Haemoconcentration as a Treatment Goal in Heart Failure: Ready for Prime Time?

One of the most remarkable evidences of biological defense is the preservation of intravascular volume and pressure. This is often done against difficult odds, and therefore with varying degrees of success. Richard Gorlin, MD (1961)1 For the centuries that heart failure has been described by clinicians, the centrality of hypervolaemia in its pathogenesis has remained undisputed. Indeed, when patients with heart failure experienced acute exacerbations, therapeutics always aimed at volume reduction—previously with bloodletting and Southey Tubes, and then with diuretics.2 One might therefore anticipate that clinicians have turned this considerable experience into a streamlined methodology to assess patient volume status that is both accurate and precise. Unfortunately, nothing could be further from the truth: volume status on the hospital wards today is assessed via a mishmash of physical exam findings, imaging studies, and laboratory tests that are integrated by the treating physician whose appraisals fall unambiguously under the purview of clinical gestalt. If pressed, many clinicians might counter that invasively determined filling pressures provide us with the ‘gold standard’ for assessment of volume status in heart failure. This is a very common misconception. While it is true that there are correlations between elevations in cardiac chamber pressures and excess volume, these are far from robust.3 The reason for this is that fluid homeostasis in the heart failure patient is an incredibly complex process with multiple components of perturbed pathways and regulatory networks. For example, excess volume in these patients exists in a multitude of interdependent and overlapping compartments such as total body water, extracellular water, intracellular water, blood volume, plasma volume, red cell volume, venous blood volume, stressed blood volume, and unstressed blood volume; these are in continuous flux in response to interventions and disease state. Furthermore, if we were able to measure each of these at the bedside—which we are not—the combinatorics alone would be impractical for clinical decision-making. Furthermore, intravascular pressure elevations result from increases in either vascular volume, vascular tone, or both. This complex relationship between pressure and volume in heart failure was illustrated in an elegant study by Guazzi and colleagues where they exposed subjects to a brief episode of high rate ultrafiltration (500 mL/h) until the right atrial pressure was reduced by 50%. This ∼6-h intervention resulted in an average decrease in blood volume of ∼20% (Figure 1A).4 Over the next 48 h they noted that filling pressures remained reduced and at a constant level (Figure 1B), but blood volume rebounded completely back to baseline (Figure 1A). The missing variable explaining the disconnect between pressure and volume was, of course, vascular (venous) tone, evidenced by the large reductions in norepinephrine and plasma renin activity (Figure 1C and D). It is for these very reasons that heart failure patients who have undergone apparently effective diuresis often remain with significant volume overload despite feeling well and appearing ‘euvolaemic’ to the treating clinician. Miller and Mullan demonstrated this very fact by showing that total blood volume assessed via radiolabelled albumin—a gold standard measure of intravascular volume—was only modestly reduced prior to discharge in heart failure patients who had undergone aggressive diuresis and large reductions in body weight.5 It was the above physiology that motivated investigators to look into metrics such as haemoconcentration that could query changes in blood volume. The concept here is that, regardless of what happens to filling pressures or vascular tone, if the amount of water in the intravascular space is reduced, the concentration of large molecules that are obligatorily intravascular (e.g. albumin or red blood cells) will increase in parallel. This hypothesis was first tested in heart failure patients by our group in 2010, when we noted that among 336 patients enrolled in the ESCAPE trial, those who experienced in-hospital haemoconcentration received higher loop diuretic doses, a higher incidence of worsening renal function, but enjoyed substantially improved survival.6 These findings were reproduced in a number of different populations ranging from single-centre studies to post-hoc analyses of large trials such as PROTECT and EVEREST.7-9 Interpretation of these observations was challenging since it was unclear whether the survival advantage resulted from improved decongestion, or it was merely identifying patients who were healthier and thus more diuretic responsive. To delve deeper into this question, we shifted our focus on the timing of haemoconcentration. If haemoconcentration was simply a signal for a ‘healthier’ heart failure patient, the timing of this finding should not have mattered: patients with both early and late haemoconcentration would fare equally well. Alternatively, if it were a signal for adequacy of decongestion, only those who retained a degree of haemoconcentration at the time of discharge due to refilling from extravascular stores—patients in whom intravascular volume overload did not recur upon de-escalation of diuretics—would show improved survival. We tested this hypothesis in a cohort of 845 heart failure patients and found that patients with late vs. early haemoconcentration had similar baseline characteristics, cumulative in-hospital loop diuretic administered, incidence of worsening of renal function, and a similar degree of maximum haemoconcentration. However, at the time of discharge, the majority of early haemoconcentration patients no longer had evidence of haemoconcentration, suggesting that