Timing of Initiation of Kidney Replacement Therapy in Acute Kidney Injury in the Critically Ill Patient

Abstract

Acute kidney injury (AKI) represents one of the most frequent complications in critically ill patients. In recent years, mortality rates have exceeded 50%, and 10% of them require kidney replacement therapy (KRT). Since the 60’s, the question of when to start KRT has been raised, classically the time of temporality when life-threatening is present (hyperkalemia, metabolic acidosis, uremia, and fluid overload); Therefore, in the last decade, the possibility of early initiation was raised as a strategy to achieve better outcomes. Current evidence shows that the timing of late onset dialysis has the same results as the strategy of early begining. We will also review the considerations in relation to renal capacity and demand generated by the acute pathology in a critically ill patient and the set of variables to make better decisions.

1. Introduction

Classically, acute kidney injury (AKI) is defined as a decrease in glomerular filtration rate (eGFR) in a period of less than 3 months along with the retention of nitrogenous products and changes in the internal environment, and anuria or oliguria may or may not be present [1].

In 2004, the Acute Dialysis Quality Initiative (ADQI) group considered that staging the damage is important in the diagnosis of AKI because small increases in creatinine or decreased diuresis are associated with greater morbidity and mortality and longer hospital stay [2].

In the Acute Kidney Injury Network (AKIN) Classification System, the value of increased serum creatinine (sCr) greater than 0.3 mg/dl is added with greater diagnostic sensitivity and eliminates the Loss and ESRD stages [3]. The Kidney Disease Improving Global Outcomes (KDIGO) workgroup in 2012 [4] encompasses an elevation of sCr greater than 0.3 mg/dl or an increase greater than 50% above the baseline value in a period of 7 days; in each of the three stages, a range is placed in relation to the fold increase in creatinine in relation to the basal range, and the other parameter that is taken into account is urine production [4]. Comparison of the KDIGO criteria, creatinine has greater predictive ability than urinary volume.

After reaching consensus on the definition of AKI by the Acute Disease Quality Initiative (ADQI) group, the AKI incidence in the general ward has been 20% [5], with rates increasing by approximately 10% per year [6]. The incidence of AKI in the intensive care unit (ICU) has been 30 to 50%, and 10% require kidney replacement therapy (KRT) [7, 8]. The hospital mortality of AKI has been 27.5% in those who did not require KRT and from 33 to 53% [9] in those who required KRT. This incidence has not changed over time probably due to more serious pathologies and increasingly long-living patients.

The development of AKI in the ICU is considered a predictor of mortality with a specificity of 71% and a sensitivity of 93% similar to the APACHE II score with a specificity of 75% and a sensitivity of 93% and where the relationship between AKI and mortality in the ICU is significant (odds ratio 3.99; 95% CI 2.125–7.481) [10].

2. Classification and biomarkers

The KDIGO guide classifies the severity of AKI into three stages based on functional criteria (sCr and diuresis), one of which is the elevation of creatinine with respect to baseline, and if the baseline value is not known, creatinine values are taken into account accordingly, depending on sex, age, and race, another variable is the reduction in urinary volume as a function of time. It is known that creatinine is not the best marker of renal function, in addition to being a late marker, and can be interpreted in different clinical situations such as states of fluid overload (FO), and the creatinine value should be corrected by the FO. There are also drugs that affect the tubular secretion of creatinine (trimethoprim-sulfamethoxazole, cimetidine, and ranitidine) and changes in the rate of sCr production (muscle hypercatabolism, the elderly).

It is important to identify predisposing and susceptibility factors that may develop AKI, such as heart failure, diabetes mellitus, liver disease, chronic kidney disease (CKD), sepsis, emergency surgery, cardiac surgery, and the use of nephrotoxic drugs. The use of biomarkers in populations at clinical risk will allow the detection of subclinical AKI in those at risk of renal complications after exposure to a noxa and allow opportunities for preventive intervention strategies to improve clinical outcomes and reduce the need for KRT [11]. The ADQI consensus conference (Table 1) recommends a combination of damage and functional biomarkers (Table 2) to improve the diagnostic accuracy of AKI and has made it possible to modify the KDIGO classification in stage 1 AKI with three subcategories (1S, 1A, and 1B) and AKI stage 2 and 3 with subcategories (2 A biomarker negative, 2B biomarker positive) (recommendation grade B) [12, 13].

	Structural injury
	No damage	Damage
Stable GFR	Not here; No loss of function no cellular damage	Subclinical AKI; Cellular damage; Without loss of function; Conserved creatinine; Positive damage response biomarkers; Positive cell regulation biomarkers
Decreased GFR	Functional here; Loss of function; Without cellular damage; Elevated creatinine; Negative tubular function biomarkers	Established here; Cellular damage; With loss of function; Elevated creatinine; Positive damage response biomarkers; Positive cell regulation biomarkers

Table 1.

AKI classification.

The ADQI consensus conference recommends a combination of damage and functional biomarkers to improve the diagnostic accuracy of AKI and have allowed modification of the KDIGO classification in stage 1 AKI with three subcategories (1S, 1A, and 1B) and stages 2 and 3 from AKI with subcategories (2A biomarker negative and 2B biomarker positive) [12].

Biomarkers glomerular filtration	Tubular function biomarkers	Tubular damage biomarkers	Damage response biomarkers	Cellular regulation biomarkers
Creatinine; Cystatin	FEU, FENA; Furosemide stress test	Urinary sediment.; Albuminuria	MCP1, IL6, IL10, IL18, KIM-1, LFABP	NGAL; TIMP2-IGFBP7

Table 2.

Biomarker IN AKI.

Biomarkers are classified into glomerular filtration, tubular function, tubular damage, injury response, and cell regulation [12].

Urinary insulin-like growth factor-binding protein (IGFBP-7) and tissue inhibitor of metalloproteinase (TIMP-2), neutrophil gelatinase-associated lipocalin (NGAL), interleukin-6, interleukin-10, interleukin-18, Kidney Injury Molecule-1 (KIM-1), liver-type fatty acid-binding protein (L-FABP), Fractional Excretion of Sodium (FENa), and Fractional Excretion of Urea (FEUrea).

3. Considerations at the start of kidney replacement therapy

The question of when to start KRT dates back to 60 years, Parsons F. et al. described a retrospective cohort of 17 patients with sepsis of surgical origin with AKI and who underwent intermittent hemodialysis with BUN values >180 mg/dl and had an 88% mortality. One year later, the same work group reported another cohort of 17 patients with AKI due to sepsis with daily BUN increases of 30 mg/dl in which early intervention in dialysis treatment decreased mortality to 25% [14].

The KDIDO-2012 guideline recommends initiate KRT emergently when life-threatening changes in fluid, electrolyte, and acid-base balance exist (Not Graded). The clinical context should be considered, and trends from laboratory tests should be evaluated, rather than single nitrogen values (Not Graded). These appraisals are far from being practical and we hope that, with the updated evidence of recent years and the forthcoming recommendations from the KDIGO, they will provide a clearer guideline [15].

The decision to start KRT should not only be based on kidney function or KDIGO stage, but many aspects mentioned below should also be evaluated (Figure 1).

Figure 1.

Considerations at the start of kidney replacement therapy. To consider the initiation of KRT, patient comorbidities, severity of acute illness, severity of AKI, and the expressed will of the patient, family, human resources, and available machines and costs are taken into account [16].

The kidneys have a finite capacity; when the demands exceed the renal capacity in the context of AKI, an imbalance with non-renal organ dysfunction is generated. Renal demands and capacity should be assessed taking the following considerations:

1. Chronic diseases: Chronic heart disease, chronic lung disease, CKD and chronic liver disease present loss of functional reserve and have lower tolerance to fluid overload, solute overload and inflammatory response leading to clinical worsening.

2. Severity of acute disease: elevated vasoactive inotropic score (VIS), SOFA, APACHE and SAPS III scores, refractory hemodynamic instability, macromicrovascular alterations, mitochondrial damage, oxidative stress, inflammatory response, capillary leak syndrome and vasoactive dependence express the degree of severity of acute pathology and in these clinical conditions, loss of renal functional capacity is likely not tolerated.

3. The degree of fluid and solute overload (acids, K, urea myoglobin, and other unmeasurable uremic toxins) generates organ dysfunction in other organs with less functional reserve.

4. The sCr and diuresis are used as markers of renal function; they are not specific and have many limitations in their interpretation, besides being late markers. In the evaluation of renal capacity, the trend of elevated nitrogen levels is important. The use of biomarkers (NGAL, TIMP-IGFB 7) and other tools such as the furosemide stress test (FST) allow the assessment of renal capacity, but they are still far from perfect as markers of damage (Figure 2).

5. The initiation of KRT should also consider the risks inherent to the treatment, such as: catheter-associated infections, worsening hemodynamic instability, and nutrient elimination associated with dialytic therapy.

6. It is important to know the wishes of the patient and family, to evaluate with the medical team whether dialysis is part of a therapeutic plan of the multidisciplinary team and that it does not represent a futile measure, and to consider the availability of nursing staff, machines, and costs [16].

Figure 2.

Kidney capacity and demand. The relationship of renal capacity and demands refers to the state of functional reserve of the kidneys. They are not able to withstand the severity of acute illness, fluid overload, and metabolic overload, where previous comorbid conditions may be a factor in a lower tolerance to the development of AKI [16].

4. Management of complications of acute kidney injury and classic indications for kidney replacement therapy

4.1 Hyperkalemia

In AKI, the incidence of hyperkalemia is 13 to 24% and hypokalemia 11 to 17%; dyskalemia values (<3.5 to >5.5 mmol/L) are associated with longer hospital stay and mortality [17, 18]. The prevalence of hyperkalemia increases linearly with the fall in glomerular filtration rate and with the severity of AKI, 3.4% occurring without AKI, 8.8% in AKI stage 1, 17% in AKI stage 2, and 32.2% in AKI stage 3 [19]. It may also be secondary to intracellular potassium release (rhabdomyolysis, tissue injury, and hemolysis) and altered transfer to the intracellular space (metabolic acidosis, insulin deficit) [18].

Independent of AKI, there are factors that contribute to the development of concomitant hyperkalemia in AKI, such as the use of nonsteroidal anti-inflammatory drugs, renin-angiotensin-aldosterone inhibitors (RAASI), and potassium-sparing diuretics; diabetes mellitus (DM); heart failure (HF); chronic kidney disease (CKD); severe tissue breakdown; and potassium intake in crystalloid solutions or oral potassium supplements [20].

The KDIGO Conference on Potassium Controversies recommended cardiac monitoring and 12-lead electrocardiogram (ECG) for potassium levels >6.0 mmol/L. There is evidence to support that severe hyperkalemia levels do not necessarily exhibit ECG changes, as will be described later [21].

Mortality increases in a linear relationship with increasing potassium, K = 4.5–5.0 mmol/L with odds ratio (OR) 1.25; K = 5.0–5.5 mmol/L with OR 1.42; K = 5.5–6.0 mmol/L with OR 1.67; K = 6.0–6.5 mmol/L with OR 1.63; K > 6.5 mmol/L with OR 1.72 where the risk of mortality is higher. Hyperkalemia, if not decreased by >1.0 mEq/L within 48 h after measurement, predicts death [22].

There is important evidence in a meta-analysis [23] with more than 1 million patients with dyskalemia that was associated with all-cause, cardiovascular mortality and shows a U-shaped curve for hypokalemia (3.0 mmol/L) with HR 1.49 and in hyperkalemia (>5. 5 mmol/L) with HR 1.22. Likewise, values considered within the normal range for healthy subjects do not seem to be as innocuous for patients with AKI, since potassium values ≥4.8 mmol/L are associated with higher all-cause mortality with an adjusted HR of 1.12. Decreased eGFR was a strong risk factor for hyperkalemia (>5.5 mmol/L); also, the interaction of hyponatremia with hyperkalemia was associated with higher 30-day (HR 1.15) and 90-day (HR 1.16) mortality. Values of “normokalemia” seem to have an impact on mortality in patients with AKI, but these findings need to be confirmed in prospective studies [24]. Li et al., in a retrospective observational study in adjusted multivariate analysis, shows the risk of all-cause mortality for K values 4.1–4.79 mmol/L with HR 1.45; K 4.8–5.49 mmol/L with HR 2.15 and for K > 5.5 mmol/L with HR 2.34 [25].

