Computer models in bedside physiology
Zhang, Y.
Publication date
2013
Link to publication
Citation for published version (APA):
Zhang, Y. (2013). Computer models in bedside physiology.
General rights
It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulations
If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.
91
Chapter 6
Dynamics of Pulse Pressure Variability and the Difficulty of Predicting
Fluid Responsiveness.
Yanru Zhang and John M. Karemaker
(to be submitted)
Abstract
Background: In the intensive care setting intravenous fluid is often administered to prevent or correct hypotension. However, not all patients will respond favourably to a fluid load. Measurement of respiration induced pulse pressure variability (PPV) is currently under study as method to predict ‘fluid responsiveness’ (FR). In this study we tested PPV as predictor for FR in healthy (awake) test subjects and compared their responses to a computer model of the circulation. We expected that PPV in the test subjects would poorly correlate to the obtained increase in cardiac output (CO) due to spontaneous variability in PPV and circulatory control, and in the model the correlation would be almost perfect. Methods and results: We re‐analyzed recordings in 6 healthy test subjects from an earlier study where they had received ~ 1.5 l of saline infusion in 20 minutes as part of a space mission simulation (HDT’88). We performed spectral analysis of pulse pressure and of PPV. Both showed a large contribution of VLF and LF variability to total variance. Due to the deviant results in one test subject, average PPV measured over 5 minutes correlated poorly to the obtained increase in CO. The computer model behaved well in this respect, even better when autonomic outflow was diminished as simulation of sedation or anesthesia. Conclusions: PPV is a complex measure when used in awake subjects. It shows large moment‐to‐moment variability due to low‐frequency oscillations in the cardiovascular and respiratory control systems. Its capacity to predict volume responsiveness, or use in automated systems for fluid loading, is restricted to stable, heavily sedated or anesthetized and ventilated patients. In those cases the circulation reduces to the simple textbook situation.93
Introduction
The teachings of many great physiologists have shaped our understanding of the circulation. Notably the work of Arthur Guyton (1919 – 2003), by his experimental work and computer modeling (6, 14, 15, 17, 18), and in particular by his influential textbook since 1956 (16), is still the basis for many of today’s clinical interventions. He stressed the importance of venous return, central venous pressure and mean circulatory filling pressure to understand what made the circulation go round (19, 20). In that light, it came as a surprise that static data like right atrial pressure do not help much to guide volume infusion in the case of a failing circulation (22, 25). Intravenous volume, if correctly given to critically ill patients will improve ventricular function and survival rate after surgery (13, 28). However, overfilling may put the lungs, heart and other organs at risk (3, 34). Over the last 20 years, attention has shifted to dynamic parameters derived from continuous arterial blood pressure like pulse pressure and stroke volume variation (PPV and SVV) as better predictors of volume responsiveness. These variations are measured over the course of one or more respiratory cycles, working on the assumption that PPV and SVV bring the mechanical effect of respiration on ventricular filling to light, thereby answering the question whether there is room for increased diastolic filling to improve cardiac function. However, in view of the multitude of supervening regulatory mechanisms and real world ‘noise’, these measures have shown to be applicable only in sedated patients on mechanical ventilation (25, 26). Under those circumstances, many of the normal influences are no longer at work, or at least are much less prominent. We undertook the present study to first, test the dynamic properties of PPV in spontaneously breathing, awake, healthy test subjects who received a massive, quick infusion of saline. Next, we used a computer model of the (regulated) circulatory system, to simulate these same events in the ‘awake’ condition of the model. Then we purposely reduced the effectiveness of the models’ autonomic nervous outflow, to simulate the condition of a sedated or anesthetized patient. The computer model, by design, has just as much ‘noise’ as the scientist allows it to have. Our hypothesis was that the model (like the patient under anesthesia) allows using PPV as an accurate probe of cardiac function, while in the real world, in awake subjects, normal variability virtually precludes this. For these purposes we re‐analyzed continuous blood pressure recordings from 6 healthy (male) test subjects who had received a large infusion of saline in a short period of time in thesetting of a space‐mission simulation study (HDT’88, (2)). Next, we adapted our earlier developed computer model of the circulation(39) to simulate the effects of this same volume infusion, with and without preceding volume loss, both in the ‘awake’ and ‘sedated’ condition.
