Effects of early dialysis on the outcomes of critically ill patients with acute kidney injury: a systematic review and meta-analysis of randomized controlled trials

The appropriate timing for initiating renal replacement therapy (RRT) in critically ill patients with acute kidney injury (AKI) remains unknown. This meta-analysis aims to assess the efficacy of early initiation of RRT in critically ill patients with AKI. The Pubmed, Embase and Cochrane databases were searched up to August 13, 2019. Only randomized controlled trials (RCTs) comparing the effects of early and late RRT on AKI patients were included. The primary outcome was 28-day mortality. Eleven RCTs including 1131 and 1111 AKI patients assigned to early and late RRT strategies, respectively, were enrolled in this meta-analysis. The pooled 28-day mortality was 38.1% (431/1131) and 40.7% (453/1111) in the patients assigned to early and late RRT, respectively, with no significant difference between groups (risk ratio (RR), 0.95; 95% CI, 0.78–1.15, I2 = 63%). No significant difference was found between groups in terms of RRT dependence in survivors on day 28 (RR, 0.90; 95% CI, 0.67–1.25, I2 = 0%), and recovery of renal function (RR, 1.03; 95% CI, 0.89–1.19, I2 = 56%). The early RRT group had higher risks of catheter-related infection (RR, 1.7, 95% CI, 1.01–2.97, I2 = 0%) and hypophosphatemia (RR, 2.5, 95% CI, 1.25–4.99, I2 = 77%) than the late RRT group. In conclusion, an early RRT strategy does not improve survival, RRT dependence, or renal function recovery in critically ill patients with AKI in comparison with a late RRT strategy. However, clinicians should be vigilant because early RRT can carry higher risks of catheter-related infection and hypophosphatemia during dialysis than late RRT.

unnecessary clearance of important medications, delayed recovery of renal function and increased costs 13,14 . Several randomized controlled trials [15][16][17][18][19][20][21][22][23] (RCTs) were conducted to find the optimal timing of RRT for critically ill AKI patients, but no consistent results were found. In 2018, one large RCT 24 focused on patients with septic shock and severe AKI in the IDEAL-ICU trial found no significant difference in 90-day mortality between patients with early and delayed initiation of RRT. Their findings were consistent with those in another multicenter RCT by the AKIKI study group 22 , but were contrary to the findings of a recent single-center RCT in the ELAIN trial 23 . All of these findings indicate uncertainty about the usefulness of early RRT in critically ill patients. Although this issue had been discussed in one meta-analyses 25 recently, we aimed to conduct an updated systematic review and meta-analysis of RCTs to assess the efficacy of early initiation of RRT in critically ill AKI patients.

Materials and Methods
Study search and selection. This systematic review and meta-analysis were conducted according to the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement (Supplemental Table 1). All clinical studies were identified by a systematic review of the literature in the PubMed, Embase, and Cochrane databases until August 13, 2019 using the following Mesh terms -"earl*", "accelerat*", "acute kidney", "acute renal", "anuria", "oliguria", "acute renal failure", "anuria", "oliguria", "organ failure", "dialy*", "renal replacement", "hemodialysis", "hemofiltration", "hemodiafiltration", "RCT*" and "random*". We excluded observation studies, case reports or case series, studies enrolling pediatric patients, and conference abstracts, and therefore, only RCTs that compared the clinical efficacy of early RRT and late RRT for critically ill adult patients with AKI were included. In addition, we searched all references in the relevant articles and reviews for additional eligible studies. Two reviewers (Chang & Wang) searched and examined publications independently to avoid bias. When they disagreed, another author (Lai) resolved the issue. The data included authors, year of publication, study design and duration, study population, sites of study, disease severity, indications for early RRT, and outcomes. Ethics board approval and patient consent were not required due to the nature of a systematic review. This meta-analysis was performed according the guidelines of Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA).
Definitions and outcome. The primary outcome was 28-day mortality and secondary outcomes included recovery of renal function, RRT dependence among survivors and adverse events. Data analysis. We used the Cochrane Risk of Bias tool to evaluate the quality of enrolled studies and the risk of bias 26 . The statistical analysis was conducted using the software Review Manager, version 5.3. The degree of heterogeneity was evaluated with the Q statistic generated from the χ 2 test. The proportion of statistical heterogeneity was assessed by the I 2 measure. Heterogeneity was considered significant when the p-value was less than 0.10 or the I 2 more than 50%. The fixed effects model and the random effects model were applied when the data was homogenous and heterogeneous, respectively. Pooled risk ratios (RR) and 95% confidence intervals (CI) were calculated for outcome analyses. Funnel plot was used to probe for publication bias. A p-value <0.05 was set as the threshold of statistical significance. Sensitivity analyses were conducted by excluding or subgrouping studies to reduce the potential confounding effects of patient population, RRT modality, study design, and study sample size.
In the four studies 16,17,20,23 that only enrolled surgical patients, the early RRT group had a lower risk of mortality than the late RRT group (RR, 0.52; 95% CI, 0.27-0.99, I 2 = 77%). There were no significant differences in terms of mortality between groups in the analysis of other subgroups, including mixed study populations, study sites, study designs, portion of patients with sepsis and RRT modality ( Table 2).