Mortality is related to not only the severity of hyperkalemia but also the prolonged duration of hyperkalemia and lack of correction of the disorder, rapid increases in serum potassium, massive cell turnover, and underlying heart disease. In a study of 408 patients excluding patients with end-stage renal disease (ESRD) and on hemodialysis (HD), hyperkalemia of prolonged duration was associated with concomitant AKI (OR = 3.88; p = 0.03), metabolic acidosis (OR = 4.84; p < 0.01), tissue necrosis (OR = 4.55; p < 0.01), and potassium supplements (OR = 5.46; p < 0.01), and all of them are associated with higher hospital mortality [26].

Mortality is attributed to changes in heart rhythm and its consequences such as hemodynamic instability, myocardial ischemia, and sudden cardiac death [27]. Under normokalemia conditions, the cardiomyocyte cell membrane is polarized (resting potential at −90 mV). In moderate hyperkalemia, the cell membrane partially depolarizes, bringing the resting potential closer to the threshold potential for action potential initiation, where fast sodium channels are activated, increasing excitability and conduction velocity, and it electrocardiographically manifests as a T-wave spike. In severe hyperkalemia (K ≥ 6.5 mmol/L), voltage-dependent inactivation of sodium channels and activation of internal rectifying potassium channels lead to reductions in conduction velocity, and cells enter a state refractory to excitation and is expressed electrocardiographically with widening of complexes and blockades driving [28, 29].

ECG findings do not have a relationship with hyperkalemia levels in patients with AKI and ESRD [30]; ECG is an insensitive tool for the diagnosis of hyperkalemia even with K values between 6 and 7.2 mmol/L and with K values >7.3 mmol/L the predictive value is minimal; non-traditional electrocardiographic changes are also described [31].

In a prospective study [32] of 77 patients that included both AKI and CKD, 94% associated with a fall in eGFR and 70.9% had metabolic acidosis, 10.4% developed mild (5.5–5.9 mmol/L), moderate 40.3% (6–6.4 mmol/L), and severe hyperkalemia 49.3% (>6.5 mmol/L); electrocardiographic changes occurred in 74.6% of patients, the most frequent findings being atrial fibrillation (13.4%), peaked T wave (11.9%), widened QRS, and prolonged PR (10.5%); none of these electrocardiographic findings were more significant with greater severity. The sensitivity in the electrocardiographic detection of hyperkalemia was 0.28 for the emergency physician and 0.36 for the cardiologist. The diagnosis improved with K values >6.5 mmol/L and in another study with values >7.8 mmol/L. The lack of sensitivity in the electrocardiographic diagnosis in these study groups may be biased by the slow increases in K in ESRD and the use of hemodialysis. The electrocardiographic findings may be masked by the effect of fluid overload and other electrolyte disorders such as concomitant hypocalcemia. In this sense, hypocalcemia can induce QT prolongation and widening of T waves and may mask the changes in T wave morphology caused by hyperkalemia [33, 34].

There is no conclusive evidence that the electrocardiogram can guide the treatment of hiperkalemia and electrolyte monitoring is necessary.

In a prospective cohort of patients with AKI, it was identified that 60.8% developed dyskalemia, and 8 evolution groups were identified according to the kalemia characteristics. Group 7 was characterized by K values that were normal on hospital admission and increased to hyperkalemia; group 8 were those that never returned to a normal kalemia value and corresponded to uncorrected hyperkalemia. In both groups, the increase in mortality was identified, 37% for group 7 and 63% for group 8, and the latter group had a higher risk (40%) of requiring KRT [35].

When hyperkalemia is associated with arrhythmias and/or hemodynamic instability despite medical treatment, the initiation of intermittent hemodialysis or sustained low-efficiency dialysis (SLED) or continuous kidney replacement therapy (CKRT) is indicated. In situations where rapid and prolonged K generation is present, such as in rhabdomyolysis, tissue lysis syndrome and tissue ischemia, where hyperkalemia values are persistent and the K load exceeds the elimination capacity by CKRT, intermittent hemodialysis is recommended [36].

It is important to consider that the longer the duration of hyperkalemia and the in-ability to correct itself, the higher the mortality and we see that mortality is associated with extreme values of dyskalemia.

4.1.1 Management of acute hyperkalemia

There is limited evidence, and the existing is based on the recommendation of experts or studies with a small number of patients for the analysis of the treatment of acute hyperkalemia in AKI, and many of the studies of this disorder come from patients with CKD (Table 3).

	Mechanism	Onset (min)	Duration (hours)	Effect on potassium plasma level	Dose	Side effects
Calcium gluconate 10%	Membrane stabilization	1–3 min	1–2	None	10–20 ml iv within 5 min	Hypercalcemia; Digoxin toxicity
Calcium chloride 10%	Membrane stabilization	5 min	1–2	None	10–20 mL; iv within 5 min	Tissue necrosis
Hypertonic sodium 20%	Membrane stabilization	5–10 minutes	2	None	10–20 mL IV within 5 min	Hypernatremia; Volume overload
Regular insulin/dextrose	Shift K in to the cells	10–20 minutes	4–6	−0.7 ± 0.6 mEq/L at 60 min	Regular insulin; 5 or 10 U; +; Glucose level > 200 mg/dl, none dextrose.; Glucose 100–200 mg/dl, supplemental 25 gr dextrose 50% (50 ml).; Glucose level < 200 mg/dl, supplemental 50 gr dextrose 50% (100 ml) at 60 min	Hypoglycemia; Hyperglycemia
Sodium Bicarbonate	Shift K in to the cells	90 min	2–4	−0.47; ±0.31 mmol/L at 30 min	100 mL of 8.4% iv at 30 min.; either; 125–250 ml of 4.2% iv at 30 min.	Hypernatremia; Volume overload; Hypocalcemia
B2 – Agonist	Shift K in to the cells	30 min	2–6	−0.5; ±0.1 mmol/L at 60 min	0.5 mg IV; either; 10 mg nebulization over 15 min	Tachycardia, tremors
Patiromer	K removal	180-420 min	12–24	0.75 mmol/L at 48 hours	8.4–25.2 g per day	Gastrointestinal intolerance and hypomagnesemia
Zyrconium	K removal	120 min	24–48	−0.67 mEq/L at 48 hours	10 g one to three times a day	Gastrointestinal side effects and edema
Furosemide	K removal	15 min	2–3 hours	−0.3 mmol/L at 3–6 hours	40–80 mg IV	Volume depletion
Hemodialysis	K removal	immediately	3 hours	−1 mmol/L at 60 min	3–4 hours on modality intermittent hemodialysis	Arrhythmias

Table 3.

Drugs for the management of hyperkalemia [37, 38].

4.1.1.1 Membrane stabilizers

Calcium gluconate and calcium chloride

The recommendation for the use of calcium gluconate is based on expert opinion; there are no randomized studies.

Calcium gluconate increases intracellular calcium entry and binds to calcium-dependent calmodulin and protein kinase II (CaMKII), which allows sodium channel activation, leading to intracellular sodium entry, thus restoring action potential in phase 0 dV/dt. Also, hypertonic physiological solution increases extracellular sodium and allows greater intracellular displacement of sodium, managing to increase the speed of the action potential; it is an option as a membrane stabilizer [18, 39]. There are no randomized studies available showing a clear benefit of hypertonic saline.

Membrane stabilizers should be started early, after identification of hyperkalemia >6.5 mmol/L or when some ECG change is evident in the context of hyperkalemia. The recommended dose is from 10 ml (2.3 mmol of Ca 2+) to 20 ml of 10% undiluted calcium gluconate, administered over 3 to 5 minutes; if ECG changes persist, it can be administered again later 5 min; the effect lasts between 30 to 60 minutes [40]. Another option as a membrane stabilizer is calcium chloride, which has three times the concentration of Ca 2+ (6.8 mmol of Ca 2+) than calcium gluconate, but tissue damage secondary to skin extravasation has been described.

In an observational and prospective study of 111 patients with a mean K > 7.1 mmol/L, 243 pathological ECGs were found, of which in 79 cases, they were identified as major rhythm disorders (AV block, sinus bradycardia, right and left bundle block, escape beats, atrial fibrillation, ventricular tachycardia), and of these only 9 cases improved with calcium gluconate in a dose of 10 ml at 10% and was repeated up to three times. Calcium gluconate may only be effective in major rhythm disorders with limited evidence and did not cause statistically significant improvement in non-rhythmic ECG disorders [41].

4.1.1.2 Intracellular redistribution of potassium

Insulin-dextrose

The insulin binds to glucose transporter type 4 receptor on skeletal muscles and allows the translocation of intracellular Na + -K + ATPase to the cell membrane and induces the transfer of potassium from the extracellular space to the intracellular space. Insulin stimulates the phosphorylation of FXYD1 (phospholigen) by atypical protein kinase C, increasing the Na-K ATPase Vmax for potassium transfer into the intercellular space [42].

Regular insulin allows a reduction of K in a range of 0.5 to 0.9 mEq/L and dextrose alone in a range of 0.2 to 0.6 mEq/L; current evidence recommends both treatments as first line of treatment [18].

In a randomized crossover study in 10 patients per group on chronic hemodialysis, in the glucose infusion alone group (100 ml of 50% glucose) versus the glucose + insulin group (100 ml of 50% glucose + 10 IU of regular insulin), K Values were measured at 60 minutes and there was a significant decrease in K 0.83 ± 0.53 mmol/l (p < 0.001) in the combined treatment compared to the glucose alone group (0.50 ± 0.31 mmol/l) [43].

In a retrospective cohort study of 174 critically ill patients with concomitant hyperkalemia in AKI or CKD, two groups were evaluated, those receiving 5 IU of regular insulin and those receiving 10 IU of regular insulin. Both groups received 25 g of dextrose along with intravenous insulin and 1 hour after receiving insulin a second dose of 25 g of dextrose was administered and a third dose of 25 g of dextrose if blood glucose was less than 70 mg/dl and hourly blood glucose was measured and K control were performed after 2 or 3 hours of insulin administration. The rate of hypoglycemia at the sixth hour was higher in the 10-unit insulin group 19.5% and 9.2% in the 5-unit group (p = 0.052); in severe hypoglycemia, there was no difference between the groups. The reduction K was similar (−0.8 ± 0.7 mEq/L in the 5-unit group vs. −0.7 0. ± 6 mEq/L in the 10-unit group (p = 0.430)) [17]. (p = 0.052) [17]. Regular insulin doses of 5 units reduce K levels without difference with higher doses and with fewer hypoglycemic events.

In a systematic review of 11 studies, different doses of insulin are described. In eight studies, 10 units of regular insulin were administered and of these, in 5 studies the method of administration was a bolus and in two studies as an infusion over 15 to 30 minutes, and in another three studies regular insulin was administered 5 IU/Kg equivalent to 20 units in 60 minutes. Regarding the glucose dose, it was 25 g in six studies, 30 g in one study, 40 g in two studies, 50 g in one study, and 60 g in another study. The incidence of hypoglycemia was 30% in the group that used 25 g of glucose; when 60 g of glucose was administered, no hypoglycemia was reported. When a bolus dose of 10 units of regular insulin was used, the serum potassium decrease at 60 minutes was 0.78 ± 0.25 mmol/L, and with the administration of insulin infusion of 20 units over 60 minutes, the decrease in serum potassium was 0.79 ± 0.25 mmol/L, with no significant difference between both groups (p = 0.98) [44].

