Renal replacement therapy in intensive care unit: a narrative review Authors: Kada Klouche, Stéphane Gaudry, Jerôme Naudin, Alexandre Boyer, Elise Cuquemelle, Etienne Javouhey, Maud Injeyan, Sandrine Jean, Fabrice Lesage, Mehran Monchi, Luc Morin, Mehdi Oualha, Lamia Ouanes Besbes, Jean Pierre Quenot, Bruno Ranchin, Frédérique Shortgen, Alexandre Lautrette, Théophile Gaillot Published: 2026 ABSTRACT Acute kidney injury (AKI) is common in hospitalized patients, and its incidence is rising sharply in intensive care units. It is associated with significant morbidity and mortality due to a profound change in its epidemiological profile - multifactorial in origin, often septic, and associated with other organ failures. The mortality rate reaches 30–50% in the most severe forms, particularly when AKI requires renal replacement therapy (RRT). Temporary RRT, when indicated, must be part of an overall therapeutic approach that also includes support for other vital functions, particularly cardio-circulatory, respiratory and nutritional. The current diversification of these methods means that they must be well mastered and the potential of each of them in terms of cost, efficacy, limitations and risks must be fully understood before a therapeutic decision can be taken. This narrative review was written by an expert panel and presented to the members of the jury of the French consensus conference to help in the development of guidelines for the use of RRT in critically ill adult and pediatric patients, excluding cases related to acute intoxication. Introduction Acute kidney injury (AKI) is common in critically ill patients and is associated with significant morbidity and mortality. Approximately 5–6% of these patients develop severe AKI requiring renal replacement therapy (RRT) [[1], [2], [3], [4], [5]]. RRT remains a major challenge in intensive care unit (ICU) and its use is associated with high mortality rate, which can reach 30–50% [2,6]. In adult patients, the diversification of RRT modalities requires a thorough understanding of each technique, including their respective costs, efficacy, limitations, and risks, which are essential for informed therapeutic decision-making. In pediatric intensive care, RRT differs significantly from adult practice due to the wide variability in patient size, evolving physiology, and distinct disease profiles. The French Intensive Care Society (Société de Réanimation de Langue française, SRLF) wished to update the latest french guidelines on RRT, which dated back ten years. It called upon the SRLF’s reference and evaluation committee to organize these recommendations in the form of a consensus conference. A jury panel, neutral on the subject, was selected to draft these recommendations, based on the work of adult and pediatric experts, whose position papers were presented during a public meeting. The SRLF/GFRUP ((Groupe Francophone de Réanimation et d’Urgence Pédiatrique) consensus conference on the RRT in ICU was published in 2025 in AIC journal [7]. The committee members defined seven generic questions, each of which was divided into several sub-questions. Two experts were appointed for each generic question, one with experience in the field of adult patients and the other in pediatric intensive care. The experts analyzed the literature and drafted a text for the panel members responsible for discussing their assigned question, including the most recent scientific data, as well as their opinions and arguments. This document summarizes the arguments presented by the experts at the consensus conference. Its purpose is to clarify the basis for the guidelines resulting from the consensus conference, highlighting areas of certainty and emphasizing uncertainties that could be the subject of future research. Indications and timing of RRT Adults Clinical practice typically distinguishes between urgent and non-urgent indications for RRT initiation. Urgent indications -usually life-threatening conditions- include refractory severe hyperkalemia, refractory severe metabolic acidosis, pulmonary edema refractory to diuretic administration and serious complications of hyperuremia (i.e. pericarditis) [4]. It is difficult to define precise clinical and biological threshold values for initiating RRT, as very few randomized controlled trials (RCTs) have addressed this issue. Table 1 summarizes these threshold values in the main RCTs [[8], [9], [10], [11], [12], [13]].Table 1Comparison of randomized controlled trials on the timing of Renal Replacement Therapy (RRT) in acute kidney injury.Table 1AKIKI [8]IDEAL-ICU [10]STARRT-AKI [9]ELAIN [17]AKIKI 2 [18]PatientsICU patients receiving invasive mechanical ventilation and/or catecholamine infusionICU patients with early-stage septic shockICU patientsPost operative ICU patients (cardiac surgery 47%, abdominal surgery 34%, trauma 12%)ICU patients receiving invasive mechanical ventilation and/or catecholamine infusionNumber of centers3129168139Early RRT initiation criteriaKDIGO stage 3FAILURE stage of RIFLEKDIGO stage 2 or 3KDIGO 2KDIGO 3 oliguria for more than 72 h or serum urea > 40 mmol/lDelayed RRT initiation criteria•Serum potassium ≥6.0 mmol/L•Acute pulmonary edema due to fluid overload responsible for severe hypoxemia, despite diuretic therapy•Pure metabolic acidosis or mixed acidosis with pH ≤ 7.15•Blood urea nitrogen >112 mg /dL (Urea>40 mmol/L)•Oliguria or anuria lasted for >72 h•Serum potassium ≥6.5 mmol /L•Extravascular fluid overload that was refractory, to diuretics, with pulmonary edema•Metabolic acidosis with pH ≤ 7.15•No renal function recovery after 48 h•Serum potassium ≥6.0 mmol/L•Evidence of severe respiratory failure and clinical perception of volume overload•pH ≤ 7.20 or serum bicarbonate ≤12 mmol/L•Persistent AKI for ≥72 h after randomization.•KDIGO 3More-delayed strategy, RRT initiation until mandatory indication (noticeable hyperkalaemia or metabolic acidosis or pulmonary oedema) or until serum urea reached 50 mmol/l.Patients (n)6194882927231278Time difference between delayed and early RRT initiation (h)5244252172Percentage of patients free of RRT in the delayed group (%)4938389Number of RRT-free days: 12 days (IQR 0–25) in the delayed 10 days (IQR 0–24) in the more-delayed (p = 0·93).Survival (primary outcome)Day 60 mortality 150/311 vs 153/308 (p = 0.79)Day 90 mortality 138/246 vs 128/242 (p = 0.38)Day 90 mortality 643/1465 vs 639/1462 (p = 0.92)Day 90 mortality Early 44/112 vs delayed 65/119 (p = 0.03)Day 60 mortality 60/137 delayed vs 77/141 more-delayed (p = 0·071). Multivariable analysis, risk factors associated with 60-day mortality: more-delayed strategy (HR 1·65; 95% CI 1·09–2·50, p = 0·018),Renal function recovery•Earlier adequate diuresis resumption in the delayed group (P < 0.001)•RRT dependence at day 60:•2% in the early group and 5% in the delayed; p = 0.12•RRT dependence at day 28: 13% in the early group and 12% in the delayed; p = 0.89•RRT dependence at day 90: 10 % in the early group and 6% in the delayed; relative risk: 1.74; 95% CI, 1.24 to 2.43•RRT dependence at day 90:•Early 11.8%•Late 14.8%•P = 0.62RRT dependence at day 60: Delayed 4% More delayed 2% P = 0.62 Table 1Comparison of randomized controlled trials on the timing of Renal Replacement Therapy (RRT) in acute kidney injury.Table 1AKIKI [8]IDEAL-ICU [10]STARRT-AKI [9]ELAIN [17]AKIKI 2 [18]PatientsICU patients receiving invasive mechanical ventilation and/or catecholamine infusionICU patients with early-stage septic shockICU patientsPost operative ICU patients (cardiac surgery 47%, abdominal surgery 34%, trauma 12%)ICU patients receiving invasive mechanical ventilation and/or catecholamine infusionNumber of centers3129168139Early RRT initiation criteriaKDIGO stage 3FAILURE stage of RIFLEKDIGO stage 2 or 3KDIGO 2KDIGO 3 oliguria for more than 72 h or serum urea > 40 mmol/lDelayed RRT initiation criteria•Serum potassium ≥6.0 mmol/L•Acute pulmonary edema due to fluid overload responsible for severe hypoxemia, despite diuretic therapy•Pure metabolic acidosis or mixed acidosis with pH ≤ 7.15•Blood urea nitrogen >112 mg /dL (Urea>40 mmol/L)•Oliguria or anuria lasted for >72 h•Serum potassium ≥6.5 mmol /L•Extravascular fluid overload that was refractory, to diuretics, with pulmonary edema•Metabolic acidosis with pH ≤ 7.15•No renal function recovery after 48 h•Serum potassium ≥6.0 mmol/L•Evidence of severe respiratory failure and clinical perception of volume overload•pH ≤ 7.20 or serum bicarbonate ≤12 mmol/L•Persistent AKI for ≥72 h after randomization.