Synchronous Rhythmical Vasomotion in the Human Cutaneous Microvasculature during Nonpulsatile Cardiopulmonary Bypass

The Human Investigation Committee of Yale University School of Medicine (New Haven, CT) approved the study protocol. After obtaining written informed consent, seven male patients and one female patient aged 53 to 76 yr undergoing elective coronary artery bypass grafting with hypothermic NP-CPB were enrolled in the study. Exclusion criteria were the presence of symptomatic peripheral vascular disease, diabetes mellitus, or diagnosed autonomic neuropathy. The anesthetic technique was similar for all eight patients and consisted of weight-related doses of sufentanil, midazolam, and pancuronium bromide, with a volatile anesthetic (isoflurane 0.5%) used during NP-CPB. NP-CPB was conducted with a membrane oxygenator (Gish Vision, Irvine, CA) and a nonpulsatile roller pump (Cobe, Arvada, CO), using crystalloid prime. Pump flow rates were kept between 2.2 and 2.5 l · min⁻¹· m⁻², mean arterial pressure (MAP) was maintained between 50 and 70 mmHg by use of phenylephrine hydrochloride or sodium nitroprusside as required, and hematocrit was maintained between 20% and 25% during NP-CPB. Patients were cooled to a core (bladder) temperature (t°C) of 30–32°C, with active rewarming to 37°C at the end of NP-CPB.

The LDF probes (Periflux 2B; Perimed, Stockholm, Sweden) were applied on the forehead and palmar surface of the index finger contralateral to the arterial catheter site to quantify the cutaneous microcirculatory flux of erythrocytes, a measure of blood flow within the 1-mm³sampling volume. The LDF methodology and validation techniques have been reviewed in detail elsewhere. 19LDF, MAP (via  radial artery catheter), and t°C signals were simultaneously recorded at a sampling frequency of 250 Hz using an analog-digital converter and custom data acquisition and analysis software (SnapMaster v3.02; HEM Data Corp., Southfield, MI). Pump flow rate was manually recorded every minute and with every change during NP-CPB; systemic vascular resistance (SVR) was calculated using the formula [(MAP − central venous pressure)/(pump flow)]× 80. Hematocrit was measured in arterial blood before NP-CPB and at four time points during NP-CPB (minutes 8, 25, 40, and at 10 min after the onset of systemic rewarming) using a blood gas analyzer (Stat Profile 5; Nova Biomedical, Waltham, MA).

The study period was divided into 12 standardized segments: pre-CPB (immediately before the start of NP-CPB); minutes 0–1, 1–3, 4–6, 7–9, 11–13, 17–19, 20–22, 25–27, 30–32, and 40–42 during NP-CPB; and 10 min after the onset of systemic rewarming. Each segment was analyzed with respect to changes in overall values (time-domain analysis), as well as for the development of organized oscillatory activity (approximate entropy determination and frequency-domain analysis).

For each measured variable (MAP, LDF^forehead, LDF^finger, and t°C), the instantaneous values within each study segment were averaged to obtain one data point per segment. From these, ensemble mean ± SE across the eight subjects was subsequently calculated, resulting in an overall time series for each variable.

A model-independent statistic quantifying regularity (orderliness) in serial data were computed for each study segment as the logarithmic likelihood that patterns in the continuous laser Doppler waveforms that were similar remained similar on subsequent incremental comparisons. 20,21This was accomplished with a custom-made C program, following a previously described algorithm 21that used the length of the epoch (N), the number of previous values used for the prediction of the subsequent value (m), and a filtering level (r). The noise filter (r) was normalized to the SD of the N amplitude values. For each study segment, we used m = 2, r = 20% of the SD of the amplitude values, and N = 900, because theoretical considerations suggested these parameters as optimal. 20,21The first 3 min of NP-CPB were not included in analysis due to the wide fluctuations in LDF. Average values across the eight subjects were subsequently calculated for each study segment.

The frequency and spectral power of LDF and MAP oscillations were characterized for each study segment by auto power spectral density (APSD) analysis, using a Parzen window with a frequency resolution of 0.005 Hz. The mean value of the signal was subtracted from each time point to eliminate its influence on the APSD. 22Power spectra were separated into LF (0.04–0.10 Hz) and HF (> 0.10 Hz) bands. The spectral power in each frequency band was calculated by integrating the area under the APSD curve and was expressed in normalized units (NU), which represented the ratio (in percent) of each power component to total spectral power, to minimize the effect of changes in total power on assessments of LF and HF power. 8,22When oscillatory activity occurred simultaneously in LDF^forehead, LDF^finger, or MAP, we computed the cross-power spectra, coherence function, and phase spectra to assess the power interchanged and the phase delays between pairs of signals, as previously described. 10All frequency-domain analyses were performed using commercially available software (SnapMaster v3.02, HEM Data Corp.).

