Timing of Initiation of Renal Replacement Therapy in Acute Kidney Injury: A Systematic Review and Meta-analysis

Abstract

Objective: The aim of the study is to summarize the effects of timing of initiation of renal replacement therapy (RRT) on mortality. Methods: A systematic search for randomized controlled trials (RCTs) and other clinical studies was performed without language restriction in PubMed, Web of Science, and Embase. We estimated pooled relative risk ratios (RRs) and 95% confidence intervals (CIs) using fixed effects model or random effects model as appropriate. Heterogeneity, publication bias, and subgroup analyses were conducted. Results: We analyzed the date extracted from 15 studies (3 RCTs, 2 prospective, and 10 retrospective comparative cohort studies) with a total of 2955 patients. Overall, 51.0% (772/1514) patients died in the “early” RRT group compared with 58.0% (836/1441) in the “late” RRT group. The pooled RR was 0.71 (95% CI: 0.59, 0.86), but the heterogeneity existed (p < 0.00001). Subgroup analysis based on modality did not record heterogeneity across trials. In continuous RRT (CRRT) group (n = 607), patients treated with “early” CRRT suggested a significant decrease in mortality compared with those in “late” CRRT group (27.8% vs. 43.0%) and the RR was 0.69 (95% CI: 0.56, 0.84) without evidence of heterogeneity (I₂ = 33%, p = 0.18). In intermittent hemodialysis (IHD) group (n = 115), the RR was 0.26 (95% CI: 0.15, 0.45) without evidence of heterogeneity (I₂ = 0%, p = 0.50). In the mixed group, heterogeneity existed. Conclusion: “Early” CRRT and “early” IHD both could reduce the mortality of patients with acute kidney injury compared with “late” CRRT or IHD.

• timing of initiation
• renal replacement therapy
• acute kidney injury

Although the renal replacement therapy (RRT) technology has been improved greatly and has been widely used in acute kidney injury (AKI) patients, the mortality caused by AKI was still very high in intensive care unit (ICU) during the last several decades.¹ Recently, it is controversial whether “early” RRT can reduce the mortality of patients with AKI compared with “late” RRT. One prospective randomized trial² could not show a significant survival benefit for early institution of RRT in patients with ARF, but the other one³ suggested that the early start dialysis therapy might be preferable for improvement of survival of the patients with AKI. Thus, we undertook a systemic review and meta-analysis to obtain current estimate on the effects of initiation time of RRT on mortality of patients with AKI.

SUBJECTS AND METHODS

Literature Search Strategy

We conducted a systematic search for clinical trials of PubMed, Web of Science, and Embase from 1990 to August 2011, using the following keywords together with RRT: timing, initiation, AKI, and ICU. There was no language restriction on the research. Details of the searching process are shown in Figure 1.

Figure 1. Literature search strategy.

Inclusion and Exclusion Criteria

The inclusion criteria were as follows: (1) randomized controlled trials (RCTs), prospective or retrospective studies; (2) comparing mortality and other clinical outcomes of “early” and “late” RRTs of patients with AKI. We excluded studies not accessible to full research data. Review, case report, letter, and editorial were excluded.

Data Extraction

Two authors (Xuan Wang and Wei Jie Yuan) independently reviewed each report found by the searching strategy, and results obtained by them were compared and discussed to reach an agreement. The study selection, data extraction, and reporting of results were all based on the Preferred Reporting Items for Systematic reviews and Meta-Analyses checklist.⁴ Data extracted included basic characteristics (publication year, study design, mean age, gender, total number of patients, and biochemical parameters at the initiation of RRT) and main characteristics (modality, patient population, and definition of “early” and “late” RRT). The main outcome was the mortality.

Statistical Analysis

Pooled estimates of the risk ratio (RR) and relevant 95% confidence interval (CI) were obtained by using random effect model or fixed effect model. Heterogeneity across trials was evaluated with Q and I² statistic. I² values ranged from 0% to 100%: 0% suggested no observed heterogeneity, 25–49% suggested low, 50–74% moderate, and ≥75% high heterogeneity.⁵ p-Value <0.1 was defined as significant heterogeneity. If heterogeneity existed, a random effect model was used to assess the overall estimate. Otherwise, a fixed effect model was chosen.

