Physiology of Renal Replacement Therapy and Ultrafiltration

Renal replacement therapy moves water and selected solutes across an artificial or biological barrier. Its effects depend on transport mechanism, membrane properties, fluid composition, treatment time, and the patient's changing solute distribution. Understanding those components separates clearance from fluid removal and a machine setting from the treatment actually delivered.

50.1 Replacement Therapy Substitutes for Selected Renal Functions

The native kidney combines filtration with tubular transport, metabolic activity, and endocrine signaling. Extracorporeal exchange does not reproduce that integrated organ. Removing urea, changing potassium, and achieving negative fluid balance are distinct accomplishments; none establishes complete replacement of renal function.

In a typical venovenous circuit, blood leaves a vascular catheter, passes through a pump and hollow-fiber filter, and returns to the venous circulation. The membrane separates blood from dialysate or collected ultrafiltrate. Dialysate flows on the nonblood side; replacement fluid enters the blood circuit to replace much of the water removed during hemofiltration. Standardized nomenclature helps distinguish these flows and their different purposes. 1

Clearance expresses solute removal normalized to the concentration supplying the device. Let Ṁremoved be net removal rate in mmol/h and Cpatient the representative patient concentration in mmol/L. Clearance K is expressed in L/h.

For an illustrative removal of 30 mmol/h from a circulating concentration of 20 mmol/L, K is 30/20 = 1.5 L/h. This is the hypothetical volume completely cleared per hour, not the volume of blood extracted or the amount of patient water removed. At a different concentration, the same clearance produces a different mass-removal rate.

Net removal must account for solute entering as well as leaving. A potassium-containing replacement solution can return potassium while the filter removes it. Comparing only potassium in the waste bag would overstate the patient's loss. For a solute absent from all incoming circuit fluids, the accounting is simpler.

Blood access, membrane function, and uninterrupted operation are prerequisites for the intended exchange. Anticoagulation can preserve circuit function while introducing its own systemic and metabolic effects. The treatment is therefore a dynamic physiological intervention rather than an isolated filter specification.

50.2 Diffusion Follows a Concentration Gradient

Diffusion transfers solute down its chemical-potential gradient. Low dialysate concentration favors movement of a retained blood solute across the membrane. The direction can reverse for another substance: bicarbonate or an electrolyte may move toward blood when its gradient favors entry.

A simplified transfer relationship combines the membrane and adjacent fluid layers. Let K₀A be the overall mass-transfer area coefficient for a particular solute in L/h, incorporating effective exchange area and transport resistance. Let ΔCmean be an appropriate mean blood–dialysate concentration difference in mmol/L. Diffusive transfer rate Ṁdiff is in mmol/h.

The gradient varies along the filter, so a single inlet difference generally cannot replace the appropriate mean. Blood loses solute while dialysate gains it. Countercurrent flow maintains a favorable gradient along more of the membrane than would complete mixing into one shared compartment.

Small, readily diffusible solutes can approach equilibration with slowly moving dialysate. In that setting, dialysate throughput strongly influences clearance. Increasing dialysate flow eventually produces less additional benefit when membrane transfer or solute delivery from blood becomes limiting. Increasing blood flow likewise has diminishing returns if the dialysate side or membrane is the bottleneck.

K₀A is solute-specific and condition-dependent. Membrane thickness, effective area, molecule size, boundary layers, and blood exposure alter transfer. A value measured with a simple laboratory solution does not guarantee the same performance in a patient. Depner and colleagues modeled cross-dialyzer urea extraction in the HEMO study and found important differences between nominal and inferred blood flow and between in-vitro and in-vivo membrane performance. This was maintenance-dialysis evidence, rather than a direct intensive-care prescription study. 2

Diffusive exchange need not remove a large net volume of patient water. A dialyzer can move substantial solute while fluid accounting maintains near-zero patient balance. Concentration gradients and pressure-driven water flow describe different transport components even when both occur in the same device.