Data were expressed as mean ± SEM. Changes from baseline in LDF^forehead, LDF^finger, MAP, SVR, t°C, as well as in the APSD^MAP, APSD^forehead, and APSD^fingerwere tested by H-test (Kruskall-Wallis);P < 0.05 was considered statistically significant. As previously recommended, 23if one or more time- or frequency-domain indices for a given analysis had a highly skewed distribution (i.e. , skewness coefficient ≥ 1.00), logarithmic transformation of these indices was performed to provide an approximately normal distribution for the purpose of statistical analysis.

Both univariate and multivariate generalized estimating equations (GEE) were constructed using LDF and APSD as dependent variables and MAP, SVR, t°C, and hematocrit as covariates (using SAS v.8; SAS Institute, Cary, NC), with P < 0.05 accepted as statistically significant. The GEE method takes into account the potential correlation among measurements in the same subject between time points by estimating the covariance structure of the data and adjusting the significance levels accordingly. This provides greater power to find an effect in studies with relatively large numbers of observations per subject. 24Since LDF, APSD, and all independent variables were highly correlated with time, the multivariate GEE analysis excluded time as a covariate to avoid redundancy.

Onset of NP-CPB consistently resulted in a transient increase in LDF^forehead(mean, 4 min; range, 3.5–5 min), followed by a decrease to approximately 50% below baseline values for the remainder of the NP-CPB period (fig. 1A). Conversely, LDF^fingerremained near baseline during NP-CPB (fig. 1B). The time courses of simultaneously recorded hemodynamic variables (MAP, SVR) and t°C are displayed in figures 1C and D. By univariate GEE analysis, LDF^foreheadwas inversely correlated to MAP (P = 0.004) and SVR (P = 0.002), directly correlated to t°C (P = 0.09), and not correlated to hematocrit. In contrast, univariate models for LDF^fingerwere nonsignificant for all of the aforementioned independent variables.

The average approximate entropy of LDF^foreheadsignal was significantly reduced during NP-CPB, from 1.2 at baseline (pre-CPB) to less than 0.5 after 40 min of NP-CPB (fig. 2A;P < 0.001 by univariate GEE model). The increase in regularity, as reflected by a decrease in approximate entropy, occurred early and persisted throughout the measurement period. On average, approximate entropy of LDF^foreheaddecreased by 0.17 units for every 10 min during NP-CPB. In contrast, the average approximate entropy of LDF^finger, after an initial decrease, showed an upward trend during much of NP-CPB (fig. 2B).

Although short epochs of regular activity alternating with highly irregular ones were present in the LDF^foreheadduring the baseline period (pre-CPB), the increased regularity of LDF^foreheadwith NP-CPB was associated with visible self-sustained oscillations (fig. 3A) in all eight subjects, with a median onset time of 5.25 min (range, 4.5–16 min). Power spectral analysis identified the frequency of forehead flow oscillations at 0.13 ± 0.03 Hz (within the HF band;fig. 3B).

Finger skin blood flow displayed oscillatory variations in only five of eight patients (fig. 3C). These occurred in the LF band, with a median frequency centered at 0.07 Hz (range, 0.04–0.09 Hz;fig. 3D). Simultaneously recorded MAP was nonoscillatory, characteristic for NP-CPB (figs. 3E and F).

During NP-CPB, APSD^foreheadshowed a progressive gain of spectral power in the HF band, from 17.5 NU in the first minute after NP-CPB onset to 89.1 NU after 40 min of NP-CPB (fig. 4A). Univariate GEE models for HF power in the APSD^foreheadconfirmed direct associations with MAP (P < 0.001) and SVR (P < 0.001) and inverse associations with LDF^forehead(P < 0.001) and t°C (P = 0.002). In the multivariate model, all of these covariates retained their significance. No association was found between APSD^foreheadand hematocrit.

There were no statistically significant changes during NP-CPB in the LF spectral power of APSD^finger(fig. 4B). Univariate GEE analysis for LF power in the APSD^fingeridentified a direct association with t°C (P = 0.05), inverse associations with LDF^finger(P < 0.001), MAP (P < 0.001), and SVR (P = 0.04), and no association with hematocrit. In a multivariate model, only the relation with LDF^fingerand MAP remained significant.

No relation could be established between the amplitude and the frequency of spontaneous LDF oscillations at the forehead and finger sites. Whenever simultaneous oscillations in LDF^forehead(fig. 5A) and either LDF^finger(fig. 5B) or MAP (fig. 5C) were detected, cross-spectral analysis confirmed the independence of HF band oscillations in the LDF^forehead(fig. 5D). In contrast, when present, LF oscillations in LDF^fingerand in MAP were highly coherent, with a positive phase lag (i.e. , LDF leading MAP) of π/6 rad, suggestive of peripheral (microvascular) origin with upstream transmission (figs. 5E and F).