A meta-analysis of 10 retrospective cohort studies (n = 3437) with low or moderate risk of bias evaluated standard doses (10 IU regular insulin) vs. alternative doses (5 IU regular insulin or 0.1 IU/kg). K reduction was similar in the two groups (mean difference − 0.02 mmol/L, 95% CI –0.11–0.07, I2 = 53%), with no difference in hospital mortality (OR 1.03, 95% CI 0.58–1.81, I2 = 0%); in the alternative dose group there was a lower risk of hypoglycemia (OR 0.55, 95% CI 0.43–0.69, I2 = 8%); and severe hypoglycemia (OR 0.41, 95% 0.27–0.64) [37].

The use of dextrose and insulin constitutes the first line of treatment for intracellular K redistribution; when comparing 5 units vs. 10 units of regular insulin, the K decrease values are similar, and with fewer hypoglycemia events in the first group, the recommended dose of 50 ml dextrose 50% (25 g) is a safe regimen to avoid hypoglycemia [18].

In case of severe hyperkalemia, a dose of 20 units of regular insulin with 60 g of glucose (200 ml 30% dextrose) for 60 minutes is recommended, and an alternative in non-severe hyperkalemia is a dose of 10 units of regular insulin, and 25 g (50 mL 50% dextrose) to glucose is a safe regimen to avoid hypoglycemia.

Sodium bicarbonate

Evidence for the use of sodium bicarbonate is limited and heterogeneous including patients with ESRD, AKI, and various therapeutic interventions for the management of hyperkalemia.

Sodium bicarbonate allows intracellular entry of sodium through the Na + /H + (NHE) exchanger; higher intracellular sodium activates sodium-potassium adenosine triphosphatase (NaK + ATPase) and allows transfer of potassium from the extracellular space to the intracellular space [18].

Studies on sodium bicarbonate show that it is not able to rapidly reduce serum potassium levels with optimal efficacy, and the onset of its action can take several hours; therefore, it is not recommended as first-line treatment [18].

There is controversy regarding its usefulness, due to the small decrease in K values <1 mmol/L and slow onset. Some predictors of a better response to the use of bicarbonate are mentioned, such as the presence of hyperkalemia >6 mmol/L, metabolic acidosis with pH < 7.35 or bicarbonate <17 mmol/L, AKI and the use of bicarbonate doses >120 mEq [36]. There are clinical situations where its use should be avoided, such as in organic acid acidosis (diabetic ketoacidosis, lactoacidosis) and hypervolemia [45].

In a retrospective study of 106 patients, of which 38 patients correspond to the sodium bicarbonate group (regular insulin and bicarbonate) and 68 patients to the control group (regular insulin without bicarbonate), the primary objective was to compare the absolute reduction in serum potassium between initial concentrations and at 2-hour intervals, up to 8 hours. Median initial potassium concentration was higher in the bicarbonate group (6.6 vs. 6.1 mmol/L (p = 0.009)); the lowest potassium concentration was reached within the first 8-h period, with no difference between both groups (bicarbonate group 5.35 mmol/L) vs. (control group 5.15 mmol/L (p = 0.255)). We saw that adding bicarbonate did not improve the absolute reduction of K. It must be clarified that this study has several limitations and randomized controlled trials are required to confirm the efficacy of bicarbonate [46].

The different concentrations of 1.4% or 8.4% sodium bicarbonate have the same capacity to reduce potassium values, even a transient increase of K in patients with CKD is mentioned, the lack of evidence should be considered due to a small number of patients in the studies [47].

Potassium homeostasis and management of dyskalemia in kidney diseases: conclusions from a Kidney Disease: Improving Global Outcomes (KDIGO) Controversies Conference recommends a bicarbonate dose of 1 amp of 50 ml of 8.4% over 15 minutes. The European Resuscitation Council Guidelines recommend sodium bicarbonate 1 mmol/kg IV in case of a metabolic acidosis (pH < 7.2) and/or in cardiac arrest [21, 48].

In the BICAR-ICU study [49], 4.2% sodium bicarbonate in patients with severe metabolic acidosis (pH < 7.2) reported lower K values and less hemodialysis requirement when compared with the control group. The use of hypertonic sodium bicarbonate (100–250 mL of 8.4% sodium bicarbonate over 20 min) is also recommended in patients with severe metabolic acidosis and AKI, who have a contraindication to the use of calcium gluconate. It is important to mention that sodium bicarbonate has side effects, such as fluid overload and hypocalcemia; the aforementioned electrolyte disorder is crucial for cardiac contractility, for which its monitoring and replacement are important.

Beta-agonists

Beta-adrenergic agonists activate pancreatic β receptors, causing increased insulin secretion, which activates Na-K-ATPase and stimulates K movement into cells [18, 50].

The standard doses are 10 and 20 mg for the inhaled forms and between 0.5 and 2.5 mg intravenously; the effect begins between 15 and 30 minutes regardless of the formulation used. K decreases were 0.3 mmol/L with 10 mg and 0.6 mmol/L with 20 mg, with a duration of effect of 2 hours [51].

In a randomized, double-blind, placebo-controlled trial of few patients (n 17) with ERSD, there was an increase in K by 0.1 mmol/L in the first minute after inhalation; when compared to placebo (p < 0.001), the increase in K was transient without generating arrhythmias and was attributed to the β 1-activity of these drugs. Also, the values of glycemia, insulin, and heart rate increase after 5 minutes of salbutamol inhalation [51].

The paradoxical elevation of K by beta-agonists is slight and transient; it is not known if they have a clinical effect in critically ill patients and raises doubts about the use of these drugs.

4.1.1.3 Elimination of potassium

Diuretics

Loop diuretics act by inhibiting the NKCC2 channel on the apical surface of the thick cells of the ascending limb of the loop of Henle, generating natriuresis and kaliuretic effect. The use of furosemide in AKI is useful for the management of fluid overload and as a stress test to assess the progression of sustained AKI and to determine the need for hemodialysis; the use of diuretics in non-responders should not delay the use of other types of therapies for the management of hyperkalemia [18, 52].

New resins

Patiromer is a non-absorbable and sodium-free potassium-chelating polymer. It is not ideal for the management of severe acute hyperkalemia, since the decrease in serum K is 0.21 ± 0.07 mmol/L in 7 hours and 0.75 mmol/ L in 48 hours; its use is focused on patients with CKD and ESRD [18, 38].

Sodium zirconium cyclosilicate (zs-9)

ZS-9 is a crystal that is selective for potassium and ammonium ions, and they exchange sodium for potassium. A systematic review and meta-analysis of data from phases II and III clinical trials found that ZS-9 produced a decrease in K of −0.17 mEq/L 1 hour after dosing, and − 0.67 mEq/L at 48 hours; at this time, it is not recommended for severe acute hyperkalemia [18, 53].

In a review of seven randomized controlled trials, it shows that the use of IV insulin + dextrose and salbutamol in any of the pharmacological presentations are the most effective in lowering serum potassium. Evidence on the benefits of IV sodium bicarbonate is limited [54].

In severe hyperkalemia and refractory to medical treatment, KRT is indicated; the details of the indication and timing of KRT will be discussed following.

Kidney replacement therapy.

In a large retrospective study of AKI complications in critically ill patients in the multivariate logistic regression model adjusted for AKI-related complications, KRT was found to be associated with increased hospital survival (OR 0.75; 95% CI, 0.58 to 0.96), and KRT decreased mortality by 45% in patients with hyperkalemia [19].

The causes of death in AKI attributed to hyperkalemia, metabolic acidosis, volume overload and uremia with the onset KRT allowed correction of these alterations, was associated with improved hospital survival (OR, 0.75; 95 % CI, 0.58 to 0.96) [19].

In RCTs such as ELAIN and STARRT, late onset of CKRT was associated with K values >6 mmol/l and/or ECG abnormalities; in the AKIKI study, the value was K > 6 mmol/L (or > 5.5 mmol/L without improvement despite medical treatment); in the IDEAL-ICU, the K values were higher (> 6.5 mmol/L), and 4% had a K of 7 mmol/L for the start of dialysis treatment in a delayed strategy or there were no significant differences between the groups in other adverse events, nor in mortality. Medical treatment for hyperkalemia could prevent or delay the initiation of KRT in patients with AKI and avoid complications of dialysis therapy and additional costs [55, 56, 57, 58].

4.2 Metabolic acidosis

Acidemia is the accumulation of protons in the plasma, whose expression is a low blood pH. Severe acidemia is defined when the pH is <7.20, but it is not a universally accepted term. Metabolic acidosis is classified based on time, as acute (hours to days) and chronic (weeks to months), and based on the presence of an anion gap (AGMA) and no anion gap (NAGMA). In AKI, metabolic acidosis is caused by decreased renal synthesis of bicarbonate and decreased excretion of nonvolatile acids in the urine [59].

Metabolic acidosis has an impact on several systems. At the cardiac level, Ca 2+ transport decreases through the SR Ca 2+ -ATPase (SERCA) channels, the ryanodine receptor (RyR) and the Na + /Ca 2+ (NCX) exchanger, having a negative inotropic effect, due to less coupling and decreased sensitivity of troponin C regulatory sites to Ca2+. Also, in acidosis, it conditions a delayed β-adrenergic response and decreases the affinity of β-adrenergic receptors for vasoactive (norepinephrine, epinephrine) and inotropic agents, decreasing the responsiveness to catecholamines [60].

The impact of the drop in pH generates a lower affinity of hemoglobin for oxygen at the tissue level (Bohr effect), shifting the oxygen dissociation curve to the left; acidosis also decreases cerebrospinal pH and could be responsible for the confusional state. Altered pH suppresses lymphocyte function; the chemotactic and phagocytic capacity of leukocytes is reduced, and the proinflammatory response of macrophages is increased, all of which sets the stage for the development of infections. The impact is also expressed in energy and enzymatic exhaustion, stimulating apoptosis and cell death. These mentioned alterations are of multifactorial origin in the context of sepsis, ischemia reperfusion, and AKI [61].

Not all acidosis is the same and it also depends on the disease that causes the acidosis, since the type of acid determines the risk of mortality. In an observational cohort study of 851 critically ill patients, it was identified that 64% developed acidosis and mortality was 45%. When the subgroup analysis was performed, mortality was higher in the lactic acidosis subgroup (56%) when compared to strong ion gap (SIG) acidosis (39%) and hyperchloremic acidosis (29%) [62].

The treatment of acidosis should be aimed at treating the cause that generates it. The administration of bicarbonate has the objective of reversing the deleterious effects of severe acidosis, and the complications related to its use must be weighed. Sodium bicarbonate is reported to cause a transient decrease in blood pressure and cardiac output and hypocalcemia, sensitizing the heart to abnormal electrical activity [63].

The use of sodium bicarbonate is a common medical practice in critically ill patients and has not demonstrated a clear benefit. Next, we will review the evidence that limits the general use of bicarbonate and is limited to particular cases (Table 4).

Bicarbonate is indicated in patients with hyperlactatemia with circulatory shock with pH < 7.2 and AKI in the initial phase of resuscitation (Level of evidence: B); In severe non-anion gap metabolic acidosis (NAGMA) (Level of evidence: B).; There is no evidence to recommend its use in a general way in cardiac arrest unless caused by hyperkalemia; there is no clear evidence for its use in diabetic ketoacidosis and rhabdomyolysis [64].

Table 4.

Recommendations for the use of bicarbonate IN different clinical scenarios.

In a prospective, observational, multicenter study, 155 critically ill patients with severe acidemia (single metabolic acidemia or mixed respiratory and metabolic acidemia) with pH values <7.2 within the first 24 hours of ICU admission were analyzed. It showed that the incidence of acidosis in the ICU was 6%; 90% of the patients studied required mechanical ventilation and vasopressors; 20% required KRT within the first 24 hours, and the mortality rate was 57%. In non-survivors, SOFA, SAPSII, anion gap, base excess, lactatemia scores and length of ICU stay were higher and associated with higher mortality (p = <0.01). Regarding the development of AKI or the need for KRT, there was no difference between survivors and non-survivors. Administration of bicarbonate at different concentrations in the first 24 hours did not improve prognosis [65].

