INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

IN NORMAL SUBJECTS, THE RELATIONSHIP between cerebral blood flow (CBF) and mean arterial blood pressure (MABP) is dominated by an autoregulatory mechanism that tends to maintain CBF relatively constant, despite changes in MABP in the range 60-150 mmHg (34). Earlier studies of cerebral autoregulation have not taken into account the latency of the autoregulatory response, only considering the net change in CBF resulting from changes in MABP, usually induced by vasoactive drugs (22, 27). More recent work has analyzed the CBF response to dynamic changes in MABP (2, 27). Investigations of dynamic cerebral autoregulation usually estimate CBF from CBF velocity (CBFV) measurements performed with transcranial Doppler ultrasound because of its superior temporal resolution.

The most common approaches for dynamic studies have been the induction of a sudden drop in MABP by the release of inflated thigh cuffs (2) or the use of spontaneous fluctuations in MABP in 5- to 10-min recordings at rest, coupled with frequency domain analysis techniques to describe the linear dynamic relationship between CBFV and MABP (4, 17, 21, 29, 32, 42). Each method has its advantages and limitations. Although baseline measurements can be seen as the most physiological situation to study cerebral autoregulation, it requires a much longer recording time than the thigh cuff technique, and its accuracy is dependent on the amount of spontaneous power in the frequency range below 0.2 Hz (32, 43). On the other hand, the inflation of thigh cuffs up to 30 mmHg above systolic ABP, before rapid deflation, is rather painful, and most investigators have reported a need for repeated tests to improve accuracy (20, 39, 41). As a consequence, the total duration of the assessment becomes comparable to the baseline approach, and the sympathetic activation produced by pain might also influence results. Above all, it is not clear whether the thigh cuff method and the frequency domain technique applied to baseline recordings provide the same physiological insight into cerebral autoregulation function. The CBFV response to the pressure drop after thigh cuff release has been modeled by a second-order differential equation proposed by Aaslid (39). From this model, it is possible to extract an autoregulatory index (ARI) ranging from 0 (absence of autoregulation) to 9 (best autoregulation) (39). A second parameter, critical closing pressure, can also be extracted from Aaslid's model, but its interpretation and significance have not been established (33, 39). From the frequency domain analysis of baseline recordings, it is possible to estimate the CBFV impulse and step response functions to changes in MABP (29, 32, 42). Direct comparison of these results with thigh cuff data has been hampered by the difficulty of performing fast Fourier transform (FFT) spectral analysis of short and nonstationary segments of data, as represented by the 30-s window containing the sudden drop in arterial blood pressure (ABP). As an alternative, CBFV impulse responses estimated from baseline recordings have been used to predict the time course of the CBFV transient during thigh cuff maneuvers (29, 42). Although this exercise indicates a reasonable degree of fit, it is still not clear whether the thigh cuff test might be providing additional information because of its much larger pressure changes (>15 mmHg) than observed during spontaneous fluctuations in ABP. Significant pressor and depressor changes in MABP can also be produced by other maneuvers, and some of these, like the Valsalva maneuver and the cold pressor test, have been proposed as tests of cerebral autoregulation (35, 38). Again, it is not clear whether the direction of ABP change, its rate of change, or the effects of these maneuvers on the autonomic nervous system might lead to physiological differences in dynamic autoregulatory responses.

We tested the hypothesis that CBFV responses to ABP transients might be dependent on the type of maneuver adopted to produce pressor or depressor changes by studying a group of normal individuals subjected to the thigh cuff test, Valsalva maneuver, cold pressor test, lower body negative pressure, and hand grip test, as well as continuous baseline recordings at rest.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects and Measurements

The ABP, dual-channel CBFV Doppler shift signals, end-tidal CO₂, and ECG were continuously recorded on digital tape for subsequent analysis (Sony PC 108M). After 30 min of rest, signals were recorded for two 5-min periods with subjects breathing normally at rest. Subsequently, five different tests were performed in random order. In each case, recording was started 60 s before the test and continued for 60 s after the test to allow readings to return to baseline values. Subjects were asked to lie still and refrain from talking or inadvertently raising intrathoracic/abdominal pressure during the tests.