Methods and test subjects
Experiments in healthy test subjects. We used recordings from a space‐mission simulation study that has been conducted in 1988 in the DLR‐Institute of Cologne, Germany (HDT’88) and has extensively been published elsewhere (2). In short, six young (21‐34 yrs) male, healthy subjects went through a 16 days baseline period, 10 days 6º head‐down tilted space simulation and 7 days recovery. During those periods a number of interventions were repeatedly applied to test the condition of the cardiovascular and fluid regulatory systems, of which the most important for the present paper is an infusion of saline (22 ml/kg over 20 minutes) (11, 21). The DLR Institutional Research Review Committee had approved the study and all subjects participated after written, informed consent. The senior author of the present paper (JMK) participated as one of the PI’s in that study; together with his co‐worker at the time, D.J. ten Harkel, they recorded non‐invasive Finapres® continuous finger blood pressure during all experimental interventions (36, 37). The digital records of beat‐by‐beat systolic, diastolic and mean blood pressure and heart rate have been kept on file and were re‐analyzed here. Cardiac output measurements were made by way of the foreign gas rebreathing method, carried out by Drs. H. Schulz and A. Hillebrecht (11, 21) who had made the results of their measurements available to us. Analysis of blood pressure tracings: after visual inspection of the records for outliers and artifacts, we chose suitable periods of some 5 minutes baseline recording before the main intervention and 5 minutes at the peak of the intervention for spectral analysis of PPV. PPV was measured as: 100*(maximal pulse pressure – minimal pulse pressure)/(average pulse pressure) in a sliding window, covering a period of 10 heartbeats (cf. appendix ).95
Power spectral density curves were computed using the spectral techniques as described in (7, 43) and sampled in the usual frequency bands as proposed by the Task Force (10): very low frequency‐VLF (0.003‐0.04 Hz), low frequency LF (0.04‐0.15 Hz) and high frequency or HF (0.15‐0.4 Hz). The latter is broadly ascribed to respiratory variability. Model description The computer model is an extension of the circulation models to simulate the effects of gravity by van Heusden et al (39), and the model to simulate cardiopulmonary resuscitation by Babbs (1) and adapted in Zhang and Karemaker (44). In short, the circulation model includes four heart chambers, the systemic and pulmonary circulations. In the systemic circulation, the thoracic aorta feeds the arteries to the upper body, kidneys, intestines, liver, and lower body. The upper body returns its blood to the superior vena cava (SVC), the organs below heart level return it to the abdominal vena cava (AVC), the intestinal venous outflow via the liver. Both venous flows join in the right atrium, where the systemic circulation ends and the pulmonary circulation starts. The whole circulation model includes 15 compartments; the basic unit is a compliance in series with a resistance. Blood pressure in the model has (baroreflex driven) ANS control to influence heart rate, cardiac contractility, vascular compliances and unstressed volumes and systemic vascular resistance. The respiration effect that we added here is simulated as a varying pleural pressure that influences intrathoracic pressure, as in Figure 1. All organs in the thorax undergo this pleural pressure, with negligible effect on the ones with relatively stiff walls, i.e. the left ventricle, thoracic aorta, right ventricle during systole and pulmonary artery during systole. The model can be fitted to individual data by inputting a specific subjects’ information, including height, weight, (resting, supine) HR, blood pressure etc. Figure 1. Pleural pressure is added as voltage source into unit of intrathoracic organs that is elastic: right atrium, right ventricle during diastole, pulmonary veins, pulmonary arteries during diastole, left atrium.PPV affected by anesthesia, volume loss & infusion During surgery patients will often be under general anesthesia or at least sedated. Consequently, both sympathetic and parasympathetic activity are lowered or even have completely ceased. The model simulates this condition by an extra gain factor in the (mainly baroreflex driven) output of the autonomic nervous system (ANS) between zero and one. A gain of one implies normal ANS‐output of the baroreflex, the patient is conscious; at a gain of zero, ANS activity has totally stopped; at gains between zero and one different levels of sedation or anesthesia are present. The model simulates volume loss by connecting a ‘bleeding’ branch in the arterial compartment as in Figure 2A, which is uncoupled when the lost volume reaches its preset value. Volume infusion is simulated by connecting a current source to the right atrium as in Figure2B until the preset infusion value is reached. Figure 2 A: Volume loss is simulated as a low resistance branch, which connects the arterial compartment to ground (0 V); the value of the resistance sets the amount of current that will flow when the switch is closed; B: Volume infusion is simulated as a low resistance branch, which connects in the same fashion a high pressure (voltage) source to the right atrium.