In the BICAR-ICU a multicenter, open-label, randomized controlled, phase 3 trial, a study of 389 patients, of whom 61% had sepsis (195 in the bicarbonate group and 194 in the control group), included patients within 48 hours of ICU admission presenting with severe acidemia (pH <7.2, PaCO2 ≤ 45 mm Hg, and sodium bicarbonate concentration ≤ 20 mmol/L) and an arterial lactate concentration of 2 mmol/L or more total Sequential Organ Failure Assessment (SOFA) score of 4 or more. The bicarbonate group received 125–250 ml in 30 min, with a maximum of 1000 ml within 24 h of inclusion. When assessing mortality on day 28 and failure of ≥1 organ on day 7 (primary composite outcome), there were no significant differences (p = 0.24) and in those who developed AKI on the AKIN 2 and 3 scale, the use of bicarbonate improved survival at 28 days (p = 0.017) and failure of ≥1 organ on day 7 (p = 0.014) and decreased the need for KRT (p = 0.0283). This study shows several weaknesses and requires a trial with a larger number of patients to obtain more conclusive results [49].

In another RCT of 1718 patients diagnosed with sepsis and metabolic acidosis (pH < 7.3 and bicarbonate <20 mmol/L) randomized into two groups, 500 patients in the bicarbonate group and 1218 patients in the control group, bicarbonate infusion had no effect in reducing mortality (p = 0.67), but it is beneficial in reducing mortality risk in patients with KIDGO AKI stage 2 or 3 and pH < 7.2 by 22% (p = 0.032) [66].

In two meta-analyses of five randomized controlled trials (RCTs) of AKI secondary to cardiac surgery (CSA-AKI) as a preventive measure, bicarbonate was administered at a bolus dose of 0.5 mmol/kg of body weight, diluted in 250 ml of dextrose to 5% for 1 hour immediately after induction of anesthesia and then continuous infusion of 0.15 mmol/kg/h diluted in 1000 ml of 5% dextrose for 23 hours. There were no differences in the development of CSA-AKI between the patients in the sodium bicarbonate group and the control group; there was no difference in mortality at 30 and 90 days, no differences in the need for KRT, nor in the days of stay in ICU between both groups. It should not be recommended for the prevention of CSA-AKI, and the perioperative infusion. Should be administered with caution due to the risk of fluid overload [67, 68].

It should be considered that the injury mechanisms in CSA-AKI and in sepsis are different and it is possible that the response and benefits of the use of bicarbonate are also different. We hope that future studies will make it possible to individualize the therapy in each group.

KRT is indicated for severe metabolic acidosis refractory to medical treatment in the setting of AKI. In a retrospective study of 1815 AKI patients undergoing CKRT who were assessed for pH trajectory in five groups during CKRT: 1st group normal pH; 2nd group, suboptimal pH trajectory of 7.3 initially and approached pH 7.4; the 3rd group, recovering from acidosis with pH 7.2 to 7.3; 4th group, tendency to worsen acidosis with pH 7.3 to 7.2; and 5th group, uncorrected pH trajectory and less than 7.2 despite CKRT. AKI due to sepsis was more common in the 3rd group; the SOFA and APACHE II scores, and the requirement for mechanical ventilation were higher in groups 4 and 5. In the multivariate analysis, mortality increased from the first to the fifth group, showing higher mortality from the 3rd to the 5th group (3rd group 74.2%, 4th group 78.2%, and 5th group 82.2%) despite the start of CKRT. The measurement of CRP values >30 mg/dl occurred in a greater proportion in the patients of groups 4 and 5 of metabolic acidosis. It is important to consider that the inflammatory response triggered by different pathways, conditions the origin of oxidative stress, and mitochondrial and endothelial damage that can be a point of no return in patients with sepsis, septic shock, and AKI. We see that with the support of CKRT, mortality did not decrease. This work has several limitations, and randomized clinical trials are required to ascertain whether pH correction during CKRT improves the survival rate of patients with metabolic acidosis [69].

There is no consensus with which pH or bicarbonate value hemodialysis should be started. In the RCTs in recent years, they use different values. In the ELAIN and AKIKI study, they use a pH value <7.15 in the context of pure metabolic acidosis or mixed acidosis, and it represented 21% of the CKRT indications in the second study. In the IDEAL-ICU study, the cut-off value was pH < 7.15 and base deficit >5 mEq/l or HCO3 − < 18 mEq/l, and acidosis represented 13.4% of CKRT indications. In the STARRT-AKI study, the absolute criterion for CKRT was considered in the presence of severe acidemia and metabolic acidosis, with a pH ≤ 7.2 or HCO3 − < 12 mmol/l, and this criterion was present in 16.6% of those who required CKRT. Patients with severe metabolic acidosis and stage 2 or 3 AKI should be managed with bicarbonate with the based of delayed dialysis therapy that avoids complications associated with early dialysis therapy and which does not confer a survival benefit [55, 56, 57, 58].

4.3 Fluid overload

The administration of intravenous fluids is one of the cornerstones of the resuscitation of critically ill patients, with the aim of optimizing hemodynamics, increasing stroke volume, improving organ perfusion, and supplying the O₂ supply. Fluid resuscitation has been empirical; there is no ideal administration strategy free of complications. The Surviving Sepsis Campaign recommends the use of a minimum of 30 mL/kg (ideal body weight) of intravenous crystalloids and was based on observational studies [70].

Fluid resuscitation carries a narrow line between insufficient resuscitation and fluid overload. In the past, hydration was based on clinical parameters and basic measurements (“blood pressure, edema, arterial blood gases, chest Rx, central venous pressure (CVP) and pulmonary capillary wedge pressure (PCWP)” which have low sensitivity, specificity and are late [71]. Currently, for a better assessment, dynamic tests are used to assess the change in stroke volume after a maneuver that increases or decreases venous return (preload) through echocardiography at the patient’s bedside, pulse contour analysis (PCA), and trans-pulmonary thermodilution (TPTD). It is mentioned that only 50% [72] of hemodynamically unstable patients respond to volume; according to the Frank-Starling curve, increasing the preload allows the left ventricular (LV) stroke volume to be increased until the optimal preload is reached, achieving a stroke volume at a constant plateau. Resuscitation must be individualized and guided in order to improve preload and increase stroke volume with fluid challenge and avoid excessive use of fluids in nonresponders [73, 74].

Fluid overload is the consequence of excess fluids in resuscitation, maintenance solutions, drug dilutions, nutrition, and use of blood products. Fluid overload (FO) was defined as the ratio between cumulative fluid balance (L) and initial body weight (kg) at admission and expressed as a percentage. It is considered that a FO > 10% at the time of AKI development has an adjusted odds ratio (OR) for associated death of 3.14 (95% CI, 1.18 to 8.33); for this reason, its importance in the assessment lies in the critical patients (Table 5) [75].

% Fluid overload = (total fluid in − total fluid out) /admission body weight × 100)

Table 5.

% Fluid overload.

It is important to differentiate that the positive cumulative balance that reflects higher admissions than discharges without necessarily generating fluid overload; the fluid overload itself reflects the degree of fluid accumulation in tissues (pulmonary congestion and/or edema) and is expressed as a percentage [76].

The clinical condition with a positive cumulative fluid balance will have an impact with multiple adverse effects such as the development and progression of AKI, need for KRT, mechanical ventilation, intra-abdominal hypertension, and capillary leak syndrome and with a higher probability of overload-associated mortality (OR 2.07) [77].

How is fluid overload generated?

The development of FO is due to the close balance between a deterioration in cardiac function, the vasoplegic state that does not improve fluid supply, the AKI in a state of oliguria/anuria, the inflammatory response in sepsis or other proinflammatory clinical condition, and endothelial and glycocalyx damage, which are relevant factors in the development of FO.

A key component is the endothelium of all blood vessels; it is covered by glycocalyx or Endothelial Surface Layer (ESL), which is made up of a layer of polysaccharides that forms a network of molecules that generate a state of equilibrium between the vascular wall and the plasma [78].

The glycocalyx is made up of: (Figure 3)

1. Proteoglycans: These are centrally arranged proteins with a domain attached to the basement membrane through glypicans, syndecans, and other proteoglycans. They are usually soluble and released into the bloodstream with mimecans, perlecan, and biglycan.

2. Glycosaminoglycans: It is a grid made up of multiple members such as: chondroitin sulfate, heparan sulfate, dermatan sulfate, keratan sulfate, and hyaluronic acid; these are attached to the proteoglycans in the form of side chains and serve as binding sites for other proteins such as albumin, anticoagulant factors (antithrombin III, Heparin cofactor II, Thrombomodulin, and tissue factor pathway inhibitor), and antioxidants (superoxide dismutase).

   There are patterns of sulfation, deacetylation, and epimerization in the disaccharides that bind glycosaminoglycans, the structural variation of glycosaminoglycans, and their effect on binding to specific proteins generates a modification in glycocalyx function, thickness, charge modification, transcellular permeability, and paracellular.

3. Glycoproteins: They are of three types:

   • L and P type selectins that bind to leukocytes activated by cytokines (IL-1, TNF alpha, lipopolysaccharides).

   • Integrins that interact with laminin and fibronectin and subcellular matrix.

   • Immunogglogulins: I-CAM 1-2, V-CAM-1, PE-CAM-1 that increase the adhesion of leukocytes and platelets and allow diapedesis, respectively.

     Von Willebrand Factor platelet receptor: Allows the Ib-IX-V, VWF, and platelet complex union.

Figure 3.

Structure of the glycocalix—capillary leak syndrome. (A) Structure of the glycocalyx. (B) Modified Starling’s law. (C) Capillary leak syndrome [78, 79, 80].

The glycocalyx plays an important role in homeostasis between the vascular wall and the blood, dampens shear forces, prevents albumin loss, prevents leukocyte and platelet adhesion, locally regulates coagulation, and has antioxidant properties (superoxide dismutase) and vasodilator (release of nitric oxide) [78].

The difference of the pressures of the hydraulic conductivity and the reflection coefficient of the proteins: Jv/A = Lp[(Pc − Pi) − σ(πC− πi)], which, according to Starling’s studies, shows that the hydrostatic pressure capillary (Pc) is 40 mm Hg and the capillary oncotic pressure (πc) is 25 mm Hg) at the arterial end of the capillaries, allowing fluid to leak from the endothelium into the interstitium; when the Pc (15 mm Hg) at the venous end will be less than πc, it will allow the reabsorption of liquid from the interstitium into the blood vessel. This theory has been modified, and the participation of the glycocalyx is incorporated through the subglycocalyx space located above a tight junction in the intercellular space, which has subglycocalyx hydrostatic pressure (Pg) and subglycocalyx oncotic pressure (πg) typical of the subglycocalyx space; the pressures that oppose liquid filtration toward the interstitium are πC – πg, which must be greater than πi, which equalizes the rate of lymphatic drainage and thus prevents interstitial edema (Figure 3). Alteration of the endothelial barrier allows albumin to move into the interstitium, increasing πi, and fluid moves into the interstitium of different organs. Crystalloids increase Pc and reduce πc, which translates into increased transcapillary filtration forces, while colloids also increase Pc but do not decrease πc. In inflammatory states, albumin allows better intravascular volume expansion, but it is not retained intravascularly and loses its oncotic effect, and its administration did not improve oxygenation in ARDS in a sustained manner, nor mortality (Figure 3) [79, 81].

