Perioperative assessment of volume status: measurement of preload is not measurement of preload responsiveness

Perioperative haemodynamic care of patients aims at maintaining an adequate oxygen supply to the tissues. In the presence of sufficient oxygen carrying capacity in the blood, tissue oxygen delivery will mainly depend on cardiac output. The determinants of cardiac output are heart rate, the ventricular loading conditions (preload and afterload) and the myocardial contractility. Assessment of these different determinants of cardiac function in the clinical setting remains challenging, and continuing efforts are directed towards the design of a monitoring tool that allows for correct estimation of these different variables. Preload optimization is regarded as a prerequisite for adequate cardiac performance. Physiologically, preload is defined as the end-diastolic myocardial fibre length or fibre tension. In the intact heart, preload can be defined as the ventricular wall stress at end-diastole. This ventricular wall stress is determined by ventricular end-diastolic volume, end-diastolic pressure and wall thickness, and their interaction is expressed by the Laplace equation. End-diastolic wall stress, however, cannot be measured in the clinical setting and, therefore, different surrogate markers have been proposed to assess preload. Although, physiologically, ventricular volumes appear to most closely approximate muscle fibre length, historically, the pressure correlates [left ventricular (LV) end-diastolic pressure, left atrial pressure, pulmonary capillary wedge pressure (PCWP), right atrial pressure and central venous pressure (CVP)] were first used to determine cardiac preload. Their accuracy to reflect cardiac preload, however, is critically dependent on the compliance of the ventricle, the integrity of the mitral valve, the presence of normal pulmonary conditions, the integrity of the pulmonary and tricuspid valves and the presence of normal right heart function. Because of all these limitations, pressure surrogates only poorly reflect LV preload. With the advent of transoesophageal echocardiography, a clinically applicable continuous assessment of LV dimensions became available; however, the final calculation of LV end-diastolic volumes is based on a number of mathematical assumptions (e.g. the Simpson algorithm) necessitating measurements of LV end-diastolic areas at different levels. In addition, such determinations are time consuming, and the quality of data critically depends on the operators' skills. Alternatively, systems have been designed that provide an estimation of left and right ventricular end-diastolic volumes using the thermodilution technique. It seems that these latter volumetric indices correlated well with echocardiographic measurements and better reflected preload compared with pressure correlates such as PCWP and CVP [1–4]. In the current issue of the European Journal of Anaesthesiology, Della Rocca et al.[5] compared volumetric preload indices measured by pulmonary artery thermodilution [continuous right ventricular end-diastolic volume index (cRVEDVI)], transoesophageal echocardiography [left ventricular end-diastolic area index (LVEDAI)] and by measurement of the PCWP and the CVP. This study was performed in liver transplant patients. In order to minimize the role of nonhomogeneity, patients with heart disease, other than the expected consequences of the liver disease, were excluded. The authors observed a good correlation between stroke volume index (SVI) and cRVEDVI and between SVI and LVEDAI, but less so between SVI and, respectively, PCWP and CVP. These observations led them to conclude that cRVEDVI and LVEDAI provide a better reflection of preload than the filling pressures. Although the conclusions on the filling pressures are in line with what is already known and to be expected based on the physiological considerations outlined above, the conclusions based on the volumetric data are open for discussion. When comparing different methods to assess a variable, it is expected that these methods are compared with a generally accepted ‘gold standard’ used in the assessment of that particular variable. In the absence of such a gold standard, comparisons of methods may suffer from methodological limitations. Della Rocca et al.