97
Modeling & simulation We used the model to simulate the following situations: First simulation series: 1. subject supine, in awake, stable condition, then a volume of 500 ml of ‘whole blood’ is infused. The model does not have an interstitial fluid compartment, only an effective circulating volume, therefore this infusion of ~ 10% of circulating volume is comparable to the 1.6 l saline that was infused in the test subjects. Saline will distribute over the entire extracellular fluid volume. 2. same as the first simulation, but a volume loss of 500ml is induced, followed by the volume infusion of 500ml Second simulation series: 1. subject under anesthesia, volume infusion of 500 ml, subject wakes up. 2. Subject under anesthesia, loss of 500ml, variable infusion volumes of 500 ml and more, subject wakes up.Results
Infusion experiments in test subjects. We restricted our analysis to the second infusion test in the pre‐head down tilt period. The subjects had already undergone one such test, so they knew what was coming. One exception here: the rebreathing data for the second control period infusion of test subject P24 have been lost, for him we had to take the first test. To prevent the additional challenge of acute head down tilting, subjects were in the horizontal, supine position, which is more comparable to the clinical situation. Figure 3 shows an example from the resting control period before the second infusion test in subject P24. The recording shows beat‐by‐beat values of systolic and diastolic blood pressure, heart rate and, in panel B, pulse pressure and pulse pressure variability. Pulse pressure is showing conspicuous very‐low‐frequency oscillations one minute apart that appear exaggerated in pulse pressure variability. Moreover, small variations in amplitude of the respiratory oscillations in pulse pressure appear as low‐frequency oscillations at around 5/min in PPV. This property of the PPV‐algorithm is shown in Figure 4 for all subjects: the bars denote the relative amounts of VLF, LF and HF, both for pulse pressure (Figure4A) and PPV (Figure4B). Obviously there is still some power left in the respiratory band of PPV, however, this should be considered ‘noise’ from looking at the spectral curves – no clear peaks remain visible in PPV‐spectra, although respiratory peaks are, of course, very prominent in the spectra of pulse pressure. About 50% of PPV‐variability in all subjects is accounted for by VLF. Figure 3 A stretch of recording from the resting period before infusion (#2) in P24. A: systolic and diastolic pressures (red bar and lines), blue line: HR. B: Pulse pressure (red line), Pulse pressure variability (blue line).99
Figure 4 Relative amounts of VLF, LF and HF in % of total variability in the control period before infusion. A: Pulse pressure, B: Pulse pressure variability. Table 1 summarizes the results of the infusion experiments and Figure 5 shows cardiac output as function of PPV in the control period. It should be noted that the infusions were on average 1.6 litres in 20 minutes (22 ml/kg) as opposed to the usual clinical fluid challenge of 500 ml (9, 27, 32, 35, 38). This PPV has been measured over a longer control period than that shown in Figure 3, which resulted in a more stable estimate of average PPV. In spite of the large infusion volume, the CO increase in 4 test subjects is less than what is usually required to call a subject ‘volume responsive’ i.e. less than 15% (24). These 4 have a supine PPV of less than 9%. One exception, though: subject P29 with low PPV (8.6%) reacted with a CO‐increase of 46%. Subject P25 with a PPV of 11.8% scored 29% increase. Remarkably, PPV –on average‐ only diminished from 8.5% pre‐infusion to 7.7% post. Pulse pressure, on average, did not change.Table 1. Results of infusion experiments; test subjects & model simulation
control period end of infusion
Subject PPV CO MAP SVR PPV CO MAP SVR
(%) (l/min) (mmHg) (m.u.) (%) (l/min) (mmHg) (m.u.)