In patients, sepsis, COVID-19, acute pancreatitis, and burned within the multiple etiologies, have a common denominator that is the inflammatory response, where TNF alpha, heparinase, hyaluronidase, thrombin, reactive oxygen species (ROS), metalloproteinase 15, and the Atrial Natriuretic Peptide (ANP) are described as molecules that damage the integrity of the glycocalyx, expose the selectins and integrins of the endothelium, allow the adhesion of leukocytes and platelets, and increase hypercoagulability and loss of endothelial integrity with capillary leakage, altering blood flow and altering the transport of oxygen to tissues and organs. In experimental studies syndecane-1, heparan sulfate are used as a biomarker of endothelial damage and disease severity [82].

Fluid overload has a negative impact on lung function due to the presence of extravascular water that alters gas exchange, increases the work of breathing, reduces compliance, decreases PaO2/FiO2, and decreases blood oxygen content. In a randomized study of 1000 patients, liberal vs. conservative hydration was compared in patients with acute respiratory distress syndrome (ARDS), where the cumulative balance was positive (6992 ± 502 ml) in the liberal strategy group and the conservative strategy (−136 ± 491) (p < 0.001). The conservative strategy had more free days without mechanical ventilation (14.6 ± 0.5 vs. 12.1 ± 0.5) and shorter ICU stay, but no statistical differences were found in mortality [83].

Considering the deleterious effects of FO in a multisystemic way, in medical management, with the current evidence, the use of restrictive measures in fluid resuscitation is fundamental, and a strategy that should be used is the ROSE protocol and deresuscitation [77].

Similar results are found in a systematic review and a meta-analysis evaluating the efficacy of conservative fluid strategies in adults and children with ARDS, sepsis, or systemic inflammatory response syndrome (SIRS); there were no differences in mortality with the conservative strategy vs. liberal (p = 0.83); duration in days of ventilation was shorter in the conservative group (10.13 vs. 12.6 days, p = < 0.05). In the FACCT study, a multicenter, randomized, and controlled trial of 1000 ARDS patients who are intubated and receiving positive pressure ventilation, two hydration strategies are compared in two groups, the conservative group (n = 503) and the liberal group (n = 497); there were no differences in 60-day mortality (p = 0.30), the duration of mechanical ventilation was shorter (10.37 vs. 13.59 days (p < 0.001)), shorter ICU stay within the first month (1.4% vs. 11.2% (p < 0.001)), CV failure-free days within first week 3.9% vs. 4.2% (p = 0.04), and furosemide daily dose on day 7 is higher in the conservative group (137 mg vs. 87 mg) (p < 0.001). In a post hoc analysis of the FACTT study, Liu et al. show that the incidence of AKI is lower in the conservative strategy group and in only one study of this meta-analysis, it shows that the conservative strategy decreased the days of KRT dependence (p < 0.05) [83, 84].

In the REVERSE-AKI trial, a multinational, randomized, and controlled study of 100 critically ill patients with AKI according to KDIGO criteria who received fluid resuscitation was randomized into two groups, the restrictive fluid management group (n = 49) and the usual care group (n = 51). The primary outcome as cumulative fluid balance at 24 (−416 ml)(p = 0.004) and 72 h was lower in the restrictive group (−1080 ml) (p = 0.033); cumulative fluid balance at ICU discharge /day 7 was less in the restrictive group (−2166 ml) (p = 0–043); fewer patients in the restrictive group required KRT (13% vs. 30%) (p = 0.043); there were no differences in the duration of AKI [85].

In a retrospective cohort of 172 patients of whom 21.8% had FO, the median fluid accumulation in the FO group was 8.6% (12,644 mL) versus 0–4% (5976 mL) in the non-FO group. In the multivariate analysis, the predictor variables for the development of FO were surgery prior to ICU admission (OR 2.35), septic shock (OR 2.05), need for mechanical ventilation on ICU admission (OR 1.56) and planned ICU admission (OR 1.70), baseline lactate (OR 1.28) on day 3 of admission to the ICU. The Random Forest and Boruta, Classification Decision Tree, and Fast and Frugal Tree models show that the highest risk of FO is found in patients with lactate ≥2.6 mmol/L, bicarbonate <19.0 mmol/L, surgery prior to admission, baseline creatinine >156 μmol/L (> 1.7 mg/dl), and APACHE IV of ≥36. This proposal of FO phenotypes based on the variables described is interesting, but it has limitations as it is a retrospective, single-center study with possible bias [76].

In recent years, pulmonary ultrasound has gained a lot of ground in the assessment at the patient’s bedside and has been very useful in the diagnosis of pulmonary congestion. In a prospective cohort of 50 critically ill patients with different pathologies, 4 parasternal territories were evaluated with ultrasound (US) with a score of 0 to 32 according to the degree of pulmonary congestion, and extravascular lung water index (ELWI) measurements were also performed; a close correlation was observed between the ultrasound score and the EVLWI (Spearman’s r = 0.91, p < 0.0001). A US score > 18.5 correlates with a severely increased EVLWI >15, with a sensitivity of 92.3% and a specificity of 94.6% (AUC ROC = 0.9636). The correlation of chest radiographs with the EVLWI was weak; therefore, US is an excellent alternative to evaluate FO in the ICU [86].

The harmful effect of FO on the kidneys begins with the use of unbalanced solutions rich in chlorine that generate the release of adenosine that causes vasoconstriction of the afferent arteriole; a decrease in renal perfusion is described, seen in magnetic resonance with the use of 0.9% saline [87]. FO causes interstitial edema in the kidneys when the lymphatic drainage capacity is exceeded, and the kidney, being an encapsulated organ, does not have space to expand, consequently an increase in intracapsular pressure is generated, known as “renal tamponade,” leading to altered tissue oxygenation worsening of venous congestion which reduces the transrenal pressure gradient for renal blood flow (RBF) and reduces arterial blood flow affecting glomerular filtration rate and also impacts renal recovery [88]. Rapid infusion of solutions increases atrial pressure, allowing the release of ANP, which is described as one of many determinants in the degradation of endothelial glycocalyx-syndecan-1 [89].

Liberal resuscitation with crystalloids in terms of total volume as well as volume administration rate are risk factors in patients with sepsis, pancreatitis, burns, and polytraumatized patients for increased intra-abdominal pressure (IAP). When PIA values are ≥12 mmHg in a sustained manner, it is defined as intra-abdominal hypertension (IAH), and when sustained PIA values are >20 mmHg, it may or may not compromise the Abdominal Perfusion Pressure (APP = TAM-PIA); if the values of APP are <60 mmHg, it is associated with new organ dysfunction or failure, and this clinical situation is called abdominal compartment syndrome (ACS) [80].

The increase in intra-abdominal pressure elevates the diaphragm, which causes an increase in intrathoracic pressure, central venous pressure, and pulmonary capillary pressure, and reduces cardiac output (CO). The increase in CVP allows the increase in intracranial pressure, generating a decrease in cerebral blood flow. The IAH produces compression of the inferior vena cava; it also reduces renal blood flow and, together with the decrease in cardiac output, are factors for the development of AKI. The impact of IAH extends to the intestinal level with decreased intestinal capillary perfusion, causing ischemia, bacterial translocation, and increased release of cytokines [90].

The management of IAH/ACS has three pillars: measuring the magnitude of intra-abdominal pressure, identifying the development of organic dysfunction, and identifying the etiology. Management is focused on resolving the medical or surgical cause; the use of diuretics is a treatment option, but the decrease in cardiac output and resistance to diuretics are factors that limit their usefulness. KRT is an alternative and has an effect on the decrease in intra-abdominal pressure, described in a retrospective cohort study of nine critically ill patients where changes in IAP, global end-diastolic index (GEDVI), and EVLWI were evaluated through the PiCCO monitor in those receiving dialysis, Sustained low-efficiency daily diafiltration (SLEDD), or continuous veno-venous hemofiltration (CVVH) with the objective of achieving a negative cumulative balance. It was observed that Δ IAP decreased per dialysis session in a range of −1.4 mmHg (p < 0.0001); decrease in GEDVI after dialysis was 830 ml/m² (range 628 to 1199 ml/m²) (p = 0.008). The reduction in EVLWI was very modest at 1 mL/kg or 65 mL for a mean fluid loss of 1.9 L. This slight reduction in EVLW is said to be due to greater extravascular water mobilization from other regions than from the lungs, and the presence of capillary leak syndrome perpetuates the movement of fluid into the pulmonary interstitium. There are many limitations to this work, ours being a small, heterogeneous group with cardiogenic and noncardiogenic pulmonary edema [88, 91].

The use of diuretics is an important alternative in the management of FO. In a multicenter, randomized, controlled trial of 59 fluid-overloaded adult patients admitted to the ICU (radiographic evidence of pulmonary edema, clinical signs of volume overload in association with elevated central venous pressure > 16 mm Hg or pulmonary wedge pressure capillary >16 mm Hg). They were randomized into two arms of furosemide administration, bolus group (n = 27) vs. continuous infusion group (n = 32); diuresis in 24 hours was similar 5.3 L vs. 5.4 L (p = 0.64); in the bolus group a higher dose of furosemide (24.1 mg/h) was required versus 9.2 mg/h in the infusion group (p = 0.0002) in the first 24 hours; and after 24 hours, there were no differences. Mean urine output per furosemide dose was higher in the continuous infusion group (31.6 ml/mg versus the 18 ml/mg bolus group) (p = 0.014) in the first 24 hours, and after 24 hours, there were no differences; the values in serum creatinine between the beginning and at the end of the therapy with furosemide did not have differences [92].

Extrapolating from the results of the DOSE study in patients with acute decompensated heart failure, we compare the administration of loop diuretics in continuous infusion versus intermittent bolus and low versus high doses in terms of achieving symptom improvement and deterioration of renal function at 72 hours. There were no statistically significant differences in the global assessment of symptoms in the intermittent bolus or continuous infusion groups, or between the low and high dose groups, and the high dose tended to be better than the low dose (p = 0.06). High-dose use resulted in an increase in creatinine >0.3 mg/dl at 72 hours as a positive effect of decongestion compared with the low-dose group, but there was no difference in kidney injury rates at 60 days (4 vs. 9%) [93].

In critical patients with resistance to loop diuretics, the dose can be increased exponentially or sequential blockade of the tubules with acetazolamine, thiazda, and/or spironolactone with the aim of improving the diuretic response, another option widely used empirically is the use of furosemide and human albumin [94, 95].

In a meta-analysis of 13 studies of heterogeneous quality, where they recruited 422 patients and the administration of furosemide with albumin increased the mean diuresis rate to 31.45 ml/hour (p = <0.01), and increased natriuresis by 1.76 mEq/hour (p = <0.01). In the subgroup analysis, it was observed that those with albumin values <2.5 gr/dl; the use of higher doses of albumin (>30gr) had a better effect in terms of diuresis and natriuresis. A trend of better diuresis is observed in eGFR lower than 60 ml/min/1.73 m2 without statistical relevance (p = 0.1) [96].

In cases of significant FO, acute pulmonary edema refractory to diuretics with PaO2/FiO2 < 300 mmHg, a resource in this clinical scenario, is Slow Continuous Ultrafiltration (SCUF) and/or CKRT depending on the clinical requirement, and these are alternatives that have been extensively studied with conflicting results. In a retrospective cohort study of 98 critically ill patients who required ECMO, 85% of them developed AKI, and of these, 49% required CKRT; in those who had FO > 10% 72 hours after connection to ECMO and developed severe AKI, 90-day mortality was higher when compared to those who did not develop AKI (HR 2.2; 95% CI, 1.3 to 3.8; p = 0.004) and those in the subgroup requiring CKRT had higher mortality (p = 0.029), and it is observed that CKRT does not ensure a negative fluid balance, but helps to achieve a less positive balance [97].

