KDIGO Clinical Practice Guideline for Acute Kidney Injury
Recommendations and Rationale

• 5.1.1: Initiate RRT emergently when life-threatening changes in fluid, electrolyte, and acid-base balance exist. (Not Graded)
• 5.1.2: Consider the broader clinical context, the presence of conditions that can be modified with RRT, and trends of laboratory tests—rather than single BUN and creatinine thresholds alone—when making the decision to start RRT. (Not Graded)

Rationale

While no RCTs exist for dialysis for life-threatening indications, it is widely accepted that patients with severe hyperkalemia, severe acidosis, pulmonary edema, and uremic complications should be dialyzed emergently. In the absence of kidney function, and when therapeutic measures that promote the intracellular shift of potassium (such as correction of acidosis with bicarbonate, glucose and insulin infusion, and beta-2 agonists) are exhausted, an excess of potassium can only be eliminated with RRT. On the other hand, when intermittent dialysis is used after these therapeutic interventions, the extracorporeal removal of potassium will be reduced and the post-treatment rebound of serum potassium will be more pronounced.⁵²⁶

Metabolic acidosis is a frequent clinical problem in patients with severe AKI. Although the discussion as to when metabolic acidosis in critically ill patients should be corrected is outside the scope of this guideline, metabolic acidosis associated with AKI can usually be corrected with bicarbonate and should rarely require urgent dialysis if not accompanied by volume overload or uremia.⁵²⁷ As the pH and bicarbonate values to initiate dialysis for metabolic acidosis are not supported by evidence, no standard criteria for initiating dialysis for acidosis exist. A variety of poisons, drug overdoses, and toxic compounds (e.g., salicylates, ethylene glycol, methanol, metformin) can contribute to acid-base problems and also lead to AKI. In these circumstances, RRT may also facilitate removal of the offending drug.^528-530

Only one RCT has evaluated the effect of timing of initiation of RRT on outcome. Bouman et al.⁵³¹ randomized 106 critically ill patients with AKI to early vs. late initiation of RRT. The early initiation group started RRT within 12 hours of oliguria ( < 30 ml/h for 6 hours, not responding to diuretics or hemodynamic optimization), or CrCl < 20 ml/min. The late-initiation group started RRT when classic indications were met. The study did not find differences in ICU or hospital mortality, or in renal recovery among survivors, but was clearly too small to allow for definitive conclusions (Suppl Table 30).

The remaining data come from observational studies. The association of early initiation of dialysis with survival benefit was first suggested by case series with historical controls conducted in the 1960 s and 1970 s.^532-535 In these studies, levels of blood urea or BUN were used to distinguish early vs. late start of dialysis. However, these studies mostly combined early start with more-intensive dialysis and late start with less-intensive dialysis. More recent studies have continued the trend focusing on BUN as a biomarker for starting RRT. Single-center observational studies that were restricted to AKI after trauma⁵³⁶ and coronary artery bypass surgery^537,538 suggested a benefit to RRT initiation at lower BUN concentrations. A prospective multicenter observational cohort study performed by the Program to Improve Care in Acute Renal Disease (PICARD) analyzed dialysis initiation—as inferred by BUN concentration—in 243 patients from five geographically and ethnically diverse clinical sites. Adjusting for age, hepatic failure, sepsis, thrombocytopenia, and SCr, and stratified by site and initial dialysis modality, initiation of RRT at higher BUN ( > 76 mg/dl [blood urea > 27.1 mmol/l]) was associated with an increased risk of death (RR 1.85; 95% CI 1.16–2.96).⁵³⁹ In a prospective multicenter observational study conducted at 54 ICUs in 23 countries, timing of RRT was stratified into ‘‘early’’ or ‘‘late’’ by median urea at the time RRT started (24.2 mmol/l [BUN 67.8 mg/dl]), and also categorized temporally from ICU admission into early (less than 2 days), delayed (between 2–5 days), or late (more than 5 days). Timing by serum urea showed no significant difference in mortality. However, when timing was analyzed in relation to ICU admission, late RRT was associated with greater crude mortality (72.8% late vs. 62.3% delayed vs. 59% early, P = 0.001) and covariate-adjusted mortality (OR 1.95; 95% CI 1.30–2.92; P = 0.001). Overall, late RRT was associated with a longer duration of RRTand stay in hospital, and greater dialysis dependence.⁵⁴⁰ It is, however, not clear whether AKI occurring later in the course of ICU stay has the same pathophysiology and prognosis than AKI present on or early after admission. The most recent study on this subject is the analysis of surgical ICU patients with AKI, showing that late initiation of RRT (defined as RIFLE-I or -F) was an independent predictor of mortality (HR 1.846; CI 1.07–3.18).⁵⁴¹

