Stroke volume optimisation: is the fairy tale over?

Advanced haemodynamic monitoring is espoused as a means to achieve optimisation of cardiac stroke volume during surgery. It is believed that individuals are ‘normovolaemic’ when their stroke volume is at the plateau of the Frank-Starling curve in the supine position 1. This definition by Truijen et al. appears reasonable since when stroke volume is thus located in awake subjects, cardiac output and oxygen delivery are such that maximal venous oxygen saturation is established 2. Since the curve is flat here, an increase in preload will not increase stroke volume. According to the same model, patients are hypovolaemic when their stroke volume is on the steeper, ascending leftward part of the Frank-Starling curve. Proponents of goal-directed therapy believe it is possible to exploit this clinically with a fluid challenge: a minimal response suggests that stroke volume is on the plateau. If the stroke volume increases 10% or more within five minutes, then the patient is assumed be on the steep upward part of the curve and to have been ‘fluid responsive’ (as opposed to no increase which would place the patient on the flat, ‘normovolaemic’, portion of the curve). Stroke volume variability and related arterial wave-derived parameters are based on the same concept: the stroke volume response to intermittent positive pressure ventilation-induced changes in intrathoracic pressure during each respiratory cycle is inversely related to circulating volume. When the stroke volume is on the plateau of the Starling curve, arterial pressure swings are minimised; when hypovolaemia is present, variability increases 3. The terminology used to describe goal-directed therapy in the literature is inconsistent and confusing 4. For the purposes of this editorial, we use the term ‘stroke volume optimisation’ to indicate simple interventions with fluid and/or vasopressors to maintain stroke volume; ‘goal-directed therapy’ is a broader term that encompasses stroke volume optimisation as well as a wide variety of therapies aimed at increasing global oxygen delivery. The stroke volume optimisation paradigm is as follows. Metabolic demands of surgery create a growing oxygen demand that drives an increasing cardiac output requirement, and conventional haemodynamic signs fail to reveal the deficit 5. In the absence of adequate peri-operative fluid loading, occult hypovolaemia results: a circulating volume deficit of as little as 10% may lead to splanchnic vasoconstriction and reduced oxygen delivery 6. Accordingly, intestinal complications are amongst the most common after major surgery 7. Minimally invasive technology is therefore used continuously throughout surgery to monitor for any deficit in functional circulating volume. Most ‘flow based’ monitors derive stroke volume: some use the pulse power or contour of the arterial waveform, others use Doppler ultrasound to measure blood velocity in the descending thoracic aorta. Algorithms incorporating fluid challenges are used to maintain stroke volume throughout, thereby providing incremental benefit through the avoidance of postoperative oxygen debt. All the early fluid responsiveness studies were positive: protocol patients had substantially better outcomes than controls 5, 8-11, even though the primary intervention differentiating the groups appeared to be marginal: the infusion of about 500 ml colloid during surgery, with no postoperative intervention at all. Endorsement of this technology by the National Institute for Health and Care Excellence (NICE) on the basis of this initial evidence base may have been premature 12, 13. Contemporary surgical outcomes have been improved by the systematic delivery of peri-operative care known as ‘enhanced recovery’ 14. Prominent editorials and reviews 4, 13, 15, 16 have called for trials of the specific impact of stroke volume optimisation within such programmes. Several recent studies have now done so, all with neutral 17, 18 or detrimental results 19-22. A paper by Godfrey and colleagues published in this issue of Anaesthesia provides some translational insight into why this might be 23. Godfrey et al. adopted Truijen et al.'s definition of normovolaemia as the stroke volume when awake in the supine position, and set out to investigate fluid responsiveness in this state. In 17 healthy, hydrated volunteers, they employed a passive leg raise manoeuvre to increase preload, and used transthoracic echocardiography to measure stroke volume response at four time points within the following five minutes. Six participants were excluded from analysis due to inadequate Doppler measurements, leaving 11 subjects, five of whom (45%) had a peak stroke volume response in excess of 10%. This observation challenges the assertion that an increase of this magnitude in response to a fluid challenge always correctly classifies an individual as being hypovolaemic. Apparently