Thigh cuff test. Wide blood pressure cuffs were attached to both legs and inflated 40 mmHg above systolic pressure for 45 s. The cuffs were rapidly deflated, and signals were recorded for another 2 min.

Lower body negative pressure. The subject's lower body, from the iliac crest downward, was placed in a custom-made metal box. A domestic vacuum cleaner was attached to the box to generate suction to produce a gradual reduction in systolic ABP of 10-20 mmHg for 90 s. The vacuum source was then rapidly disconnected to allow the pressure to quickly equilibrate.

Cold pressor test. The subject's left hand was placed in a bowl of water at 4°C for 45 s to induce a pressor response and then removed and dried while the subject remained still.

Isometric hand grip. Subjects were asked to grip with their left hand a rolled-up slightly inflated sphygmomanometer cuff at 30% maximal voluntary compression for 2 min. The grip pressure was visually displayed to aid compliance. After 2 min, subjects were asked to release the grip suddenly, leading to a dynamic MABP drop.

Valsalva maneuver. In the reclined-supine position, subjects were asked to blow into a syringe with an integral constant-bleed device that was attached to a pressure transducer. An intrathoracic pressure of 40 mmHg was maintained for 15 s; on release, a dynamic increase in MABP was obtained during phase IV of the Valsalva maneuver. This maneuver was repeated three times; it was the only test performed more than once.

Data Analysis

The ABP signal was calibrated at the start of each recording, and all signals were visually inspected for artifacts or noise. Narrow spikes on the CBFV signals were removed by linear interpolation, and the four signals were low-pass filtered with a zero-phase eighth-order Butterworth digital filter with a cutoff frequency of 20 Hz. The beginning and end of each cardiac cycle were detected from the upstroke of the arterial pulse pressure wave, and the MABP and mean CBFV (MCBFV) were calculated for each cycle. Linear regression was used to model the instantaneous CBFV-ABP relationship for each cardiac cycle, and the resistance-area product (RAP) was estimated as the inverse of the regression slope, as described previously (10, 12, 31). Ectopics occurring during maneuvers led to the rejection of the data, as did more than one ectopic per 30 s for the baseline recording. Occasional ectopics could be marked and removed by linear interpolation. Spline interpolation was used to resample the data at 0.2 s to create a uniform time base.

From the 10-min (2 recordings at 5-min intervals) baseline recordings, two additional sets of data were extracted containing spontaneous MABP and MCBFV transients. Positive spontaneous MABP transients were extracted from the first series of baseline recordings by marking the largest such transient by visual inspection, thus obtaining one positive spontaneous transient for each subject. A similar procedure was adopted to mark the largest negative spontaneous transient in the second set of baseline recordings. For all the maneuvers, transients, and baseline recordings, a 1-min epoch (300 samples) was extracted for further analysis. For baseline recordings, a random initial position was used to extract the 1-min epoch. For the Valsalva maneuver, the second of the three maneuvers recorded was used for analysis. For the five maneuvers and the two spontaneous transient records, the maximum derivative of the MABP pressor or depressor change was used as a marker, and the beginning of the epoch was established as 20 s before this fiducial point. An automatic algorithm was used to extract the MABP change around the fiducial point (P) and the corresponding change in MCBFV (V). The start (P_start and V_start) and end points (P_end and V_end) of each (P and V) change were also detected. Representative temporal patterns of the MABP and MCBFV transients for each maneuver and baseline measurements were obtained by calculating the mean and SE for each sample point along the 1-min ensemble. From the ensemble averages, the mean value of RAP (RAP_M) was calculated during a 10-s interval, starting 20 s before P_start. For the Valsalva maneuver, a different time interval was used, because relatively stable values of RAP were observed only in the interval from 12 to 2 s before P_start at the beginning of phase IV (20). For each maneuver, RAP_M represents the prevailing RAP before the MABP transient takes place.

For all recordings, the MABP and MCBFV 1-min epochs were normalized by the mean value and subtracted by 1. The resulting zero-mean signals, reflecting relative changes in ABP and CBFV, are represented by P(t) and V(t), respectively. These signals were decimated by using only one of every three samples to increase the whiteness of P(t) (24, 28), the effective sampling interval then becoming 0.6 s.

