Wait and see approach for dialysis in acute kidney injury

Abstract

Timing of dialysis initiation in critically ill patients is controversial. The STARRT-A KI trial reports that an accelerated initiation strategy did not improve 90- day survival and increased dialysis dependency compared with a standard approach in which patients had greater fluid accumulation and metabolic complications at initiation but 38% avoided dialysis.

Kidney replacement therapy (KRT) is frequently applied in critically ill patients with acute kidney injury (AKI)¹. However, considerable variability exists in the timing of initiation of KRT in the absence of life-threatening indications. A key issue is balancing the risk of an unnecessary procedure with the risk of not addressing a potentially treatable condition. In the past few years, three randomized controlled trials (RCTs) that compared early versus delayed approaches for KRT initiation based on AKI severity and associated complications have reported conflicting results (TABLE 1; Supplementary Table 1). The ELAIN study² showed a survival benefit with early versus late KRT, whereas the AKIKI³ and IDEAL-ICU⁴ RCTs did not replicate this finding.

Table 1 |.

Large randomized controlled trials investigating optimal timing of KRT in AKI

Characteristic	ELAIN2	AKIKI3	IDEAL-ICU4	STARRT-AKI5
Patients (n)	231	620	477	2,927
Setting	95% surgical	80% sepsis	Septic shock	67% medical
Criteria for early KRT	≤8 h of stage 2 AKI & NGAL ≥150 ng/ml	≤6 h of stage 3 AKI	≤12 h stage 3 AKI	≤12 h of stage 2 AKI
Criteria for delayed KRT	≤12 h of stage 3 AKI & NGAL ≥150 ng/ml	Urgent criteriaa	48 h of stage 3 AKI	AKI for 72 h or conventional criteriab
Primary outcome	90-day mortality	60-day mortality	90-day mortality	90-day mortality
Mortality: early KRT (%)	39*	49	58	44
Mortality: delayed KRT (%)	55	50	54	44
Mortality: patients on KRT in delayed group (%)	N/A	62	68	NA
Patients in delayed group who did not receive KRT (%)	9.2	49	38	38
Dialysis-dependence at 90 days (%)	13 (early) versus 15 (delayed)	2 (early) versus 5 (delayed) at 60 days	2 (early) versus 3 (delayed)	10.4 (early) versus 6.0 (delayed); RR 1.74; 95% CI 1.24 to 2.43

More details of the ELAIN, AKIKI, IDEAL-ICU and STAART-A KI trials are provided in Supplementary Table 1.

^*P = 0.03.

^aSevere hyperkalaemia (>6 mmol/l), severe pulmonary oedema refractory to diuretics, severe acidosis (pH <7.15), serum urea >40 mmol/l or oligo-anuria >72 h.

^bSerum potassium level ≥6.0 mmol/l, pH ≤7.20 or serum bicarbonate level ≤12 mmol/l, severe respiratory failure (paO₂/FiO₂ ≤200) and clinical perception of volume overload. AKI, acute kidney injury; KRT, kidney replacement therapy; NA, not available; NGAL, neutrophil gelatinase-associated lipocalin.

Additional evidence is now provided by the STARRT-AKI RCT in critically ill patients with AKI KDIGO stage 2 or higher⁵. This study compared an accelerated KRT strat-egy (initiation ≤12 h after eligibility criteria were met) with a standard approach in which KRT was postponed unless conventional indications were met (TABLE 1) or AKI lasted for >72 h. The trial design excluded patients with advanced chronic kidney disease (CKD; estimated glomerular filtration rate (eGFR) <20 ml/min/1.73 m²), uncommon causes of AKI, kidney transplants and those who needed emergent dialysis. Patients who the attending physicians considered to require either urgent dialysis or deferral of dialysis were also excluded. Consequently, 67% of 11,852 eligible patients were excluded from the trial; 2,927 of 3,019 enrolled patients were included in the modified intention-to-treat analysis

The trial groups were well-balanced at randomization. Two-thirds of patients were admitted for a medical condition, 44% had CKD, 58% had sepsis and their mean age was 65 years. The mean Sequential Organ Failure Assessment (SOFA) score was 11.7 ± 3.6, 45% of patients had oligoanuria and the mean cumulative fluid accumulation was 3% of body weight. Only 62% of patients in the standard group received KRT compared with 97% of those in the accelerated group. The median time from KRT eligibility to initiation was 6.1 h in the accelerated group and 31.1 h in the standard group. Among patients who were dialysed in the standard group, two-thirds started dialysis because of a conventional indication, 24% owing to AKI >72 h and 10% for reasons that are unclear. For clinicians, it could be interesting to determine whether the percentages of oliguria and fluid overload differed between patients who did or did not receive KRT in the standard group and whether a diuretic challenge was helpful in predicting need for KRT.