50.3 Ultrafiltration Moves Water and Convection Carries Solute

Ultrafiltration is water movement across a membrane driven by effective transmembrane pressure. Convection is solute transport accompanying that bulk water movement. The two are linked, but water permeability does not guarantee equal permeability to every dissolved molecule.

For a simplified membrane segment, let Quf be water flow in L/h and LpA its hydraulic conductance in L/(h·mmHg). Pb and Pd are blood-side and nonblood-side hydrostatic pressures; πb and πd are oncotic pressures, all in mmHg. The dimensionless protein reflection coefficient σprotein represents how effectively retained protein generates an opposing osmotic force.

The equation describes an idealized local relation. Protein concentration rises as water is removed from blood, while hydrostatic pressure changes along the fibers. A machine's displayed transmembrane pressure generally estimates an average hydrostatic separation from available pressure sensors; it does not directly measure every local effective gradient or the opposing oncotic pressure. Local filtration and backfiltration can coexist despite a positive average.

Solute passage is described by a sieving coefficient. In a simplified well-mixed model, let S equal ultrafiltrate concentration divided by representative filter plasma-water concentration. S approaches one for a freely passing solute and is much smaller for a strongly retained substance. Convective clearance relative to that reference concentration is approximately:

For illustrative Quf of 2 L/h and S of 0.8, Kconv is 0.8 × 2 = 1.6 L/h. This describes the selected solute under the model assumptions; another molecule can have a different clearance through the same filter.

In a small clinical transport study, Brunet and colleagues found different relationships between flow and removal for small solutes and β₂-microglobulin, with effects of filter type and predilution. The findings support mechanism-specific assessment rather than a universal claim that one transport method removes all molecules better. 3

Patient clearance also depends on the concentration actually reaching the membrane. Predilution, protein binding, and concentration changes along the filter can separate patient-referenced clearance from this simplified filter calculation.

50.4 Membrane Selectivity and Adsorption Change What Is Removed

A membrane's transport profile depends on pore distribution, chemistry, thickness, and interaction with blood. Molecular weight is useful but incomplete: shape, charge, binding, and access to the pore also matter. Increasing permeability to larger molecules can increase removal of desired targets while permitting loss of useful constituents.

Protein binding limits immediate access to many solutes. A small molecule attached to albumin behaves differently from its freely dissolved fraction. Removing free solute can be followed by dissociation from protein, so extraction evolves during treatment. A high total plasma concentration does not guarantee proportionately high membrane removal.

Adsorption captures substances on a surface rather than requiring their appearance in ultrafiltrate or dialysate. Filter material may adsorb some compounds, while dedicated hemoperfusion cartridges provide additional binding surfaces. Capacity is finite and selectivity depends on surface chemistry, available sites, competing substances, and concentration. Early extraction therefore need not remain constant as a cartridge becomes occupied.

Effluent analysis detects molecules carried into waste fluid but can miss those retained on the device. Conversely, cartridge extraction does not establish a durable systemic concentration reduction. Production, tissue release, redistribution, and changing binding can offset removal.

Schädler and colleagues tested a cytokine adsorber in a randomized septic-patient trial. The device removed interleukin-6 across the cartridge, yet the study did not demonstrate the intended reduction in systemic interleukin-6 relative to control. It did not establish a clinical benefit from cytokine extraction. 4

Protein deposition, concentration polarization near the membrane, clotting, and loss of usable area can reduce exchange before complete circuit failure. Neither pressure requirements nor effluent volume alone establish how efficiently every solute is removed.

Unintended removal matters alongside target removal. Drugs, nutrients, and useful proteins may be affected according to their specific properties. This physiology supports reassessment of exposure and balance; it does not provide a universal drug-adjustment factor from the modality name alone.

50.5 Predilution Trades Filter Conditions Against Concentration

During hemofiltration, replacement fluid can enter before or after the filter. Postfilter replacement preserves the undiluted inlet concentration but allows more hemoconcentration within the fibers. Prefilter replacement dilutes blood before filtration, reducing both cell and protein concentration at the membrane while lowering the concentration of patient solute available for removal.

