Table 3. Summary of clinical practice guidelines for starting RRT in critically ill patients with AKI
Can CRRT Improve Outcome?
There is controversy about whether CRRT itself may modify patient outcomes or whether, as a supportive therapy in the setting of high illness severity, it is largely a surrogate for the effect of critical illness on outcome. Observational data have shown that receiving any RRT per se may increase mortality risk among ICU patients with AKI [6, 14, 15]. These studies have compared outcomes among patients with AKI who received or did not receive RRT. While these data have methodological limitations common to observational studies, such as differences in populations studied (i.e., case-mix, illness severity), residual confounding by indication and uncontrolled bias (i.e., provider practice variation, information bias), an honest speculation does arise about whether, in selected circumstances, CRRT may contribute hazard to critically ill patients with AKI if not carefully selected or if perceived to have only marginal benefit [16]. Patient, provider, and institutional-level factors may interact to confound the observed association between RRT and outcome; however, these factors are often not accounted for in observational data. Similarly, many of these studies may have included patients in whom CRRT was never likely to improve outcome. For example, high utilization of CRRT for patients with very low survival probability (i.e., advanced chronic illness or severe acute illness) can represent an important source of bias, as these patients will shift effect estimates of risk of CRRT toward poor outcome [17]. Alternatively, inclusion of patients with less severe AKI and marginal indications for CRRT, where the likelihood of survival and kidney recovery was high regardless of CRRT, will also confound effect estimates of the hazards of CRRT [18]. In this circumstance, it is conceivable that the risk and/or harm associated with RRT per se could potentially outweigh benefit. Interestingly, additional observational data among critically ill patients with conventional indications for CRRT have shown starting RRT may improve survival [19, 20].
Recent Trials Examining RRT Initiation Strategies in AKI
Two high-profile trials of strategies for starting RRT in critically ill patients with severe AKI were reported. The Early Versus Late Initiation of RRT In Critically Ill Patients with Acute Kidney Injury (ELAIN) trial was a single-center randomized trial including 231 critically ill patients that evaluated whether early RRT, defined as starting RRT within 8 h of fulfilling KDIGO stage 2 AKI, would improve patient survival as compared to delayed RRT, defined as starting RRT within 12 h of developing KDIGO stage 3 AKI or upon an absolute indication ensuing (Table 4) [21]. All patients in the early group and 91% in the delayed group received RRT, with a median difference in time to initiate being 21 h (interquartile range 18–24). The early RRT intervention conferred a 15.4% absolute reduction in 90-day mortality compared with delayed RRT. Early RRT was found to have greater RRT independence, shorter duration of RRT, and reduced duration of hospitalization. Early RRT also showed an early reduction in 2 pro-inflammatory mediators (interleukin-6, interleukin-8).
The Artificial Kidney Initiation in Kidney Injury (AKIKI) trial was a multicenter randomized trial that evaluated whether a delayed strategy of starting RRT improved survival among 620 critically ill patients with severe AKI (Table 4) [7]. The early strategy started RRT within 6 h of fulfilling KDIGO stage 3 AKI and the delayed strategy started RRT only in response to the development of conventional indications. The early strategy did not improve 60-day mortality; however, RRT utilization was significantly different between the strategies, with only 51% in the delayed strategy receiving RRT compared to 98% in the early strategy. The median difference for starting RRT between strategies was 57 h (interquartile range 25–83). In the delayed strategy, RRT-free days were greater and there was a reduced incidence of catheter-related bloodstream infections. There was no difference in key secondary outcomes including ventilator and vasoactive-free days through day 28, ICU length of stay, hospital length of stay, and RRT dependence at day 60.
The ELAIN and AKIKI trials are important achievements for critical care nephrology and effectively disproved the notion that well-designed trials comparing RRT initiation strategies in the ICU were not feasible. However, there are issues that clinicians should consider when determining how to interpret these findings and incorporate them into clinical practice. First, though these trials were the largest to date to examine RRT timing in critically ill patients with AKI, both were underpowered to detect small but potentially clinically important differences in mortality. AKIKI was designed to detect a 15% absolute reduction in mortality. While conceivable that a delayed strategy may translate into avoidance of RRT-related complications, such survival differences are implausible large. Similarly, ELAIN estimated a sample size based on a 55% mortality at 90-days, assuming an 18% absolute reduction in mortality. Though it demonstrated a mortality reduction with early RRT, ELAIN had a Fragility Index of only 3, implying instability and imprecision in its effect estimate. Second, the thresholds for starting RRT differed between these studies, with the early RRT arms of both trials and the delayed RRT group in ELAIN needing to fulfil KDIGO staging for AKI. This use of relatively static criteria for triggering RRT may have contributed to practice misalignment for starting RRT in both groups of ELAIN and the early strategy of AKIKI. This would raise the speculation on whether a proportion of patients who entered AKIKI and ELAIN were individuals for whom RRT would not be considered in usual practice. This highlights one of the challenges of assessing timing of initiation of RRT in critically ill patients with AKI. The absence of validated clinical or laboratory tools to reliably discriminate patients with a high likelihood of progressing to receive RRT from those in whom recovery is imminent limits the possibility to minimize exposure to unnecessary RRT. AKIKI would imply, at a minimum, that a conservative strategy of “watchful waiting” and starting RRT in response to complications or persistent AKI may be acceptable. Two large multicenter RCTs that add knowledge to this important clinical dilemma are yet to be reported (ClinicalTrials.gov identifiers: IDEAL-ICU [NCT01682590]; STARRT-AKI [NCT02568722]).
Table 4. Summary of published randomized trials evaluating the timing of initiation of RRT in AKI (adapted from [22])
Conclusions
The optimal time to start RRT in critically ill patients with AKI remains a vexing clinical dilemma for nephrology and critical care clinicians. Available evidence has highlighted the challenges and complexity of protocolizing the timing of RRT initiation strategies. In the absence of a reliable clinical tool to predict which patients will worsen and are likely to receive RRT, a patient-centered “personalized” approach that encompasses careful consideration of the overall trajectory, integrating baseline clinical information, illness acuity, burden of organ dysfunction, along with trends in physiological and laboratory data, rather than relying on absolute or arbitrary threshold laboratory values, is