Effects of assisted and variable mechanical ventilation on cardiorespiratory interactions in anesthetized pigs

After approval by the local Animal Care Committee, eight female juvenile pigs (32–42 kg) were anesthetized intravenously (propofol 2–7 mg kg⁻¹ h⁻¹, sufentanil 0.3–1.5 µg kg⁻¹ h⁻¹, and, for the controlled MV modes only, atracurium bromide 1 mg kg⁻¹), and tracheally intubated with a cuffed tube (8.0 mm inner diameter). An indwelling catheter was inserted in the right femoral artery and a pulmonary artery catheter (Swan-Ganz catheter, Edwards Lifesciences, Irvine, CA, USA) advanced through an introducer sheath placed in the left external iliacal vein. An esophageal balloon catheter (Cardinal Health, USA) was introduced in the esophagus, in order to measure the esophageal pressure, a surrogate of intra-pleural pressure. After intubation, the animals were mechanically ventilated using an EVITA XL ventilator (Dräger, Germany) with PSV mode until a stable respiratory pattern was reached (approximately 30 min).

Then, as illustrated in figure 1, the animals underwent 30 min of two controlled and two assisted MV modes in randomized order according to a Latin square design: PCV, random variable PCV (PCV_noisy), PSV, and random variable PSV (PSV_noisy). During PSV and PSV_noisy each inspiration was triggered when the initial spontaneous inspiratory airflow of the animal exceeded a predefined threshold (5–7 L min⁻¹, the minimum value that avoided non-respiratory-related triggering was selected for each animal), whereas during PCV and PCV_noisy it was initiated artificially by the ventilator at a fixed rate (set as the average respiratory rate found for the last preceding period of PSV or PSV_noisy). At the onset of inspiration the ventilator started to apply a given positive driving pressure to the subject airways, and continued until a condition for the termination of inspiration occurred. For assisted MV such a condition was the airflow falling below a certain level of the peak inspiratory airflow (15–25%), whereas for controlled MV was reaching a preset duration of inspiration. These conditions were adjusted to achieve in all MV modes an inspiration/expiration ratio of 0.3, compatible with the values found for PSV in the clinical setting (Tokioka et al 1989), since differences in such a ratio could affect RSA (Strauss-Blasche et al 2000). Finally, the inspiration was followed by passive expiration. The inspiratory driving pressure was tuned cycle-by-cycle by an automatic control system (Beda et al 2010) to achieve a target mean tidal volume (V_T) of 12 mL kg⁻¹ in all modes, a random (white noise-–no correlation with preceding cycles) breath-by-breath V_T variability of 30% for PCV_noisy and PSV_noisy (Spieth et al 2009a), and no variability for PCV and PSV. The value of mean V_T was chosen to allow lower respiratory frequency, which is expected to increase RSA. Furthermore, such mean V_T is commonly used during general anesthesia in non-injured lungs. Throughout the experimental protocol, anesthetic agents were administered at constant infusion rates, the inspired fraction of O₂ was 0.4, and the positive end-expiratory pressure was 5 cmH₂O. Details of the MV and anesthesia are summarized in table 1.

Zoom In Zoom Out Reset image size

The following signals were acquired synchronously during the last 20 min of each MV mode with sampling frequency of 2000 Hz, using a data acquisition card (NI USB-6210, National Instruments, USA) connected to a PC: airway pressure (P_AW) and esophageal pressure (P_E), using pressure transducers (163PC01D48-PCB, Sensortechnics GmbH, Germany); one non-standard derivation of ECG using a biosignal amplifier (LP511, Grass Technologies, USA); and femoral arterial blood pressure measured invasively using disposable transducers (DTXPlus, Becton Dickinson, USA). Airflow was acquired continuously from the ventilator to the PC through serial interface (sampling frequency of 180 Hz), and synchronized off-line with the other signals (through automatic time alignment, using maximal covariance, between the airway pressure signal acquired using the data acquisition system and the airway pressure signal that could also be retrieved from the ventilator). At the end of each period of MV, arterial and mixed-venous blood gases, cardiac output, and blood temperature were measured (the last two using an M1097 monitor, Agilent, Germany).

The onset of inspiration and expiration were automatically detected in the airflow signal, and then manually edited. For PSV and PSV_noisy, the data regarding the 5% longest respiratory cycles of each task were excluded from further analysis, to avoid the confounding effect of episodic prolonged expiratory pauses. For the remaining respiratory cycles, V_T, respiratory period, and inspiration/expiration ratio were calculated, and the peak and mean values of airway and transpulmonary pressure (where the latter is the difference between airway and esophageal pressure) computed.