•KDIGO 3More-delayed strategy, RRT initiation until mandatory indication (noticeable hyperkalaemia or metabolic acidosis or pulmonary oedema) or until serum urea reached 50 mmol/l.Patients (n)6194882927231278Time difference between delayed and early RRT initiation (h)5244252172Percentage of patients free of RRT in the delayed group (%)4938389Number of RRT-free days: 12 days (IQR 0–25) in the delayed 10 days (IQR 0–24) in the more-delayed (p = 0·93).Survival (primary outcome)Day 60 mortality 150/311 vs 153/308 (p = 0.79)Day 90 mortality 138/246 vs 128/242 (p = 0.38)Day 90 mortality 643/1465 vs 639/1462 (p = 0.92)Day 90 mortality Early 44/112 vs delayed 65/119 (p = 0.03)Day 60 mortality 60/137 delayed vs 77/141 more-delayed (p = 0·071). Multivariable analysis, risk factors associated with 60-day mortality: more-delayed strategy (HR 1·65; 95% CI 1·09–2·50, p = 0·018),Renal function recovery•Earlier adequate diuresis resumption in the delayed group (P < 0.001)•RRT dependence at day 60:•2% in the early group and 5% in the delayed; p = 0.12•RRT dependence at day 28: 13% in the early group and 12% in the delayed; p = 0.89•RRT dependence at day 90: 10 % in the early group and 6% in the delayed; relative risk: 1.74; 95% CI, 1.24 to 2.43•RRT dependence at day 90:•Early 11.8%•Late 14.8%•P = 0.62RRT dependence at day 60: Delayed 4% More delayed 2% P = 0.62 Comparison of randomized controlled trials on the timing of Renal Replacement Therapy (RRT) in acute kidney injury. •Serum potassium ≥6.0 mmol/L Serum potassium ≥6.0 mmol/L •Acute pulmonary edema due to fluid overload responsible for severe hypoxemia, despite diuretic therapy Acute pulmonary edema due to fluid overload responsible for severe hypoxemia, despite diuretic therapy •Pure metabolic acidosis or mixed acidosis with pH ≤ 7.15 Pure metabolic acidosis or mixed acidosis with pH ≤ 7.15 •Blood urea nitrogen >112 mg /dL (Urea>40 mmol/L) Blood urea nitrogen >112 mg /dL (Urea>40 mmol/L) •Oliguria or anuria lasted for >72 h Oliguria or anuria lasted for >72 h •Serum potassium ≥6.5 mmol /L Serum potassium ≥6.5 mmol /L •Extravascular fluid overload that was refractory, to diuretics, with pulmonary edema Extravascular fluid overload that was refractory, to diuretics, with pulmonary edema •Metabolic acidosis with pH ≤ 7.15 Metabolic acidosis with pH ≤ 7.15 •No renal function recovery after 48 h No renal function recovery after 48 h •Serum potassium ≥6.0 mmol/L Serum potassium ≥6.0 mmol/L •Evidence of severe respiratory failure and clinical perception of volume overload Evidence of severe respiratory failure and clinical perception of volume overload •pH ≤ 7.20 or serum bicarbonate ≤12 mmol/L pH ≤ 7.20 or serum bicarbonate ≤12 mmol/L •Persistent AKI for ≥72 h after randomization. Persistent AKI for ≥72 h after randomization. •KDIGO 3 KDIGO 3 •Earlier adequate diuresis resumption in the delayed group (P < 0.001) Earlier adequate diuresis resumption in the delayed group (P < 0.001) •RRT dependence at day 60: RRT dependence at day 60: •2% in the early group and 5% in the delayed; p = 0.12 2% in the early group and 5% in the delayed; p = 0.12 •RRT dependence at day 28: 13% in the early group and 12% in the delayed; p = 0.89 RRT dependence at day 28: 13% in the early group and 12% in the delayed; p = 0.89 •RRT dependence at day 90: 10 % in the early group and 6% in the delayed; relative risk: 1.74; 95% CI, 1.24 to 2.43 RRT dependence at day 90: 10 % in the early group and 6% in the delayed; relative risk: 1.74; 95% CI, 1.24 to 2.43 •RRT dependence at day 90: RRT dependence at day 90: •Early 11.8% Early 11.8% •Late 14.8% Late 14.8% •P = 0.62 P = 0.62 The severity of hyperkalemia is characterized more by its cardiac effects (significant conduction abnormalities) than by a specific value. A serum potassium concentration >6.0 or >6.5 mmol/L is often used, particularly in high-risk situations such as rhabdomyolysis or tumor lysis syndrome. In these cases, failure to respond to medical interventions may require urgent RRT [4]. The severity of metabolic acidosis was defined in the BICAR-ICU multicenter randomized trial [14] as a pH ≤ 7.20, PaCO2 ≤ 45 mmHg, and serum bicarbonate <20 mmol/L. Four hundred patients with severe metabolic acidosis and total Sequential Organ Failure Assessment score ≥4 or an arterial lactatemia ≥2 mmol/l were randomly assigned to receive either no bicarbonate (control arm) or bicarbonate infusion (bicarbonate arm) to maintain the arterial pH above 7.30. Randomization criteria were stratified among three pre-specified strata: age with a threshold of 65, presence of sepsis, and presence of acute kidney injury (Acute Kidney Injury Network Score of 2 or 3). The intervention group (195 patients) received one or more infusions of 4.2% sodium bicarbonate. Survival at D28 were 46% [95% CI: 40–54] in the control group and 55% [95% CI: 49–63] in the bicarbonate group (194 patients), p = 0.08. In the pre-specified AKI stratum of patients, survival at D28 was higher (37% [95% CI: 28–48] vs. 54% [[45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65]]; p = 0.028) in the bicarbonate arm. Renal-replacement therapy was more frequently used in the control arm (100/194 patients, 52%) as compared to the bicarbonate arm (68/195 patients, 35%); p < 0.01. In the more recent BICARICU-2 open-label clinical trial involving 640 critically ill patients with a total SOFA score >4 or an arterial lactatemia ≥2 mmol/l within 48 h after ICU admission and with the following criteria within 6 h before enrollment: pH ≤ 7.20, bicarbonatemia ≤20 mEq/L, PaCO2 ≤ 45 mm Hg, and stage 2 or 3 KDIGO AKI, sodium bicarbonate treatment did not significantly reduce 90-day mortality (195/ 314 patients, 62.1% in the bicarbonate group vs 193/313, 61.7% in the control group; p = 0.91), but did reduce the use of RRT (recommended for patients meeting at least 2 of the following criteria 24 h after enrollment: urine output less than 0.3 mL/kg/h during 24 h, pH ≤ 7.20 despite resuscitation, and kalemia >6.5 mEq/L): 35% vs. 50%, absolute difference −15.5%, 95% CI −23.1% to −7.8% [15]. The degree of fluid accumulation has been strongly correlated with mortality in observational data [16], but no causal link has been established. The single-center randomized ELAIN trial [17] involved 231 critically ill postcardiac surgery patients with AKI KDIGO stage 2 and plasma neutrophil gelatinase–associated lipocalin level >150 ng/mL and compared early (within 8 h of diagnosis of KDIGO stage 2; n = 112) or delayed (within 12 h of stage 3 AKI or no RRT; n = 119) initiation of RRT. Its showed that early initiation of RRT significantly reduced 90-day mortality (39.3%) compared with delayed initiation of RRT (54.7%; hazard ratio [HR], 0.66 [95%CI, 0.45 to 0.97]; difference, −15.4% [95%CI, −28.1% to −2.6%]; p = 0.03). In fact, patients with refractory fluid overload and pulmonary edema actually benefited from early initiation of RRT in this study. Indeed, the occurrence of pulmonary edema is often a determining factor in the initiation of RRT in most RCTs. However, its severity is rarely determined by the degree of hypoxemia, given that a PaO₂/FiO₂ ratio < 200 has been used as a threshold in the STARRT-AKI trial [9]. Plasma urea levels were evaluated in the open-label, multicenter, prospective, randomized, controlled AKIKI 2 trial [18]. Its included 278 critically ill patients with KDIGO stage 3 AKI, defined as having oliguria for more than 72 h or a serum urea level >40 mmol/L. Patients were then randomly assigned (1:1) either to a delayed strategy group in which RRT was initiated immediately after randomization, or to a delayed strategy group in which the initiation of RRT was postponed until the onset of an urgent indication (severe hyperkalemia, metabolic acidosis, or pulmonary edema) or until the serum urea level reached 50 mmol/L. Multivariate analysis revealed a higher 60-day mortality rate in the “more delayed” group (OR 2.16 [95% CI, 1.17–4.01], p = 0.014), suggesting that RRT should be initiated when serum urea level exceeds 40 mmol/L in the absence of signs of renal recovery (i.e., resumption of diuresis). A secondary analysis showed that the more delayed strategy was associated with later awakening from coma, suggesting that hyperuremia or accumulation of uremic toxins contribute to the multifactorial encephalopathy seen in critically ill patients [19]. There is no threshold value for serum creatinine that determines the need for RRT. At last, 3 large RCTs (4,034 patients) [[8], [9], [10]] evaluated early versus delayed RRT initiation strategies in patients without urgent RRT indications (life-threatening conditions). In the early strategy, RRT was initiated within hours of randomization (i.e., after the onset of AKI severity criteria as defined by the KDIGO or AKIN classifications) [20], whereas in the delayed strategy, it was postponed until predefined criteria were met (in particular, life-threatening conditions and the persistence of oligo-anuria or AKI). Mortality at D60 or D90 did not differ between treatment groups in any of the trials. A meta-analysis of individual patient data, including nine studies, showed similar results [21]. However, in the AKIKI 2 trial [18], a more delayed strategy without a time limit for the duration of oligo-anuria was associated with higher 60-day mortality. Delayed strategies in the AKIKI [8] and IDEAL-ICU [10] were associated with a greater number of days without RRT (at day 28) compared to the early group, but STARRT-AKI found no significant difference at D90 between the two strategies [9]. A faster recovery of adequate urine output and an earlier spontaneous decrease in plasma creatinine were observed in the delayed treatment group [8], while the early treatment group was more often dependent on RRT at day 90 (RR 1.74; 1.24–2.43) [9]. Pediatrics The 2016 French expert recommendations [22] suggested that RRT should be initiated in children with a fluid overload of more than 20%. Two meta-analyses reported a statistical association between fluid overload and mortality [23,24]. However, causality remains to be proven due to the lack of RCTs. A recent European study involving 340 ICU pediatric patients revealed that an average fluid overload of only 4.4% was sufficient to initiate RRT [25]. To date, there is no absolute threshold for RRT initiation. Selection criteria for different RRT modalities Adults Intermittent and continuous RRT are usually opposed [26,27]. The different techniques should be compared based on the speed of metabolic control, the risk of osmotic shift and the rate of fluid removal. Low-intensity modalities cause gradual osmotic changes, allow for low clearance rates, and therefore require continuous treatment in order to achieve metabolic goals. Hybrid techniques, typically pIRRT [28] combine the advantages of intermittent techniques without the disadvantages of continuous techniques (Table 2).Table 2Theoretical advantages et disadvantages of the most common Renal Replacement Therapy modalities.Table 2AdvantagesDisadvantagesIntermittent high clearance extracorporeal RRT (conventional hemodialysis, high volume hemodiafiltration)•Rapid control of metabolic disorders and poisoning•Possible anticoagulant free session•More patients treated•Lesser cost of disposables•Transition from continuous technics•Patient mobilization•Rapid osmotic shifts•Rapid fluid removal•Complex implementation in the absence chronic dialysis center•Low flexibility if nephrology team requiredProlonged low clearance extracorporeal RRT (pIRRT, hybrid methods)•Low risk of osmotic shift•Slow fluid removal•Possible anticoagulant free session if < 12 h•Transition from continuous technics•Patient mobilization•Lesser cost of disposables•More patients treated•Possible with hemodialysis or continuous RRT machines•Best settings under evaluated•Few data on regional citrate anticoagulation•Few data on drugs management•Risk of hypokalemia and hypophosphatemiaContinuous extracorporeal RRT (CRRT)•Low risk of osmotic shift•Slow fluid removal•Control of fluid balance•Can be combined with other extracorporeal supports•Management of anticoagulants•High cost•Patient immobilization•Risk of hypokalemia and hypophosphatemia•Risk of hypothermiaContinuous peritoneal dialysis•Simplicity•Low cost•Low risk of osmotic shift•Slow fluid removal•Anticoagulant free•Insufficient dialysis dose in hypercatabolic state and in obese patients•Limited and uncontrolled rate of fluid removal•Increase in intra-abdominal pressure Table 2Theoretical advantages et disadvantages of the most common Renal Replacement Therapy modalities.Table 2AdvantagesDisadvantagesIntermittent high clearance extracorporeal RRT (conventional hemodialysis, high volume hemodiafiltration)•Rapid control of metabolic disorders and poisoning•Possible anticoagulant free session•More patients treated•Lesser cost of disposables•Transition from continuous technics•Patient mobilization•Rapid osmotic shifts•Rapid fluid removal•Complex implementation in the absence chronic dialysis center•Low flexibility if nephrology team requiredProlonged low clearance extracorporeal RRT (pIRRT, hybrid methods)•Low risk of osmotic shift•Slow fluid removal•Possible anticoagulant free session if < 12 h•Transition from continuous technics•Patient mobilization•Lesser cost of disposables•More patients treated•Possible with hemodialysis or continuous RRT machines•Best settings under evaluated•Few data on regional citrate anticoagulation•Few data on drugs management•Risk of hypokalemia and hypophosphatemiaContinuous extracorporeal RRT (CRRT)•Low risk of osmotic shift•Slow fluid removal•Control of fluid balance•Can be combined with other extracorporeal supports•Management of anticoagulants•High cost•Patient immobilization•Risk of hypokalemia and hypophosphatemia•Risk of hypothermiaContinuous peritoneal dialysis•Simplicity•Low cost•Low risk of osmotic shift•Slow fluid removal•Anticoagulant free•Insufficient dialysis dose in hypercatabolic state and in obese patients•Limited and uncontrolled rate of fluid removal•Increase in intra-abdominal pressure Theoretical advantages et disadvantages of the most common Renal Replacement Therapy modalities. •Rapid control of metabolic disorders and poisoning Rapid control of metabolic disorders and poisoning •Possible anticoagulant free session Possible anticoagulant free session •More patients treated More patients treated •Lesser cost of disposables Lesser