Subgroup analysis was performed to account for heterogeneity. Despite potential advantages of an improvement in patients’ outcomes, using continuous RRT (CRRT) compared with intermittent hemodialysis (IHD) remains controversial. Two trials have studied the effects of CRRT versus IHD on the mortality of patients with AKI,^6,7 so we specified subgroup analysis based on the modality studies with all patients using CRRT listed in the CRRT group and studies with patients using the modality of IHD made up of the IHD group and in the remaining studies, a part of patients used CRRT, the others might use IHD or peritoneal dialysis (PD) or sustained low-efficiency dialysis (SLED), which were included in the mixed group.

Sensitivity analyses were performed to assess the effects of selected measures of study quality and clinical factors on heterogeneity. The influence of individual study on the conclusion was estimated by deleting the study from the analysis and noting the degree to which the pooled effect size changed. We considered a study influential if the exclusion changed our conclusion or the effect estimate by at least 20%.

A funnel plot was used to assess the presence of publication and other reporting biases by plotting the standard error against the log RR. The shape of the funnel plot was symmetrical, indicating no significant publication bias. All analyses were performed using the software Review Manager 5.0 (Cochrane Collaboration, Oxford, UK).

Assessment of Methodological Quality

For the three RCTs, we use special criteria (sequence generation, concealment allocation, blinded assessment, incomplete outcome data, and free of selective reporting) to assess their quality, rather than any scale with regard to the limitation of this method.⁸ The remaining 12 nonrandomized trials were evaluated in a descriptive fashion similar to a previous study,⁹ including the reported criteria for RRT initiation, comparability of intervention/control arms, and a description of dropouts.

RESULTS

Study Selection

Our search strategy yielded 1238 citations shown in Figure 1. We retrieved 47 citations for detailed evaluation, of which 32 were excluded. Fifteen studies,^2,3,10–22 which included 3 RCTs, 2 prospective and 10 retrospective comparative cohort studies, met the inclusion criteria. All these studies were written in English.

Characteristics of the Studies

The basic characteristics of the studies were reported in Table 1. A total of 2955 patients were enrolled. Table 2 showed the modality of RRT, population, and the definition of “early” and “late” RRTs. Patients in seven studies^{2,3,13–15,18,19} used CRRT and patients in two studies^10,17 used IHD. In the remaining studies,^{11,12,16,20–22} a part of patients used CRRT modality, and the others used IHD or SLED or PD modality. The study conducted by Bouman et al.² included three treatment strategies of early high-volume hemofiltration, early low-volume hemofiltration, and late low-volume hemofiltration, and the early high-volume hemofiltration strategy was excluded from our analyses.

Table 1. The basic characteristics of studies included in meta-analysis.

						Creatinine		Urea
First author	Publication year	Study design	Mean Age	Male (%)	Total number of patients	Early	Late	Early	Late
Bouman^[2]	2002	Randomized	69	59	71	5^a	6^a	NR	NR
Durmaz^[10]	2003	Randomized	56	77	44	304	286	22.0	18.0
Sugahara^[3]	2004	Randomized	64	64	28	256	265	NR	NR
Liu^[11]	2006	Prospective	56	68	243	301	415	16.9	41.0
Bagshaw^[12]	2009	Prospective	62	65	1237	230	396	15.0	38.8
Gettings^[13]	1999	Retrospective	45	79	100	148	238	15.2	33.7
Elahi^[14]	2004	Retrospective	70	75	64	328	379	23.9	26.8
Dermirkilic^[15]	2004	Retrospective	NR	78	61	NR	NR	NR	NR
Andrade^[16]	2007	Retrospective	43	NR	33	583	548	73.9	82.8
Manche^[17]	2008	Retrospective	65	NR	71	233	404	14.4	35.2
Iyem^[18]	2009	Retrospective	63	37	185	186	256	19.5	24.3
Shiao^[19]	2009	Retrospective	66	58	98	292	336	24.6	29.2
Carl^[20]	2010	Retrospective	54	67	147	442	514	23.6	48.9
Chou^[21]	2011	Retrospective	65	67	370	298	300	28.0	30.0
García-Fernández^[22]	2011	Retrospective	68	59	203	139	111	NR	NR

Note: ^aBouman et al. reported creatinine clearance (mL/min).