The time series of the intervals between consecutive R-peaks in the ECG, and of systolic, diastolic, and mean arterial pressure (SAP, DAP, and MAP respectively), and pulse pressure (PP = SAP – DAP), were automatically estimated from the ECG and arterial pressure signals and manually edited to avoid artifacts, as described elsewhere (Beda et al 2007). For each respiratory cycle, the difference between the longest and the shortest R–R interval was taken as an estimate of RSA amplitude, while the differences between the largest and the smallest value of SAP, MAP, DAP, and PP were taken as estimates of the amplitude of their respiratory-related oscillations. Also, for each respiratory cycle, the inspiratory (expiratory) gradient of RSA was estimated as the slope of the line interpolated from the subset of the time series of R–R intervals starting and/or ending within the inspiration (expiration), as described elsewhere (Carvalho et al 2009). The average gradients were then calculated considering all the cycles belonging to each task. The same processing was applied to compute the inspiratory/expiratory gradients of SAP, DAP, and PP.

We estimated the transfer function and the coherence (after spline interpolation and resampling with a sampling frequency of 4 Hz) using the Welch modified periodogram method (with windows of 1024 samples, overlapping 50%) for the following time series: airway pressure → R–R interval, and SAP → R–R interval. The first transfer function provides information about the direct influence of respiration during MV on R–R interval duration, while the second was used to assess the characteristics (or the absence) of baroreflex modulation of RSA. For this purpose the average phase shift (in radians) and the average gain of the transfer functions in the respiratory frequency range (high-frequency (HF): 0.15–0.5 Hz) were computed, considering only those frequencies for which the estimated coherence was above a threshold value. The latter was estimated using the surrogate data approach detailed by Porta et al (2002), an improvement compared to the common practice of choosing an arbitrary value (typically 0.5). For the transfer function X → Y, a positive (negative) phase shift indicates that X oscillations precede (follow) those of Y. The average phase shifts in the HF band were calculated also as delays (in seconds) after applying the transformation delay(f) = phase(f)/(2 ⋅ π ⋅ f) for each frequency f.

To assess quantitatively the presence of significant CVC, for all subjects and tasks the time series of the interval between the last R-peak in each respiratory cycle and the start of the next inspiration was computed, and the median proportional Shannon entropy (SH_α) of the last 200 samples was computed using the same method reported by Tzeng et al (2003). Then, a significance threshold was computed using the Monte Carlo approach: the values of SH_α of 1000 series of 200 random samples with a uniform distribution were computed, and the value T of the 5% percentile of the resulting distribution of SH_α was estimated. For each animal in each task, if SH_α < T the CVC was considered statistically significant and analyzed qualitatively by observation of the synchrograms (i.e. plotting one dot for each R-peak, with the respiratory cycle number the R-peak belongs to as abscissa, and the corresponding interval between the R-peak and the previous onset of inspiration as ordinate-–for an example, see figure 5), such that CVC can be observed as the presence of horizontal 'bands' of dots, as described elsewhere (Tzeng et al 2003).

Values are reported as mean ± SD. The differences between the MV modes in the mean value of the cardiovascular and respiratory parameters were assessed using two-way repeated measures ANOVA (factor 1: controlled/assisted MV; factor 2: non-variable/variable MV) and post-hoc paired t-tests. Results were considered statistically significant when p < 0.05.

Table 2 reports the hemodynamic and ventilatory variables. R–R interval, SAP, DAP, MAP, cardiac output, minute ventilation, respiratory rate, and inspiration/expiration ratio did not differ significantly between assisted and controlled MV, or between variable and non-variable ventilation. SAP was slightly higher during PCV than PCV_noisy (mean difference = 4.9 mmHg), and blood temperature was slightly higher during assisted compared to controlled MV (mean increase = 0.4 °C). Arterial and venous levels of CO₂ and O₂ were similar for all MV modes, with the exception of a higher partial pressure of O₂ in arterial blood for variable MV (mean difference of approximately 11 mmHg), and a partial pressure of CO₂ in venous blood for assisted compared to controlled MV (mean difference of approximately 8 mmHg). The V_T variability was higher during variable versus non-variable and assisted versus controlled MV, while mean V_T differences were <5% on average. Mean and peak values of P_AW, but not transpulmonary pressure, decreased significantly during assisted compared to controlled MV.