cost of disposables •Transition from continuous technics Transition from continuous technics •Patient mobilization Patient mobilization •Rapid osmotic shifts Rapid osmotic shifts •Rapid fluid removal Rapid fluid removal •Complex implementation in the absence chronic dialysis center Complex implementation in the absence chronic dialysis center •Low flexibility if nephrology team required Low flexibility if nephrology team required •Low risk of osmotic shift Low risk of osmotic shift •Slow fluid removal Slow fluid removal •Possible anticoagulant free session if < 12 h Possible anticoagulant free session if < 12 h •Transition from continuous technics Transition from continuous technics •Patient mobilization Patient mobilization •Lesser cost of disposables Lesser cost of disposables •More patients treated More patients treated •Possible with hemodialysis or continuous RRT machines Possible with hemodialysis or continuous RRT machines •Best settings under evaluated Best settings under evaluated •Few data on regional citrate anticoagulation Few data on regional citrate anticoagulation •Few data on drugs management Few data on drugs management •Risk of hypokalemia and hypophosphatemia Risk of hypokalemia and hypophosphatemia •Low risk of osmotic shift Low risk of osmotic shift •Slow fluid removal Slow fluid removal •Control of fluid balance Control of fluid balance •Can be combined with other extracorporeal supports Can be combined with other extracorporeal supports •Management of anticoagulants Management of anticoagulants •High cost High cost •Patient immobilization Patient immobilization •Risk of hypokalemia and hypophosphatemia Risk of hypokalemia and hypophosphatemia •Risk of hypothermia Risk of hypothermia •Simplicity Simplicity •Low cost Low cost •Low risk of osmotic shift Low risk of osmotic shift •Slow fluid removal Slow fluid removal •Anticoagulant free Anticoagulant free •Insufficient dialysis dose in hypercatabolic state and in obese patients Insufficient dialysis dose in hypercatabolic state and in obese patients •Limited and uncontrolled rate of fluid removal Limited and uncontrolled rate of fluid removal •Increase in intra-abdominal pressure Increase in intra-abdominal pressure RRT modality and impact on hemodynamic tolerance Intradialytic hypotension is mainly related to rapid osmotic shifts and high rate of fluid removal [29]. Four small RCTs were conducted to specifically evaluate intradialytic hemodynamic tolerance [[30], [31], [32], [33]] and reported similar tolerance between continuous and pIRRT (SLED). A reduction in the incidence of hypotension over 24 h was observed in 30 septic patients undergoing CRRT compared to IRRT [31]. In addition, 80 randomly assigned postoperative patients showed significantly better tolerance (secondary endpoint) with CRRT compared to IRRT, despite a higher net volume removed in the CRRT group [34]. Nevertheless, meta-analyses concluded that there was no superiority in terms of hemodynamic tolerance of continuous compared to intermittent techniques, even in hemodynamically unstable patients [35,36]. RRT modality and impact on dialysis disequilibrium syndrome (DDS) RRT can lead to changes in intracranial pressure, causing cerebral oedema, whether continuous or intermittent; but rapid osmotic changes and a high bicarbonate load increase the risk. In patients with acute brain injury, a limited number of observations reported more frequent DDS with conventional IRRT compared to CRRT [37]. When IRRT is used, an increase in sodium dialysate concentration is advised. Hybrid modalities could address the objectives of patients with acute brain injury, but conclusive evidence is lacking [34]. RRT modality and impact on survival An initial study demonstrated a significantly higher mortality rate with CRRT compared to IRRT [38], but subsequent RCTs have failed to demonstrate a significant difference in mortality between the two modalities [13]. Indeed, a network meta-analysis including 30 RCTs [35] concluded that there was no difference in mortality between CRRT and IRRT (RR, 1.04; 95% CI, 0.93–1.18; low certainty), and potentially higher in-hospital mortality with CRRT compared to pIRRT (RR, 1.06; 95% CI, 0.85–1.33; low certainty). Three post hoc analyses of large RCTs exploring the timing or dose of RRT evaluated patient outcomes based on the first modality applied [[39], [40], [41]]. Better survival with IRRT (all modalities combined) was observed compared to CRRT in one study, with a significant difference in the less severe patients [39], while others reported similar survival with both modalities [40,41]. The results of the “ICRAKI” study—a multicenter, randomized, controlled clinical trial currently underway, designed to assess whether conventional intermittent hemodialysis is non-inferior to CRRT—will shed new insight into the debate regarding RRT modalities [42]. Yet, CRRT has not been compared with pIRRT in RCTs, and further research is needed in this area. RRT modality and impact on renal recovery Observational studies report a greater dependence on dialysis in patients managed by IRRT, compared to CRRT, but most of these are inconclusive [43]. Three post hoc analyses of large RCTs assessing the risk of dialysis dependence have produced contradictory results [[39], [40], [41]]. Place of peritoneal dialysis Since the development of extracorporeal CRRT, peritoneal dialysis (PD) has been gradually abandoned for critically ill patients. However, it remains a reliable alternative due to its simplicity in low-income countries or during times of crisis [44,45]. Despite technical improvements, a number of limitations remain, including insufficient dialysis dose in hypercatabolic and obese patients, and insufficient ultrafiltration rates. Additional RRT sessions were required in 25%–50% of patients when PD did not achieve the desired goal. During PD, the average serum urea level did not decrease significantly in most studies [46,47]. Also, the consequences of PD on respiratory mechanics and its feasibility in the prone position must be evaluated. The optimal RRT modality for use in critically ill adult patients is far from established and can only be clearly identified after additional randomized studies. Pediatrics To date, no RCTs have compared CRRT with IRRT or pIRRT for managing AKI [48,49]. Current practice is based on observational data and expert opinion [25,[50], [51], [52], [53]]. CRRT is preferred over IRRT. During the initial period of shock, the increased risk of instability requires careful monitoring, particularly in smaller and more severely ill children, in whom the extracorporeal volume may exceed 10–15% of blood volume [[54], [55], [56], [57]]. Currently available RRT systems dedicated to this specific population (weight <10 kg) offer good hemodynamic tolerance [[58], [59], [60], [61]]. Two recent studies have compared mortality with historical cohorts from the 2001–2005 American “ppCRRT” registry and showed improved survival [62,63]. However, this could be due to advancements in patient care rather than the RRT techniques themselves. Dialysis dependence at day 90 is rarely mentioned in the PICU literature. In the WE-ROCK study [64], dialysis dependence at D90 was observed in 9.4% of patients and was associated with late RRT initiation and cardiac comorbidities. The results of the Italian registry dedicated to low-birth-weight children [58], showing incomplete renal recovery in a significant proportion of survivors, raised ethical considerations related to long-term quality of life [65]. Despite the lack of RCTs in children, PD remains a valuable option, especially in cases of difficult