
0.41
0.47
0.35
0.46
25
36.10 14.56
34.11 13.72
1.87 0.22
2.37 0.28
0.89 0.65
0.91 0.76
0.77 0.45
0.84 0.56
50
27.12
20.83
0.46
0.54
0.78
0.86
0.60
0.67
75
32.55
30.31
0.88
1.37
0.82
0.88
0.70
0.75
Valid
35
35
35
35
37
37
37
37
Missing
SD Minimum Maximum Percentiles
Right
N
2
2
2
2
0
0
0
0
Mean
27.92
22.82
0.78
0.83
0.72
0.79
0.56
0.64
Median
27.41
21.62
0.62
0.57
0.72
0.81
0.58
0.65
SD
13.99
11.54
0.52
0.50
0.11
0.10
0.14
0.14
Minimum
-6.53
-2.93
0.27
0.25
0.47
0.60
0.33
0.39
Maximum
57.44
50.55
1.90
1.94
0.90
0.93
0.78
0.88
Percentiles
25
22.60
19.26
0.40
0.37
0.66
0.74
0.41
0.54
50
27.41
21.62
0.62
0.57
0.72
0.81
0.58
0.65
75
33.33
26.62
1.15
1.19
0.78
0.86
0.68
0.73
Mx mean velocity index
two measurements had to be removed because of coherence less than 0.53 in the LF range. Figure 4a–c shows the Bland–Altman plots comparing phase shift, gain, and Mx measured by Finapres and arterial catheter. The differences did not vary in a systematic way over the range of measurement.
Discussion There is considerable methodological diversity among studies measuring DCA. TCD combined with Finapres has been the method of choice for the non-invasive assessment of DCA. Among acutely ill patients in the intensive care setting, however, vasoconstriction often limits the use of non-invasive finger plethysmography, and direct intravascular blood pressure measurement is used for blood pressure monitoring. Thus it is important that arterial approaches also be available for the assessment of cerebral autoregulation in this population [5]. Precision and accuracy comparing Finapres to invasive arterial assessment has been demonstrated in measuring blood pressure [17]; however, it has been less certain whether these different techniques may be used interchangeably to measure cerebral autoregulation. Lavinio et al. compared the autoregulatory index calculated as the correlation coefficient between fluctuations of CPP and
123
CBFV with a completely non-invasive approach using Finapres instead of CPP. For the calculation of CPP they measured intravascular radial artery blood pressure and intracranial pressure [26]. This implied two assumptions: (1) negligibility of ICP changes on DCA calculation and (2) accuracy of Finapres blood pressure slow-wave estimation. Although the non-invasive Mx satisfactorily described DCA in comparison to the invasive approach (R = 0.755), limits of agreement calculated according to Bland–Altman were fairly wide (±0.36). It remained unclear, whether this was due differences in blood pressure measurement or the result of ICP changes. In our study, we found good agreement between the two methods as blood pressure sources. There was no significant difference in mean ABP assessed with Finapres and arterial catheter with a mean bias of only 1.9 mmHg. Coherence between the two signals was excellent for frequencies above 0.05 Hz. The drop in coherence for very low frequencies has been described before and might be due to greater variability in mean ABP with Finapres in this frequency range. Our study thus confirms the results of others that showed similar beat-to-beat blood pressure measurements comparing Finapres to intra-arterial catheter [27, 28]. Importantly, we found a good correlation between the estimates of DCA obtained with Finapres and arterial
Neurocrit Care (2014) 20:60–68
A
C
65
B
D
Fig. 2 Average coherence (a), phase shift (b), gain (c), and Mx (d) between mean ABP and CBFV for Finapres (solid line) and arterial catheter (dashed line). Gray areas represent LF range (0.06–0.12 Hz)
catheter. However, there was a significant difference in absolute values for mean phase shift and Mx. Since both were calculated using the same CBFV signal, any difference measured is attributable to differences in the ABP signals. As measured with the arterial catheter, the Mx was on average 0.08 points higher (suggesting poorer autoregulation) when compared to Finapres. It is possible that slightly increased noise in the Finapres signal led to the lower correlation coefficient. It becomes more difficult to explain the differences in phase shift, which was on average 4.6° lower (suggesting poorer autoregulation) with arterial catheter when compared to Finapres. Our findings are in agreement with the only other study testing the effects of invasive ABP measurement on parameters of DCA. Sammons et al. [28] analyzed phase shifts obtained via Finapres and intra-aortic catheter in 27 patients undergoing coronary catheterization. The absolute values obtained in their study using arterial catheter were 5.7° lower than values obtained with Finapres. They suggested that this might be due to the difference in vascular compliance between aorta and peripheral circulation, leading to the bias in autoregulation parameters. We found a similar bias and all our measurements were obtained from the radial artery making it less likely that amplification of the pulse pressure waveform from aorta to the peripheral
circulation is an important factor. However, it is still possible that the slightly more distal location of the Finapres may have contributed to the difference in phase shift, as this site has been shown to have a greater effect on rhythmic changes in arteriolar tone, thus resulting in increased phase shift [29]. Sammons et al. further hypothesized that medication effect might have played a role, as 76 % of their patients were taking beta-adrenergic antagonists. Our results do not support this hypothesis as we found similar discrepancies, but only 24 % of the patients in our study were taking beta-adrenergic antagonists. In addition, the phase shift bias was not significantly different among those taking medication compared to those without (5.6° vs. 3.7°, p = 0.819). An alternative explanation is that Finapres is not able to precisely reproduce the dynamic changes in the low frequency range as well as invasive monitoring. Omboni and colleagues found that Finapres overestimated systolic blood pressure powers also in the very low-frequency range, which may have resulted in the increased phase shift. They hypothesized that these discrepancies between finger and intra-arterial data are due to an intrinsic defect of the Finapres method. Although absolute differences in measurements of autoregulation may not matter in patients being followed with one method or the other as individual
123
66
Neurocrit Care (2014) 20:60–68
A
B
C
Fig. 3 Scatterplot with regression line of phase shift (a) gain (b), and Mx (c) measured with Finapres and arterial catheter. Triangles represent values from the left MCA, circles those from the right MCA Table 3 Bias and correlation coefficients of parameters of autoregulation using Finapres and Arterial catheter Parameter
Left phase shift (°) Right phase shift (°)
Bias ABPFin-Art
4.17
SD
7.21
95 % CI of the Difference Lower
Upper
0.18
8.17
p value
Correlation coefficient
0.042
0.70
p value
0.004
5.10
6.77
1.36
8.85
0.011
0.88