The amplitude of RSA more than doubled from controlled to assisted MV modes (figure 3(a)). A similar effect was found also for the amplitude of the respiratory-related DAP oscillations, while that of SAP, MAP, and PP did not change significantly among MV modes (figures 3(b) and (c)). Also, during assisted MV, the gradient of R–R interval duration was negative during inspiration and positive during expiration, while during controlled MV a coherent pattern among the subjects could not be identified (figure 3(d)). Conversely, the gradient of SAP was positive during inspiration and negative during expiration in all MV modes (figure 3(e)). No coherent pattern of PP or MAP along the respiratory cycle was found among the animals in any of the MV modes.

Zoom In Zoom Out Reset image size

Figure 4 illustrates the average gain, phase shift, and coherence of the transfer functions between P_AW and R–R interval, and between SAP and R–R interval, during PCV and PSV (the results are very similar to those of PCV_noisy and PSV_noisy, respectively, and hence are not reported). Figure 4 also reports the corresponding average values of gain and phase shift in the HF band for all MV modes (figures 4(g)–(j)). The gain and phase shift of both transfer functions changed considerably between assisted and controlled MV. Specifically, at respiratory frequencies (i.e. HF) R–R interval and SAP oscillations were in phase during controlled MV, while during assisted MV R–R interval oscillations preceded those in SAP of more than 1 s (p < 0.001, figure 4(j)), resulting in the two being almost in counter-phase (i.e. delay larger than –π/2, p < 0.048). Similarly, HF oscillations of R–R interval and P_AW are in phase during controlled but not assisted MV (p = 0.007, figure 4(i)), when P_AW significantly leads R–R interval. The estimated gain in the HF band between P_AW or SAP and R–R interval increased from controlled to assisted MV (p < 0.026, figures 4(g) and (h)).

Zoom In Zoom Out Reset image size

Figure 5 illustrates the effect of different MV modes on CVC in a representative animal. The synchrograms (figure 5(a)) show that during controlled MV CVC is absent or very limited (as R-peaks are not organized in horizontal bands), while it is clearly present during assisted MV, resulting in separate horizontal bands of dots. Accordingly, privileged locations for beat occurrence within the respiratory cycle could be observed during assisted compared to controlled MV (figure 5(b)). Similar patterns were observed in six out of eight animals investigated. The value of SH_α decreased significantly from controlled to assisted MV (p = 0.035) implicating a more 'organized' synchrogram. SH_α was lower than the significance threshold (i.e. presence of statistically significant CVC) in two cases for PCV, three cases for PCV_noisy, and seven cases for PSV and PSV_noisy.

Zoom In Zoom Out Reset image size

Our results indicate that during controlled MV a variable pattern of V_T does not increase RSA amplitude. This observation differs from what reported by Mutch et al (2005), which might be related to differences between the experimental protocols. First, Mutch et al (2005) applied simultaneous changes of respiratory period and V_T (to maintain minute ventilation constant breath-by-breath). Second, they employed considerably higher respiratory rates (approximately 20 min^–1), which can influence RSA per se (Berntson et al 1993). Third, the use of different anesthetic agents with distinct cardiovascular and autonomic effects (Barash et al 2005) may have also influenced the results.

An additional new finding is that variable V_T also does not influence RSA amplitude during assisted MV.

CVC was practically absent during controlled MV in our animals, in agreement with a previous report in humans (Larsen et al 1999). However, we showed for the first time that assisted MV restores a certain level of CVC. Hence, the overall effect of assisted MV goes beyond the simple increase of RSA, by reintroducing some level of synchronization between the cardiovascular and the respiratory system (see discussion above). It is interesting to note in this context that a loss of coupling between communicating organs has been suggested as a mediator of multiple organ dysfunction syndrome, and re-coupling may be a necessary step for recovery (Buchman 2002). Furthermore, variable ventilation did not improve CVC, in contrast to the speculations of Mutch et al (2005).

During MV, the level of RSA clearly depends greatly on the mode (controlled or assisted) employed. Our results suggest that RSA during controlled MV could be significantly modulated by mechano-electric feedback in the heart, while central drive is suppressed. Hence, the commonly used interpretation that RSA reflects parasympathetic activity, based primarily on investigations of spontaneously breathing subjects, is probably not applicable in this condition. Consequently, previous speculations that changes in the parasympathetic modulation are responsible for the variation in RSA during controlled MV between different levels of sedation (Unoki et al 2009), settings of MV (Van de Louw et al 2010a), and managements of anesthesia (Galletly et al 1994, Latson et al 1992), might be reconsidered taking into account the possible direct effects of MV on cardiovascular oscillations, such as the mechano-electric feedback in the heart due to variations in hemodynamic and intra-thoracic pressure (Beda et al 2011).