The cutaneous microcirculatory network, like other body tissues, exhibits spontaneous oscillations in diameter and flow (vasomotion) ranging from periodic, nearly sinusoidal, to highly irregular, chaotic patterns. 25,26Although well documented for more than half a century, the origin, fundamental mechanisms, and significance of these fluctuations remain debated, with some researchers hypothesizing that skin blood flow is exclusively under local control 27–29and others suggesting both central adjustments and baroreflex-induced changes. 16,30,31Traditionally, autonomic neural responses at the microvascular level have almost exclusively been attributed to adrenergic pathways, resulting in oscillations with sympathetic frequency response characteristics (i.e. , VLF and LF bands); analysis of oscillations in the HF range typically has been confounded by variability in other cardiovascular signals, which often are synchronous with respiration-induced changes in heart rate, stroke volume, and blood pressure. 7–10 

The current study is the first to document the development of HF oscillatory activity in the microvasculature in the absence of cardiac contractions and respiration. Furthermore, the increasingly regular oscillations (at 0.13 ± 0.03 Hz) in the forehead microcirculatory flow were independent of oscillations in LDF^fingeror MAP. Prior studies have identified the emergence of oscillations in systemic arterial pressure in the absence of natural heartbeat in experimental animals 32–34and humans, 35,36referred to as vasomotor waves or idioperipheral pulsations. These oscillations always have been in the VLF and LF range and have been attributed to baroreceptor-induced variations in sympathetically mediated vascular resistance. 32–36The current documentation of LF oscillations in the microvasculature of a region with rich adrenergic innervation (finger) is consistent with these previous studies that focused on arterial pressure. However, the identification of HF oscillations has not previously been reported in the absence of cardiac and respiratory factors. Their presence supports our hypothesis as to their peripheral origin. In light of the confirmation that similar HF oscillations of the forehead microvasculature (0.14 ± 0.02 Hz) during systemic infusion of phenylephrine were abolished by atropine, 14,18we propose that the comparable activity in the current investigation constitutes a mechanism of cholinergic oscillatory control at the level of the microvasculature.

In the absence of extrinsic control, vascular smooth muscle cells contract and relax independently, such that neighboring vessels typically oscillate out-of-phase, thus offsetting one another in the summated laser Doppler signal. 37,38The emergence of periodic self-sustained oscillations suggests that the different oscillating microvessels are phase-locked by a control mechanism that causes an entire local microcirculatory network to oscillate in synchrony at a single frequency, thereby coordinating the fluctuations of the 10–60 capillaries in the 1-mm³sampling volume of the laser Doppler probe. 10,19The different fundamental frequencies of the periodic oscillations observed in the two microvascular regions studied herein are consistent with the frequency response characteristics of the predominant autonomic supply to the respective region (i.e. , adrenergic for finger microvasculature, cholinergic for forehead microvasculature). It has been suggested that oscillatory perfusion may confer specific physiologic advantages over steady state flow. 26,37–39Both theoretical and experimental models have suggested that the onset of periodic vasomotion results in an increase in the effective diameter of a single vessel. 26,39Moreover, the impedance of a vessel whose diameter varies sinusoidally is lower than that of a constant-caliber vessel with the same average diameter. 39Thus, consistent with Poiseuille's determination that changes in flow are proportional to the fourth power of the radius, the potential physiologic significance of rhythmical vasomotion include enhancement of microcirculatory flow and mass transport and promotion of lymphatic drainage. 26,37–39Such phenomena would likely be beneficial during pathophysiologic conditions of compromised tissue perfusion. In addition to their development during NP-CPB described in this study, HF microvascular oscillations have been observed during clamping of a regional feeding artery, 40,41in the rat cerebral cortex during norepinephrine infusion and hypoperfusion, 42in the feline mesentery during vasoconstriction, 43and in the human forehead microcirculation during systemic infusion of phenylephrine 14,18and hyperventilation. 44 

Our data suggest that nonpulsatile perfusion was associated with different microvascular oscillatory responses, not only in terms of fundamental frequencies, but also in regularity of oscillatory behavior. LF oscillations in LDF^fingerand MAP, as may be caused by baroreflex-mediated variations in sympathetic nerve activity, 32–36were an inconsistent finding (in only five of the eight studied patients) and did not display increasingly regular patterns during NP-CPB. In contrast, there was a progressive increase in regularity of HF oscillations in the forehead in every patient, suggesting an active homeostatic control mechanism that becomes increasingly more efficient in synchronizing neighboring oscillating microvessels during NP-CPB. Here, a distinction needs to be made between regularity, which describes the recurrence of patterns within the LDF signal, and power spectra, which describe the relative magnitudes of oscillatory components. Therefore, increased regularity of a signal cannot be accounted for simply by an increase in the amplitude of the oscillations. In the case of forehead HF oscillations, both phenomena seem to occur during NP-CPB, as the decrease in approximate entropy is accompanied by a progressive gain in HF spectral power.

