Physician, Critical Care Department, Hospital Israelita Albert Einstein – São Paulo – Brazil
National Director of Hospital Services, DaVita – São Paulo
Student of Postgraduation Evidence-Based Medicine, Federal University of São Paulo – São Paulo, Brazil
MBA in Healthcare Management, Hospital Israelita Albert Einstein
Director, Cardio-Renal-Metabolic Committee, Brazilian Society of Nephrology
A 58-year-old woman with a background of ischaemic cardiomyopathy (LVEF 30%) and stage 2 CKD is admitted to the ICU following emergency laparotomy for perforated sigmoid diverticulitis. She received 6 litres of crystalloid intraoperatively and a further 4 litres in the first 12 hours of ICU admission. She is intubated and mechanically ventilated, vasopressor-dependent (noradrenaline 0.25 mcg/kg/min), and in the context of evolving septic shock.
On day 2, the team documents a cumulative positive fluid balance of +11 litres. Chest radiograph shows bilateral pulmonary infiltrates; lung ultrasound reveals bilateral confluent B-lines in all six zones with subpleural consolidation at the bases. SpO₂ on FiO₂ 0.6 is 91%. Urine output has fallen to less than 200 mL in the preceding 24 hours. Creatinine has risen from her baseline of 180 µmol/L to 394 µmol/L. Lactate is 3.8 mmol/L, pH 7.21, bicarbonate 14 mmol/L, potassium 6.1 mmol/L.
Point-of-care echocardiography demonstrates a dilated, poorly contracting left ventricle with a D-shaped interventricular septum consistent with elevated right heart pressures. The inferior vena cava is plethoric and non-collapsible. Passive leg raising produces no meaningful change in pulse pressure or velocity time integral.
The nephrology and ICU teams agree that kidney replacement therapy with ultrafiltration is indicated. However, the haemodynamic phenotype—a non-preload-responsive, fluid-overloaded patient with biventricular dysfunction and vasopressor dependence—raises immediate concern about ultrafiltration tolerance. The team is debating on how to prescribe, monitor, and adapt fluid removal safely in a patient with almost no circulatory reserve.
(Click / Tap on Questions to Reveal Content)
Ultrafiltration tolerance in critically ill patients is not determined by the degree of fluid overload, but by the ability of the cardiovascular system to preserve effective circulating volume and tissue perfusion during intravascular volume reduction. In practice, this becomes evident very early: some profoundly overloaded patients tolerate fluid removal well, while others become unstable with minimal ultrafiltration.
I have found it most useful to approach this through four interdependent physiological systems: vascular refilling, cardiac response, venous capacitance, and systemic vascular resistance. When these systems are intact, ultrafiltration is generally tolerated. In the ICU, however, they are frequently impaired simultaneously.
Vascular refilling is often unreliable in sepsis due to endothelial dysfunction, glycocalyx disruption, and increased permeability. Even in patients with marked edema, intravascular volume may not be adequately maintained during ultrafiltration.
Cardiac reserve is commonly limited. In particular, diastolic dysfunction and right ventricular impairment make cardiac output highly sensitive to small reductions in preload. In addition, dialysis itself may induce myocardial stunning, which I have seen precipitate instability early in the session, sometimes before meaningful fluid removal has even occurred.
The venous system also plays a critical role. In health, unstressed volume can be mobilized through sympathetic tone. In vasoplegic or sedated patients, this mechanism is blunted, leaving the circulation more dependent on absolute volume.
Finally, impaired vascular tone—especially in septic shock—limits the ability to maintain arterial pressure during fluid removal. These patients often develop hypotension despite ongoing vasopressor support, reflecting a failure of vascular compensation rather than simple volume depletion.
Taken together, ultrafiltration intolerance reflects a failure of integrated cardiovascular reserve rather than hypovolemia per se. This is fundamentally different from chronic dialysis populations, where compensatory mechanisms are more preserved and responses to fluid removal are more predictable .
A negative passive leg raising test indicates that cardiac output is not preload-responsive, suggesting that reductions in preload are less likely to produce a preload-dependent fall in cardiac output.
However, I have learned to be cautious in interpreting this as reassuring.
PLR interrogates only one component of the haemodynamic response. While a positive test clearly identifies patients at risk of preload-mediated instability, a negative test does not exclude intolerance to ultrafiltration. In my experience, many patients who are PLR-negative still develop hypotension, often through mechanisms unrelated to preload dependence.
These include impaired vascular tone, inadequate venous recruitment, limited cardiac reserve, and insufficient vascular refilling. Moreover, haemodynamic state is not static; a patient may transition from preload independence to dependence as ultrafiltration progresses.
For this reason, I tend to interpret a negative PLR as excluding a specific mechanism, rather than as evidence of global haemodynamic stability. It is informative, but incomplete .
One of the most challenging situations in ICU practice is the coexistence of severe congestion and profound haemodynamic fragility. These patients clearly need decongestion, yet are unable to tolerate it.