vascular access, such as in newborns, after cardiac surgery, and when other approaches are contraindicated [25,49,50,53,66]. Fourteen percent of children enrolled in the European AKI registry were treated with PD [25,49].To maximize the effectiveness of PD, international guidelines provide clear recommendations [45,67]. The choice between different types of RRT seems to be influenced by economic factors. In Europe, countries with lower gross domestic product are more likely to use PD as first-line treatment [66]. In low-income countries, the cost of equipment and specialized personnel remains the major obstacle [50]. Dialysis dose Adults The relationship between dialysis dose and survival remains widely debated. Continuous renal replacement therapy In the 2000s, two single-center studies suggested improved survival in patients treated by a CRRT dose ≥35 ml/kg/h compared to 25 ml/kg/h [68,69]. However, other prospective studies have shown no real benefit with a higher RRT dose. In 106 post-cardiac surgery AKI patients randomized into 3 CRRT groups, i.e. early high volume (48 ml/kg/h), early low volume (20 ml/kg/h) and late low volume, no significant differences were found between groups in terms of mortality and recovery of renal function [70]. Tolwani et al. [71] also reported no significant clinical difference in a study involving 200 patients divided into 2 groups of CRRT doses: 20 versus 35 ml/kg/h. Most importantly, the multicenter study published by the Acute Renal Failure Trial Network (ATN) randomly compared two dialysis dose in 1124 patients from 27 centers [12]. The "intensive dialysis" group included a predilutional CVVHDF at 35 ml/kg/h, or a daily IHD (6 sessions/7days with a Kt/v/session of 1.2−1.4). The "less intensive" dialysis group benefited from a predilutional CVVHDF at 20 ml/kg/h, or an IHD at 3 sessions/7days with a Kt/V/session of 1.2−1.4. There was no significant difference in either mortality at D60 or recovery of renal function between the 2 groups (53.6 vs. 51.5%, intensive vs. less intensive). A similar study, carried out exclusively with CVVHDF, with 747 patients randomized to 40 ml/kg/h or 25 ml/kg/h led to the same conclusions, i.e. similar 90-day mortality, but with a significantly higher frequency of hypophosphatemia in the high dose group [11]. A 2017 Cochrane review concluded that there was no significant benefit on mortality at 30 and 90 days and kidney recovery function from intensifying the dialysis dose [72]. The more recent Meta-analysis of 7 RCT, in which 1884 patients (51.2%) were randomized to standard intensity (prescribed dose: 20−25 ml/kg/h) and 1798 (48.8%) to higher intensity RRT (prescribed dose: 35−48 ml/kg/h), concluded that all-cause mortality at 28, 60 and 90 days did not differ between groups [73]. More high intensity RRT patients remained dialysis dependent at D28 (RR1.16 95% CI 1.00–1.33, p = 0.05). In another meta-analysis including 7 RCTs and one prospective study, 2970 patients were treated by hemofiltration with low dose (20–36 ml/kg/h) or high dose (35–85 ml/kg/h) [74]. The pooled results showed no significant difference in ICU, hospital and D90 mortalities. A systematic review of 4 RCTs, including 470 septic patients, evaluated the effects of high-volume compared with standard-volume HF [75]. No differences regarding to D28 and ICU mortalities or hospital lenght-of-stay and renal function recovery were observed. Hypokalemia and hypophosphatemia were more commonly observed in high volume patients (Table 3) [76].Table 3Studies comparing dialysis dose during continuous renal replacement therapy (CRRT).Table 3First authorYearNo of patientsstudyComparatorResultsRonco [67]2000425randomizedCVVHF post-dilution: 20 vs 35 vs 45 ml/kg/hSignificantly higher survival with 35−45 vs 20 ml/kg/h, surgical patientsBouman [69]2002106randomizedCVVHF post-dilution, timing and doses in ml/kg/h: early 48, early 20, delayed 20.no difference in mortality or Renal Recovery (post cardiac surgery)Saudan [68]2006206randomizedCVVHF (25 ml/kg/h) vs CVVHDF (addition of 18 ml/kg/h)Better survival at day 28: 39% vs 59% with the higher dose, p = 0.03Tolwani [70]2008157randomizedCVVHDF predilution. doses in ml/kg/h, prescribed 20 vs 35 (delivered 17 vs 29)No significant difference in day 30 mortality or renal function recoveryVA/NIH Acute Renal Failure Trial Network [12]20081124randomizedCVVHDF 35 vs 20 ml/kg/h (RRT initiation not formalized)No significant difference in day 60 mortality or renal function recoveryRENAL Replacement Therapy Study Investigators [11]2009747randomizedCVVHDF 40 vs 25 ml/kg/hNo significant difference in day 90 mortalityUchino [76]20131349retrospective7 subclasses of CRRT doses, from <10 to >35 ml/kg/hNo significant difference in hospital mortality between the 7 classesClark [74]20144704 RCTs/ Meta-AHigh Volume CVVHF vs standard, in septic AKINo significant difference in day 28 mortality or renal function recoveryBorthwick [71]2017200Meta-AIntensive vs StandardNo significant difference in day 30 and day 90 mortality or renal function recoveryLi [73]20172970Meta-A (8 studies)CVVHF 20−36 vs 35−85 ml/kg/hNo significant difference in ICU, Hospital or day 90 mortalityWang [75]20183682Meta-A (7 RCT)CRRT (5), IRRT (1), both (1): 20−25 vs 35−48 ml/kg/hNo significant difference in day 28 or day 90 mortality or RRT dependanceCVVH/D/F: continuous venovenous hemo/dia/filtration; IRRT: Intermittent Renal Replacement Therapy. Table 3Studies comparing dialysis dose during continuous renal replacement therapy (CRRT).Table 3First authorYearNo of patientsstudyComparatorResultsRonco [67]2000425randomizedCVVHF post-dilution: 20 vs 35 vs 45 ml/kg/hSignificantly higher survival with 35−45 vs 20 ml/kg/h, surgical patientsBouman [69]2002106randomizedCVVHF post-dilution, timing and doses in ml/kg/h: early 48, early 20, delayed 20.no difference in mortality or Renal Recovery (post cardiac surgery)Saudan [68]2006206randomizedCVVHF (25 ml/kg/h) vs CVVHDF (addition of 18 ml/kg/h)Better survival at day 28: 39% vs 59% with the higher dose, p = 0.03Tolwani [70]2008157randomizedCVVHDF predilution. doses in ml/kg/h, prescribed 20 vs 35 (delivered 17 vs 29)No significant difference in day 30 mortality or renal function recoveryVA/NIH Acute Renal Failure Trial Network [12]20081124randomizedCVVHDF 35 vs 20 ml/kg/h (RRT initiation not formalized)No significant difference in day 60 mortality or renal function recoveryRENAL Replacement Therapy Study Investigators [11]2009747randomizedCVVHDF 40 vs 25 ml/kg/hNo significant difference in day 90 mortalityUchino [76]20131349retrospective7 subclasses of CRRT doses, from <10 to >35 ml/kg/hNo significant difference in hospital mortality between the 7 classesClark [74]20144704 RCTs/ Meta-AHigh Volume CVVHF vs standard, in septic AKINo significant difference in day 28 mortality or renal function recoveryBorthwick [71]2017200Meta-AIntensive vs StandardNo significant difference in day 30 and day 90 mortality or renal function recoveryLi [73]20172970Meta-A (8 studies)CVVHF 20−36 vs 35−85 ml/kg/hNo significant difference in ICU, Hospital or day 90 mortalityWang [75]20183682Meta-A (7 RCT)CRRT (5), IRRT (1), both (1): 20−25 vs 35−48 ml/kg/hNo significant difference in day 28 or day 90 mortality or RRT dependanceCVVH/D/F: continuous venovenous hemo/dia/filtration; IRRT: Intermittent Renal Replacement Therapy. Studies comparing dialysis dose during continuous renal replacement therapy (CRRT). CVVH/D/F: continuous venovenous hemo/dia/filtration; IRRT: Intermittent Renal Replacement Therapy. There are few data comparing a lower dialysis dose of 15 ml/kg/h with a dose >25 ml/kg/h. A multicenter