P24 8.3 7.72 100.6 0.78 7.7 8.43 107.0 0.76 P25 11.8 6.82 68.2 0.60 11.0 8.82 70.7 0.48 P26 6.7 7.99 72.7 0.55 8.4 8.96 68.2 0.46 P27 6.9 7.45 80.5 0.65 5.0 7.81 80.9 0.62 P28 8.6 7.50 74.4 0.60 7.3 8.43 73.7 0.52 P29 8.6 4.65 73.4 0.95 6.9 6.79 76.1 0.67 Average 8.5 7.02 78.3 0.69 7.7 8.20 79.4 0.59 Model Awake 11.1 6.80 101.3 0.89 6.3 7.37 91.1 0.74 Sedated 16.5 6.75 86.3 0.77 8.6 7.95 87.3 0.66
Values after infusion as % of control:
CO MAP SVR P24 109 106 97 P25 129 104 80 P26 112 94 84 P27 105 100 96 P28 112 99 88 P29 146 104 71 Average 119 101 86 Model Awake 108 90 83 Sedated 118 101 86 MAP = mean arterial pressure, measured from Finapres finger pressure. m.u. = medical units, i.e. mmHg/(ml/sec).
101
Figure 5 CO after infusion as % of control value, as function of the pulse pressure variability in the control period. Squares denote test subjects in the study, round symbols results from simulation experiments as described in the text. The infusions did not raise blood pressure much: between + and ‐6% as shown in Table 1. The remainder of the CO‐increase was, consequently, accounted for by a fall in systemic vascular resistance (SVR). This was particularly large in subject P29 with the 46% increase in CO: he had a 39% drop in SVR (and only 4% rise in mean BP). This test subject has low resting CO and high SVR compared to the other subjects anyway, as shown in the table. This pattern was the same in the other saline infusion tests, in the earlier one in the control phase, as well as the HDT‐ and the recovery phase.Modeling experiments First simulation series: model in the ‘awake’ condition 1. volume infusion: 500ml (Figure 6) The model had been fitted, first, to the supine resting data of a healthy, male subject. After initial stabilization a volume expansion was started at t=300 s, completing the set 500 ml in about 50 s. The model shows a drop in HR and MAP but an increase in pulse pressure, PPV is almost halved. The model data have been entered in Table 1 and Figure 5 to be compared with the healthy subjects. It shows that the model response is slightly different from the subjects: it has higher resting PPV at a low increase in CO due to infusion.
103
Figure 6 Simulation of volume expansion in a healthy subject. At t = 300 s an infusion is started which ends after ~ 50 s at a volume of 500ml as indicated in panel B. From above down: Beat to beat aortic blood pressure (A), heart rate (B), pulse pressure (C) and pulse pressure variability (D).
2. volume loss of 500 ml, followed by infusion of 500ml (Figure 7)
At t=200 s the model loses 500 ml in 20 s. At t=300 s this volume is repleted as in the previous case. The figure shows that PPV roughly doubles at a decreased pulse pressure after volume loss. After re‐infusion the baseline is restored.
105
Figure 7 Simulation of the same subject as in Figure 6. At t = 200 s a volume of 500ml is lost within 20 s, at t = 300 s this is repleted as indicated in panel B. From above down: Beat to beat aortic blood
Second simulation series: model in the ‘anesthetized’ condition 1. volume infusion: 500ml, subject wakes up (Figure 8) For the simulation of autonomic effects by anesthesia we assumed vagal and sympathetic activity to drop to ¼ of the normal, awake condition. In Figure 8 anesthesia is started at t=100 s, (models may have a new condition within seconds or less): BP drops and PPV is immediately increased compared to baseline. At t=300 s an infusion is started, totalling a volume of 500 ml in about 50 s. Awakening is at t=500 s. When the autonomic outflow is altered by the anesthesia, the model behaves differently: now the hypervolemia does no longer lead to a decrease in MAP, systolic pressure is not defended by the vagal baroreflex as it was in the awake state. Obviously, after anesthesia the final condition is equal to that of Figure 6. We have entered the values for the anesthetized model in table 1 and Figure 5 as well. Comparison with the test subject values now shows a more ‘normal’ reaction to volume infusion: MAP does not decrease, systolic pressure is not as heavily defended by the vagal baroreflex as it was before, in the ‘awake’ model. Still, the values for the model are shifted to higher values of PPV than those for the test subjects.
