Background

Acute kidney injury (AKI) is a major issue in the intensive care unit (ICU) and is strongly associated with high morbidity and mortality. In fact, despite its potential to be reversed, several studies performed in different clinical settings confirmed that occurrence of AKI is independently associated with in-hospital mortality and negative short- and long-term outcomes of critically ill patients [1,2,3]. Moreover, early development of AKI during the ICU stay has been shown to be associated with a reduced probability of being alive or having good Health-related quality of life (HRQoL) one year after ICU admission [4].

Given the possible severe implications of this condition, in the last few decades researchers mainly focused their attention on the pathogenesis of AKI and on its prompt recognition, leading to the development of a series of different biological markers with higher sensitivity and specificity for the occurrence of renal impairment [5, 6]. These markers play a fundamental role in the early diagnosis and treatment of AKI [7]. Nonetheless, their potential role in improving patients’ outcome is still debated [8].

In fact, actual indications for renal replacement therapy (RRT) in the ICU require the development of severe clinical manifestations of renal impairment, such as volume overload unresponsive to medical therapy, hyperkaliemia or major electrolyte disturbances, acidosis or uremic complications such as pericarditis or encephalopathy. Whether an “earlier” initiation of RRT might be effective in improving survival of critical ill patients affected by AKI is still debated. Unfortunately, previous meta-analyses on this topic were unconclusive due to the paucity of pubblished data [9,10,11]. Recently, a large, high-quality randomized clinical trial (RCT), the IDEAL-ICU trial was published in NEJM [12]. In this multicenter trial, 488 adults with septic shock and severe AKI were randomized to receive RRT within 12 h (early strategy) of documented failure stage or after a delay of 48 h (late strategy). Nearly all patients in the early-strategy group received RRT while approximately 30% of patients in the delayed-strategy group did not receive RRT since they had spontaneous recovery of renal function. The IDEAL-ICU trial was stopped early for futility. The primary outcome of mortality at 90 days did not differ between patients who received early versus late initiation of RRT (58% vs. 54%; P = 0.38). Furthermore, no benefits were seen from early initiation of RRT in secondary outcomes. Results of previous larger studies were conflicting, Infact, one major trial (ELAIN) showed a 90-day mortality benefit, while another (AKIKI) did not show a benefit at 60 days. The ELAIN trial was smaller, conducted almost exclusively in postoperative AKI patients, and the difference in timing between early versus late initiation of RRT was less than 24 h. Therefore, we decided to perform an updated meta-analysis of all the RCTs ever performed on “earlier” initiation of RRT versus standard RRT in critically ill patients to evaluate its effect on outcome of critically ill patients with AKI.

Methods

Search strategy

Pertinent studies were independently searched in BioMedCentral, PubMed, Embase, and the Cochrane Central Register of clinical trials by three investigators. The full PubMed search strategy aimed to include any RCTs ever performed on “earlier” initiation of RRT in critically ill patients with AKI. In addition, we employed backward snowballing (i.e., scanning of references of retrieved articles and pertinent reviews) and contacted international experts for further studies with no language restriction.

Study selection

The following inclusion criteria were used for potentially relevant studies: studies performed on critically ill patients; random allocation to treatment (“earlier” initiation of RRT versus later/standard initiation). The exclusion criteria were non-adult patients, duplicate publications and lack of data on all of the following: mortality, survival with dependence on RRT, ICU stay, hospital length of stay (HLOS). References were first independently examined at abstract level by three investigators, with divergences resolved by consensus. If potentially pertinent, articles were retrieved as full articles. Two investigators independently assessed compliance to selection criteria and selected studies for the final analysis, with divergences resolved by consensus.

Data abstraction and study

Baseline and outcome data were independently abstracted by three investigators (Tables 1 and 2). At least two separate attempts at contacting original authors were made in cases of missing data. The co-primary endpoints of the present review were mortality at the longest follow-up available and survival with dependence on RRT. Secondary endpoints were ICU stay (days) and HLOS (days). Adverse effects (bleeding complications, blood transfusions, catheter-related bacteremia, rhythm disturbances, thrombocytopenia) were also analysed. Subanalysis were performed on the subgroup of patients who underwent cardiac surgery and on general ICU patients.

Table 1 Study characteristics
Full size table
Table 2 Primary and secondary outcomes, adverse events and sensitivity analyses
Full size table

The internal validity and risk of bias of included trials was appraised by two independent reviewers according to the latest version of the “Risk of bias assessment tool” developed by The Cochrane collaboration [13]. Publication bias was assessed by visually inspecting funnel plots. Sensitivity analyses were performed by sequentially removing each study and reanalyzing the remaining dataset (producing a new analysis for each study removed) and by analyzing only data from studies with low risk of bias.

Data analysis and synthesis

Computations were performed with Review Manager version 5.2. Hypothesis of statistical heterogeneity was tested by means of Cochran Q test, with statistical significance set at the two-tailed 0.10 level, whereas extent of statistical consistency was measured with I2, defined as 100% X (Q-df)/Q, where Q is Cochran’s heterogeneity statistic and df the degrees of freedom. Binary outcomes from individual studies were analysed to compute individual and pooled odds ratio (OR) with pertinent 95% confidence interval (CI), by means of Mantel-Haenszel method and with a fixed-effect model in case of low statistical inconsistency (I2 < 25%) or with random-effect model (which better accommodates clinical and statistical variations) in case of moderate or high statistical inconsistency (I2 > 25%). To evaluate if the small study effect will have an influence on the treatment effect estimate, in case of evidence of between-study heterogeneity (I2 > 25), we compared the results of both fixed and random effect models. Sensitivity analyses were performed by sequentially removing each study and reanalysing the remaining dataset (producing a new analysis for each study removed) and by analysing only data from studies with low risk of bias. Statistical significance was set at the two tailed 0.05 level for hypothesis testing. Unadjusted p values are reported throughout. This study was performed in compliance with The Cochrane Collaboration and Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [13,14,15].

