Background
Acute kidney injury (AKI) is common in patients with critical illnesses admitted to intensive care units (ICUs) [
1]. The most critical patients with kidney injury may need immediate renal replacement therapy (RRT; dialysis) since AKI is potentially accompanied with lethal complications, such as severe fluid overload, electrolyte disturbances, and acidemia [
2]. However, when exactly to initiate RRT in the absence of compellingly lethal complications of AKI in critically ill patients remains unknown [
3‐
5].
The question of whether to utilize accelerated or standard initiation of RRT has been long debated within the past two decades. Previously published related meta-analyses reported potential benefits of accelerated RRT in a subset of patients, but the conclusions have not been widely accepted due to the heterogeneity in the studies and limited number of patients in the randomized controlled trials (RCTs) [
4,
6]. It had previously been concluded that early initialization of dialysis for AKI could be beneficial for surgical patients and in the setting of continuous renal replacement therapy (CRRT) [
4]. However, initializing CRRT early has not shown a definitive benefit of patient survival and kidney recovery when compared to intermittent dialysis in other reports [
7,
8].
Recently, the largest-to-date, multicenter RCT study focusing on this issue has been published, and it recruited patients globally [
9]. Therefore, herein, we combined all the available RCT data and conducted a systematic review and meta-analysis to investigate whether accelerated or standard initiation of RRT in critically ill AKI patients is beneficial in terms of several outcomes, including mortality, free of dialysis, dialysis dependence and also scrutinize their subgroup analyses.
Methods
Search strategy and selection criteria
We reported the meta-analysis according to the Preferred Reporting Items of Systematic Reviews and Meta-Analyses (PRISMA) statement [
10] and used Cochrane methods [
11]. We prospectively submitted the systematic review protocol for registration on PROSPERO [CRD42020201466] (Additional file
1: Appendix 6).
Data sources and search strategy
Electronic searches were performed on PubMed (Ovid), Medline, Embase, Cochrane library, and Cnki.net from inception to July 20, 2020. The search strategies are listed in Additional file
1: Appendix 1. We screened references by titles and abstracts and included related studies for further analysis. Reference lists of related studies, systematic reviews and meta-analyses were manually examined to identify any additional publications relevant to our analysis. Both abstracts and full papers were selected for quality assessment and data syntheses.
Inclusion and exclusion methods
We enrolled RCT studies with the following inclusion criteria: (1) studies that clearly specified participants were randomized into either control or experiment group; in (2) literature search using MeSH terms or free-texts words with acute kidney injury, renal replacement therapy, as well as the words characterized with initiation; (3) participants included critical patients with AKI who were at least 18 years of age and were not previously on dialysis; (4) assessed at least one of these outcomes: free of RRT rate, in-hospital mortality, 28-day and 90-day mortality rates after hospital discharge, and dialysis dependence rate after hospital discharge. We excluded articles that did not clearly define the timing of RRT initiations, that included participants younger than 18 years of age, lacked outcomes aforementioned, and participant randomization was not clearly defined. Full-text papers were selected for quality assessment and data syntheses.
Study selection and data extraction
Two investigators (Ying-Ying Chen and Heng-Chih Pan) independently reviewed the search results and identified eligible studies. Any resulting discrepancies were resolved by discussion with a third investigator (Chih-Chung Shiao). All relevant data were independently extracted from the included studies by two investigators (Ying-Ying Chen and Heng-Chih Pan) according to a standardized form. Extracted data included study characteristics (leading author, publication year, patient enrollment, sample size, events, duration of follow-up (weeks), the National Clinical Trial number) and participants’ baseline (age (years), gender (%), comorbidities, severity of the illness). When available, odds ratios and 95% confidence intervals (CIs) from the cohort or case-controlled studies were extracted. Other a priori determined parameters were the type of ICU setting (surgical /mixed or medical), criteria used for AKI and severe AKI diagnosis, cohort size, presence of sepsis, study quality, and the proportions of patients on mechanical ventilation. The baseline characteristics of included studies are illustrated in Table
1. The primary outcome was in-hospital mortality rates, while the secondary outcomes were free of RRT and RRT dependence. The survivors who did not need RRT at the end of the study were defined as being free of dialysis; others who were kept on dialysis were considered dialysis dependent. We also evaluated the 28-day and 90-day mortality rates after hospital discharge. Any disagreements were resolved by discussion with the investigator (Vin-Cent Wu).