Two different methods were used to estimate the linear dynamic relationship between MABP and MCBFV. The first approach used the mathematical model proposed by Aaslid (39) to estimate the ARI during thigh cuff maneuvers. This model can also be fitted to baseline recordings or to MABP transients produced by other means (28, 33). In each case, the 1-min P(t) signal was used with the equations given by Tiecks et al. (39) to predict the corresponding V(t) transient. Ten preestablished sets of model parameters were tested, and the set corresponding to the best fit, as determined by the correlation coefficient between the model-predicted and measured V(t) signals, was used to determine the final value of ARI, which is scaled to vary between 0 (absence of autoregulation) and 9 (best possible autoregulation) (28, 39). As shown by Tiecks et al., each value of ARI corresponds to an MCBFV response to a hypothetical step change in MABP. The time derivative of this step response corresponds to the MCBFV impulse response to changes in MABP [i_T(t)], as described previously (28). The amplitudes of the impulse and step responses were adjusted to minimize the mean square error (MSE) between the model-predicted response and the velocity data.

The second approach adopted to model the dynamic relationship between P(t) and V(t) is the moving-average Wiener-Laguerre representation reported previously (24, 28). In this case, a linear analysis was performed by calculating only the first kernel of the Volterra expansion (24). The Laguerre expansion used eight terms with a value of  = 0.2 (24, 28). Single-value decomposition was used to estimate the coefficients of the Laguerre expansion leading to the velocity impulse response [i_W(t)] (28). The number of lags was 16. The corresponding step response [s_W(t)] was obtained by integrating i_W(t) along time.

Performance of the Aaslid and Wiener-Laguerre models was assessed by comparing the predicted model responses [V_M(t)] with the real data [V(t)]. In each case, the model response was calculated with the convolution operation by use of the impulse response and the input signal [P(t)]

(1)

where i(m) can be i_T(t) or i_W(t) and L is the number of lags.

Statistical Analysis

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A total of 56 subjects (31 men), 62 ± 15 (SD) yr of age (range 23-80), were studied. Clinical systolic and diastolic blood pressures were 132 ± 12 mmHg (range 97-160) and 78 ± 7 mmHg (range 59-91), respectively. Body mass index was 26.3 ± 3.6 kg/m² (range 18.0-35.2), and resting heart rate was 63 ± 10 beats/min (range 45-96). Transcutaneous measurements of PCO₂ indicated stable intrasubject values between recordings, but its temporal resolution was not adequate to detect changes taking place during maneuvers. In a subgroup of 12 subjects, during baseline recordings, positive and negative spontaneous transients, the hand grip test, and end-tidal CO₂ did not show any significant fluctuations or trends. For lower body negative pressure, however, end-tidal CO₂ decreased significantly from 39.6 ± 2.9 mmHg before vacuum was applied to 36.2 ± 2.4 mmHg before the return to atmospheric pressure.

Similar results were obtained for the right and left MCAs for the five maneuvers, two transients, and baseline recording; therefore, recordings from the two sides were pooled. The total number of recordings obtained for each maneuver is given in Table 1. In five subjects, CBFV was recorded in only one side. The most common reasons for rejection in other cases were an insufficient P change, lack of a simultaneous change in CBFV, or a nonsignificant model correlation coefficient for the Wiener-Laguerre impulse response.

Figure 1 shows the coherent averages of all recordings available for each subject. For the baseline recording, random fluctuations are averaged, and a relatively flat coherent average is obtained (Fig. 1A). For the other maneuvers, a fast change in MABP is accompanied by a simultaneous change in MCBFV. In every case, the mean MCBFV transient returns to its original value before MABP, suggesting an active autoregulation. Although different maneuvers/transients produced different P changes (Fig. 1), roughly similar temporal patterns were obtained in some cases, for example, between the thigh cuff test and negative spontaneous transients, cold pressor test and positive spontaneous transients, and lower body negative pressure and Valsalva maneuver. A comparison of mean V with the mean P change is presented in Fig. 2, showing surprisingly similar slopes for most maneuvers, the main difference being the start and end points for MABP and MCBFV.