107
Figure 8 Simulation of the same subject as in Figure 6. At t = 100 s anesthesia is instituted by diminishing autonomic outflow to 0.25 of the normal value. At t = 300 s an infusion of 500 ml is given as in Figure 6. Anesthesia is stopped at t = 500 s. The interventions are indicated in panel B. From above down: Beat to beat aortic blood pressure (A), heart rate (B), pulse pressure (C) and pulse pressure variability (D).2. volume loss of 500 ml, infusion: 500ml, subject wakes up (Figure 9) At t=200 s the volume loss starts, stopping after 10 s, when 500 ml has been lost. After an initial jump PPV follows the slow compensation that is set in by the (crippled) sympathetic nervous system, however, the earlier level is never reached. After the 500 ml (whole blood) volume expansion that started at t=300 s and completes within ~ 50 s, BP, HR and PPV get back to their levels before volume loss (the first 100‐200 s). At t=500 s the subject is back to consciousness; first BP shows an overshoot, then pressure settles back to baseline (0‐100 s), all values return to normal. In this simulation PPV seems a good reflection of cardiac preload, or volume status.
109
Figure 9 Simulation of same subject as in Figure 6. Anesthesia at t = 100 s as in Figure 8; at t = 200 s a volume of 500 ml is lost and repleted at t = 300 s as in Figure 7. The interventions are indicated in panel B. From above down: Beat to beat aortic blood pressure (A), heart rate (B), pulse pressure (C) and pulse pressure variability (D).3. volume loss of 500ml; infusion: 1500 ml, subject wakes up (Figure10) In this simulation, obviously, the circulation became overfilled. When we followed PPV as probe for volume status, infusion could have been stopped at the moment that the pre‐bleeding value of PPV had been reached. Stressing the outcome shown in Figure 10, where PPV reaches the earlier value at the time when the volume loss had been compensated by the infusion (at the arrow in panel D of Fig.10). When the infusion continues beyond that level, PPV drops further, BP starts to rise above pre‐bleeding values and right atrial pressures and pulmonary pressures rise as well, sometimes to unwanted high pressures. After return from anesthesia in an overfilled state, of course, BP overshoots to higher than normal values. This pattern is dramatically shown here, where the infusion only was stopped when a volume of 1500 ml had been reached. Since the infusions in the model were by ‘whole blood’ rather than by saline, the simulation had a tendency to run out of normal boundaries at the high infusion values.
111
Figure 10 Simulation of same subject as in Figure 6. Anesthesia at t = 100 s as in Figure 8; at t = 200 s a volume of 500 ml is lost and at t = 300 s an infusion is started, stopping at 1500 ml after 150 s. Anesthesia is stopped at t = 500 s. The interventions are indicated in panel B. The arrow in panel D points to the moment where the pre‐bleeding PPV is reached again. From above down: Beat to beat aortic blood pressure (A), heart rate (B), pulse pressure (C) and pulse pressure variability (D).Discussion
“Don’t fuss, just fill” or paraphrases of this corollary have been taught to generations of ICU clinicians. However, the side effects of filling more than necessary became also well‐known by recent studies inspired by Evidence Based Medicine: slower wound healing, danger of pulmonary edema, longer duration of hospital admission (4, 23, 29, 34). These newer insights made a strategy for optimal volume control necessary. In that vein, measurement of pulse pressure variation has been cheered as the new wonder of ingenuity and has been the source of disappointment (5). The present study was undertaken to test if PPV predicted volume response in awake subjects receiving a large amount of intravenous saline within a short period. Furthermore, our computer model of the circulation was put into action to mimic the use of PPV in sedated, ventilated patients, where its use in automated infusion has been proposed (31). The volumes of the respective infusions (test subjects/computer model) may not be directly compared: saline has a much larger distribution volume than whole blood: in the HDT’88 study haematocrit by the 1.6 l infusion was lowered by about 10% at the completion of infusion, to return back to normal values in about 2 hours post‐infusion (33). ‘Infusion’ of whole blood into