Table 1
Characteristics of included comparative studies
| Multi-center mixed | Netherland | 71 (35/36) | Yes | 68.4 | 59.2 | NA | N/A | Within 12 h after fulfilling the following criteria: UOP < 30 mL/h and CCr < 20 mL/min on 3-h sample | When the patient fulfilled the conventional criteria for RRT: BUN > 40 mmol/L, K > 6.5 mmol/L or severe pulmonary edema (CVP or pulmonary artery occlusion pressure of 16 mm Hg and lung edema with positive end expiratory pressure of 10 cm H2O and PO2/FIO2 ratio of 150 mm Hg | CRRT | In-hospital mortality, In-ICU mortality, 28-day mortality |
| Single-center, surgical | Japan | 28 (14/14) | No | 64.5 | 64.3 | 57.0/38.5/NA/NA | N/A | Within 12 h of UOP < 30 mL/h or < 750 mL/day | After 12 h of UOP < 20 mL/h or urine output < 500 mL/day | CRRT | 14-day mortality |
| Multi-center, mixed | Canada | 100 (48/52) | Yes | 63.1 | 72.0 | 53.0/39.0/11.0/NA | 56.0 | Within 12 h of fulfilling eligibility: at least two of the following: twofold increase in serum Cr from baseline; UOP < 6 mL/kg in the preceding 12 h; whole-blood NGAL > 400 ng/mL | Severe hyperkalemia (> 6 mmol/L); severe pulmonary edema; severe metabolic acidosis (serum bicarbonate < 10 mmol/L) | Mixed (IHD/CRRT/SLED) | In-hospital mortality; In-ICU mortality; 90-day mortality |
| Single-center, surgical | Germany | 231 (112/119) | Yes | 67.0 | 63.2 | 81.8/19.5/41.6/40.7 | 32.5 | KDIGO Stage 2 AKI (within 8 h) and plasma NGAL > 150 ng/mL | KDIGO stage 3 | CRRT | 30-/60-/90-day mortality |
| Multi-center, mixed | France | 619 (311/308) | Yes | 66.2 | 65.4 | 53.0/26.3%/9.0/9.7 | 79.8 | KDIGO Stage 3 AKI (within 6 h) | Severe hyperkalemia (> 6 mmol/L); severe pulmonary edema refractory to diuretics; severe acidosis (pH < 7.15); serum urea > 40 mmol/L; oligo-anuria > 72 h | Mixed (IHD/CRRT) | 60-/90-day mortality |
| Single-center, mixed | Thailand | 40 (20/20) | Yes | 66.8 | 55.0 | NA | 72.5 | AKI, any RIFLE stage | Severe metabolic acidosis (pH < 7–20); severe hyperkalemia (> 6 2 mmol/L); severe pulmonary edema refractory to diuretics; persistent oliguria or anuria; serum urea > 40 mg/dL | CRRT | 28-day mortality |
| Multicenter, mixed | Thailand | 118 (58/60) | Yes | 67.1 | 46.2 | 44.9/24.6/NA/NA | 58.5 | Furosemide stress test-nonresponsive patients (UOP < 200 mL in 2 h) (initiation within 6 h); AKI (any stage of KDIGO) | Serum urea > 100 mg/dL; severe hyperkalemia (> 6 mmol/L); severe metabolic acidosis (pH < 715); severe pulmonary edema | CRRT | 28-day mortality |
| Multicenter, mixed | France | 477 (239/238) | Yes | 69.0 | 60.7 | 57.8/30.5/8.2/15.6 | 100.0 | Within 12 h after documentation of RIFLE-F | Severe hyperkalemia (> 6 5 mmol/L); severe pulmonary edema refractory to diuretics; severe metabolic acidosis (pH < 715); no renal function recovery after 48 h | Mixed | 90-day mortality |
| Single-center, mixed | China | 142 (71/71) | No | 56.7 | 58.5 | 37.3/45.8/NA/NA | 100 | Within 24 h of ICU entry | After 48 h of ICU entry | CRRT | 28-day mortality |
| Multicenter, mixed | Multiple | 2927 (1465/1462) | Yes | 64.7 | 68.0 | 55.9/30.7/13.9/43.9 | 57.7 | Within 12 h of fulfilling eligibility: at least two of the following: twofold increase in serum Cr from baseline; UOP < 6 mL/kg in the preceding 12 h; whole-blood NGAL > 400 ng/mL | Until the development of one or more of the following criteria: a serum potassium level > 6.0 mmol/L, a pH < 7.20 or a serum bicarbonate < 12 mmol/L, severe respiratory failure based on a ratio of the partial pressure of arterial oxygen to the fraction of inspired oxygen of < 200 and clinical perception of volume overload, or persistent AKI for at least 72 h after randomization | Mixed | 90-day mortality |
Quality assessment
The Cochrane risk of bias tool was used for quality assessment of RCTs [
12]. The following domains were assessed: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting and other bias. The criteria for rating study quality were as follows: high risk study (2 or more items rated as high risk of bias); low risk study (5 or more items rated as low risk and no more than one as high risk); moderate risk study (all remaining situations) [
13].