A simplified dilution model makes this tradeoff explicit. Let Qw be the incoming solute-bearing plasma-water flow and Qpre the prefilter replacement flow, both in mL/min. Assume replacement contains none of the reference solute, mixing is complete, and the solute remains in the modeled plasma-water compartment. The filter inlet concentration relative to the patient's concentration is:

Choose Qw of 100 mL/min and Qpre of 25 mL/min. The ratio is 100/(100 + 25) = 0.8. If the filter provides 2 L/h clearance relative to its diluted inlet concentration, patient-referenced clearance is approximately 2 × 0.8 = 1.6 L/h. The same filtered volume removes less patient solute because each liter reaching the membrane is more dilute.

This example is a plasma-water mixing model, not a universal blood-flow correction. Whole-blood pump flow includes red cells and a nonwater fraction. Urea can exchange with red-cell water, while other solutes have different cellular access. An exact calculation must match the flow compartment, concentration convention, replacement composition, and transport kinetics.

Water removal within the fibers raises hematocrit and protein concentration. A filtration fraction describes filtered flow relative to an appropriate incoming plasma or plasma-water flow, with predilution included consistently. Dividing ultrafiltration by unqualified whole-blood flow can conceal the degree of plasma concentration.

Higher concentration can increase viscosity, oncotic opposition, and membrane deposition. Predilution can mitigate those conditions but does not guarantee a particular circuit lifetime. Anticoagulation, access function, membrane characteristics, and illness also matter. No single filtration-fraction threshold is a universal assurance of effective or safe operation.

50.6 Effluent Flow Is a Surrogate for Delivered Clearance

Effluent is fluid leaving the nonblood side of the device. In hemodiafiltration it contains spent dialysate plus total ultrafiltrate. Effluent flow can approximate clearance for a freely passing small solute when its concentration in effluent approaches the relevant patient concentration and incoming fluids contain none of that solute.

The approximation can be checked directly. Let Qeff be measured effluent flow in L/h, Ceff effluent urea concentration in mmol/L, and Cp representative patient urea concentration in mmol/L. Assume fresh circuit fluids contain no urea and adsorption is negligible.

For Qeff of 2 L/h, Ceff of 18 mmol/L, and Cp of 20 mmol/L, the removal rate is 2 × 18 = 36 mmol/h. Clearance is 36/20 = 1.8 L/h. The machine's 2-L/h effluent flow would overestimate this measured clearance if treated as an exact equivalent.

Time off treatment adds a separate loss of delivery. If the circuit provides that clearance for 20 hours during a 24-hour interval and provides none for the remaining four hours, integrated clearance is 1.8 × 20 = 36 L. The time-averaged value is 36/24 = 1.5 L/h. With concentrations held constant for this illustration, urea removal totals 36 × 20 = 720 mmol. Changing concentrations require time-resolved mass accounting.

Claure-Del Granado and colleagues measured blood and effluent urea during predilution hemodiafiltration in critically ill patients. Effluent-based estimates exceeded measured delivered clearance, with declining filter performance contributing to the discrepancy. The observed difference is evidence for checking delivery, not a fixed correction percentage for every circuit. 5

Downtime may arise from procedures, access problems, clotting, alarms, or filter changes. Pump displays, actual collected volume, solute saturation, and operating time answer different questions. Normalizing effluent to body weight does not eliminate these distinctions, and the selected weight itself may be affected by fluid accumulation.

Measured clearance also requires representative, contemporaneous samples. A postfilter or diluted circuit sample cannot automatically replace the patient's supplying concentration. Residual native-kidney clearance remains an additional contribution.

50.7 Body Distribution and Rebound Separate Plasma Change from Removal

The filter samples a circulating compartment, while much of a solute may reside elsewhere. Delivery from tissues depends on blood flow, membrane transport, binding, and intracellular exchange. A rapid plasma decline can therefore outpace equilibration with the rest of the body.