Table 2. The main characteristics of studies included in meta-analysis.

				Definitions of early and late
First author	Year	Modality	Population	Early	Late
Bouman^[2]	2002	CVVH	Cardiac surgery/medical	RRT within 12 h if urine output <30 mL/h	Urea >40 mmol/L or K >6.5 mmol/L
Durmaz^[10]	2003	IHD	Cardiac surgery (CABG)	Postoperative sCr increased by 10%	Postoperative sCr increased by 50% or urine output was <400 mL/24 h
Sugahara^[3]	2004	CVVH	Cardiac surgery	Urine output <30 mL/h	Urine Output <20 cc/h
Liu^[11]	2006	CRRT/IHD	Medical, surgery	Urea <27.1 mmol/L	Urea >27.1 mmol/L
Bagshaw^[12]	2009	CRRT/IHS	Medical, surgical	Urea <24.2 mmol/L	Urea >24.2 mmol/L
Gettings^[13]	1999	CRRT	Trauma	Urea <21.4 mmol/L	Urea >21.4 mmol/L
Elahi^[14]	2004	CVVH	Cardiac surgery	Urine output <100 cc in 8 h	K >6 mmol/L, Cr >250 mmol/L
Dermirkilic^[15]	2004	CVVHDF	Cardiac surgery	Cr >400 μmol/L, K >5.5 mmol/L	Oliguria
Andrade^[16]	2007	IHD/SLED	Medical (ARDS/sepsis)	On admission	At 24 h
Manche^[17]	2008	IHD	Cardiac surgery	Hyperkalemia	Urine output <0.5mL/kg/h
Iyem^[18]	2009	CVVH	Cardiac surgery	RRT on admission	After 48 h when anuric
Shiao^[19]	2009	CVVH	Surgery/trauma	RIFLE criteria (risk)	RIFLE injury (failure)
Carl^[20]	2010	CRRT/IHD	Medical (sepsis)	Urea <35.7 mmol/L	Urea >35.7 mmol/L
Chou^[21]	2011	CRRT/SLED	Sepsis	RIFLE criteria (risk)	RIFLE injury (failure)
García-Fernández^[22]	2011	CRRT/IHD	Cardiac surgery	≤3 days after cardiac surgery	>3 days after cardiac surgery

Note: RIFLE Criteria (Risk): Increase in serum creatinine by 1.5 times or urine output <0.5 mL/kg/h × 6 h; RIFLE Criteria Injury: Increase in serum creatinine by 2 times or urine output <0.5 mL/kg/h × 12 h; RIFLE Criteria (Failure): Increase in serum creatinine by 3 times or urine output <0.3 mL/kg/h × 24 h.

Figure 2. Risk ratio of mortality for the individual studies and pooled analysis.

Effects of Initiation Time of RRT on Mortality

Figure 2 shows the mortalities of individual studies and pooled analysis. The statistic p for heterogeneity is lower than 0.00001 (Figure 2); thus, we used the random effects model. Overall, 51.0% (772/1514) patients died in the “early” RRT group compared with 58.0% (836/1441) in the “late” RRT group. The pooled RR was 0.71 (95% CI: 0.59, 0.86), indicating a statistically significant beneficial effect of “early” RRT on mortality. But internal heterogeneity existed.