Furthermore, care should be taken when using indexes based on RSA as outcome predictors in mixed populations of patients including mechanically ventilated and spontaneously breathing subjects, as in the work by Schmidt et al (2005). When taking the results of the present and previous works (Freyschuss and Melcher 1975, Koh et al 1998, Van de Louw et al 2008) together, it seems that the transition from spontaneous to assisted to controlled ventilation is associated with a progressive decrease in RSA. If the plausible assumption is made that the severity of the clinical condition and the level of sedation influence the choice of the ventilation strategy (i.e. sicker and deeply sedated patients are more likely to receive controlled MV), a significant correlation between outcome and RSA could reflect the use of different ventilation strategies and anesthesia management, rather than pathological alterations in autonomic function.

It is questionable if the increased heart-rate variability and RSA amplitude found during assisted compared to controlled MV should be interpreted as a beneficial effect of the former. The discussion above suggests that they may simply reflect the activity of distinct physiological mechanisms depending on the MV mode considered. This adds a further argument to the increasing criticism of the use of the amplitude of heart-rate variability to assess the 'health' of autonomic regulation (Parati et al 2006).

The importance of RSA/CVC for improving gas exchange through better ventilation/perfusion matching proposed by Hayano et al (1996) and supported by other investigations (Giardino et al 2003) was questioned in recent studies, based on experimental evidence that there is no change in the ventilatory equivalents of CO₂ (ratio between CO₂ production and minute ventilation) and O₂ (ratio between O₂ consumption and minute ventilation) between conditions with significantly different levels of RSA/CVC (Hayano et al 1996, Tzeng et al 2009). However, these indexes are imperfect measures of pulmonary gas exchange efficiency (Tharion and Subramani 2011). Nevertheless, our results are in line with the conclusions in the former works: the two-fold increase in RSA and the significant CVC during assisted compared to controlled MV was not paralleled by a significant improvement in oxygenation. It is worth noting that our study is the first to provide evidence of this kind using the more appropriate direct measurement of blood oxygenation.

Several limitations in our study should be acknowledged. First, while the MV settings were kept as close as possible to those generally adopted in clinical routine (e.g. elective surgery), inspiration/expiration ratio during controlled MV was kept relatively low compared to common practice, in order to guarantee comparability with assisted MV. The resulting short inspiratory times (of the order of 1 s) might have favored the faster mechano-electric feedback in the heart in modulating RSA, while penalizing the slower baroreflex response.

Second, respiratory rates were kept relatively low in order to investigate RSA, resulting in levels of arterial CO₂ higher than the physiological range. However, since we used a crossover design this limitation is not likely to play an important role. Furthermore, Sasano et al reported no significant differences in RSA between end-tidal partial pressure of CO₂ of 40 mmHg (i.e. 'physiological condition') and 50 mmHg (compatible with the values found in our study, since partial pressure of CO₂ in arterial blood is expected to be 1–5 mmHg higher) (Sasano et al 2002). This, together with the depressed responsiveness to CO₂ associated with the sedation regime (Barash et al 2005, Poyhonen et al 2004) and the level of pH within the physiological range (see table 2), suggests that lower values of arterial CO₂ would not have influenced the results obtained.

Third, a muscle paralyzing agent was administered during controlled ventilation to avoid spontaneous breathing activity. However, previous findings suggest that the dosages of atracurium used in this study have no effect on the autonomic nervous system (Hughes and Chapple 1981).

Fourth, the interpretation of phase shift computed from transfer function estimation is ambiguous by definition, since a positive shift of α rad (0 ≤ α ≤ 2π, input leading output) can always be interpreted as a negative shift of α – 2π rad (output leading input). Applying this concept to time delays, the negative average delay between SAP and R–R during assisted ventilation, which we estimated to be more than 1 s, could as well be 'recalculated' as SAP leading R–R with a delay of >5.2 s. However, such positive delay seems physiologically unlikely, because it is significantly larger than the average respiratory period during assisted MV (<4.5 s), which would imply a baroreflex modulation in the HF acting only on the next respiratory cycle. Nevertheless, due to the limited literature available regarding the baroreflex delay in pigs, we cannot completely exclude this hypothesis.