Portal de Periódicos da CAPES

Automated Assessment of Fluid Responsiveness in Mechanically Ventilated Patients

2008; Lippincott Williams & Wilkins; Volume: 106; Issue: 4 Linguagem: Inglês

10.1213/ane.0b013e318167abe5

1526-7598

Azriel Perel,

Blood Pressure and Hypertension Studies

Fluid administration to anesthetized and critically ill patients is one of the most commonly used therapeutic measures. The immediate goal of fluid administration is to augment cardiac preload and, in turn, increase the cardiac output (CO). Yet, in the many studies that have examined the immediate hemodynamic effects of fluid loading, CO was repeatedly shown to increase in only about half of the patients, even though in most studies fluid loading was judged to be “clinically indicated.” The sobering reality that emerges from these persistent findings raises doubts about our ability to accurately assess fluid responsiveness at the bedside. The uncertainty that characterizes current decision-making about fluid administration is because we lack robust physiological variables that will successfully predict the response to fluid loading. Clinical examination, arterial blood pressure, heart rate, and even central venous and pulmonary artery occlusion pressure, have repeatedly been shown to be poor predictors of fluid responsiveness, and to be unable to differentiate between patients who respond to intravascular volume loading (responders) and patients who do not (nonresponders). Even more accurate measures of preload, such as the global end-diastolic volume, the left ventricular (LV) end-diastolic area, and the right ventricular end-diastolic volume, are mediocre predictors of fluid responsiveness, since the relationship of any static “preload” variable to the response of the CO to fluid loading depends on the elusive slope of the LV function curve. A dynamic approach to assess fluid responsiveness is offered by measuring the effect of the decrease in venous return on the CO during a mechanical positive-pressure breath. During the last 20 yr, an increasing number of publications have described the usefulness of variables such as the systolic pressure variation (SPV), pulse pressure variation (PPV, ΔPP), stroke volume variation (SVV), plethysmographic waveform variation, and a variety of other variables, which reflect the hemodynamic changes that occur during mechanical ventilation. For many of these variables, this “respiratory variation” has been shown to be a better predictor of fluid responsiveness than commonly measured static preload variables. However, the penetration of functional hemodynamic variables into mainstream clinical practice has been exceedingly slow.1 This may be because, until recently, assessing these variables required manual measurements of the pressure or plethysmogram changes. Some specialty monitors for advanced hemodynamic monitoring have made these variables available. These include the PiCCOplus (Pulsion Medical Systems, Germany) and the LiDCOplus (LiDCO Ltd., UK), which continuously display PPV and SVV, and the FloTrac CO monitor (Edwards, USA), which displays the SVV. Three articles in this issue of the journal present new methods for the automated estimation of respiratory-induced variations in the arterial pressure and in the plethysmographic waveform, which are implemented in more “conventional” monitors.2–4 These articles confirm that the interest in functional hemodynamic variables continues to grow, but they also highlight some of the problems associated with their automated measurement. Two of the articles describe novel methods for the automatic continuous measurement of the PPV.2,3 Auler et al.'s method uses a mainstream capnographic signal to identify individual breaths, so that the PPV value for each breath is calculated,2 whereas Cannesson et al. use a complex algorithm, which calculates a mean PPV value.3 In the third article, Cannesson et al. describe the plethysmographic variability index (PVI), which is a measure of the respiratory variations in the pulse oximeter plethysmogram waveform amplitude.4 From the brief descriptions of these methods, it seems that rather than “automate” the measurement of manually measured variables, they create new variables that are not equivalent to those already described. Auler et al.'s PPV2 is claimed to be different than the PPV available in other monitors, which has been repeatedly studied,5 in that it displays the PPV induced by a single mechanical breath rather than a mean over an arbitrary period. Cannesson et al.3 indicate that the automated PPV algorithm tested, which is calculated and averaged over four cycles of 8 s, is not expected to be exactly the same as the manually measured PPV. Indeed, the agreement between the manual and the automatic PPV was found to be weak, and the threshold values for predicting fluid responsiveness were different (10% and 12% for the automatic and manual PPV, respectively).3 In the third article,4 the PVI is calculated using a different formula than the ones used previously by the same authors and by others6–9 to calculate the respiratory variations in the plethysmographic signal. Generalizing the results of these studies is thus compounded by the fact that measurements obtained by one specific device or method cannot be applied to similar variables from other manufacturers. This is especially true regarding the automatic measurement of the variations in the plethysmographic signal. In a recent editorial in this journal,10 Feldman has already noted that different pulse oximeters use different nonstandardized proprietary plethysmographic signal-processing algorithms, requiring individual validation for each variable in each monitor. Other factors that have to be considered when interpreting the plethysmographic waveform variations are the uncertain role of the venous element of the signal, its dependency on the site of measurement, and