Definition
Accelerated initiation and standard initiation were defined as relatively earlier versus later hemodialysis according to each study. This study was to investigate the effectiveness of earlier rather than later dialysis. Instead of identifying the point in time of early or late dialysis, we have standardized the terminology in this manuscript to refer to all relatively early dialysis timing as accelerated initiation.
Outcomes
The primary outcome was all-cause mortality. Secondary outcomes were dialysis dependence and being free of RRT rate after hospital discharge.
Subgroup analysis
We hypothesized that the following factors could have high impact on patient outcomes observed among different studies: patient population (surgical vs. mixed/medical), disease severity (Sequential Organ Failure Assessment (SOFA) score ≥ 11 vs. < 11), RRT modality (CRRT vs. intermittent hemodialysis [IHD]/mixed), diabetes mellitus prevalence (≥ 35% or < 35%), discrepancy in interval between accelerated and standard initiation time (difference ≥ 24 vs. < 24 h), using comprehensive AKI definitions (with or without including urine amount criteria), single versus multicenter, and sepsis prevalence.
Data synthesis and statistical analysis
Overall summary log odds ratios (ORs) and 95% CIs were calculated by the Mantel–Haenszel method. The fixed-effects model was used to pool the results of the RCTs. Statistical heterogeneity was assessed by the chi-square test and the I2 statistic with a p < 0.05 or I2 > 50% was an indication of substantial heterogeneity. In the case of considerable heterogeneity (I2 > 50% or p < 0.05), we performed a sensitivity analysis to detect the influence of a single study on the overall estimate by omitting one study in turn and pooling the remaining ones. In the subgroup analysis, we performed meta-regression to assess the interaction between variables and the timing of RRT initiation on mortality and RRT dependence. Any potential publication bias was assessed by visual assessment of the funnel plots constructed.
We then did the trial sequential analysis (TSA), as well as the sequential monitoring boundaries. The conventional nonsuperiority boundaries were calculated assuming significance levels of 0.05, and a power of 80%. The a-spending boundaries were also calculated using significance levels of 0.01 and 0.05 and the O’Brien-Fleming multiple testing procedure [
14]. In order to calculate the neutrality zone, we chose a risk ratio reduction of 20%, because of its compatibility with many trials in the ICU [
15], and its representation of an absolute mortality difference of around 10% to 15%, which we considered to be a reasonable effect size. Furthermore, funnel plots were used to evaluate the possibility of publication bias. We used STATA (Version 16, Stata Corp. 2019. College Station, TX: Stata Corp LP) software for the meta-analysis. TSA version 0.9.5.5 (reviewed in November 2016) b software was used for these analyses the cumulative effect of randomized trials on mortality.