“deferring treatment to avoid KRT might be worthwhile”

The primary outcome of 90-day mortality was comparable between the accelerated (43.9%) and standard groups (43.7%), without differences in pre-specified subgroups (for example, sex, baseline eGFR, severity of illness, surgical admission, sepsis and geographical region). Causes of death were also similar between groups, although four deaths due to hyperkalaemia were reported in the standard group and no deaths due to this cause in the accelerated group.

The STAART-AKI results are similar to those of the AKIKI and IDEAL-I CU trials in that delaying KRT resulted in a substantial proportion of patients (≥38%) not receiving dialysis (TABLE 1).These findings suggest that deferring treatment to avoid KRT might be worthwhile. However, whether this

“we do not currently have reliable tools to predict the need for KRT in patients with severe AKI”

strategy can be universally applied depends on why these patients avoided dialysis. In STARRT-AKI, these details are not provided; however, dialysis avoidance could be due to kidney recovery within 72 h or to death, which was reported in 12.5% of patients who did not receive KRT in the standard group. Identification of the characteristics and outcomes of non-dialysed patients could help to guide treatment decisions. In addition, information on the causes of death in non-dialysed patients would help to determine whether these could have been altered by KRT. Observational studies have found that oligoanuria and greater fluid overload at KRT initiation confer a worse prognosis^6,7. In STARRT-AKI, median urine output was lower (350 ml versus 453 ml) and fluid overload was more common (6.7% versus 3.1%) at KRT initiation in the standard group than in the accelerated group; however, these differences did not translate into differences in outcomes. In the AKIKI and IDEAL-ICU trials, patients who received dialysis in the delayed groups had higher fluid overload and metabolic complications at dialysis initiation and higher mortality than those who were dialysed in the early group (TABLE 1).

Any treatment strategy requires consideration of complications and long-term outcomes. More adverse events were reported in the accelerated group (23%) than in the standard group (17%; RR, 1.40; 95% CI 1.21–1.62), mainly due to higher incidences of KRT-related hypotension (8.7% versus 5.6%) and hypophosphataemia (7.5% versus 4.2%). However, these data are not provided for the per-protocol analysis in the standard group, which lowers the rate of KRT-associated complications by basing them upon all patients instead of on dialysed patients. At 90 days, dialysis dependence was also higher in the accelerated group (10.4% versus 6.0%; RR 1.74; 95% CI 1.24–2.43) as was the risk of rehospitalization (RR 1.23; 95% CI 1.02–1.49).

The STARRT-AKI trial adds to the growing body of evidence that we do not currently have reliable tools to predict the need for KRT in patients with severe AKI. We are likely addressing the wrong question by trying to identify an optimal starting point for dialysis instead of determining which patients are unlikely to need therapy. In STARRT-AKI, physicians were uncertain about the need for KRT in one-third of potentially eligible patients. In the standard group, 25% of patients started therapy within 19 h, overlapping with the timing of dialysis in the accelerated group. Moreover, 24% were dialysed owing to AKI >72 h; in the absence of other criteria for dialysis initiation these patients could potentially have been deferred longer or have avoided dialysis. If we are going to focus on withholding dialysis for as long as possible, we must define parameters to guide the start and end of the waiting period, how long we can wait and the consequences on outcomes, while considering the heterogeneity of AKI and process of care⁸. The results of the reported trials clearly indicate that AKI severity and complications are inadequate in answering these questions. We need dynamic assessment of quantifiable measures to track an individual patient’s course for personalized management. One potential approach to individualize treatment is to quantify the mismatch between the demand placed on the kidneys and their capacity^9,10.

The STARRT-AKI investigators are to be congratulated on a well-designed and well-executed trial that provides new insights into a common clinical problem. Their data, combined with those of other trials, should spur us to complement clinical judgement regarding the need for or avoidance of KRT initiation with additional tools, such as demand–capacity mismatch, biomarkers of kidney damage and function, and the furosemide-stress test. Ultimately, the goal is to provide timely dialysis interventions that support patients and improve their outcomes.