extravascular volume overload remained when diuresis was de-escalated. Importantly, late haemoconcentration was associated with significantly improved survival whereas early haemoconcentration conferred no mortality benefit.10 A global theme that has emerged from the haemoconcentration literature thus far suggests the following: haemoconcentration tracks with more aggressive diuresis, worsening of renal function, and substantially better survival. However, simply having haemoconcentration at one moment in time is not good enough as only patients with sustained haemoconcentration appear to derive benefit. That said, the bases for the above findings are retrospective or post-hoc analyses of clinical trials, with substantial limitations. In the current issue of the journal, Breidthardt and co-authors performed the most comprehensive study of haemoconcentration to date.11 A total of 1019 patients hospitalized for acute heart failure at the University Hospital in Basel, Switzerland, were prospectively enrolled in a study with serial protocol-driven blood collections. Haemoconcentration was defined via a simultaneous increase in three out of four of haemoglobin, haematocrit, total protein, and albumin levels at any time during hospitalization. The reason for a multimarker strategy such is this is due to the fact that each one of these markers individually has relatively poor precision. However, the random variation of each marker is independent of the other, thus requiring several markers to move simultaneously in the same direction greatly improves the signal to noise ratio. Breidthardt's findings in the current study were exactly along the lines of what has been reported previously: greater decongestion, worse kidney function, but dramatically improved outcomes. Moreover, they made a similar observation that the timing and persistence of haemoconcentration modulates the associated survival.12 This comprehensive validation of prior studies in a rigorous manner goes a long way to add credence to the assertion that the published associations with haemoconcentration are real rather than spurious. So does this mean that we should now enthusiastically incorporate haemoconcentration into our daily clinical practice? Conventional wisdom would mandate a strong recommendation against adopting an epidemiological signal into clinical practice based only on observational findings. We have learned time and time again that despite strong epidemiological signals (i.e. people with higher HDL levels live longer), therapeutic manipulation of these parameters often does not result in the same. However, there are a few reasons to consider addition of haemoconcentration to our daily practice despite the above. First, haemoconcentration is the only widely available inexpensive tool that allows us to query change in intravascular volume. Secondly, the majority of parameters we use on a daily basis clinically to evaluate volume status are abysmal with respect to data fidelity, and their use has never been shown to be beneficial in a prospective trial. Finally, many of the key measures on which we anchor our clinical decision-making, such as fluid and weight loss, do not consistently track with improved outcomes in epidemiological studies.13 So, whereas the data surrounding haemoconcentration are by no means conclusive, comparison of this metric with other standard clinical parameters greatly elevates it. As a result, cautious incorporation into our appraisal of whether a patient has been adequately decongested may be reasonable. This recommendation is somewhat consistent with the current US and European guidelines that recommend close evaluation of laboratory parameters along with physical exam findings to gauge adequacy of volume removal; however, precise mention of haemoconcentration is presently absent.14, 15 That said, there are important caveats surrounding how we should incorporate haemoconcentration into clinical practice. Even if haemoconcentration were the perfect measure for change in blood volume—which it is not—it remains only a single dimension of the many parameters that define a patient's volume and congestion status. To that end, Darawsha and colleagues recently demonstrated that standard physical examination parameters of congestion not only correlated poorly with haemoconcentration, but had independent and additive prognostic value.16 Furthermore, haemoconcentration only gives us relative rather than absolute information about a patient's blood volume, a parameter which is in equilibrium with other fluid compartments. Therefore, the presence or absence of haemoconcentration does not indicate we have reached an endpoint, simply that we have moved the patient in the right direction. Other barriers to the integration of haemoconcentration into clinical practice include a lack of clear definitions, specific target goals, and unclear interactions with various parameters such as renal function, medications, and different heart failure phenotypes. As a result, haemoconcentration cannot be recommended as a stand-alone tool or endpoint in a treatment algorithm, but rather it should represent a contribution to how we formulate our clinical gestalt about whether a patient is ‘wet’ or ‘dry’. Where shall we go from here? We believe that the current evidence makes haemoconcentration stand out as an important addition to our approach to volume assessment, in large part due to the severe shortcomings of the company it will keep. To move forward, deeper dives into the unknowns of volume and pressure homeostasis in heart failure will be necessary, and this will require additional prospective research. However, given the profound impact of volume overload in heart failure, successes in these endeavours are likely be transformative. NIH Grants, K23HL114868, and L30HL115790. Conflict of interest: none declared.