Discussion
In a systematic review of 10 RCTs including 4753 critically ill patients with severe AKI, we did not find significant survival benefits (28-day nor 90-day mortality) in patients who underwent accelerated versus standard RRT. For those critical AKI patients who underwent CRRT treatment or were in the surgical ICU setting, accelerated RRT showed survival benefits as well as more free of dialysis. However, the relative risk of dialysis dependence increased in accelerated RRT group, when those AKI patients were non-CRRT and of high disease severity groups. To our knowledge, this is the most comprehensive systematic review to date that included the highest number of RCTs and the largest number of critically ill AKI patients.
Even though the literatures addressing this comparison were highly heterogeneous, our funnel meta-regression analysis showed only limited publication bias. Our TSA showed a constant result and low risk of bias among these randomized studies comparing the impact on mortality between accelerated versus standard initiation of dialysis in critically ill AKI patients. Furthermore, the total number of patients is enough to achieve a confident conclusion because the Z-curve did cross the neutrality line from the TSA.
Being free of dialysis
Contrary to the previous reports, we did not find a significant effect of standard dialysis leading to a higher rate of being free of RRT when all patient populations were considered; however, in critically ill AKI patients with higher disease severity or who underwent CRRT treatment, there were higher rates of being free of dialysis among the survivors.
Some large observational studies, that only included patients who were receiving RRT, suggest that CRRT is an independent predictor of renal recovery among survivors [
23,
24]. CRRT could permit slow but continuous removal of solutes and water, thereby conferring better hemodynamic tolerability. Relative hemodynamic stability during CRRT sessions, compared to intermittent dialysis, could mitigate occult kidney injury. We showed the evidence to elucidate the impact of choice of therapy on this outcome from RCTs [
25].
Dialysis dependence
Another conclusion to result from this meta-analysis is that about 90.8% (89.7% in accelerated and 91.9% in standard group) of the survivors with severe AKI in our survey did not undergo dialysis at the end of the studies, due to the fortunate spontaneous recovery of renal function from managing underlying etiologies and/or accompanying co-morbidities which occurred within hours to days.
To our knowledge, our study is the largest systematic review to date to address the question of optimal timing regarding RRT initiation and its impact on patient survival and dialysis dependence that included several recent elegant large RCTs [
7‐
9,
19]. Furthermore, our finding is consistent in 73.3% of RCTs when watchful waiting strategy was adopted into the study design. For the sake of all-cause mortality, it is startling to find 90.8% of the survivors to avoid the potential risks of an extracorporeal support technique if they do not actually need it, not to mention the savings to be yielded in terms of medical cost, time, man-power and capacity of the ICU.
A delayed strategy for the initiation of RRT could allow time for stabilization of the patients’ condition, thus enabling starting of RRT when the patients were more hemodynamically stable, or even precluding the need for such therapy if the renal function recovers spontaneously that contributed to free of dialysis. The results of the BICAR-ICU study [
26] suggested that optimizing medical treatment can avoid RRT for severely metabolically acidotic AKI cases, and thus reducing mortality. Moreover, the time needed for most inotropes to reach their efficacy is less than 28.1 h, which is the average difference in timing between these accelerated versus standard initiation. This implicated that the patients who would have to utilize RRTs were at a more hemodynamically stable state. Although we still lack miracle drug therapies capable of blocking or reversing severe AKI, accelerated RRT might not be the solution to all critical AKI patients.
Subgroup analyses
Previous investigation had looked at special populations such as sepsis and found no significant difference in survival of AKI patients according to strategy for the initiation of RRT [
27]. In the current study, we further tried to explore the clinical impact of some other potential factors, such as different study settings, disease severities, diabetic percentage, and dialysis discrepancy time less than 24 h. In our subgroup analyses, we found no survival differences between accelerated versus standard RRT initiation time even after multivariate adjustment. We also found that the values of combining serum creatinine and urine output (which are components of the modern RIFLE and KDIGO criteria), as well as percentages of the septic patients had no significant impact on patient outcomes, and therefore they should not be used to decide on the time of RRT initiation.
There are some possible explanations for the discordance and heterogeneity among different studies. Using varying AKI definitions and different AKI stage criteria for RRT initiation could account for part of the observed heterogeneities. For most of the previously reported cohort studies, the differences in pre-intervention study groups contributed to the heterogeneity of the results, and therefore made the systematic reviews difficult to interpret.