retrospective study (1349 patients) was conducted in 14 Japanese ICUs and 54 ICUs in other countries [77]. CRRT intensity was classified into 7 subclasses: <10, 10–15, 15–20, 20−25, 25–30, 30–35, and >35 ml/kg/h. Severe AKI patients treated with low-intensity CRRT (14.3 ml/kg/h) did not have worse outcome compared to those with currently standard intensity (20.4 ml/kg/h). A recently published single-center retrospective cohort study (494 patients) showed however an increase in 90-day mortality in the below-median (<13.2 ml/kg/h) dose group (HR, 1.73 [95% CI, 1.19–2.51], P = 0.004) [78]. Intermittent hemodialysis In a single center study, 106 AKI patients were randomized to daily or alternate day IRRT with weekly dialysis dose Kt/V of 5.8 ± 0.4 vs. 3 ± 0.6 respectively [79]. Mortality was significantly reduced (28% vs. 46%, p = 0.01), as the time to recovery of renal function in the daily IRRT group (9 ± 2 vs. 16 ± 6 days, p = 0.001). In the multicenter ATN study there was no significant difference in either mortality at D60 or recovery of renal function between the "intensive dialysis" group (6 IRRT sessions/7d) and the "less intensive" group (3 sessions/7d); Kt/V/session of 1.2−1.4 [12]. The number of intradialytic hypotensive episodes and the Kt/V delivered per session (1.3) were relatively high in the less-intensive group. A post hoc analysis of 246 patients from the ATN study treated solely with IRRT and randomized to receive 3 or 6 sessions per week with a delivered Kt/V ≥ 1.2/session showed that low intensity group was associated with higher number of RRT-free days (mean difference 2.5 days, 95% CI 4.79−0.27days, p = 0.028) and higher likelihood for renal recovery through D28 (OR 0.49, 95% CI: 0.28−0.87, p = 0.016) [80]. Furthermore, 156 AKI patients randomly assigned to receive standard [dosed to maintain uremia at 20−25 mmol/L] or intensified dialysis [uremia<15 mmol/L] did not showed any differences for survival by D14 (70.4% vs 70.7%) or D28 (55.6% vs 61.3%), or for renal recovery by D28 (60.0% vs 63.0%) [81]. The main studies are shown in Table 4.Table 4Studies comparing dialysis dose in intermittent renal replacement therapy (IRRT).Table 4First authordatenstudyComparatorResultsSchiffl [78]2002160randomizedIRRT 1/day vs 1/2 days, Kt/V per week 5.8 vs 3Significantly lower mortality and higher renal recovery with daily IRRT, (higher ultrafiltration rates and more hypotension with 1 IRRT/2 days)VA/NIH Acute Renal Failure Trial Network [12]20081124randomizedIRRT/pIRRT (Kt/V 1.2−1.4): 1 per day vs 3 per weekNo significant difference for mortality at day 60 and for renal recovery; more hypotension with less frequent IRRTFaulhaber [80]2009156randomizedStandard vs intensive according to plasma urea post IRRTNo significant difference for mortality at day 14, or 28 and for renal recoveryVijayan [79]2017246Ancillary ATNIRRT 3 per week vs 6 per weekNo significant difference for mortality, improving renal recovery with 3/week but higher ultrafiltration ratespIRRT: prolonged intermittent renal replacement therapy. Table 4Studies comparing dialysis dose in intermittent renal replacement therapy (IRRT).Table 4First authordatenstudyComparatorResultsSchiffl [78]2002160randomizedIRRT 1/day vs 1/2 days, Kt/V per week 5.8 vs 3Significantly lower mortality and higher renal recovery with daily IRRT, (higher ultrafiltration rates and more hypotension with 1 IRRT/2 days)VA/NIH Acute Renal Failure Trial Network [12]20081124randomizedIRRT/pIRRT (Kt/V 1.2−1.4): 1 per day vs 3 per weekNo significant difference for mortality at day 60 and for renal recovery; more hypotension with less frequent IRRTFaulhaber [80]2009156randomizedStandard vs intensive according to plasma urea post IRRTNo significant difference for mortality at day 14, or 28 and for renal recoveryVijayan [79]2017246Ancillary ATNIRRT 3 per week vs 6 per weekNo significant difference for mortality, improving renal recovery with 3/week but higher ultrafiltration ratespIRRT: prolonged intermittent renal replacement therapy. Studies comparing dialysis dose in intermittent renal replacement therapy (IRRT). pIRRT: prolonged intermittent renal replacement therapy. Conclusion During CRRT, a dialysis dose below 20−25 ml/kg/h, and even as low as 20 ml/kg/h, does not worsen mortality at D28, D90, or the recovery time of renal function compared to a higher dose. However, there are insufficient data to support the use of a dose of CRRT < 20 ml/kg/h. In addition, recognizing that the prescribed dose is not always fully administered, a total prescribed effluent dose of 20–30 mL/kg/h would be appropriate. During IRRT, a Kt/V of 3.9/week may be recommended. The pre-dialysis serum urea level should be maintained below 30 mmol/l, with a serum urea reduction rate >60% for IRRT [82], or an average level of 20 mmol/l for CRRT. A higher dose of RRT may be indicated in cases of hyperkaliemia, severe acidemia, elevated ammonia levels, or acute cases of increased catabolism (e.g., septic shock, rhabdomyolysis, tumor lysis syndrome, etc.) [83]. Once severe metabolic disorders improve, the RRT dose should be reduced to return to the target dose. Pediatrics European guidelines suggest a CRRT dose of 30–35 mL/kg/h [84]. Several pediatric studies confirm compliance with these recommendations [50,85], reporting prescribed doses ranging from 30 to 40 mL/kg/h. A recent meta-analysis of pediatric cohort data suggested a reduction in mortality in septic children receiving CRRT doses of 50–100 mL/kg/h [86]. As IRRT is rarely used in PICUs, no studies have specifically evaluated its optimal conditions of use. Different treatment regimens can be considered, ranging from systematic daily use to use every other day, or based solely on metabolic indications [50,67,87]. Prescription, setting and monitoring of each RRT modality Adults Hemodialysis water treatment Ultrapure dialysate and a high water quality are essential for the effectiveness and biocompatibility of the extracorporeal circuit, as well as for the overall safety of the treatment [88]. No studies have compared the complications or efficacy of systems that generate dialysate from a mobile osmosis unit or a looped distribution of osmosis water. In the study by Klouche et al. [89], the use in ICU of on line-intermittent hemodiafiltration with ultrapure water proved to be feasible and safe. Intradialytic hemodynamic tolerance, automated settings, Special cases Intradialytic hypotension occurs in 30–70% of IRRT sessions and in 19–43% of CRRT sessions [89]. These hypotensive episodes have been associated with increased mortality [90,91] and reduced renal perfusion, compromising renal function recovery [92]. IRRT sessions lasting 3–4 h, with high solute clearances, can cause rapid changes in plasma osmolality and water movements facilitate intradialytic hypotension. The ultrafiltration rate, which sometimes exceeds plasma refilling, may also contribute to this phenomenon [29]. Despite the lack of evidence of the effectiveness of one or more interventions, it is suggested that the combination of a higher sodium concentration in the dialysate [93], an ultrafiltration rate adapted to the patient's hemodynamic status and inflation [94,95], and a lower dialysate temperature [96] could reduce the incidence of per-IRRT hemodynamic instability in ICU patients [93,97,98]. The use of the sodium profile of the dialysate could reduce episodes of intradialytic hypotension. In the literature, a number of adjustments to the IRRT are suggested [99]. Filters, membranes Randomized studies evaluating the survival of AKI patients according to the type of membrane used do not allow definitive conclusions