View larger version (32K): [in this window] [in a new window]   Fig. 1.   Coherent averages of mean arterial blood pressure (ABP, solid line) and mean cerebral blood flow velocity (CBFV, dashed line) for baseline recordings (A), positive spontaneous transients (B), negative spontaneous transients (C), thigh cuff maneuver (D), lower body negative pressure (E), cold pressor test (F), hand grip test (G), and Valsalva maneuver (H). N, number of data segments used in each average. For simplicity, error bars represent only the largest SE of each average.

View larger version (18K): [in this window] [in a new window]   Fig. 2.   Change in mean CBFV and mean ABP from the foot (+) to the peak () of the main transient of each maneuver. For the baseline recording (BSL, dotted line), the excursion has been represented around the mean values of mean ABP and CBFV of the 1-min epoch of the mean ± SD of the spontaneous fluctuations in mean ABP and mean CBFV. The largest excursion observed is for the Valsalva maneuver (VAL, dashed line). PST, positive spontaneous transients; NST, negative spontaneous transients; THC, thigh cuff test; LBP, lower body negative pressure, CPT, cold pressor test; HDG, hand grip test.

Fitting the Wiener-Laguerre and Aaslid models to the MABP-MCBFV dynamic relationship produced the mean impulse responses depicted in Fig. 3. In every case, impulse responses estimated separately for the right and left MCAs resulted in very similar patterns for the mean impulse response. The lowest correlation coefficient between the right and left mean impulse responses was obtained for the negative spontaneous transients (r = 0.976, P < 10⁵). No significant differences were observed between mean values of ARI for the two sides for any maneuver. The sign test applied to the correlation coefficient and MSE showed superior model fitting with the Wiener-Laguerre approach for all five maneuvers, the two transients, and the baseline recording (P < 10⁵). The Durbin-Watson tests were consistently below the critical limit (d < 1.7) for the eight situations studied, indicating that residuals were not random. In general, the Wiener-Laguerre response shows a higher value than the Aaslid model at time 0 and also a more negative trough at 0.6 s. With the exception of the thigh cuff, cold pressor, and hand grip tests, the Wiener-Laguerre impulse response shows an overshoot of ~1.2 s.

View larger version (24K): [in this window] [in a new window]   Fig. 3.   Impulse responses (IR) and SE obtained with the Wiener-Laguerre method (solid line, ) and Aaslid model (dashed line, +) for baseline recordings (A), positive spontaneous transients (B), negative spontaneous transients (C), thigh cuff maneuver (D), lower body negative pressure (E), cold pressor test (F), hand grip test (G), and Valsalva maneuver (H).

The corresponding step responses are represented in Fig. 4. For time 0, the Wiener-Laguerre step response is always greater than the corresponding Aaslid model response, but the difference between the two models decreases beyond time 0. As with the coherent averages, the drop of the step responses beyond time 0 reflects an active autoregulation.

View larger version (27K): [in this window] [in a new window]   Fig. 4.   Mean step responses and SE obtained with the Wiener-Laguerre method (solid line, ) and Aaslid model (dashed line, +) for baseline recordings (A), positive spontaneous transients (B), negative spontaneous transients (C), thigh cuff maneuver (D), lower body negative pressure (E), cold pressor test (F), hand grip test (G), and Valsalva maneuver (H).

The similarity between the temporal patterns of the mean impulse responses of the five maneuvers, two transients, and baseline recording displayed in Fig. 3 was confirmed by correlation analysis. For i_W(t), the minimum correlation between any two responses was 0.961 (P < 10⁵) for the comparison of the thigh cuff test vs. the Valsalva maneuver. The correlation matrix was calculated using all 16 lags, with marginally higher correlations when the number of lags was limited to eight. For the Aaslid model, the minimum correlation between any two mean impulse responses was 0.997 (P < 10⁶).