Results

Study characteristics

Database searches, snowballing, and contacts with experts yielded a total of 657 articles (updated October 15th 2018). The flow chart to select the final 10 manuscripts trials [12, 16,17,18,19,20,21,22,23,24] is detailed in Fig. 1. Excluding 563 non-pertinent titles or abstracts, we retrieved in complete form and assessed 94 studies according to the selection criteria. 84 studies were further excluded because of our prespecified exclusion criteria. (Fig. 1).

Fig. 1
figure 1

Flow-chart for study selection

Full size image

The 10 included trials randomized 2214 patients, 1073 to earlier initiation of RRT and 1141 to later initiation. (Table 1) Clinical heterogeneity was mostly due to setting and criteria for early and late initiation of RRT. (Table 1) Indeed three trials were performed in patients who underwent cardiac surgery [19, 20, 23] while the other seven were performed in general ICU patients [12, 16,17,18, 21, 22, 24] (Table 1).

Quantitative data synthesis

Overall analysis showed that early initiation of RRT does not improve outcome of critically ill patients with AKI. In fact, no differences in mortality (Fig. 2) and survival with dependence on RRT were observed between groups. (Fig. 3) Results were confirmed at sensitivity analyses and the funnel plot illustrated in the Additional file 1. (Table 2; Additional file 1: Figures S1 and S2).

Fig. 2
figure 2

Forest plot for mortality

Full size image
Fig. 3
figure 3

Forest plot for survival with dependence on RRT

Full size image

On the contrary, early initiation of RRT was associated with a significant reduction in HLOS. (Additional file 1: Figures S3 and S4) Nonetheless results were not confirmed at sensitivity analyses. (Table 2) Visual inspection of funnel plots did not identify a skewed or asymmetrical shape for the primary endpoints (Additional file 1: Figures S5 and S6).

No differences in occurrence of adverse events were observed (Table 2).

Discussion

Our meta-analyses suggests that early initiation of RRT does not improve clinically relevant outcomes of critically ill patients with AKI. In fact, mortality and survival with dependence on RRT don’t differ between patients who received early RRT and patients who received standard treatment. Moreover, although we found an overall significant reduction in HLOS in the subgroup of patients who received early RRT, these positive results were not confirmed in the high-quality studies. In addition, we didn’t find a subgroup of patients in which early initiation of RRT could me more beneficial since outcome did not improve both in cardiac surgery patients and general ICU patients.

Our results diverge from the results of recent meta-analyses on this topic, while confirm the results of the less recent meta-analyses performed by Wierstra et al. [9,10,11]. Nonetheless, the conclusions of Wiestra at al. were weaker since were based on fewer, lower quality studies and didn’t include the most recent, high-quality trials published in the last year. We updated their results with three recently published manuscripts, [12, 16, 17] thus increasing the number of patients by more than 100% (up to 2214 overall randomized patients included in our meta-analysis) and allowing to have more robust data. Our results are consistent with the results of another recent meta-analyses performed by Feng et al. [25]. Nonetheless, Feng et al. limited their analyses to mortality and ICU and hospital length of stay, without considering adverse effects. Moreover they did not perform any subanalyses on general ICU patients or cardiac surgery patients, therefore drawing weaker conclusions. On the contrary, Moreira et al. found an increased risk of catheter-related bloodstream infection when renal replacement therapy was initiated early. Moreira FT, et al. [10] our study doesn’t confirm this finding.

Although our meta-analysis includes all the randomized clinical trials ever published on early vs late RRT and two large, recent, high-quality RCTs, the optimal timing of initiating RRT remains unclear. Actually, we couldn’t add great new findings to previous published meta-analyses. A reasonable explanation for this is that our study is still underpowered for mortality. Moreover, the analyzed studies were conducted over a wide range of time, during which the management of AKI patients has greately changed. In fact, in the last decade the Kidney Disease Improving Global Outcomes (KDIGO) Clinical Practice Guideline contributed to standardize AKI treatment. This means that the more recent studies published after 2010 failed to show a significant survival benefit from early RRT treatment, while a reduction in mortality was shown by older studies. Initiation of RRT, to some extent, depends on creatinine level and urine output, namely, the KDIGO criteria. Therefore, one of the main limitations of our meta-analysis and of all the performed and ongoing trials is the lack of definition of “early” versus “late” criteria, that varied among the included studies and may have led to great differences in the requirements for RRT and their therapeutic impact. Larger, well conducted RCTs should be performed to further clarify this issue. Actually, there is another ongoing RCT that will probably provide additional informations on the optimal timing of starting RRT in critically ill patients admitted to general ICU (STARRT-AKI, NCT02568722). Unfortunately, given the previous reported limitations, this trial will not probably allow to draw definitive conclusions on the optimal timing of starting RRT in critically ill patients.

Conclusions

Our meta-analysis supports the notion that early initiation of RRT in critically ill patients with AKI does not provide a clinically relevant advantage when compared with standard/late initiation. Large, multicenter RCTs are warranted to clarify the optimal timing of starting RRT. Based on the limitations of the data available for our analysis, future work in the following areas is desirable: (1) stardardized definition of “early” and “late” initiation of RRT; (2) special populations such as the septic shock patients or post cardiac surgery patients; (3) an assessment of the performance of the different RRT modalities and dosage options.