[5] analysed the volumetric data cRVEDVI, LVEDAI and continuous right ventricular ejection fraction with respect to SVI. SVI was obtained in two ways: conventional thermodilution and continuous cardiac output measurement. The measured SVI was then used in conjunction with the independently measured continuous right ventricular ejection fraction to calculate the cRVEDVI. As end-diastolic volume is stroke volume divided by ejection fraction, it follows that cRVEDVI must correlate with SVI, and, similarly, autocorrelations will exist between ejection fraction and SVI and between cRVEDVI and SVI. In other words, when variables are derived from each other, it automatically follows that they are highly correlated due to their mathematical coupling. Some authors have attempted to minimize the effects of such mathematical coupling by measuring the different variables by independent techniques such as for instance the Fick and thermodilution methods [6,7]; however, one can expect both techniques to be highly correlated, as the principle underlying all indicator techniques is the Fick principle. Della Rocca et al.[5] also assessed LV filling by transesophageal echocardiogram (TEE)-derived measurements of the LVEDAI and found a good correlation between SVI and LVEDAI. Although both variables are indeed obtained with different methodologies, they are still physiologically coupled (Frank–Starling relationship: stroke volume depends on end-diastolic volume). In other words, variables that are either mathematically or physiologically interdependent are coupled and will by definition be highly correlated. Therefore, comments and conclusions based on observations of correlations between such variables may be extremely hazardous. Historically, assessment of variables reflecting ventricular preload have been used to guide fluid therapy; however, in recent years, it has increasingly been recognized that preload does not necessarily reflect ‘preload responsiveness’ [8,9]. Indeed, it seems that about half of the ICU patients are unable to respond to fluid loading by a substantial increase in cardiac output [10]. This phenomenon is explained by the shape of the Frank–Starling relationship that consists of two distinct parts: on the initial steep limb, stroke volume is highly dependent on end-diastolic volume, whereas on the terminal flat portion, stroke volume will hardly change with an increase in end-diastolic volume. Consequently, if the heart is working on the initial part of the curve, volume administration will result in a significant increase in stroke volume. When the heart is working on the flat portion of the curve, however, a similar volume administration will hardly increase stroke volume. In other words, a given value of ‘cardiac preload’ can be associated with the presence of preload reserve in cases of normal cardiac function, but with the absence of preload reserve in cases of impaired cardiac function. Actually, from a clinical point of view, we are mainly interested in identifying those patients who might benefit from an additional volume loading; hence, we need to know on which portion of the Frank–Starling curve the patient's heart is working. This information cannot be simply derived from a single measurement of a LV end-diastolic volume or pressure-derived variable. Indeed, all of the classical preload measures, including the CVP, the PCWP, the LV end-diastolic dimensions, the early to late diastolic wave ratio, and the LV ejection time duration fail to discriminate between responders and nonresponders to fluid therapy [10–18]. Instead, prediction of fluid responsiveness necessitates provocative interventions, that is, volume challenges to assess the ability of the heart to respond to an increase in circulating volume by an increase in stroke volume [10–23]. Such assessment is referred to as ‘functional haemodynamic monitoring’ [24]. The principle of such assessment is to induce a change in cardiac preload and to observe its effects on cardiac output or any of its surrogates. Initially, this was assessed by observing the effects on cardiac function of administration of a fixed amount of fluid; however, as this method constitutes an irreversible method of fluid administration, it potentially carries the risk of adverse effects secondary to overloading of the patient. Therefore, alternative methods have been developed in recent years that allow for such an assessment without the administration of additional fluids. Basically, these methods constitute reversible interventions such as passive leg raising or are based on the analysis of respiratory variations of haemodynamic signals such as stroke volume variation, pulse pressure variation, or caval vein diameter variations. These latter approaches have been shown to provide reliable information on the volume responsiveness of patients, provided that the variables (stroke volume or surrogate markers of stroke volume) are displayed and analysed in a beat-to-beat manner. With these considerations in mind, study methodologies should be adapted to the actual experimental hypothesis. If the experimental hypothesis relates to the assessment of volume responsiveness of a patient's heart, studies on static preload correlates are not appropriate. Instead, such studies should aim to determine which preload-derived variable is the best to perform functional haemodynamic monitoring and to assess volume responsiveness.