Strengths and limitations
The strength of our present analysis rests on our extensive literature search on related RCTs. We used standard Cochrane protocols and had the largest cumulative RCTs study sample size to date in comparison with the previous reports. We only focused on the RCTs that had a reasonable quality with limited differential dropout based on the assigned treatment arm. One of the differences that our study has in comparison with previous reports is the inclusion of the recently published elegant RCT studies with large patient numbers and global multi-national inclusion in patients’ recruiting [
9], especially its effect on free of dialysis. These recently published RCTs that included watchful waiting strategies were not included in prior meta-analysis; this accounts for the differences in our results from those of earlier authors [
28‐
31]. The strength of our meta-analysis also lies in comprehensive data search with subgroup analyses across several clinical scenarios. We adapted the GRADE approach to rate the certainty of evidence [
32].
Protocolizing the optimal timing of RRT for all AKI patients may be too crude and imprecise, in the era of modern personalized medicine. It could be more important to give clinicians reliable information about when to initialize RRT in certain precisely defined patient groups. The negative result on the primary endpoint turns out to be hiding among a high level of heterogeneity in terms of disease progression, that could not be accurately predicted by the staging of AKI at the time of inclusion.
In our systematic review, we found no further information regarding the other factors associated with mortality, and therefore, we cannot comment on differences in the outcomes on the basis of one single intervention, i.e., accelerated or standard dialysis initiation. There were only two trials that exclusively enrolled surgical patients, of which one of them were entirely from cardiovascular surgery. Furthermore, no trial standardized the dialysis modality or dose delivered during initializing RRT. We were not able to access the unpublished reports, e.g., negative results of accelerated RRT which might have biased our results. Although our funnel meta-regression and Cochrane Collaboration’s tool analysis showed a limited publication bias (supplementary figures), the bias is always difficult to ascertain with a small sample size of the included studies. Finally, the definition of “accelerated” RRT was variable and may have unduly influenced pooled effect estimates. The timing of RRT defined by traditional markers was relatively late which may influence the effectiveness of the early treatment. In our TSA, we included trials of patients without severe AKI, which yielded enough information size to conclude that accelerated RRT probably does not benefit patients (due to the Z curve crossing the neutrality line). Nonetheless, our conclusion yielded from studies that consisted of different study designs and different clinical scenarios. Of note, we just raised the possibility of the timing of RRT defined by traditional markers was relatively late, which may influence the effectiveness of the early treatment. Our intention was to investigate whether receiving RRT earlier than the traditional timing would influence its effectiveness. Further research efforts are certainly needed for the pursuit of better precision medicine. It could be more fruitful to investigate if different etiologies of AKI (pre-renal vs. renal vs. obstructive, cardiogenic shock, hypovolemic shock, sepsis-related, etc.) affect outcomes of accelerated versus standard RRT; and to evaluate if the efficacy of CRRT fits into various underlying causes of AKI in critically ill patients. These issues can be incorporated into the design of future RCTs in evaluating the optimal timing and modalities of RRT for critically ill AKI patients, in order to reach a new horizon and higher success rate of treatment in this field. Moreover, further investigations into improvement in treatments to resuscitate the patients’ hemodynamic stability and to manage each underlying mechanisms of AKI might contribute to mitigate the current extremely high mortality rate of these critically ill patients with RRT-requiring AKI.
Conclusion
Accelerated dialysis initiation in critically ill patients with severe AKI does not decrease mortality, alter the possibility from free of dialysis, or mitigate dialysis dependence among survivors as compared with the standard RRT initiation strategy. However, in patients who underwent CRRT treatment or were in the surgical ICU setting, accelerated RRT could benefit the possibility of survival and free of dialysis. However, accelerated RRT initiation could be associated with higher risk of dialysis dependence when those severe AKI patients were treated with non-CRRT modality or were of high disease severity.
Acknowledgements
The authors would like to thank the staff of the Second Core Lab in the Department of Medical Research in National Taiwan University Hospital for technical assistance. The authors thank Dr. Jeff Chueh of Glickman Urology and Kidney Institute, Cleveland Clinic, Cleveland, OH, USA, for his critical comments.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.