normovolaemic individuals can do this too. It has been suggested before in this journal that the oesophageal Doppler monitor may represent the emperor's new clothes 24, but the primary problem may lie with our understanding of the relevant physiology 13. Where the tailoring of fluid therapy for the individual patient is concerned, it appears that one size most certainly does not fit all. Is the fairy tale over, then, for stroke volume optimisation? On the face of it, Godfrey et al.'s discovery may account for the unimpressive results from the most recent studies. Enhanced recovery programmes seek to bring patients to the operating theatre in a normovolaemic state. Measurement of tissue perfusion during contemporary bowel surgery suggests that intra-operative hypoxia is rare 25. Where deviations from normovolaemia are very small, then achievement of a gain requires precision. If the assumptions underpinning our current construct of stroke volume response are flawed, then the model is no good for making subtle adjustments. In 60 anaesthetised patients about to have elective (mastectomy, prostatectomy, abdominal) surgery, Bundgaard-Nielsen et al. 26 used oesophageal Doppler to measure the stroke volume response to a colloid fluid challenge. More than two thirds of subjects were fluid responsive. This work has been interpreted as suggesting that a large proportion of patients arrive in the anaesthetic room with a functional intravascular volume deficit. But perhaps this is incorrect. The study of Challand et al., conducted in elective colorectal patients, has been justifiably criticised because the attending anaesthetist administered an excessive amount of fluid (a mean of 17 ml.kg−1.h−1 during surgery), yet on top of this, an algorithm aimed at stroke volume maximisation allowed the investigator to give a further 1.3 litres of colloid on average to the intervention group 20. An excess of fluid given at the wrong time is detrimental. The volume expansion effects of a fluid load are context sensitive: when radiolabelled albumin is given to maintain circulating volume in bleeding patients, 90% remains intravascular, whereas in normovolaemic patients two thirds is lost to the interstitium 27. Godfrey et al. also emphasise the heterogeneity of clinical response to a fluid challenge: the mean increase in stroke volume after passive leg raise was 5.7% (95% CI −13 % to +24.5 %). Set against this backdrop, the current guidance that we give 200 ml fluid and wait five minutes for a 10% increase in stroke volume appears arbitrary and probably wrong. The association of the ‘newer’ respiratory arterial waveform variation parameters to diagnose fluid responsiveness has been rather more rigorously quantified 3. Where the pulse pressure variability signal is concerned, there is a ‘grey’ zone, in other words a range of values (9–13%) within which we cannot be certain whether an individual patient will be fluid responsive or not. This is seen in around 25% of patients receiving mechanical ventilation under general anaesthesia before major surgery 28. It appears also that the concordance between different monitors when used to measure stroke volume in anaesthetised adult patients is rather modest 29. However, there are several important caveats to consider when appraising Godfrey et al.'s paper 23. First, it does not refute that one might see an stroke volume improvement > 10% in hypovolaemic patients. It is likely that participants in earlier studies of goal-directed therapy in colorectal, cardiac and orthopaedic trauma surgery were relatively hypovolaemic at the time of induction of anaesthesia – so current evidence supports stroke volume optimisation in such patients. In acutely ill patients with the potentially disordered haemodynamics and ‘leaky capillaries’ of a systemic inflammatory response, cardiac output measurement to judge fluid resuscitation remains an attractive proposition. In individual patients who are at risk of major intra-operative haemorrhage, it is reassuring to have access to flow-based monitoring. However, evidence is scarce in these settings due to methodological difficulties. Second, most current stroke volume optimisation algorithms incorporate parameters such as central venous pressure or corrected flow time to show that the intravascular volume is replete, and to help the user avoid fluid excess. No such stopping thresholds are tested in Godfrey et al's experiment. Third, it is not clear how much can be inferred about haemodynamic physiology in anaesthetised patients receiving intermittent positive pressure ventilation from a study conducted on awake volunteers. The authors make the intriguing point that most patients having major surgery are allowed to wake at the conclusion of surgery, and that fluid therapy should be aimed at having them in the normovolaemic state when they do so. Finally, Godfrey et al's work does not actually tell us whether passive leg