obliteration of the plethysmographic signal during severe vasoconstriction (in 7% of the patients according to Cannesson et al.4). Regarding the PVI itself, the authors themselves claim that it is unable to “distinguish between changes induced by mechanical ventilation from changes induced by any other phenomenon” and requires “steady conditions.”4 The two articles that describe new methods for calculating the PPV claim that, hitherto, its measurement required the simultaneous recording of the airway pressure.2,3 This is not accurate since, as stated before, the PPV is continuously measured by other monitors by analyzing the arterial pressure waveform alone. In addition, using mainstream capnography for the calculation of the PPV, as described in Auler et al.'s article,2 is a limitation to the use of their method, as side-stream capnography is much more prevalent in clinical practice. Furthermore, although the PPV is considered to be superior to other functional hemodynamic variables, it is only marginally better than the SPV11,12 while being much more difficult to follow from the arterial pressure waveform on the monitor's screen.13 Real-time assessment of PPV is therefore difficult, if not impossible, in the absence of automation. Like all variables that are based on the difference between the maximal and minimal beat during the respiratory cycle, the PPV includes the early inspiratory augmentation of the LV stroke volume, the ΔUp,14 which is not related to fluid responsiveness. In addition, the ratio of the stroke volume to the pulse pressure decreases during hypovolemia because of decreased aortic compliance, causing the PPV to change considerably more than the SPV and the SVV during hypovolemia.15 Finally, having the mean of the maximal and minimal beat in the denominator of the PPV makes less physiological sense than having the mean pulse pressure or the pulse pressure during apnea in its stead.15 The interpretation of the results of these and other studies of fluid responsiveness must consider some common physiological and methodological problems. All functional hemodynamic variables are influenced by the magnitude of the tidal volume (even when it varies from 8 to 10 mL/kg), by the positive end-expiratory pressure level, by the compliance of the chest wall (especially when results of intact and open-chest conditions are pooled3), and by the presence of increased intraabdominal pressure. Similarly, other important factors include the use of different fluid loads (20 mL/kg2 versus 500 mL3), considering fluid responsiveness to be an “all-or-none” phenomenon (by using a fixed change (15%) in CO to separate responders and nonresponders), and the use of a PPV threshold of 13% as a “gold standard” (even though it was found to range between 10% and 17% in previous studies). Last but not least, excluding patients with impaired cardiac function from validation studies of functional hemodynamic variables2,3 limits the ultimate clinical benefit of these studies, since anesthetized patients with impaired LV function may often be fluid-responsive12,16 and since careful fluid administration is especially important in these patients. All the above considerations and reservations do not mean that the goal of automating the respiratory-induced pressure or plethysmographic variations should not be pursued. Within their respective limitations, the evidence in the literature demonstrates that functional hemodynamic variables offer immediate, dynamic, and essential information about cardiovascular function. Discussing the relative merits of the various variables and methods of measurement is important, but is secondary to the fact that any and all of these variables perform better than static preload variables in predicting fluid responsiveness. Until the automatic measurement of these variables becomes widely available, the clinician should observe the various analog waveforms and their respiratory variations, and consider the fact that a variation of more than 10% in the arterial pressure or plethysmographic waveform is suggestive of fluid responsiveness. Obviously, the key to a more frequent, convenient, and accurate use of these variables, especially during anesthesia where mechanical ventilation is fully controlled in most cases, lies with their successful automation. The articles of Auler and Cannesson and their colleagues2–4 are therefore welcome as they herald the promise of this approach to managing fluid balance. Automation of these variables will almost certainly achieve a higher level of accuracy and reliability once monitors and ventilators are linked. Such linkage will facilitate automating the stimulus (mechanical breath)-response (change in LV surrogate variables) process. It may also allow the future development of standardized respiratory maneuvers for measuring functional hemodynamic variables.12 Such maneuvers might also include the automatic introduction of a short apnea period, which would allow the clinician to distinguish between the early systolic augmentation, ΔUp, and the true decline in systolic pressure related to reduced stroke volume, ΔDown.14,17 This latter variable is theoretically better than the SPV, PPV, and SVV in predicting fluid responsiveness, as it is solely dependent on the degree of the inspiratory reduction in venous return and is devoid of the confounding effect of the ΔUp.14,17,18 In summary, the new methods that are presented in this issue of the journal make the wealth of information contained in variables that respond to positive pressure variation accessible to the clinician at the bedside. At the present time, these methods lack standardization and thus may be difficult to compare with one another. However, even with their limitations, they will undoubtedly be very useful to the clinician who is using any one of them on a daily basis, independent of the fluid replacement strategy being used.