In the ELAIN study, fluid overload was defined as worsening pulmonary edema, PaO2/FiO2 < 300 mgHg or fluid balance >10% of body weight. In this study, the positive cumulative balance on the 3rd day after randomization was 2773 ml for the early group versus 2207 ml for the late group; these positive cumulative balances are similar to the restrictive hydration groups of other works, and these volumes could have little impact on mortality. The parameters used in the AKIKI study for fluid overload correspond to the presence of acute pulmonary edema due to fluid overload causing severe hypoxemia defined by the need for >5 L/min of oxygen to achieve SpO2 > 95%, or FiO2 requirement >50% in mechanically ventilated patients; these parameters can be subjective to define them at the bedside of the patient [55, 56].

In the IDEAL-ICU study, the definition of fluid overload corresponds, pulmonary edema due to fluid overload refractory to the use of diuretics; the accumulated positive balance on day 7 of hospitalization in the ICU in the early strategy group was 5570 ± 8761 (1%) and in the late strategy group was 5878 ± 7472 (4%), and there were no differences in terms of 90-day mortality between both strategies [57].

In the STARRT-AKI in the accelerated group, the cumulative fluid balance was positive (2.7 L) at the start of dialysis, and 97% of the patients in this group received KRT. In the standard group, the balance was more positive (5.9 L), and only 62% received KRT. The most common reason for starting dialysis in the standard group was volume overload with a PaO2 /FiO2 < 200, and KRT did not impact on improved survival [58].

It is possible that in the AKIKI, IDEAL-ICU, and STARRT-AKI studies, FO is less than 10%, so that KRT has little impact on better survival regardless of the timing of the start of KRT. The use of an agreed definition of FO and the use of ultrasonographic evaluation methods or EVLWI measurement techniques will allow a better interpretation of this variable and a better evaluation of its impact on mortality.

4.4 Uremia

For many years, it has been defined as a “toxic syndrome due to kidney damage associated with changes in the tubular and endocrine function of the kidney, characterized by the retention of toxic metabolites and accompanied by changes in the volume and composition of body fluids and excess or deficiency of various hormones” [98].

Urea measurement is not an ideal biomarker to assess uremia; it is a metric that is used as a proxy for authentic uremic toxins. The presence of uremic toxin has been extensively studied in CKD, but in AKI the evidence is scant.

Uremic toxins are metabolite products generated by the diet and are excreted by the kidneys under conditions of preserved renal function. AKI generates uremic toxins that rapidly affect nonrenal organs such as the brain, lungs, heart, liver, and intestines. Only 30% of CKD uremic toxins have been studied in AKI; these toxins have a biological effect with an impact on non-renal organs, renal recovery, and mortality [99]. AKI is characterized by reduced excretory capacity and leads to the accumulation of metabolic products that interact with the blood-brain barrier, intestinal microbiota, lung epithelium, and the cardiovascular system [100].

Uremic toxins are classified into three groups according to the European Working Group on Uremic Toxins (EuTox) (Table 6) [100].

Water-soluble compounds without protein binding (< 500 Da)	Protein-bound compounds; (< 500 Da)	Intermediate molecules (>500 Da)
ADMA1 Guanidine Uric acid Guanidinosuccinate Creatinine and urea Dimethylglycine N-Methyl-pyridone-carboxamide Uric acid Dimethylguanosine Nitrosodimethylamine Uridine Erythritol	Homocysteine Indoxyl sulfate p-Cresyl Homocysteine p-OH-hippurate 2-Methoxyresorcinol Indole-3-acetate Pentosidine 3-Deoxyglucosone Phenol	β2-Microglobulin Atrial natriuretic peptide Endothelin Hepcidin Interleukin-1β Interleukin-6 Tumor necrosis factor α Interleukin-18 Complement factor D

Table 6.

Classification of uremic toxins [101].

5. Water-soluble compounds without protein binding

In AKI, dimethylarginine dimethylaminohydrolase (DDAH) activity decreases, which is an enzyme produced in renal tissue that degrades asymmetric dimethylarginine (ADMA), the latter is a competitive inhibitor of NO synthase (NOS), thus reducing the formation of nitric oxide (NO), which is expressed by a decrease in urinary sodium and nitrite/nitrate excretion [101]. The kidneys under normal conditions metabolize ≈260 μmol (≈50 mg) of ADMA and excrete ≈60 μmol, during AKI the increase between 1 and 3 μmol/L [100]. The increase in ADMA is associated with endothelial damage through the p38 MAPK/caspase-3 pathway that regulates the organization of the actin cytoskeleton and intercellular junctions by interrupting VE-cadherin [102] of the systemic vascular endothelium and at the in the lung there is severe epithelial hyperpermeability and pulmonary edema [103]; contractility and heart rate also decrease, and pulmonary artery pressure increases due to loss of NO vasodilator capacity [99]. In AKI, proximal tubule injury, heat shock protein 27 (MAPK/HSP 27) is released, which activates p38 protein kinase, which is involved in increased vascular permeability, epithelial edema, and pulmonary capillary congestion [104]. It also decreases contractility and heart rate and increases pulmonary artery pressure due to loss of NO vasodilator capacity [101].

The increase in uric acid is related to the reduced excretion due to decreased eGFR; it is described that uric acid has a vasoconstrictor effect on afferent arterioles and inhibits the release of NO in endothelial cells and proinflammatory activity because it induces the release of MCP-1, ROS and activation of NF-KB and p38-MAPK [100, 104].

6. Protein-bound compounds

In models of acute tubular necrosis, proximal tubular cell injury results in loss of expression of OAT1/3, which are transcellular transporters, and loss of protein-bound uremic toxin secretion and reabsorption function. Accumulation of protein-bound uremic toxins may continue despite renal recovery [105].

In recent years, the concept of the kidney-intestine axis has gained much importance, due to the fact that dysbiosis is generated in AKI with a decrease in Lactobacilli Ruminococacceae and an increase in Enterobacteriaceae and Escherichia coli, the latter being responsible for the generation of p-precursors cresol and indole sulfate, and intestinal permeability is altered, and the inflammatory response is amplified [105].

6.1 Indoxyl Sulfate (IS)

It is a product of tryptophan metabolism that is eliminated by secretion from the proximal tubule and accumulates in AKI. At the pulmonary level, it decreases the expression of Aquaporin-5 and Na/K+ ATPase, altering alveolar clearance mechanisms and causing thickening of interstitial lung tissue. Is blocks a K+ channel, delays cardiac repolarization, and prolongs the QT interval with an arrhythmogenic effect, and also causes endothelial progenitor cell dysfunction [100]. At the renal level, it increases the formation of ROS, stops the tubular cell cycle in the G2M phase, damages the DNA of tubular cells, and activates fibroblasts that induce progression to CKD, and there is a correlation with the severity of renal injury [105, 106].

6.2 P-cresol sulfate

It originates from the p-cresol precursor, which is formed by bacterial fermentation of proteins in the large intestine and which increases the adhesion of leukocytes to the vascular wall, increases vascular permeability, and is associated with decreased cardiac contractility [105].

Hyperhomocysteinemia has been described in AKI, which aggravates mitochondrial damage and may be a factor in increasing apoptosis of renal tubular epithelial cells and perpetuating the damage [100, 107].

7. Average molecular weight molecules

AKI during tubular injury, regardless of the etiology, cytokines are released by the tubules and this conditions the migration to renal tissue of immune cells such as neutrophils, monocytes, and lymphocytes; the increased immune response causes damage to renal tissue. The released cytokines weigh in a range of 5 and 20 kDa, and there is controversy as to whether the proinflammatory state that impacts multiorgan damage should be considered as uremic toxins (Figure 4) [100].

Figure 4.

Systemic effects of uremic toxins. AKI conditions retention of uremic toxins and has a deleterious effect at the multisystemic level [100, 101, 104, 105, 106].

8. Role of the gut microbiota in acute kidney injury and inflammation

In AKI, the inflammatory response of the parenchyma determines a remote signal that is expressed in the activation of neutrophils in the lamina propria of the intestine; this leads to the dedifferentiation of circulating monocytes and generates a proinflammatory macrophage phenotype at the intestinal level; this is accompanied by inflammation, increased permeability of the intestinal wall, and bacterial or endotoxin translocation; a dysbiosis is generated with a decrease in Lactobacillus and an increase in Escherichia and Enterobacter, and this imbalance leads to the loss of short-chain fatty acids (SCFAs).) that under normal conditions fulfills pleiotropic functions in the integrity of the intestinal barrier, T-reg activation mediating immunomodulation and under dysbiosis conditions; it facilitates the activation of the innate immune response [108].

Through the increase of TGF-β and IL-6 and E. coli, Th17 lymphocytes are activated in the small and large intestine; it is also mentioned that the intestinal inflammatory response is expressed with increases in the mRNA of renal tumor necrosis factor (TNFα) and interferon (IFN-γ) and IL17, which is related to inflammatory amplification and greater severity of AKI [109]. TNFα, IFN, and IL-6 are elevated after ischemic and non-ischemic AKI these cytokines generate liver damage, which in turn amplifies the inflammatory response that impacts intestinal damage and induces IL-17A release [110].

The marker used to assess uremia for many years has been urea/BUN; the cut-off value of blood urea nitrogen (BUN) for CKRT onset originates from retrospective cohort studies of several studies in which high values or values considered as “late” onset for CKRT as values >94.5 mg/dl were related to higher mortality (p < 0.0001) [110]. In another retrospective cohort study of patients with sepsis and AKI, those requiring KRT were defined as “early” onset with BUN <100 mg/dl and “late” onset with BUN values ≥100 mg/dl. In logistic regression analysis, those in the “late” group mortality at 14 days (OR 3.6, p = 0.001), 28 days (OR 2.6, p = 0.01), and 365 days (OR 3.5, p = 0.02) was higher [111]. Vin-Cent Wu et al. in another retrospective cohort study of critically ill patients with acute liver injury and AKI who required CKRT after major surgery, it was observed that those in the “late” dialysis group with BUN >80 mg/dl had higher mortality rate (p = 0.02) and lower rate of recovery of renal function (p = 0.02) compared to the “early” dialysis group [112].

In the PICARD study, a multicenter observational study of 243 patients, it was observed that in the group with high azotemia (BUN >76 mg/dl) adjusted by variables, the relative risk (RR) of death associated with the start of dialysis at a high BUN was of 1.85 (95% confidence interval: 1.16 to 2.96) [113].

In the RCTs, the urea values for the start of CKRT were taken from the reference values of retrospective cohort studies; in the ELAIN study, in the “late” start group, BUN >46.67 mg/dl had an outcome of higher mortality at 90 days, when compared to the “early” start (p = 0.03.). In the AKIKI study, the development of uremia was defined as a BUN >112 mg/dl, criteria for starting KRT in the late group; 60-day mortality did not have significant differences with the early start group (AKI KDIGO3). The AKIKI 2 study defines a limit of very late when BUN >112 mg/dl or oliguria is >72 hours. In the multivariable analysis, the risk of mortality is 65%. If we wait a long time for the decision to start KRT, this study gives us the waiting limit [48, 49, 114].

Future studies on risk assessment of AKI development, prevention measures, use of biomarkers or modulation of the inflammatory response, management of dysbiosis, and timely initiation of KRT will be important to make better decisions on AKI management in critically ill patients.

9. Evidence at the START of kidney replacement therapy

9.1 Randomized controlled trials (RCTs)

Controversy has been generated for many years at the start of KRT in AKI, with the desire to answer this question; from 2016, several RCTs were carried out in order to make better decisions and with better outcomes, then will describe the most relevant studies.