For a hypothetical single, well-mixed compartment, let V be a fixed distribution volume in liters, C concentration in mmol/L, G generation in mmol/h, and K total clearance in L/h. Let t denote elapsed hours. Ignoring intake and other losses, mass balance is:

If G is zero, V and K are constant, and the initial concentration is C₀, the solution is C(t) = C₀ × exp(−Kt/V). The dimensionless Kt/V expresses cumulative clearance relative to the modeled distribution volume. It is not the fraction of body solute removed by simple subtraction.

With illustrative K of 6 L/h, t of 4 hours, V of 40 L, and C₀ of 20 mmol/L, Kt/V is 6 × 4/40 = 0.6. Concentration becomes 20 × exp(−0.6) ≈ 11.0 mmol/L, a fall of approximately 45%. It does not fall by 60%. This mathematical example excludes generation, changing water volume, tissue disequilibrium, and residual-function changes.

After rapid extracorporeal removal stops, solute can return from less-cleared tissues to plasma. Concentration rebound therefore need not mean that solute has entered the body or that the device failed to remove it. Early access or cardiopulmonary recirculation and later tissue equilibration can contribute on different time scales.

In a large maintenance-hemodialysis analysis, Daugirdas and colleagues measured postdialysis urea rebound and found a strong relationship with clearance relative to distribution volume, with additional patient and treatment influences. These observations demonstrate compartment effects without making their fitted equations universal for acute kidney injury. 6

Potassium, phosphate, protein-bound substances, and drugs have different distribution kinetics from urea. A favorable immediate post-treatment sample cannot certify their sustained control. Continued generation or release can also exceed ongoing removal despite a functioning circuit.

50.8 Total Ultrafiltration, Net Removal, and Patient Balance Are Different

Large filter water flux can coexist with little patient fluid removal when most ultrafiltrate is replaced. Conversely, fluid can be removed without a large replacement-driven convective treatment. Solute-support intensity and net fluid removal should therefore be described separately.

For a simplified circuit, let Quf be total filter ultrafiltration, Qr total prefilter plus postfilter replacement, and Qnet the net patient fluid-removal rate through that circuit, all in L/h. Assume no additional unaccounted circuit infusions or fluid storage.

If Quf is 2.0 L/h, replacement is 1.8 L/h, and dialysate is 1.0 L/h, net patient removal is 2.0 − 1.8 = 0.2 L/h while effluent is 1.0 + 2.0 = 3.0 L/h. Calling all 3 L/h patient fluid removal would be a major accounting error. Much of the waste volume arrived as dialysate or replacement fluid.

Over 20 operating hours, the circuit removes 0.2 × 20 = 4.0 L net. Now suppose other patient inputs total 3.6 L over the full day, urine is 0.6 L, and other counted losses total 0.4 L. Recorded whole-patient balance is 3.6 − 4.0 − 0.6 − 0.4 = −1.4 L. The circuit's 4-L net removal is not a 4-L negative patient balance.

The bookkeeping boundary must be explicit. Citrate, calcium, flushes, priming fluid retained by the patient, nutrition, drugs, and blood products may enter through different routes. Include each once. Some machine systems incorporate selected infusions in displayed balance; others require additional accounting. Blood left in a discarded circuit is another distinct loss.

Measured balance is also not a direct measurement of plasma volume. Water can redistribute among vascular, interstitial, and cellular compartments, and incompletely measured losses create error. The interaction between net removal, vascular refill, and organ perfusion is developed in Chapter 51.

These arithmetic examples describe accounting, not a removal-rate recommendation. The tolerable trajectory depends on the circulation and the changing distribution of retained fluid.

50.9 Fluid Composition Determines Electrolyte Balance

Dialysate and replacement fluid have different routes into the exchange process. Dialysate establishes gradients across the membrane. Replacement fluid directly adds water and solutes to blood, before or after filtration. Their composition is therefore an active determinant of treatment, not a neutral accompaniment to clearance.