A complete set of recordings was obtained in 11 subjects who were used to assess the effects of type of maneuver vs. the contribution of individual subjects to explain the variability of different parameters. Mean values and F-test results for P, V, ARI, model correlation coefficients, and MSE are given in Table 1. The superiority of the Wiener-Laguerre method to model the MABP-MCBFV relationship is also confirmed in this smaller set of data for all the maneuvers and transients analyzed. Although subjects contribute significantly to the variability of the correlation and MSE, higher values of F were obtained for the effect of the maneuver/transient (Table 1). Nevertheless, no significant effects from subjects were noticed for P and V, with the large differences attributable exclusively to the intervention (maneuver or spontaneous transient). On the other hand, the parameter ARI, extracted with the Aaslid model, was not influenced by the nature of the intervention (F = 0.83, not significant) but was significantly influenced by the subjects (F = 2.62, P = 0.009). The mean value of ARI for each maneuver, transient, or baseline recording, calculated for all subjects (Table 1), was not different from the means obtained for the 11 subjects selected for the two-way ANOVA (paired t-test, P = 0.97). The overall ARI for all data (n = 602 recordings) was 4.7 ± 2.9 (SD).

The dependence of the impulse and step response values at time 0 on the type of maneuver was associated with the different values attained by MABP preceding the P change (P_start), which is one of the values represented in Fig. 2. The first value of the step response (time 0) is the corresponding value of the impulse response multiplied by the sampling interval (0.6 s after decimation). For this reason, only the step response results will be mentioned. For the Wiener-Laguerre and Aaslid models, the mean maneuver or spontaneous transient step responses (Fig. 4) are inversely proportional to P_start (r = 0.93, P = 0.0004, and r = 0.85, P = 0.004, respectively). Furthermore, these mean step responses are also inversely correlated to RAP_M, with r = 0.93 (P = 0.0004) for the Wiener-Laguerre model and r = 0.77 (P = 0.0127) for the Aaslid model. Figure 5 shows the relationship of the mean Wiener-Laguerre initial step response value of each maneuver/transient to the corresponding RAP_M value. When the thigh cuff test value is removed from the linear regression, the correlation drops to 0.75 but remains significant (P = 0.026).

View larger version (16K): [in this window] [in a new window]   Fig. 5.   Amplitude of the mean step response of each maneuver at time 0 as a function of the mean resistance-area product (RAP) preceding the main mean ABP transient. Values are means ± SE. Regression line of mean values against RAP is also represented. See Fig. 2 legend for definition of abbreviations.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

CBFV responses to sudden changes in ABP can shed light into the mechanisms of dynamic cerebral pressure autoregulation. Using two different modeling approaches, we have shown that different pressor and depressor ABP transients can influence the amplitude of the CBFV dynamic responses, but the morphology of the derived step responses was surprisingly similar, despite the expected associated changes in sympathetic activation and arterial PCO₂ levels.

Methodological Considerations

Impulse and step responses of the MABP-MCBFV relationship were reported previously only for the case of baseline recordings (28, 32, 33, 42). For the thigh cuff test, it has been shown that the baseline impulse response, obtained by the classical FFT function approach, can predict the thigh cuff velocity response with reasonable accuracy, suggesting that baseline and thigh cuff test could be modeled by a similar impulse response function (28, 42). The use of the Wiener-Laguerre approach allows a direct time domain calculation of the impulse response of the thigh cuff test and other short MABP transients that are strictly nonstationary in the analysis interval and can thus overcome the main limitation of the FFT method. The results presented in Fig. 3 confirm the initial supposition that the impulse response for the thigh cuff resembles the corresponding function for baseline recordings. The most surprising finding of our study, however, is that the impulse and step response functions for several other maneuvers are also very similar. It could be argued that in all cases the Wiener-Laguerre impulse response is simply reflecting the baseline background fluctuations in MABP and CBFV that are also present in the 1-min epochs selected for analysis. However, this claim should be rejected by the observation in Fig. 1 that most of the MABP and MCBFV powers in these records are dominated by the maneuver-induced transient, therefore, being the main determinants of the resulting impulse and step responses. However, with the exception of baseline and thigh cuff test, most MABP transients (Fig. 1) are biphasic, and it is not possible to discard the possibility that MABP changes with different temporal patterns or very large amplitudes might lead to impulse responses with different shapes as a result of nonlinearities in dynamic cerebral autoregulation (28).