The ELAIN is a French study, with a single-center RCTs design, which included patients aged 18–90 years, with severe Sepsis with catecholamines (noradrenaline/adrenaline) at a dose >0.1 mcg/kg/min, with refractory fluid overload data (PaO2/FiO2 < 300 mgHg, FO >10% of body weight), with SOFA ≥2 and with Acute Kidney Injury - KDIGO stage 2 and NGAL >150 ng/dl. Two groups were randomized, when the “early” KRT onset (n = 112) was within 8 hours of AKI - KDIGO2 diagnosis and the “late” onset group (n = 119) within 12 hours of AKI-KDIGO3 diagnosis or when they met any of the absolute indications (Urea >100 mg/dl, oliguria <200 ml/12 hours or anuria, K > 6 and/or ECG abnormalities, magnesium >4 mmol, pH <7. 15, organ edema in the presence of AKI, resistance to diuretics). The primary outcome was mortality at 90 days showing lower mortality in the “early” group (39.3 %) vs. “late” (54.7%) with HR 0.66, and p = 0.03, in relation to secondary outcomes median duration of KRT, was lower in the “early” group (9 days) vs. “late” (25 days) (p = 0. 04); improved recovery of renal function at day 90 in the “early” group was 53.6% (p = 0.02) vs. 38.7% in the “late” group; median duration of mechanical ventilation was shorter in the early group (p = 0.02); the same group had shorter hospital stay (p = <0.001); in the rest of outcomes, there were no significant statistical differences. This is a single-center study with our patients mostly surgical; not all treatments were standardized, and it generates bias in internal validity. On a positive note, it is worth noting that CKRT was performed and then transitioned to SLED and intermittent haemodialysis. These results of the benefit in the “early” start are related to other similar results in patients operated on cardiac surgery, where the temporality is different with respect to other etiologies of AKI [55].

The AKIKI Trail, a French study, is a multicenter, unblinded, RCTs design, which included critically ill patients >18 years with invasive ventilation and/or catecholamine infusion, with AKI within 6 hours after validation of KDIGO stage 3 renal injury (defined by creatinine >354 μmol/liter or 3 times baseline, Anuria (<100 ml/day for >12 hours) and Oliguria (diuresis <0.3 ml/kg/h or < 500 ml/day for >24 hours) compatible with tubular necrosis. Two groups were randomized; the “early” KRT initiation group (n = 312) was within 6 hours after AKI-KDIGO3 diagnosis and the “late” KRT initiation group (n = 308) when they developed absolute indications (urea >40 mmol/L, K > 6 mmol/L, pH < 7.15; acute pulmonary edema with severe hypoxemia and oliguria/anuria >72 hours). The primary outcome was mortality at 60 days and there was no statistical difference between groups (p = 49); in the secondary outcome, the “early” group received KRT in 98% of patients unlike the other group (51%); there were more catheter-associated infections in the “early” group (10%) vs. 5% in the late group; in other outcomes analyzed, there were no significant statistical differences. A strength is that 49% of patients in the “late” group did not require KRT and recovered renal function (50). This work was not blinded, and this generates bias. It is noteworthy that 50% of the critical patients performed KRT in intermittent modality, and this is not a common practice in the clinical condition where 80% of the patients required vasoactive agents, and it is not a common practice in the country where the study was carried out. This study is not comparable to ELAIN, because a greater number of patients who were included in the AKIKI were medical causes of hospital admission and the most frequent cause of AKI was sepsis, which has a different pathophysiological context than the cardiac surgery patients in the ELAIN Trail [56].

In the IDEAL-ICU is another French study with a multicenter, unblinded RCTs design, which included adult patients with septic shock within 48 hours of vasopressor initiation and AKI defined and classified according to RIFLE criteria in grade F; randomization was performed in two branches: those who initiated KRT “early” (n = 239) within 12 hours of AKI-RIFLE(F) diagnosis and those who initiated KRT “late” (n = 238) was within 48 hours of AKI-RIFLE(F) diagnosis or in the presence of absolute indications (K > 6. 5 mmol/L, pH < 7.15, fluid overload refractory to diuretics). In mortality outcomes at 28 and 90 days, there were no significant statistical differences (p = 048 and p = 38, respectively), median days free of KRT, mechanical ventilation, vasopressors, and positive cumulative balance at 7th day; there were no differences between groups, but a higher incidence of hyperkalemia was identified in 4% (p = 0.03) in the “late” start group. Delayed onset allowed time for spontaneous recovery, and only 29% of this group did not require KRT [57].

STARRT-AKI Trail is a multinational RCTs with larger number of patients enrolled (n = 2927); participants were randomized 1:1 for accelerated versus standard KRT initiation. >18 years of age, admitted to an ICU with AKI defined as Cr > 100 umol/l in women and Cr > 130 umol/l in men and who have not decreased Cr by >27 umol/l in 48 hours and with evidence of severe AKI with at least 1 of the following 3 cases: >2-fold increase in creatinine from baseline, Cr > 354 umol/l and >27 umol/l above baseline creatinine and with urine output < 6 ml/kg in 12 hours. Two groups were randomized: “accelerated” (n = 1462) was termed when starting KRT within 12 hours after AKI-KDIGO3 diagnosis and the standard group(n = 1465), when AKI persists >72 hours from randomization and one or more of the classic indications: K > 6 mmol/L, pH < 7.2 or HCO3 < 12 mmol/L, acute pulmonary edema (P/F < 200). In the analysis of the outcomes, it is evident that in mortality at 90 days, there were no differences between both groups (p = 0.92); KRT dependence at 90 days was higher in the accelerated group (10.4%) vs. 6% (RR1.74), and adverse events such as hypotension and hypophosphatemia were higher in the accelerated group, all-cause mortality, KRT dependence at 90 days, MAKE, glomerular filtration, albuminuria, mechanical ventilation, and vasoactive-free days, no significant differences were found between the two groups. This study recruited the largest number of patients of all the RCTs and has internal and external validity, which is applicable to real-world life and has significant results at the time of decision making [58].

The AKIKI2 study is from the group of S. Gaudry et al. and is an unblinded, multicenter, prospective, randomized, and controlled trial. It included patients hospitalized in ICU who received mechanical ventilation and/or catecholamine infusion with AKI according to KDIGO definition. Randomization was performed in two groups, “delayed” group (n = 137) when KRT onset occurs with AKI-KDIGO3 with KDIGO with oliguria: diuresis <0.3 ml/kg/h or < 500 ml/d) or anuria (diuresis <100 ml/d) for >72 hours or azotemia: BUN >112 mg/dl (40 mmol/l) and 140 mg/dl (50 mmol/l) and the “more delayed” group (n = 141), those with absolute indications mentioned in the AKIKI or urea >140 mg/dl (50 mmol/l). In the primary outcome was number of days without KRT from randomization to day 28; there were no differences between groups (p = 0.93); in terms of ICU, in hospital mortality at 28 and 60 days, there were no statistical differences (p = 0.26, p = 0.71, respectively), the number of patients with renal recovery, days free mechanical ventilation, vasoactive agents, ICU-hospital stay time, dialysis dependency time, infection rates, and positive cumulative balance was similar in the two groups. In the multivariate analysis, the risk factors associated with 60-day mortality were more delayed initiation of KRT (HR 1.65), mechanical ventilation (HR 3.44), and SAPS III (HR 1.03). The strength of this study is that it places an “upper limit” for dialysis initiation as a mortality risk point [114].

9.2 Meta-analysis

In a systematic review and meta-analysis, they included 56 studies, and of these, 10 were RCTs; 4753 critically ill patients with severe AKI were included, where all-cause mortality was 45.5% in the accelerated group vs. 46.6% in the standard group (p = 0.46), without differences in mortality at 28 and 90 days; there were no differences in dependence between both groups (p = 0.08). In the analysis by subgroups, there was a greater dependence on KRT in early-onset patients with SOFA >11 and in mixed dialysis modality (CKRT and intermittent hemodialysis). The subgroup of surgical patients had lower mortality when receiving CKRT and less dialysis dependence than the subgroup of nonsurgical patients; the risk of dialysis dependence was increased in the accelerated KRT group when those patients used non-CKRT modality or had high SOFA scores. This meta-analysis was methodologically flawless and low in publication bias [115].

Another meta-analysis includes 18 RCTs with 2856 patients and focused on the timing of KRT initiation (accelerated vs. standard), and there were no differences in mortality (p = 0.9). In the subgroup of high-quality RCTs, there were no statistical differences (p = 0.7) and there were no differences in mortality at 28 and 90 days, mortality was similar in the CKRT and intermittent hemodialysis modalities, and mortality was similar in the AKI groups. In the community and in the ICU, it is reported that there was greater dependence on KRT in the accelerated start group and in four studies catheter-associated complications and infections were observed [116].

Wei-Ting Lin et al. report a meta-analysis of 11 RCTs assigned to early (n = 1131) and late (1111) groups; mortality at 28, 60, and 90 days was similar in both groups; in ICU and hospital mortality, there were no statistical differences. In four studies of surgical patients, mortality was lower in the early-onset group and is supported by other studies with similar results in the same population, but the dependence of KRT at 28 and 90 days was similar between groups and did not improve renal recovery either. Regarding adverse events, infections and hypophosphatemia were more frequent in the early-onset group. In this study, the heterogeneity of the studies was high [117]. In another meta-analysis of 15 RCTs, they included 5395 patients who showed similar results in that there were no significant differences in mortality at 28, 60, and 90 days, without differences in terms of ICU and hospital stay, with more episodes of hypotension and infections in the early-onset group [118].

Another meta-analysis of 14 RCTs where 5234 patients were recruited and compared early vs. late initiation of KRT, no differences were found in mortality at 30 (RR 1.0) and 90 days (RR 1.0); there was greater dependence on KRT; hypotension and hypophosphatemia in the early-onset group [119].

The current evidence supports the conservative approach of waiting for classic indications, considering a time limit of more than 3 days of anuria/oliguria or urea >240 mg/dl for the start of KRT, given that a longer delay increases the risk of mortality [114, 119].

10. Other non-classical indicators that should be taken into account when deciding at the START of the KRT

10.1 Patient severity

Patient severity influences the risk of developing AKI, as evidenced in a prospective cohort study of 33 surgical patients admitted to the ICU, 22 of whom had sepsis and AKI at different stages. The biomarkers SOFA, APACHE III and serum and urinary NGAL these last two biomarkers had an area under the curve (AUROC) of 0.98, AUROC of 0.885 respectively, the SOFA score together with serum and urinary NGAL reach an AUROC 0.963 to predict AKI and mortality [120].

It is a retrospective cohort of 90 patients who are divided into two groups: the survivors and the non-survivor group. In the non-survivor group at the start of KRT, the APACHE III value, vasoactive-inotropic score (VIS), and lactate were higher than in the survivor group; the APACHE III scores had an AUROC of 0.866 and the VIS AUROC of 0.796; the SOFA had an AUROC of 0.732, as predictors of mortality. In the multivariate analysis at the beginning of KRT, APACHE III had an OR 1.22, VIS an OR1.147, low MAP before KRT had OR 1.17, lactate before KRT had OR 1.55; time from diagnosis to “late” start of KRT reached OR 1.014; all were independent risk factors for mortality in AKI with KRT. Variables other than the classical ones are possible (uremia, pH, bicarbonate and K) and should be considered and help to choose the moment to start KRT in critically ill patients [121]. Regarding VIS, a retrospective cohort showed in the multivariable logistic regression analysis that VIS was associated with postoperative AKI (OR 1.19) (p < 0.001) and with the need for KRT (OR: 1.29, p = 0.007). The VIS AUROC is a good predictor of postoperative AKI (AUROC: 0.84, p < 0.001) and predictor of the need for KRT (AUROC: 0.91, p < 0.001) [122]. Further studies are required to validate these data.

Lactate is a good biomarker to assess tissue perfusion and energetic-metabolic status; in different clinical circumstances, production may be normal or increased with defective or close to normal clearance. The increase in lactate has been widely studied as a predictor of mortality in sepsis and septic patients receiving KRT [123].

It is important to mention that lactate clearance is 1379 ml/min under normal conditions and the clearance capacity for a hemofilter is on average 24.2 ml/min, and the hemofilter represents <3% of lactate clearance, which is not sufficient in patients critical and should focus on treating the primary cause of decompensation (Figure 5) [128].

Figure 5.