Ultrasound has been particularly helpful in reframing this problem. Rather than thinking in binary terms, I try to phenotype the patient. A combination of diffuse B-lines, a plethoric IVC, vasopressor dependence, and biventricular dysfunction defines a congested but poorly tolerant profile.
The key realization is that the need for fluid removal and the capacity to tolerate it are separate questions. The presence of pulmonary congestion confirms the indication, but does not inform the rate at which it can be safely achieved.
In practice, this has led me to adopt a strategy of progressive, physiology-guided decongestion. I start with very low ultrafiltration rates, often after stabilizing metabolic and haemodynamic variables, and adjust based on continuous reassessment. In many cases, I accept incomplete decongestion in the short term to avoid precipitating instability.
The goal is not rapid normalization, but gradual improvement without compromising perfusion. This shift in mindset has been critical in managing these patients safely .
Randomized trials suggest no clear superiority of one RRT modality over another in terms of mortality or renal recovery. However, at the bedside, modality clearly matters.
The difference lies in how each modality interacts with physiology. Continuous therapies allow gradual ultrafiltration and more stable haemodynamics, while intermittent therapies impose more abrupt changes.
In my practice, modality selection is guided primarily by haemodynamic phenotype. Patients with significant instability or limited cardiac reserve are much more likely to tolerate continuous therapies. Conversely, more stable patients can benefit from the efficiency of intermittent approaches.
This is less about superiority and more about physiological compatibility. Trial data inform us about averages, but individual patients rarely behave like averages .
There is no universally safe ultrafiltration rate in critically ill patients. While observational data suggest that moderate rates may be associated with better outcomes, I have found that tolerance is highly individualized.
In patients with minimal haemodynamic reserve, even low rates may be excessive. For this reason, I approach ultrafiltration as a titrated intervention rather than a fixed prescription.
I usually begin conservatively, sometimes with no net ultrafiltration initially, and increase gradually based on tolerance. Small adjustments, frequent reassessment, and close attention to early signs of instability are essential.
One important lesson is that harm from excessive ultrafiltration may occur without overt hypotension. Subclinical hypoperfusion is likely more common than we recognize.
The objective is not simply to remove fluid, but to do so without compromising organ perfusion—a balance that is often narrow and dynamic .
Point-of-care ultrasound has changed how I approach ultrafiltration. It allows me to move away from static prescriptions and toward continuous physiological reassessment.
Lung ultrasound informs the need for decongestion, but not tolerance. Cardiac assessment provides insight into forward flow and reserve. Venous Doppler helps quantify systemic congestion and, importantly, how the patient is responding to fluid removal.
What has been particularly valuable is identifying early signs of intolerance before hypotension occurs. Worsening right ventricular function, increasing venous congestion, or declining stroke volume often precede overt instability.
Perhaps the most important insight is that tolerance evolves during the session. A rate that is initially tolerated may become excessive over time. Serial reassessment is therefore essential.
Ultrasound, in this context, becomes less a diagnostic tool and more a guide to therapy .
In patients with severe hyperkalaemia and acidosis, the immediate priority is metabolic stabilization. Fluid removal, although important, is less urgent.
Continuous therapies allow us to separate these objectives. I typically prioritize solute clearance with adequate effluent flow while minimizing or deferring ultrafiltration initially.
Only after partial metabolic stabilization do I begin fluid removal, and even then, cautiously. Attempting to address both simultaneously at full intensity is, in my experience, a common cause of haemodynamic collapse.
When compromise is necessary, I reduce ultrafiltration before clearance. Hyperkalaemia and acidosis pose immediate risks, whereas fluid overload can usually be addressed more gradually.
This staged approach has proven more reliable than trying to achieve all goals simultaneously .
Renal recovery is largely determined by the initial insult, but what happens during ICU care often dictates whether recovery occurs.
What I have observed repeatedly is that recovery is lost not at the moment of injury, but during subsequent management. Recurrent hypotension, excessive ultrafiltration, and persistent venous congestion all contribute to ongoing injury.
Two opposing processes are particularly important: hypoperfusion and congestion. Both impair renal function through different mechanisms, and both must be minimized.
While trials have not demonstrated clear differences between modalities, the way therapy is delivered matters. Continuous therapies facilitate gradual decongestion and more stable haemodynamics, which may help preserve renal perfusion.
Ultimately, recovery depends less on the modality chosen and more on whether we avoid secondary injury. In that sense, kidney replacement therapy is not neutral—it can either protect or jeopardize recovery .
Final Perspective
Over time, my approach to ultrafiltration in critically ill patients has shifted from targeting fluid balance to managing physiological tolerance.
These patients rarely benefit from rapid correction. Instead, outcomes seem to improve when we accept slower trajectories, prioritize perfusion, and adjust therapy continuously.
Ultrafiltration, in this context, is not simply fluid removal. It is a controlled perturbation of a fragile system. The challenge is not deciding what to remove, but determining how much the patient can tolerate at each moment—and respecting that limit.