We previously reported that the linear Wiener-Laguerre model provides fitting of the MCBFV-MABP relationship during baseline recordings superior to that provided by the Aaslid model (28). This superiority has been confirmed in the present study and has also been shown to apply to the pressor and depressor spontaneous transients and the five maneuvers considered (Table 1). This result is not surprising given the flexibility of the Wiener-Laguerre approach in contrast to the fixed structure of the Aaslid model (28, 39). Both models have nonwhite residuals, as reflected by the Durbin-Watson test. As discussed previously, this probably reflects the presence of band-limited noise in the CBFV signal and the fact that the input signal (MABP) is also nonwhite (28). One advantage of the Aaslid model is that it provides an ARI that has been analyzed in previous studies (9, 23, 38, 39, 41). We have demonstrated that the ARI is influenced by intersubject variability but not by the type of maneuver performed. This result is not limited to the subgroup of 11 subjects: a similar distribution of mean values for each maneuver or spontaneous transient was obtained when all subjects were included (Table 1). This result has considerable importance for clinical applications, suggesting that, as far as the ARI is concerned, supine recordings at rest (i.e., baseline) can provide the same information as other maneuvers that can be rather uncomfortable to patients or can induce other physiological changes that make interpretation difficult. However, our results apply only to normal individuals and do not take into account the possibility that certain maneuvers might be more sensitive to specific cerebrovascular conditions. In a recent study of patients with acute ischemic stroke, we showed that the thigh cuff test was able to detect a deterioration of ARI, from 6.2 ± 2.3 in control subjects to 4.8 ± 3.1 in the nonaffected hemispheres and 4.1 ± 3.3 in affected hemispheres, whereas no differences could be detected using the static pressure changes induced by hand grip or thigh cuff inflation (9). Further investigations are required to compare the clinical usefulness of ARI estimated from different maneuvers.

Study Limitations

Given the relatively long duration of the measurement protocol, we reduced the stress time of certain maneuvers, such as lower body negative pressure, hand grip, thigh cuff, and cold pressor tests, compared with the usual durations adopted in autonomic nervous system studies. Consequently, the baseline or static changes in ABP and CBFV before or after (cold pressor test) the main ABP transient are not strictly comparable to values reported in the literature. The reduced duration of some of these tests might explain why the induced sudden ABP change was <5 mmHg in some cases.

Lower body negative pressure and other maneuvers are known to induce changes in PCO₂, and for this reason CO₂ was recorded transcutaneously throughout measurements. Although its temporal resolution is not appropriate to detect sudden changes in PCO₂, transcutaneous measurements provided reassurance that PCO₂ remained relatively stable in each subject during a 2-h protocol. In a subgroup of subjects, end-tidal CO₂ was also recorded during baseline, hand grip, and lower body negative pressure tests. This confirmed previous observations that lower body negative pressure induces hypocapnia (43), but there were no such trends or other significant changes during hand grip or baseline recordings. Similar measurements in all maneuvers performed would have been more relevant if differences in autoregulatory performance had been detected, but this was not the case.

Doppler ultrasound cannot provide an absolute measurement of CBF, and a stable relationship of this variable with CBFV is guaranteed only if the MCA cross-sectional area remains constant. Direct observation of the MCA during surgery has detected only minor changes in diameter produced by relatively large changes in mean ABP and PCO₂ (14). During thigh cuff tests, Newell et al. (25) recorded absolute flow in the carotid artery showing a relative change and time course similar to velocity in the MCA. Greenfield et al. (15) also measured carotid blood flow during Valsalva maneuvers and obtained phase IV transients similar to those we and others recorded with Doppler ultrasound in the MCA (10, 31). During sympathetic stimulation accompanied by sudden changes in MABP, Busija et al. (7) also showed very good agreement between changes in absolute flow, recorded with microspheres, and simultaneous measurements of velocity in pial arteries of dogs. The diameter of the large intracerebral arteries was also observed to remain unaltered after the infusion of norepinephrine in the internal carotid artery (26). Recently, Serrador et al. (36) reported that MCA diameter changes could not be detected during CO₂ manipulation or lower body negative pressures up to 40 mmHg suction. If MCA diameter is passively modified by MABP, increases in P_start will lead to increases in cross-sectional area and to velocity measurements that underestimate flow. This effect would not contribute to the inverse relationship observed between step response amplitude and P_start, however, because responses were estimated using the normalized signals [V(t)]. Furthermore, Fig. 5 presents strong evidence that as P_start is modified by different maneuvers, the amplitude of the impulse/step response varies inversely to RAP_M, which is an indicator of the slope of the CBFV-ABP relationship. With increases in P_start, autoregulation leads to vasoconstriction and to increases in RAP, representing a reduction of the slope of the instantaneous CBFV-ABP curve and, hence, reduced amplitude of the responses. Again, if the MCA is passively dilated by increases in P_start, cerebrovascular resistance will tend to be underestimated, but the relationship represented in Fig. 5 should be maintained, although the slope will be different. Finally, the suggestion that diameter might be changing during each maneuver is not supported by our results showing that large swings in pressure, as observed during the Valsalva maneuver, lead to impulse responses with the same shape as for baseline, which has relatively small fluctuations in MABP. It is extremely unlikely that changes in cross-sectional area during the MABP transient could have the exact temporal pattern so as to compensate different patterns of CBF changes to yield CBFV impulse responses that are all very similar.