to be drawn about hemodynamic stability or prognosis, except in the case of cuprophane membranes, which are associated with increased mortality [[100], [101], [102], [103], [104], [105]]. The use of membranes with a high ultrafiltration coefficient may induce backfiltration of endotoxins or microbial debris from the dialysate into the blood. Their use requires a high-quality dialysate containing less than 0.03 endotoxin units per milliliter [106]. Pediatrics More than one-third of critically ill children treated with IRRT experience intradialytic hypotension [107], which often leads to premature termination of the session and a reduction in ultrafiltration and dialysis dose [108]. Real-time blood volume monitoring to improve hemodynamic tolerance used in chronic hemodialysis [[109], [110], [111]] has been little evaluated in acute critical situations [107]. Various maneuvers such as adjusting dialysate sodium concentration, lowering the temperature of the dialysate, and pharmacological interventions (midodrine, mannitol) have not been tested in pediatrics. There are many different priming practices when connecting to the dialysis machine; the most common are those using diluted red blood cells or 5% albumin [55,58,59]. Priming with packed red blood cells carries risks of hyperkalemia and acid-base imbalances. The optimal strategy remains however to be determined [[55], [56], [57]]. Recent advances in circuit miniaturization, combined with the development of new generators, reduce the risk of poor hemodynamic tolerance [[59], [60], [61]]. The small number of dedicated pediatric devices developed by manufacturers limits the choice of membranes and generators [112,113]. Pediatric experience with the use of specialized membranes for both treatment of AKI and clearance by adsorption of cytokines or endotoxin is limited and does not allow to conclude on their usefulness [[114], [115], [116], [117], [118], [119], [120], [121], [122], [123], [124]]. RRT catheter insertion site and interdialytic locking Adults The two main insertion sites for temporary RRT catheters are the femoral and jugular veins. The subclavian site should only be used as a last resort, as it increases the risk of subclavian vein stenosis [125], which may compromise the subsequent creation of a functional arteriovenous fistula. The risk of colonization and the rate of CRBSI do not differ between femoral and jugular sites [126,127]. However, patients with a body mass index (BMI) below 24 kg/m2 may benefit more from the femoral site, whereas those with a BMI above 28 kg/m2 may be better suited to the jugular site [127]. The risk of catheter dysfunction increases progressively from the right jugular site to the femoral sites and then to the left jugular site [128,129]. The incidence of venous thrombosis associated with dialysis catheters is very low (0.5%), with no significant difference between femoral and jugular sites [127]. Between RRT sessions, an intraluminal lock solution is applied to prevent infections and catheter thrombosis, which are the main causes of catheter dysfunction. Various lock solutions include citrate, heparin, or isotonic saline, alone or in combination with antimicrobial agents. Highly concentrated citrate locks (>4%) are associated with a lower risk of catheter dysfunction than saline locks [129,130]. However, results comparing citrate and heparin locks are equivocal. Three single-center studies report fewer catheter malfunctions with 4% citrate locks than with unfractionated heparin locks with concentrations of 2,500 to 6,250 IU/mL [[131], [132], [133]]. However, another single-center RCT comparing 4% citrate lock with 5000 IU/mL heparin lock found no significant difference in the incidence of catheter thrombosis (5% vs. 2%, p = 0.14) or bleeding complications [134]. A reduction in catheter bacterial colonization was reported with the citrate lock compared with the saline or heparin lock in one study, but CRBSI rates were similar between the groups [129]. Other studies found no significant difference in catheter colonization or infection rates between citrate, heparin and saline locks [[130], [131], [132], [133], [134]]. A multicenter RCT comparing a 2-minute lock with 60% ethanol to a lock with saline after each RRT session, followed in both cases by an interdialytic lock with 100 IU/mL heparin, showed no significant difference in colonization, catheter infection, and incidence of CRBSI [135]. In ICU patients, there is no clear evidence in favor of a specific lock solution to prevent catheter-related infections. Pediatrics Studies specifically comparing thrombosis, infection, or dysfunction rates between femoral and jugular insertion in pediatric RRT remain scarce. Prolonged survival of the hemofilter has been associated with the internal jugular insertion site [51,63]. A single-center prospective study involving 55 critically ill children (n = 55) treated with CRRT reported that more than three-quarters of them developed infections, including 18 cases directly related to the catheter [136]. No association with the catheter insertion site—femoral, jugular, or subclavian—was found. Prevention of thrombosis of RRT circuits Adults RCTs comparing a regional anticoagulation with citrate to a systemic anticoagulation for CRRT in critically ill patients with a low risk of bleeding are displayed in supplementary Table S1 [[137], [138], [139], [140], [141], [142], [143], [144], [145], [146]]. Meta-analyses of RCTs comparing regional anticoagulation with citrate to a systemic anticoagulation, or a combination of heparin and protamine concluded that regional anticoagulation with citrate resulted in longer circuit life and a lower risk of bleeding, but with no significant difference in mortality rates [[147], [148], [149]]. Regional citrate anticoagulation was associated with an increased risk of moderate hypocalcemia (ionized calcium >0.9 mmol/L) and metabolic alkalosis (observed in only one meta-analysis with low certainty of evidence) [147]. Plasma magnesium levels were comparable with citrate regional anticoagulation and heparin [143,146], but a higher incidence of hypophosphatemia with citrate was reported in one RCT (15.4% vs. 6.2%), attributed to a decrease in CRRT interruption periods [146]. A higher number of new infections was also observed in patients treated with citrate anticoagulation (68% vs. 55%, P = 0.002) in a single RCT [146]. CRRT in patients at risk of bleeding, with a focus on patients with liver dysfunction There are few RCTs comparing anticoagulation strategies in patients with proven bleeding risk treated with CRRT [[150], [151], [152], [153], [154], [155], [156], [157], [158], [159]] (supplementary Table S2). A recent meta-analysis compared a strategy without anticoagulation to heparin, regional anticoagulation with citrate, or nafamostat [160]. Compared with the no-anticoagulation strategy, circuit life was not significantly different with heparin (weighted mean difference +1.55 h, 95% CI −0.60 to +3.71, p = 0.15), but it was significantly increased with citrate anticoagulation (weighted mean difference +23.01 h, 95% CI 17.39–28.62, p < 0.001). Some studies have reported that the no-anticoagulation strategy increased the frequency of bleeding and transfusions compared with citrate anticoagulation [152,161]. These results could be explained by the consumption of coagulation factors and platelets, as well as blood loss during circuit thrombosis. Citrate anticoagulation also appears to prolong circuit life compared with regional heparin-protamine anticoagulation (39.2 h vs. 22.8 h, p = 0.0037) [151]. Although citrate clearance is