the model of 1500 ml brought it to its boundaries of normal operation. Optimal cardiac performance has often been judged by the cardiac output value, since that should deliver the required oxygen at an appropriate pressure to the periphery. From the Frank‐Starling curve we know that increased diastolic filling only works up to a maximum, beyond which no more gain can be expected. Logically speaking that maximum is reached when respiratory variations in filling no longer lead to changes in stroke volume and thus to changes in pulse pressure. However, in the awake subject the relation between total circulating volume and cardiac performance is subject to many influences of which filling of the ventricle is only one. The autonomic nervous system influences cardiac inotropy, thereby ejection fraction and speed of ejection, to name but a few of the factors that co‐determine pulse pressure. The fact that about 50% of variability in PPV in the test subjects is found in VLF (cf. Figure4B) underpins this and points to an important involvement of the sympathetic nervous system and possibly slower, hormonal regulations. Earlier, Virtanen and co‐workers (41) have studied beat‐to‐beat oscillations in pulse pressure and concluded that several mechanisms contribute to its LF variability, among which serum insulin levels. Even though the subjects in this study received roughly the triple amount of a normal fluid113
could be expected if we look at their pre‐infusion PPV (Table 1, Figure 5); 4 out of 5 had PPV’s below 9% and reacted with less than 15% increase in CO. However, the numbers that have been established for patients under anesthesia have not been validated in awake subjects. Apart from the one subject who had 46% increase in CO, the others showed an almost linear relationship between PPV and CO‐increase (Figure 5). A reason why the values for PPV that we calculated here might deviate from those in the literature is the fact that we used ‘raw’ finger blood pressure values from an early research version of the Finapres (TNO‐BMI’s model 5, a forerunner of later space‐qualified versions and Ohmeda’s 2300 blood pressure monitor). In the Finapres® technique continuous blood pressure is measured in the finger arteries. Since those are very small, peripheral arteries, one may expect sharp, peaking systolic pressures, compared to more central pressure recordings. In young subjects one may observe a finger pulse amplification by a factor of 1.5 compared to more central pressures (upper arm, aorta) (42). Elaborations of the technique of continuously measured non‐invasive finger pressure make use of later research on pulse wave transmission along the lower arm and hand (12) applying that knowledge to estimate upper arm pressure, rather than giving the difficult‐to‐interpret finger pressure (this algorithm is used in the newer Nexfin® device, presently marketed by Edwards Lifesciences Corp.). A higher systolic and lower –peripheral‐ diastolic value leads to higher pulse pressures. Since PPV is measured as percentage of pulse pressure, one may well suppose that the numbers from the present study are on the low side. If one would apply this knowledge to the data points in Figure 5, the differences between measurements in the test subjects and those in the model would become less or might even disappear. As shown in Figure 4 the contribution of very low‐ and low‐frequency variability in PPV is considerable. This necessitates a sufficiently long stable period of blood pressure to find a reliable value of PPV in awake subjects. It is insufficient to just measure a few respiratory cycles and calculate the average, as it might be done in sedated or anesthetized patients (27, 30). Figs. 3B and 4 underline this dependency of PPV on oscillations slower than respiration. This finding has consequences for the treatment of patients in the pre‐operative stage: judging the – later to be expected – volume responsiveness before induction of anesthesia should not rely on a short‐lasting measurement of supine PPV. When in doubt, an actual test‐fluid challenge should be given, as advocated by (40).Our modeling also shows that PPV under conditions of low autonomic activity might be useful in an automated infusion control system (31). Since the model is very ‘Guytonian’ in nature this makes sense: Here PPV is mainly the consequence of variations in cardiac filling.