Effects of Different Maneuvers

During baseline, the MCBFV impulse response is in very good agreement with previous estimations based on longer recordings, with the FFT function approach (28, 32, 33, 42) or the Wiener-Laguerre method (28). Positive spontaneous transients have been previously analyzed in the time and frequency domain (30, 31), but, to our knowledge, the Wiener-Laguerre and Aaslid model results have not been reported. The use of negative spontaneous transients to study the effect of spontaneous depressor transients is also new. Figure 2 shows that the two modalities of spontaneous transients follow a similar but inverted "pathway" along the pressure-velocity relationship during the main P change. The resulting impulse and step responses shown in Figs. 3 and 4 suggest that the direction of the P change does not have a significant effect on the pattern of these functions. Recordings obtained at rest (baseline and positive and negative spontaneous transients) yielded the best model fittings (Table 1), which can be an important consideration in the choice of tests and standardization of clinical procedures.

Previous studies adopting the thigh cuff test for assessment of dynamic autoregulation concentrated on the ARI parameter extracted from the Aaslid model. In the present study, the ARI for all subjects was 4.7 ± 2.9 (SD). For normal subjects, other investigators reported ARI values of 4.8 ± 1.0 (39), 6.3 ± 1.1 (40), 6.2 ± 2.3 (9), 4.7 ± 1.0 (20), and 5.0 ± 1.0 (23). Differences among studies might be explained by population characteristics and the amount of ABP drop produced by cuff release. In addition, the duration of the temporal window adopted to fit the Aaslid model to thigh cuff responses and the sampling interval have not been standardized (9, 20, 23, 39, 41). Finally, the largest SD obtained in our case might be due to the fact that a single thigh cuff release was performed in each subject, in contrast to other studies, in which three to eight maneuvers per subject have been used (20, 23, 25, 39, 41).

Tiecks et al. (38) extracted indexes of CBFV changes during the Valsalva maneuver that compared favorably with ARI values obtained with the thigh cuff method in the same subjects. However, these investigators did not extract ARI values from phase IV of the Valsalva maneuver, as demonstrated in the present study (Table 1). During phase IV, the relative change V is significantly larger than the corresponding percent increase in P (Figs. 1 and 2), as reported previously (10, 15, 38). Recently, we suggested an explanation for this phenomenon based on the large values of critical closing pressure estimated during phases I-IIb and the sudden drop in this parameter during phase III (10). An elevated critical closing pressure corresponds to a very steep CBFV-ABP instantaneous relationship or, in other words, a lower RAP, as indicated in Fig. 5. Apart from their larger amplitude than for other maneuvers, the mean impulse and step responses estimated from phase IV of the Valsalva maneuver are very similar to other conditions studied, notably negative spontaneous transients and lower body negative pressure tests.