Sodium exchange interacts with water balance and tonicity. A sodium-containing fluid can contribute substantial sodium turnover even during near-zero net water removal. The eventual plasma concentration depends on both input and removal, other body exchanges, and cellular water shifts. An isovolemic circuit does not guarantee unchanged sodium balance.

Tan and colleagues measured electrolyte mass balance during a small randomized crossover comparison of replacement fluids. Equal fluid-volume exchange did not produce identical electrolyte balance, and the tested formulations generated different chloride and magnesium fluxes. These were formulation-specific short-term findings, rather than a universal property of every fluid carrying the same buffer label. 7

Potassium removal depends on blood concentration, fluid potassium, transport efficiency, and delivery from cells. Phosphate and magnesium can be lost during prolonged treatment even after an initial excess resolves. Total and ionized calcium require distinction because binding and complex formation can change independently of total calcium balance.

In an observational comparison, Broman and colleagues found fewer hypophosphatemic episodes with phosphate-containing dialysis and replacement fluids. This supports a composition effect on biochemical control, without establishing a mortality benefit or universally appropriate formulation. 8

Concentration trends require interpretation alongside nutrition, tissue breakdown, redistribution, native excretion, and extracorporeal losses. Continued high removal after production falls can change retention into depletion. Conversely, ongoing release can maintain a high plasma concentration despite substantial removal.

The practical physiological question is the net balance of each relevant species. Sterility, compatibility, and the actual fluid formulation matter; a theoretical gradient is not sufficient justification for unvalidated mixing or substitution.

50.10 Acid–Base Effects Depend on Transport and Metabolism

Renal replacement therapy can remove small acid-associated anions and supply bicarbonate or metabolizable buffer precursors. The resulting pH depends on ongoing acid generation, buffer transfer, strong-ion changes, and carbon dioxide elimination. It is not determined by urea clearance alone.

Bicarbonate can move down a dialysate–blood gradient or enter directly in replacement fluid, while bicarbonate also leaves in effluent. Net transfer requires both sides of that balance. Chloride and other electrolyte fluxes can modify acid–base conditions even when the labeled buffer concentration appears similar.

Lactate-containing fluid supplies a metabolizable anion. Its ultimate alkalinizing contribution depends on metabolism and the complete accompanying ionic balance; lactate accumulation cannot be treated as immediate bicarbonate delivery. In a small randomized crossover study, Tan and colleagues found different lactate and acid–base responses to the tested lactate- and bicarbonate-buffered replacement fluids. A positive calculated buffer-base balance did not guarantee a higher measured bicarbonate or base excess. 9

Citrate anticoagulation adds another coupled system. Citrate binds calcium, reducing ionized calcium within the circuit and impairing coagulation locally. Some citrate–calcium complexes leave in effluent; retained citrate can be metabolized, generally contributing an alkalinizing effect when delivered with sodium. Citrate and calcium balances therefore depend on circuit removal, systemic metabolism, and separate calcium inputs.

Impaired citrate metabolism can increase circulating citrate and calcium complexing, unlike excess successfully metabolized citrate producing alkalinization. Kramer and colleagues found lower citrate clearance in critically ill cirrhotic patients than in noncirrhotic comparators. Short controlled exposure does not establish prolonged-treatment safety, and routine liver tests incompletely describe metabolic capacity. 10

Buffer-related carbon dioxide must ultimately be handled by the lungs. A patient with limited effective ventilation can respond differently from one who readily eliminates additional CO₂. Acid–base and ionized-calcium changes also affect the circulation, linking this transport physiology to Chapter 51 without reducing those effects to a fluid-volume problem.

50.11 Modality Changes the Time Course of Exchange

Intermittent hemodialysis provides relatively high instantaneous solute clearance over a limited treatment period. The intervening interval permits continued generation and accumulation. A rapid concentration fall can be useful for a particular solute burden, while greater compartment disequilibrium and a compressed fluid-removal interval can create physiological challenges.