raise increases stroke volume in normovolaemic patients. The difference between mean baseline and mean peak stroke volume is not statistically significant so it remains possible that the apparent increase seen was due to the play of chance. The authors emphasise, though, that five of 11 subjects had an stroke volume response > 10%, and suggest that such a response to fluid challenge intra-operatively would prompt an unnecessary further bolus. However, preload was not the sole cardiovascular variable that changed. The statistically significant increase in heart rate and cardiac index seen, uncoupled from the stroke volume, suggests that the leg raise manoeuvre may not be a suitable experimental model for an isolated increase in preload. The act of lifting a conscious patient's legs is likely to trigger at least some contribution from adrenergic pathways, with potential changes in vasomotor tone, venous capacitance and cardiac contractility, none of which were directly measured but all of which may also have an effect on stroke volume. It is striking that previous physiology studies 30, 31 are at odds with Godfrey et al.'s observation, and instead show that under normovolaemic conditions, stroke volume does not increase in response to fluid. In truth, all we can reasonably conclude is that intra-operative haemodynamics are complex. The notion that we are manipulating a static Starling curve is almost certainly an oversimplification. Such pure manipulation of stroke volume is extremely difficult to achieve in vivo. Frank's original observations on the relationship between diastolic filling and the strength of ventricular contraction were made in an isolated, denervated frog heart 32. Individual patients in fact have not one but two ventricles, moving serially between a family of ventricular function curves (Fig. 1) – depending on cardiac contractility and afterload 33. If the passive leg raise manoeuvre transfers the awake subject on to a more ‘dynamic’ Starling curve then the stroke volume for a given preload will increase – but this is not ‘fluid responsiveness.’ Of course, the same applies during surgery. Painful stimuli, endogenous catecholamine levels, vasodilatory effects of neuraxial blockade and anaesthesia and tissue oxygen demand may vary considerably, such that it is difficult to be sure what the optimum stroke volume is at a particular moment. Evidence for goal-directed therapy shouldn't be extrapolated to answer questions about stroke volume optimisation. To illustrate this, consider that in the landmark trial by Shoemaker et al., the first to target supranormal oxygen delivery in surgical patients, only 101 of 2086 screened (4.8%) were considered sufficiently high-risk to be included in the study 37. Further, this work was based on measurement and targeting of supranormal values of oxygen delivery and cardiac index for around 48 hours after surgery, while minimally invasive devices derive stroke volume rather than directly measuring it. Even then, stroke volume is a surrogate – in the real world theatre environment, we hope that our haemodynamic interventions will optimise oxygen delivery but we make no direct measurement of tissue perfusion or oxygen utilisation. Algorithms guide fluid (including transfusion), vasopressor and possibly inotropic therapy, but these haemodynamic therapies all interact with one another in complex ways. The definitive Cochrane review on peri-operative goal-directed therapy 36 reports several secondary outcomes. From subgroup analyses it is possible to pick out six randomised controlled intra-operative stroke volume optimisation trials, comprising 573 patients. Stroke volume optimisation is clearly associated with a reduction in the number of patients with complications, but – as previously discussed – most of these studies were not conducted within enhanced recovery pathways. Is it time to tailor our approach to this differently? We and others 13, 20, 22, 38 have previously suggested that stroke volume optimisation provides no marginal benefit for aerobically fit patients having elective surgery within a contemporary enhanced recovery pathway. However, it is possible that the fidelity of the underlying algorithm has been wrong. The relationship between ‘fluid responsiveness’ and stroke volume derived by an advanced haemodynamic monitor could be more precisely elucidated with further translational studies: if we are to achieve bespoke fluid therapy then we need to be confident that the monitors can adequately characterise functional circulating volume for the individual. Then suitable algorithms need to be tested in adequately powered trials to investigate whether they achieve a better clinical outcome than ‘standard’ practice in the setting of contemporary surgery. The frog may yet turn back into a prince, but at this point in time the happy ending seems a long way away. No external funding and no competing interests declared.