Conclusions
We have shown in this study that PPV is a complex measure when used in awake subjects. It shows large moment‐to‐moment variability due to low‐frequency oscillations in the cardiovascular and respiratory control systems. Its capacity to predict volume responsiveness, or use in automated systems for fluid loading, is restricted to stable, heavily sedated or anesthetized and ventilated patients. In those cases the circulation reduces to the simple textbook situation, where cardiac output and its variations are determined by only the basic factors: cardiac filling, cardiac function and systemic vascular resistance.115
Appendix: how to compute PPV
The optimal way to compute respiratory‐related PPV is, of course, to record the variations from begin‐ to end‐expiration. This requires either a respiration signal to synchronize the detection or intricate algorithms to trace variations in the pulse pressure signal that might be attributed to respiration. To circumvent those problems often an approximation formula is used as follows: (PPmax – PPmin)/[(PPmax + PPmin)/2] or, alternatively: (PPmax – PPmin)/(PPavg) Where PPmax = maximal pulse pressure in the sampling period PPmin = minimal pulse pressure in the sampling period PPavg = average pulse pressure over the sampling period All measured either over one or more defined respiratory periods or over a fixed number of heart beats or period of time. In view of the practical applicability we chose to use the second formula over a fixed number of heart beats. To find the optimal number of beats, we ran a simulation of a simplified beat‐to‐beat circulation model of cardiac control (8, 43) where we added various respiratory frequencies to the pulse pressure. Due to the presence of sympathetic baroreflex driven feedback the model would show low frequency variability as well as the respiratory frequency. The result of these simulations is shown in Figure 12. It shows the normalized, computed PPV as percentage of what it should have been, as a function of sampling period and for various respiration intervals, from 3 seconds up to 15 seconds (20/min to 4/min). As to be expected, the shorter the sampling interval, the lower the amount of variability that is caught by the algorithm: it does not span the whole respiratory cycle. However, when the sampling period gets longer the algorithm will catch more of the low frequency variability and thereby overestimate the actual PPV. From Figure 11 we concluded that the optimal sampling period for the prevailing respiratory rates would be around 10 seconds, or 10 heart beats if HR is not too high. Longer periods result in overestimation, shorter in underestimation of actual PPV.Figure 11. Theoretical result of PPV‐measurement as function of sampling period and respiratory period. Ordinate gives the % result of what should have been measured, given the sampling period (abscissa). Curves have been computed for respiratory periods of 3, 4, 5, 6, 7, 8, 9, 10, 12 and 15 seconds, as indicated by the arrow (3s → 15s). The optimal value is found when sampling period and respiratory period are equal.
117
References
1. Babbs CF. Effects of an impedance threshold valve upon hemodynamics in Standard CPR: studies in a refined computational model. Resuscitation 66: 335‐345, 2005. 2. Head‐down tilt bedrest. HDT’88 – an international collaborative effort in integrated systems physiology. Acta Physiol Scand Suppl 604: 144, 1992. Guest editor: B. Saltin. Publ. Committee: Baisch F, Beck L, Blomqvist GC, and Karemaker JM. 3. Bendjelid K and Romand JA. Fluid responsiveness in mechanically ventilated patients: a review of indices used in intensive care. Intensive Care Med 29: 352‐360, 2003. 4. Brandstrup B, Tonnesen H, Beier‐Holgersen R, Hjortso E, Ording H, Lindorff‐Larsen K, Rasmussen MS, Lanng C, Wallin L, Iversen LH, Gramkow CS, Okholm M, Blemmer T, Svendsen PE, Rottensten HH, Thage B, Riis J, Jeppesen IS, Teilum D, Christensen AM, Graungaard B, Pott F, and Danish Study Group on Perioperative Fluid T. Effects of intravenous fluid restriction on postoperative complications: comparison of two perioperative fluid regimens: a randomized assessor‐blinded multicenter trial. Ann Surg 238: 641‐648, 2003. 5. Camporota L, Terblanche M, and Bennett D. Year in review 2007: Critical Care‐cardiology. Crit care 12: 232, 2008. 6. Cowley AW, Jr., Liard JF, and Guyton AC. Role of baroreceptor reflex in daily control of arterial blood pressure and other variables in dogs. Circ Res 32: 564‐576, 1973. 