Depulsation of the systemic circulation results in a progressive increase in sympathetic nerve activity 45,46(due to suppression of baroreflex-mediated inhibition of the vasomotor center), leading to progressive arterial vasoconstriction and increase in SVR, 46,47which was confirmed in the current study (fig. 1C). It is generally accepted that the pathophysiologic processes under nonpulsatile perfusion are primarily related to the behavior of the sympathetic pathways that regulate tissue perfusion. Using multivariate generalized estimating equations, we found increased SVR and decreased LDF to be independent predictors of HF spectral power gain in the forehead. This suggests that progressive increases in peripheral vascular resistance and altered local tissue perfusion are associated with the activation and progressive synchronization of HF oscillations in forehead microvascular flow. Toda et al.  46studied the simultaneous changes in sympathetic nerve activity and regional blood flow to vital organs and found organ-specific responses to nonpulsatile perfusion, with a decrease in renal cortical blood flow (consistent with increased renal sympathetic nerve activity), but no change in cerebral and myocardial regional blood flow. They raised the possibility that efferent vagal activity may modulate regional blood flow to certain organs in the face of increased sympathetic activity, but prior to the current study this potential process had not been explored. We speculate that the emergence and progressive synchronization of cutaneous microvascular oscillations—in a region of cholinergic innervation (forehead), at a frequency associated with parasympathetic activity (0.10–0.18 Hz), and in response to a known activator of sympathetic activity (i.e. , nonpulsatile perfusion)—may represent a manifestation of the counterregulatory processes proposed by Toda and colleagues. 46 

Thus, the potential implications for the anesthesiologist are that monitoring microvascular oscillatory phenomena may more effectively identify states of altered tissue perfusion and, perhaps more importantly, may identify a patient's ability to adapt to such alterations. Understanding and better characterizing these oscillatory control mechanisms may lead to therapeutic interventions designed to reduce end-organ injury following procedures requiring extracorporeal nonpulsatile perfusion and provide physiologic basis for the long-term function of nonpulsatile ventricular assist devices. Further studies are required to assess to what extent the changes observed in the forehead are reflective of other cholinergically innervated microvascular beds (e.g. , cerebral, coronary). However, in a porcine model of NP-CPB, Mutch et al.  demonstrated that computer-controlled biologically variable pulsatile NP-CPB that incorporated HF oscillations in flow was associated with significant improvements in cerebral oxygenation when compared with NP-CPB 48or conventional pulsatile CPB. 49Their findings emphasize the impact that preservation of biologic variability in microvascular flow during NP-CPB may have on critical end-organ perfusion. This theoretically could be achieved either by applying external biologically variable pulsations or by monitoring and modulating the adaptive mechanisms proposed in the current study.

Studies designed to assess the pathophysiologic effects of NP-CPB on microvascular control mechanisms should take into account the potential effects of pharmacologically induced alterations in vascular tone and of hemodilution on tissue perfusion. Although our study did not control for variations in the doses of vasoactive drugs during NP-CPB, this should not detract from the significance of confirming a peripheral etiology to HF microvascular oscillations, which was our primary hypothesis. Future studies designed to further investigate the potential significance of these oscillatory phenomena should be powered to allow corrections for such confounding variables.

We found no correlation between changes in hematocrit during NP-CPB and any of the time- or frequency-domain indices of cutaneous microvascular flow. This is in agreement with recent studies demonstrating that the linear relation between blood flow and blood velocity is not affected by changes in hematocrit in clinical ranges and can be explained by the Fahraeus-Lindquist effect. 50In addition, by actually measuring the flux of erythrocytes, not plasma flow, LDF is a hematocrit-independent measure of tissue perfusion within clinical ranges. 19 

In conclusion, the current documentation of HF microvascular oscillatory control during NP-CPB confirms that HF oscillations in the microvasculature may arise independently of cardiac or respiratory activity. The dynamic changes in synchronization of forehead microvascular oscillations may represent a microvascular homeostatic mechanism in response to the increased sympathetic nerve activity that occurs during nonpulsatile perfusion.

The authors thank Razvan Mociani (Senior Computer Analyst, Apache Software, Toronto, Canada), for designing the C program used to compute approximate entropy, and Joseph Mathew, M.D. (Associate Professor, Department of Anesthesiology, Duke University Medical Center, Durham, North Carolina), for critically reviewing the manuscript.