Continuous therapy spreads exchange across much of the day, usually with lower instantaneous clearance. It can better match continuous production and ongoing fluid input when it remains operational. Its name does not guarantee uninterrupted delivery: access problems, procedures, and circuit failure can materially change the daily exposure.

Prolonged intermittent therapy uses a longer session than conventional intermittent dialysis, often with lower instantaneous exchange. Its physiological position depends on actual flows, duration, interruptions, and composition rather than the label alone. The same daily removed amount delivered over different schedules need not produce the same concentration peaks, rebound, or circulatory response.

Continuous venovenous hemofiltration emphasizes convection; continuous venovenous hemodialysis emphasizes diffusion; hemodiafiltration combines both. Their small-solute clearances may be similar under suitable conditions while removal of a larger or bound solute differs. Diffusive and convective clearances cannot always be added as if they acted independently on an unchanged blood concentration.

Isolated ultrafiltration prioritizes fluid removal without the replacement-driven solute exchange of hemofiltration. It still carries freely passing solutes with water, so “fluid-only” does not mean literally zero solute removal. Peritoneal dialysis instead uses the peritoneal barrier, with solute gradients and osmotic water movement into cavity fluid; its exchange area and kinetics differ from a hollow-fiber circuit.

The older single-center CONVINT trial found no statistically significant survival difference between intermittent dialysis and continuous hemofiltration, with treatment switching and uncertainty limiting equivalence claims. 11

The 2026 multicenter ICRAKI trial changes that context. It tested whether an initial intermittent-hemodialysis strategy was noninferior to continuous therapy for major adverse kidney events by day 90, combining death, treatment dependence, and persistent creatinine elevation. Recruitment stopped early after a mortality signal; 554 patients were randomized and 535 included in the reported intention-to-treat population. The primary event occurred in 78.1% versus 71.6%, respectively, and noninferiority was not established. Death occurred in 63.1% versus 53.5%, with a mortality hazard ratio of 1.30 (95% confidence interval 1.04–1.63). The mortality finding favors the continuous initial strategy; noninferiority failure alone does not prove superiority for the composite. Early stopping limits effect-size precision; the result does not establish that every continuous technique benefits every patient or identify the causal mechanism. 12

Actual rates, duration, delivery, and patient conditions remain essential. Chapter 51 develops the cardiovascular consequences.

50.12 More Exchange Is Not Automatically Better Support

Increasing flow or treatment frequency can increase removal under appropriate transport conditions. That physiological effect does not ensure better patient outcomes. Additional exchange can also increase unwanted electrolyte, nutrient, or drug losses and expose the patient to more treatment-related disturbances.

The ATN trial randomized 1,124 critically ill patients with acute kidney injury to more- or less-intensive renal-support strategies. It varied intermittent-treatment frequency and continuous-treatment intensity while allowing modality to follow hemodynamic conditions. Sixty-day mortality was 53.6% versus 51.5%, respectively, without a statistically significant difference or improved renal recovery. This tested intensity, not randomized modality selection. 13

The RENAL trial randomized 1,508 patients to postdilution continuous hemodiafiltration with prescribed effluent intensities of 40 or 25 mL/kg/h. Ninety-day mortality was 44.7% in both groups with available primary-outcome data. These trial comparisons are neither universal prescriptions nor evidence that inadequate delivery is harmless. 14

Adequacy consequently requires more than achieving a nominal number. Consider whether the relevant solute burden is being controlled, whether acid generation and electrolyte release are changing, whether circuit composition creates new deficits, and whether fluid accounting matches the intended patient trajectory. A stable urea can reflect balanced high production and high removal rather than low metabolic burden.

Native kidney function remains part of the system. Urine output and solute excretion can recover at different rates, and changes in diuretics or intake can alter urine volume without equivalent recovery of clearance. Conversely, a functioning circuit can mask the concentration consequences of continued native failure.