reduced by nearly 50% in patients with acute liver failure or cirrhosis [162,163], citrate anticoagulation can be used safely in both pIRRT [164] and CRRT [165] using dedicated protocols. Intermittent hemodialysis in patients at risk of bleeding RCTs of anticoagulation strategies in patients at risk of bleeding treated by IRRT are shown in supplementary Table S3 [[166], [167], [168], [169], [170]]. A randomized study published in 1979 found more bleeding with regional heparin-protamine anticoagulation compared to low-dose heparin (19% vs. 10%, P < 0.05) [166]. In a RCT, more bleeding occurs with low-dose heparin compared to citrate anticoagulation (50% vs. 18%, p < 0.05) [168]. In cases of pIRRT lasting 6–12 h, premature circuit clotting occurs in 26%–50% of sessions with saline flush [171], and up to 25% of sessions with heparin (500–1000 IU/h) [172]. The use of regional citrate anticoagulation results in less premature clotting compared to heparin (6% vs. 12%, p = 0.028) [173,174] and to a strategy without anticoagulation (0.38, 95% CI: 0.17 to 0.84); and to less minor bleeding compared with heparin (0.14, 95% CI: 0.14 to 0.85) [175]. Hypocalcemia is rare (3.4%) and citrate accumulation has not been reported during citrate anticoagulation in IRRT, even in patients with severe liver dysfunction [164,173]. Pediatrics Studies comparing regional citrate anticoagulation (RCA) to unfractionated heparin (UFH) in CRRT are retrospective and non-randomized, and have various biases [[176], [177], [178], [179], [180], [181], [182], [183], [184], [185], [186], [187], [188], [189]]. The results of observational studies converge on the use of citrate, which allows for a longer circuit life compared to UFH or no anticoagulation [176,[178], [179], [180], [181], [182],187,[177], [178], [179], [180], [181], [182], [183], [184], [185], [186], [187], [188], [189]]. Metabolic complications were more common with citrate anticoagulation [176,179,185], with wide variability in the incidence of hypocalcemia (6.8–60 %), hypercalcemia (2.6–43.8 %), metabolic alkalosis (3–86.6 %), and hypomagnesemia (11.3–46.6 %). The incidence of citrate accumulation is highly variable, ranging from 0 [178,179,[182], [183], [184]] to 5.4% [176,177] due to different RRT modalities. The incidence of hemorrhagic events is lower with citrate in the largest cohorts [176,178,180,[179], [180], [181], [182],185,189]. In children with a risk of bleeding or severe liver failure, regional anticoagulation with citrate gives better results than heparin or no anticoagulation in terms of circuit life and bleeding events [[190], [191], [192]]. Weaning from renal replacement therapy Adults RRT weaning practices vary widely [193,194]. Early weaning, while reducing the inherent costs and workload, would avoid exposing the patient to the risks associated with vascular access (infections, catheter thrombosis, hemorrhage) and, above all those associated with intradialytic hemodynamic instability [8,9]. On the other hand, later withdrawal would result in a lower risk of fluid overload and better metabolic control. Predictive factors of RRT weaning Studies evaluating predictive markers for RRT withdrawal are observational, with a heterogeneous definition of withdrawal success. Consequently, the level of evidence for identifying one or more factors is rather low [[195], [196], [197]]. A urine output >436 ml/24 h (AUROC: 0.81 [95% CI: 0.77 to 0.84]) or >2330 ml/24 h with the use of diuretics (AUROC: 0.671 [0.585−0.75]) is predictive of successful weaning [195]. The predictive value of urinary flow rate [198] is better when combined with 24-h urinary creatinine [199] or creatinine clearance measured by the GFR kinetic formula [[198], [199], [200]]. A urinary urea excretion index [urinary urea (mmol/L)] x diuresis (L/24 h) / patient weight (kg)] above 1.35 mmol/kg/24 h would have an even better predictive value (AUROC of 0.96 [95% CI: 0.93 to 0.99]) [201]. In a recent before-and-after study [202], the combined use of this index (threshold >1.35 mmol/kg/24 h) reduced the duration of dialysis catheter use by 3 days. Unconventional urinary biomarkers such as NGAL, cystatin C, IL6, proenkephalin A 119–159, and HMGB1 do not improve the prediction of RRT weaning [[203], [204], [205], [206], [207], [208]]. Modality of RRT weaning: switch from CRRT to IRRT or not? The switch from continuous to intermittent RRT allows native renal function to be tested between sessions. Among 213 patients treated with CRRT for a median duration of five days and then with IRRT, 41% were still receiving norepinephrine at the time of transition to IRRT. Risk factors for hemodynamic instability during this transition were fluid balance (OR 1.04, 95% CI 1.01–1.06) and vasopressor support (OR 3.29, 95% CI 1.84–5.89), while the persistence of significant urine output was protective (OR 0.9, 95% CI: 0.82 to 0.98) [209]. In a recent randomized controlled trial involving patients with AKI treated with IRRT, the conservative strategy of using IRRT only in cases of urgent metabolic criteria such as severe fluid overload or a serum urea level >40 mmol/L reduced the number of sessions and the time to recovery of renal function (patient alive and not receiving hemodialysis for at least 14 consecutive days) compared to a conventional strategy of continuing treatment at a rate of 3 sessions per week until urine output exceeded 1 L/24 h, or creatinine clearance reached 20 mL/min, or at the nephrologist’s discretion when it is between 8 and 20 mL/min [210]. Given that only 40% of patients were in the ICU at the time of randomization and that hypoxemic patients, whether ventilated or not, were excluded, this study warrants further confirmation. Pediatrics The process of weaning from RRT remains complex and varies between PICUs, as reported in a European survey conducted by ESPNIC [50]. The main factors guiding clinicians' decisions are fluid overload management and metabolic control. The only high-quality study evaluating predictors of successful RRT discontinuation [211] identified a urine output >0.8 ml/kg/h in the six hours prior to discontinuation of RRT, although the use of diuretics weakened this association. Creatinine levels were not a determining factor in weaning, and biomarkers such as cystatin C or NGAL were not studied. However, native creatinine clearance estimated by the KeGFR formula could be a promising tool [212]. Very few studies have examined the optimal strategy for transitioning from CRRT to IRRT [213]. A recent multicenter study identified patients with optimal conditions for switching [87]: lower vasoactive agents and AKI severity scores, preserved urine output, and/or shorter CRRT duration. Conversely, transition failure was associated with an increased risk of chronic kidney disease or long-term dependence on RRT. For children weighing <6 kg, IRRT remains difficult due to circuit volume [214]. Ultimately, the choice of strategy depends on local resources and expertise. Conclusion Despite numerous technical innovations, the mortality rate associated with RRT for AKI remains high, due to a frequent association with other organ failures and the fact that it often develops in a septic context. The diversity of therapeutic modalities should enable more RCTs to determine the more optimized approach, better adapted to the type of patient and pathology, with the aim of improving survival and preserving the recovery of satisfactory renal function. Anyway, the management of AKI in the critically ill patient should henceforth follow the adage: ‘less and better renal replacement therapy’. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.