7. DeBoer RW, Karemaker JM, and Strackee J. Comparing spectra of a series of point events particularly for heart rate variability data. IEEE Trans Biomed Eng 31: 384‐387, 1984. 8. DeBoer RW, Karemaker JM, and Strackee J. Hemodynamic fluctuations and baroreflex sensitivity in humans: a beat‐to‐beat model. Am J Physiol 253: H680‐689, 1987. 9. Diebel LN, Wilson RF, Tagett MG, and Kline RA. End‐diastolic volume. A better indicator of preload in the critically ill. Arch Surg 127: 817‐821; discussion 821‐812, 1992. 10. Task Force of the European Society of Cardiology, and the North American Society of Pacing and Electrophysiology. Heart rate variability: standards of measurement, physiological interpretation and clinical use. Circulation 93: 1043‐1065, 1996. 11. Gaffney F, Buckey J, Lane L, Hillebrecht A, Schulz H, Meyer M, Baisch F, Beck L, Heer M, and Maass H. The effects of a 10‐day period of head‐down tilt on the cardiovascular responses to intravenous saline loading. Acta Physiol Scand Suppl 604: 121‐130, 1992. 12. Gizdulich P, Imholz BP, van den Meiracker AH, Parati G, and Wesseling KH. Finapres tracking of systolic pressure and baroreflex sensitivity improved by waveform filtering. J Hypertens 14: 243‐250, 1996. 13. Grocott MP, Mythen MG, and Gan TJ. Perioperative fluid management and clinical outcomes in adults. Anesth Analg 100: 1093‐1106, 2005. 14. Guyton AC. Determination of cardiac output by equating venous return curves with cardiac response curves. Physiol Rev 35: 123‐129, 1955. 15. Guyton AC. Long‐term arterial pressure control: an analysis from animal experiments and computer and graphic models. Am J Physiol 259: R865‐877, 1990. 16. Guyton AC. Textbook of medical physiology. Philadelphia,: Saunders, 1956.17. Guyton AC, Coleman TG, Cowley AV, Jr., Scheel KW, Manning RD, Jr., and Norman RA, Jr. Arterial pressure regulation. Overriding dominance of the kidneys in long‐term regulation and in hypertension. Am J Med 52: 584‐594, 1972. 18. Guyton AC, Coleman TG, and Granger HJ. Circulation: overall regulation. Annu Rev Physiol 34: 13‐46, 1972. 19. Guyton AC, Lindsey AW, and Kaufmann BN. Effect of mean circulatory filling pressure and other peripheral circulatory factors on cardiac output. Am J Physiol 180: 463‐468, 1955. 20. Guyton AC, Polizo D, and Armstrong GG. Mean circulatory filling pressure measured immediately after cessation of heart pumping. Am J Physiol 179: 261‐267, 1954. 21. Hillebrecht A, Schulz H, Meyer M, Baisch F, Beck L, and Blomqvist CG. Pulmonary responses to lower body negative pressure and fluid loading during head‐down tilt bedrest. Acta Physiol Scand Suppl 604: 35‐42, 1992. 22. Keyl C, Stockinger J, Laule S, Staier K, Schiebeling‐Romer J, and Wiesenack C. Changes in pulse pressure variability during cardiac resynchronization therapy in mechanically ventilated patients. Crit care 11: R46, 2007. 23. Lobo DN, Bostock KA, Neal KR, Perkins AC, Rowlands BJ, and Allison SP. Effect of salt and water balance on recovery of gastrointestinal function after elective colonic resection: a randomised controlled trial. Lancet 359: 1812‐1818, 2002. 24. Marik PE, Cavallazzi R, Vasu T, and Hirani A. Dynamic changes in arterial waveform derived variables and fluid responsiveness in mechanically ventilated patients: a systematic review of the literature. Crit care med 37: 2642‐2647, 2009. 25. Michard F. Changes in arterial pressure during mechanical ventilation. Anesthesiology 103: 419‐428, 2005. 26. Michard F. Volume management using dynamic parameters: the good, the bad, and the ugly. Chest 128: 1902‐1903, 2005. 27. Michard F, Boussat S, Chemla D, Anguel N, Mercat A, Lecarpentier Y, Richard C, Pinsky MR, and Teboul JL. Relation between respiratory changes in arterial pulse pressure and fluid responsiveness in septic patients with acute circulatory failure. Am J Resp Crit Care Med 162: 134‐138, 2000. 28. Mythen MG and Webb AR. Perioperative plasma volume expansion reduces the incidence of gut mucosal hypoperfusion during cardiac surgery. Arch Surg 130: 423‐429, 1995. 29. Nisanevich V, Felsenstein I, Almogy G, Weissman C, Einav S, and Matot I. Effect of intraoperative fluid management on outcome after intraabdominal surgery. Anesthesiology 103: 25‐32, 2005. 30. Preisman S, Kogan S, Berkenstadt H, and Perel A. Predicting fluid responsiveness in patients undergoing cardiac surgery: functional haemodynamic parameters including the Respiratory Systolic Variation Test and static preload indicators. Br J Anaesth 95: 746‐755, 2005. 31. Rinehart J, Alexander B, Le Manach Y, Hofer C, Tavernier B, Kain ZN, and Cannesson M. Evaluation of a novel closed‐loop fluid‐administration system based on dynamic