Connect mechanism, delivered exchange, and patient response. Unexplained concentration changes may arise from production, redistribution, sampling, composition, downtime, or membrane performance. Clinical benefit and treatment tolerance require their own assessment.

References

1. Neri M, Villa G, Garzotto F, et al. Nomenclature for renal replacement therapy in acute kidney injury: basic principles. Critical Care. 2016;20:318. doi:10.1186/s13054-016-1489-9

2. Depner TA, Greene T, Daugirdas JT, Cheung AK, Gotch FA, Leypoldt JK. Dialyzer performance in the HEMO Study: in vivo K₀A and true blood flow determined from a model of cross-dialyzer urea extraction. ASAIO Journal. 2004;50:85–93. doi:10.1097/01.MAT.0000104824.55517.6C

3. Brunet S, Leblanc M, Geadah D, Parent D, Courteau S, Cardinal J. Diffusive and convective solute clearances during continuous renal replacement therapy at various dialysate and ultrafiltration flow rates. American Journal of Kidney Diseases. 1999;34:486–492. doi:10.1016/S0272-6386(99)70076-4

4. Schädler D, Pausch C, Heise D, et al. The effect of a novel extracorporeal cytokine hemoadsorption device on IL-6 elimination in septic patients: a randomized controlled trial. PLoS ONE. 2017;12:e0187015. doi:10.1371/journal.pone.0187015

5. Claure-Del Granado R, Macedo E, Chertow GM, et al. Effluent volume in continuous renal replacement therapy overestimates the delivered dose of dialysis. Clinical Journal of the American Society of Nephrology. 2011;6:467–475. doi:10.2215/CJN.02500310

6. Daugirdas JT, Greene T, Depner TA, et al. Factors that affect postdialysis rebound in serum urea concentration, including the rate of dialysis: results from the HEMO Study. Journal of the American Society of Nephrology. 2004;15:194–203. doi:10.1097/01.ASN.0000103871.20736.0C

7. Tan HK, Uchino S, Bellomo R. Electrolyte mass balance during CVVH: lactate vs. bicarbonate-buffered replacement fluids. Renal Failure. 2004;26:149–153. doi:10.1081/JDI-120038495

8. Broman M, Carlsson O, Friberg H, Wieslander A, Godaly G. Phosphate-containing dialysis solution prevents hypophosphatemia during continuous renal replacement therapy. Acta Anaesthesiologica Scandinavica. 2011;55:39–45. doi:10.1111/j.1399-6576.2010.02338.x

9. Tan HK, Uchino S, Bellomo R. The acid-base effects of continuous hemofiltration with lactate or bicarbonate buffered replacement fluids. International Journal of Artificial Organs. 2003;26:477–483. doi:10.1177/039139880302600605

10. Kramer L, Bauer E, Joukhadar C, et al. Citrate pharmacokinetics and metabolism in cirrhotic and noncirrhotic critically ill patients. Critical Care Medicine. 2003;31:2450–2455. doi:10.1097/01.CCM.0000084871.76568.E6

11. Schefold JC, von Haehling S, Pschowski R, et al. The effect of continuous versus intermittent renal replacement therapy on the outcome of critically ill patients with acute renal failure (CONVINT): a prospective randomized controlled trial. Critical Care. 2014;18:R11. doi:10.1186/cc13188

12. Gaudry S, Boubaya M, Louis G, et al., for the ICRAKI Trial Group. Intermittent or continuous renal-replacement therapy in acute kidney injury. New England Journal of Medicine. 2026;395:1677–1690. doi:10.1056/NEJMoa2608397

13. VA/NIH Acute Renal Failure Trial Network. Intensity of renal support in critically ill patients with acute kidney injury. New England Journal of Medicine. 2008;359:7–20. doi:10.1056/NEJMoa0802639

14. RENAL Replacement Therapy Study Investigators. Intensity of continuous renal-replacement therapy in critically ill patients. New England Journal of Medicine. 2009;361:1627–1638. doi:10.1056/NEJMoa0902413