Classic and severity criteria in the critically ill patient for initiation of KRT. Classically known indications for KRT are when complications are generated in severe AKI. We mention that AKI is a pathology with systemic impact and it is important to observe other severity factors in the critically ill patient that allow us to make the best decision on when to initiate KRT [120, 121, 122, 124, 125, 126, 127].

In a prospective cohort of 186 patients with sepsis and septic AKI (S-AKI) who received CKRT-CVVHDF, serum lactate value was assessed before initiation, 24 hours after CKRT, and percentage lactate clearance. They were divided into a group of survivors and non-survivors; in this last group lactate at the beginning, at 24 hours, it was significantly higher (p = <0–001), and lactate clearance was <10% (p = 0.004). The lactate value 24 hours after starting the CKRT is associated with early mortality (48 hours) with OR 1.72 and late mortality (28 days) with OR 2.35; in those in which clearance is >10%, it is associated with lower early mortality (OR 0.114) and late (OR 0.235); the AUROC values for lactate at 24 hours predict early mortality 0.87 and late mortality 0.82; the AUROC for lactate clearance as a predictor of early mortality was 0.72 and late mortality 0.70, respectively [124].

In a retrospective cohort of 342 patients divided into three groups: early recovery, early death group, and control group, the multivariate logistic regression analysis identified factors that can predict the recovery of renal function in the first 48 hours, such as the presence of diuresis. (AUROC 0.64), SOFA <10 (AUROC 0.67), and short duration (0.3 days) between ICU admission and initiation of CKRT (AUROC 0.68); all three factors predict recovery of renal function with AUROC 0.78. In the group with early mortality, they presented SOFA values >13, SAPSIII>74 points, neurological disease (OR 9.64), use of vasopressors (OR 3.68), lactate >3.6 mmol/L with a sensitivity of 88%, and a specificity of 67% (OR 1.19).), albumin <2.2 g/dl (OR 0.52) as predictors of mortality and capable of predicting a group of very seriously ill people who do not benefit from KRT [129]. In the post hoc study of the AKIKI and IDEAL-ICU clinical trials in the stratified analysis of thirds of the SOFA score, those with SOFA >10, the CKRT did not show a decrease in mortality at 60 days [125].

In a prospective cohort of 999 patients with S-AKI requiring dialysis, three patient phenotypes were identified through the Manhattan plot of the standardized differences of the clinical characteristics evaluated. Phenotype 1 is characterized by young patients, low Charlson comorbidity index (CCI), normal renal function, low Glasgow, low Po2, low PaO2/FiO2, high lactate and SOFA, APACHE III, unlike phenotype 2 with intermediate characteristics and phenotype 3 with older patients, altered renal function, and less severe acute disease with low lactate and SOFA. Phenotype 1 presented a higher risk of mortality 73.86% and when compared with group 2 (56.57%) and group 3 (46.22%) (p < 0.001); also, a lactate >3.3 mmol/L at the start of KRT is associated with mortality (HR 1. 34), and KRT-dependent (HR 0.69) could be a predictive biomarker for dialysis initiation or SA-AKI severity, and further research is required to confirm this hypothesis [126].

In a prospective cohort of 500 patients with type 1 cardiorenal syndrome (type 1 CRS) who required KRT, serum lactate measurement may be a marker of hemodialysis withdrawal and 90-day mortality. qSOFA values were higher (>1) in the mortality group and hemodialysis dependence group; lactate values >4.2 mmol/L were associated with higher 90-day mortality (p = <0.001) and lower probability of withdrawal of dialysis (p = <0.001) in the presence or absence of sepsis [127].

In relation to what has been reviewed, it should be considered that patients with diuresis, SOFA <10, and lactate <4.2 mmol/L, are likely to have early renal recovery and do not require KRT.

10.2 Furosemide stress test (FST)

Diuretic use has been controversial in AKI; Chawa et al. show in a retrospective, prospective cohort that administering a furosemide bolus of 1 mg/kg or 1.5 mg/kg in those receiving diuretics is one way to assess functional reserve. In the presence of AKI, the use of FST with a value of <200 ml of diuresis in 2 hours is a predictive marker of AKI-AKIN III progression with AUC 0.87 (p = 0.001) with a sensitivity of 87.1% and a specificity of 84. 1% and as a predictor of the need for KRT with AUROC 0.86 and mortality with AUROC 0.70. Koyner et al. compare the use of FST and biomarkers or biomarkers alone to assess AKI progression; the combination of FST and biomarkers predicts progression to AKI-AKIN III with AUROC 0.90 and predicts the need for KRT with AUROC 0.91. This furosemide challenge is useful to differentiate those without functional reserve and consider initiation of KRT [52, 130].

In a randomized, multicenter, and controlled trial of 162 patients with ARF at any KDIGO stage without emergency indications and without contraindications, the use of FST allowed the evaluation of two groups, responders and nonresponders; the latter group was divided according to the time of KRT initiation into two groups: the early-onset group within 6 hours of ARF-KDIGO diagnosis 1,2,3 and the late-onset group when initiation is due to absolute indications. FST discriminates patients with AKI who may potentially require KRT; in nonresponders, early or late onset of KRT did not affect mortality [131].

The diagnostic accuracy of FST for AKI progression had an AUROC of 0.88 (sensitivity 0.81 and specificity 0.88); as a predictor for KRT initiation, it has an AUROC of 0.86 (sensitivity 0.84 and specificity 0.77); FST performs better as a predictor of need for KRT in AKI-KDIGO 1–2 compared to AKI-KDIGO 3 with a pooled diagnostic accuracy of 0.86 [132].

In a prospective, double-blind, and interventional cohort study of 187 patients admitted to the ICU with AKI undergoing FST, 37.5% of patients who responded to FST received CKRT, while 89.2% of patients who did not respond to FST received CKRT. On univariate analysis, platelet count was lower in the CKRT group (p = 0.04); more patients with acidosis were identified in the CKRT group (p = < 0. 05); there were more patients with AKI-KDIGO 2 and 3 in the CKRT group and a higher number of stage 1 patients in the non-KRT group; urinary volume was lower in the CKRT group (35 ml, IQR 5–143. 75 vs. 400 ml, IQR 210–890; p = 0. 000), SOFA and APACHE II were lower in the non-KRT group (p = 0.000); SOFA and APACHE II scores were higher in the CKRT group (p < 0. 05); in multivariate analysis, negative FST (diuresis <200 ml in 2 hours) was a predictor of CKRT initiation (p = 0. 032). Patients who did not respond to FST were 2.379 times more likely to initiate CKRT (p = 0. 000). Post-STF urine volume of 156 ml had an AUROC of 0.966 (sensitivity 94.85%, specificity 98.04%, p < 0.001), and SOFA (>8) had an AUROC of 0.846 predictors of CKRT initiation [133].

In a prospective, observational study of 312 patients in a medical ICU, those who developed AKI according to KDIGO diuresis criteria were subjected to receive Sequential Nephron Blockade (SBN) with initial use of Furosemide at 1 mg/kg (maximum 60 mg), followed by a maintenance dose of 10 to 20 mg/hour; adequate diuretic response was termed >0.5 ml/kg/h or > 300 ml in 6 hours and 5 ml/kg/h or > 300 ml in 6 hours, if adequate diuretic response was not achieved (<0.5 ml/kg/h or < 300 ml in 6 hours). Metolazone 10 mg was administered; those nonresponders to SBN started KRT [134]. In multivariate logistic regression analysis, those with SOFA >9 (OR 4.5), those who achieved a positive cumulative balance of 4.2 L (OR 2.82), those who required KRT (OR 1.78), and those with negative diuresis (OR 0.45) had higher mortality [134].

It is important to mention that the SOFA score is a good predictor of AKI severity, poor diuretic response, and the need for KRT. More prospective studies with larger numbers of patients are required to confirm these data.

The combination of FST and elevated urine levels of the chemokine biomarker (C-C motif) ligand 14 (CCL14) had a high predictive value (AUC ROC 0.87) as a predictor of the need for CKRT, compared to FST alone (AUC ROC 0.79) or CCL14 (AUC 0.83), p = <0.001 [135].

I believe that negative FST and the use of biomarkers along with factors determining increased demands for critical illness may be useful at the time of decision making.

10.3 Kinetic eFG

It is known that creatinine is a late marker in AKI, and its value is influenced by various variables that underestimate (fluid overload, hyperbilirubinemia, malnutrition) or overestimate the value (cimetidine, fibrates, and sulfamethoxazole and trimethoprim). Determination of eGFR in the patient with AKI with the classic eGFR formulas in CKD requires a steady state of creatinine values, a condition that is not met in AKI.

The eFG kinetic (KeFG) is a formula proposed by Chen [136] incorporates the mass equilibrium principles (generation, distribution, excretion) together with an “elapsed time” factor to assign an eGFR to each value of Cr as long as the time elapsed since the last value is known. The KeFG evaluates sudden and rapid changes in eGFR, unlike the eGFR measured by Clearance calculated, where the decrease in GFR is gradual by the Crocoft, MDRD, and CKD-EPI formulas (Figure 6).

Figure 6.

Modified see end of document. KeGFR. SSPCR is steady-state plasma creatinine; crcl is MDRD creatinine clearance. Mean PCR refers to the difference in plasma creatinine concentration. Δ time(h) is the interval in hours between two consecutive creatinine measurements and max delta pcr/day is the theoretical maximum change in plasma creatinine that can occur per day if renal function ceases completely. ΔPCr is defined as the initial creatinine subtracted from the final creatinine. MaxΔPCr/day, addresses the principle of creatinine mass balance by incorporating the volume of distribution factor (V d) [136].

In a retrospective cohort of 2492 patients, the KeGFR was incorporated with or without UO criteria (diuresis) and added to the KDIGO classification with three modified stages:

Stage 1: KeGFR 45 to 60 mL/min/1.73 m² or urine output <0.5 mL/kg/h for 6 h.

Stage 2: KeGFR 30 to 45 mL/min/1.73 m² or urine output <0.5 mL/kg/h for 12 hours.

Stage 3: KeGFR <30 ml/min/1.73 m² or diuresis <0.3 ml/kg/h for 12 h or anuria for 12 h.

In this study, the degree of agreement between KDIGO and KeGFR was very good (Cohen’s kappa with square weights = 0.77). The sensitivity of the combined KeGFR and urine criteria compared to the KDIGO criteria was 93.2%, specificity 73.0%, and accuracy 85.7%, also allowing faster AKI recognition, where a time difference in recognition between KDIGO and KeGFR at stage 1 is 5.9 hours, stage 2 is 7.2 hours, and stage 3 is 4 hours. In the logistic regression model, the prediction of mortality at 28 days by KDIGO was AUROC 0.57 and for KeGFR, an AUROC 0.60; as a predictor of the need for KRT, the KDIGO had an AUROC 0.81 and for KeGFR, an AUROC 0.80. We see that the KeGFR allows a faster diagnosis of AKI using serial creatinine determinations at time intervals and with good ability to predict the severity of AKI, with the ability to predict the need for KRT and mortality. It is possible that the use of KeGFR, FST, and biomarkers provides complementary tools for the early diagnosis of AKI and allows better decisions to be made in choosing the right moment to start KRT (Figure 7) [138].

Figure 7.

Kidney replacement therapy initiation algorithm [137].

11. Conclusions

AKI leads to increased health care costs in each country, with higher morbidity and mortality. Detection of risk factors, early diagnosis, and use of severity classification systems and biomarkers are necessary for a timely preventive and therapeutic approach.

It is important to assess trends in creatinine elevation and consider the use of KeGFR along with KDIGO, the use of biomarkers, and FST as tools that give a clear scenario of renal capacity that may not be able to support the magnitude of severity of a critically ill patient.

The controversy generated about the ideal time to initiate KRT in critically ill patients is currently supported by evidence from RCTs and meta-analyses that support waiting in the classic “late” indications, taking into account the waiting limits suggested by the AKIKI trial 2.