Traditional indications for RRT, developed for patients with advanced CKD, are not necessarily valid in the context of AKI. For instance, massive volume overload resulting from volume resuscitation may be an indication for RRT even in the absence of significant elevations in BUN or SCr. In this instance, it may be more appropriate to consider dialytic intervention in the ICU patient as a form of renal support rather than renal replacement. Indeed, some of the traditional indications for dialysis (e.g., uremic pericarditis, pleuritis, encephalopathy, coagulopathy) would be considered ‘‘complications’’ of AKI rather than indications for RRT. Additionally, the decision to start RRT should recognize the goals of therapy, keeping in mind the therapeutic potential of dialysis in general, and each dialysis modality in particular. The treatment of AKI with RRT has the following goals: i) to maintain fluid and electrolyte, acid-base, and solute homeostasis; ii) to prevent further insults to the kidney; iii) to permit renal recovery; and iv) to allow other supportive measures (e.g., antibiotics, nutrition support) to proceed without limitation or complication. Ideally, therapeutic interventions should be designed to achieve the above goals and a systematic assessment of all these factors is key to determining the optimal timing for initiating dialysis (Table 17).

There is increasing evidence that fluid overload in critical illness and AKI is associated with adverse outcomes, especially in the pediatric setting.^{83,84,542–549} Whether this is a causal relationship remains to be proven, although a randomized trial in hemodynamically stable patients with acute respiratory distress syndrome seems to suggest that it is.⁵⁴⁹ Randomizing patients according to RRT initiation on the basis of fluid status would allow this question to be answered. A secondary analysis of a randomized trial comparing IHD to CRRT showed that patients receiving RRT predominantly for solute control experienced better outcomes than those predominantly treated for volume overload. Patients dialyzed for control of both azotemia and volume overload experienced the worst outcome.⁵⁵⁰ Analysis of a multicenter observational cohort showed that mean daily fluid balance in AKI patients was significantly more positive among nonsurvivors than survivors.⁸⁴ Data from the PICARD group examining 396 ICU patients with AKI requiring RRT further supports these findings. Survivors had lower fluid accumulation at dialysis initiation compared to nonsurvivors (8.8% vs. 14.2% of baseline body weight; P = 0.01 adjusted for dialysis modality and severity score). The adjusted OR for death associated with fluid overload at dialysis initiation was 2.07 (95% CI 1.27–3.37).⁸³ These data suggest that fluid overload should be further evaluated as parameter to guide the initiation of RRT (see also Pediatric Considerations).

Other factors that might influence the decision of when to start RRT are the severity of the underlying disease (affecting the likelihood of recovery of kidney function), the degree of dysfunction in other organs (affecting the tolerance to e.g., fluid overload), the prevalent or expected solute burden (e.g., in tumor lysis syndrome), and the need for fluid input related to nutrition or drug therapy (Table 17). Early detection and accurate prediction of patients that ultimately will require RRT may allow earlier initiation in those who need it and, at the same time, prevent harm in those who do not. Recent evidence suggests a potential role for biomarkers in this field. Plasma neutrophil gelatinase-associated lipocalin was shown to have an area under the receiver operating characteristic curve of 0.82 for the prediction of RRT requirement.⁵⁵¹