We are not aware of published studies involving the sudden change in MABP produced by lower body negative pressure, cold pressor, and hand grip tests for assessing dynamic autoregulation. Zhang et al. (43) estimated the MCBFV-MABP transfer function with the FFT approach during the plateau phase of lower body negative pressure testing and concluded that autoregulation became less active with more negative values of suction down to 50 mmHg. Inasmuch as we have analyzed lower body negative pressure transients immediately after the rapid restoration to atmospheric pressure, it could be expected that the corresponding impulse and step responses would still reflect a deterioration of autoregulation, but that was not the case, as shown by the plots in Figs. 3E and 4E. Unfortunately, Zhang et al. (43) did not present impulse or step responses that could be compared with ours. Their conclusion was based on a 24% increase in the gain or amplitude frequency response at 50 mmHg lower body suction. No significant differences were observed in the phase frequency response or coherence function (43). Further work is necessary to establish whether the discrepancies between their results and ours are due to methodological differences, such as the duration and intensity of negative pressure stress, or whether cerebral autoregulation is quickly restored by the sudden interruption of suction.

During cold pressor testing, Roatta et al. (35) obtained MABP and MCBFV transients with a time course very similar to the coherent average depicted in Fig. 1F. They also observed situations in which the V change is absent or is slightly negative. In our protocol, such recordings would have been rejected, but in the majority of cases, P was >5 mmHg, and there was a simultaneous increase of MCBFV in both channels. Roatta et al. also estimated cerebrovascular resistance as MABP/MCBFV and, from its rapid increase after the P transient, concluded that sympathetic activation is more important than myogenic or metabolic mechanisms for control of dynamic autoregulation.

Sympathetic Activation

Cencetti et al. (8) analyzed spontaneous fluctuations in MABP, MCBFV, and other cardiovascular parameters and, similarly to Roatta et al. (35), concluded that sympathetic activation is more important than myogenic or metabolic mechanisms for control of dynamic autoregulation. Surprisingly, both groups of investigators have not performed a detailed analysis of the dynamic relationship between MABP and MCBFV and have also ignored evidence in the literature about the phase response of MCBFV in relation to MABP and the spectral frequency bands where autoregulation is more manifest (3, 4, 11, 13, 21, 27-29, 32, 42, 43). Furthermore, when the time course of RAP (rather than the simple ratio MABP/MCBFV) (10, 27, 31) is compared with the MABP transient, a time delay of ~2 s is always observed, showing a tight relationship between the MABP and MCBFV transients and the autoregulatory response. Unless changes in CBF or CBFV can be unequivocally associated with modulation of sympathetic activity in the absence of changes in MABP or somatosensory stimulation, there are no grounds to rule out the classical view that myogenic and metabolic mechanisms are the main agents of cerebral autoregulation in the physiological range of ABP (26). During steady-state normotension, sympathetic stimulation failed to modify CBF in dogs and cats (7). Interestingly enough, the same study measured "step responses" to sudden changes in MABP provoked by rapid occlusion of the aorta. In cats, the step response was attenuated by sympathetic stimulation, providing further evidence for the theory that sympathetic control has a protective effect on the cerebral circulation, particularly on the blood-brain barrier (7, 16). However, in light of the new results shown in Fig. 5, the reduction in the amplitude of the step responses recorded by Busija et al. (7) could simply be the consequence of an increase in RAP as induced by cold pressor, hand grip, or thigh cuff tests in our study.

Our results cannot refute the hypothesis that sympathetic activation shifts the static pressure-autoregulation curve to the right, because the P changes produced by the different maneuvers were well within the plateau region of the static curve. On the other hand, the results shown in Fig. 5 represent an alternative explanation for the apparent shift in the static autoregulation curve that could be caused not only by sympathetic activation but by any other adaptive mechanism that displaces the instantaneous pressure-flow relationship, either by changes in arterial diameter (i.e., RAP) or by modification of its pressure-axis intercept (critical closing pressure) (1, 27, 29, 31). Increases in small vessel thickness in chronic hypertension (34) or the cerebrovascular effects of arterial PCO₂ are pertinent examples (1, 16, 29).

In summary, we have not been able to detect significant differences in the shape of impulse or step responses estimated from MABP transients produced by maneuvers associated with different patterns and intensities of sympathetic nervous system activation. Nor could we observe differences due to rapid increases or decreases in ABP or between induced and spontaneous ABP changes. Nevertheless, significant differences in the amplitude of these responses were observed and could be explained by the adaptation of the instantaneous pressure-velocity relationship preceding the main ABP transient, as reflected by the mean RAP. These results should not be regarded as definitive, however, since more in-depth analyses might reveal subtle differences between the responses of different maneuvers, mainly when each subject is used as his/her own control.