Discussion
This meta-analysis of eleven RCTs with 1131 and 1111 AKI patients receiving early and late RRT, respectively, provided several significant findings. Most importantly, early RRT was not associated with a better outcome for these patients than late RRT. Overall, there was no significant difference in 28-day mortality between groups. There were no differences with different study sites (Europe, Asia or North America), study designs (single or multi-center), portion of patients with sepsis (50-<100%, or 100%) or RRT modality (IHD, CRRT, mixed). The early and late RRT groups had similar ICU-, hospital-, 60 day-and 90 day-mortality rates. All these findings are consistent with previous meta-analyses 25,[28][29][30] , and indicate that early RRT does not provide additional survival benefits for AKI patients compared with late RRT. In addition to mortality, Pasin et al. 25 ever showed that early RRT was associated with a significant reduction in length of hospital stay. However, the positive impact of early RRT on the length of hospital stay still need further confirmation in the high-quality studies.
Subgroup analysis of four studies 16,17,20,23 that only enrolled surgical patients showed that the early RRT group had a lower risk of mortality than the late RRT group (RR, 0.52; 95% CI, 0.27-0.99, I 2 = 77%). This finding is consistent with the result of a previous meta-anlaysis 31 of nine retrospective cohort studies and two RCTs showing a lower 28-day mortality rate the early RRT group (OR = 0.29, 95% CI, 0.16-0.52, p < 0.0001) than the late RRT group among critically ill patients with AKI after cardiac surgery. However, both that meta-analysis 31 and our findings in surgical patients were based on studies with very high heterogeneity. Further research with a larger number of studies and consistent results is still needed to confirm this finding in surgical patients.
We also found no differences in the recovery of renal function or RRT dependency in the early and late RRT groups. In Karvellas et al. 's meta-analysis 32 of 15 studies, early RRT was associated with greater renal recovery than late RRT. However, only two RCTs were enrolled in that meta-analysis 32 , and the quality of those heterogeneous studies varied. In contrast, the present analysis only enrolled large-scale RCTs, and our findings were consistent with other meta-analyses 29,30,33 of RCTs. These results should be more convincing than Karvellas et al.'s meta-analysis 32 . Therefore, based on current evidence, early RRT was not associated with greater renal recovery and lower dialysis dependence than late RRT for critically ill patients with AKI.
We cannot omit another important issue of RRT -safety. We evaluated the risks of several common complications during RRT. Although the incidence of most adverse events such as hemorrhage, hypotension, arrhythmia, hypokalemia and hyperkalemia were similar between groups, the early RRT group had higher risks of catheter-related infections, and hypophosphatemia than the late RRT group. Overall, our findings should remind clinicians to keep alert concerning the high risks of these two complications in early RRT for patients with AKI.
Although this meta-analysis enrolled several large-scale RCTs with a reasonable quality to enhance the level of evidence, there was one major limitation. There was relatively high heterogeneity with an I 2 value of more than 50% in the outcome analysis. These heterogeneities could be caused by significant variations in the study design, population characteristics, disease severity, timing of initiating RRT, modality of RRT, and duration of follow-up in the studies.

Conclusion
This meta-analysis suggested that early RRT does not improve the survival, RRT dependence, or renal function recovery of critically ill patients with AKI in comparison with late RRT. Early RRT was associated with a lower 28-day mortality than late RRT in surgical patients with AKI. However, clinicians should be vigilant as early RRT can carry higher risks of catheter-related infection and hypophosphatemia during dialysis than late RRT.

Data availability
The datasets used and/or analyzed in the current study are available from the corresponding author upon request.