Subgroup Analysis

To investigate the factors that may result in heterogeneity, we performed subgroup analysis based on the modality of RRT. In the CRRT group, there were seven studies with 607 patients. Patients treated with “early” CRRT had a lower mortality (27.8%) compared with those in the “late” CRRT group (43.0%), and statistical significance was reached (RR 0.69; 95% CI: 0.56, 0.84). There was no statistical heterogeneity (I² = 33%, p = 0.18). In the IHD group, there were only two studies with 115 patients. One study by Durmaz et al.¹⁰ showed no protective effect (RR 0.16; 95% CI: 0.02, 1.17), whereas the other study by Manche et al.¹⁷ suggested some benefits (RR 0.29; 95% CI: 0.18, 0.47). As there was no statistical heterogeneity (I² = 0%, p = 0.50), the two trials were pooled. Patients treated with “early” IHD suggested a significant decrease in mortality compared with those in the “late” IHD group (19.5% vs. 52.6%). The pooled RR was 0.26 (95% CI: 0.15, 0.45), confirming the beneficial effects by “early” Hiding mixed group; there were six studies with 2233 patients. The pooled RR was 0.94 (95% CI: 0.88, 1.01), indicating no favor. It was confirmed that heterogeneity existed (I² = 79%, p < 0.00001). And we did not conduct meta-regression analysis to explore the factors causing heterogeneity (Figure 3).

Figure 3. Risk ratio of mortality in subgroup patients.

Assessment of Methodological Quality

The quality of the included studies was relatively low (Tables 3 and 4). In the three RCTs, only Durmaz et al.¹⁰ described the special method of randomization, the others just mentioned that the patients were randomized into one of the groups. Double blinding and allocation concealment were impossible; all of the three studies^2,3,10 used open-label design. Of the 12 nonrandomized trials, only 4^11,12,19,21 had a prospectively assembled control group and 2 studies^12,14 accounted for withdrawals and dropouts.

Table 3. Quality assessment of the three randomized controlled trials.

First author	Sequence of generation	Concealment of allocation	Blinding	Incomplete outcome data	Free of selective reporting
Bouman^[2]	Yes	No	No	Yes	No
Durmaz^[10]	Yes	No	No	Yes	No
Sugahara^[3]	No	No	No	Yes	No

Table 4. Quality assessment of the remaining 12 nonrandomized trials.

First author	a	b	c	d	e	f	g
Liu^[11]	Yes	Yes	Yes	Yes	No	No	No
Bagshaw^[12]	Yes	Yes	No	Yes	No	No	Yes
Gettings^[13]	Yes	Yes	Yes	No	Yes	Yes	No
Elahi^[14]	Yes	Yes	Yes	No	Yes	Yes	Yes
Dermirkilic^[15]	Yes	Yes	Yes	No	Yes	No	No
Andrade^[16]	No	No	Yes	No	Yes	No	No
Manche^[17]	No	No	Yes	No	No	No	No
Iyem^[18]	Yes	No	Yes	No	Yes	Yes	No
Shiao^[19]	Yes	Yes	Yes	Yes	Yes	Yes	No
Carl^[20]	Yes	Yes	Yes	No	Yes	No	No
Chou^[21]	Yes	Yes	Yes	Yes	Yes	No	No
García-Fernández^[22]	Yes	Yes	Yes	No	Yes	Yes	No

Notes: a. Were criteria for initiation of RRT clearly defined in each group?

b. Was the measurement of criterion (or lab value) for initiation of RRT reliable?

c. Were control and intervention group comparable with respect to disease type and demographics?

d. Was control group prospectively assembled? (vs. historical, or case control)

e. Were the control and intervention group comparable with respect to disease severity?

f. Was dialysis type comparable between groups in terms of dose, solution used, filtration versus dialysis, and type of membrane?

g. Was there a description of withdrawals and dropouts?

Figure 4. Assessment of publication bias using a funnel plot.

Sensitivity and Influence Analysis

The influence of each study was examined by omission from the analysis. Omission of any study in the CRRT group did not result in a change of the conclusion. In the IHD group, there were only two studies. The study by Durmaz et al.¹⁰ showed no significant benefit of “early” IHD compared with “late” Hiding mixed group; the conclusion did not change without any one study.

Publication Bias

Publication bias was evaluated using a funnel plot (Figure 4). The shape of the funnel plot was symmetrical, indicating no significant publication bias. Because of the small number of studies that met our inclusion criteria, publication bias was of concern.

DISCUSSION

The agreeable initiation time of RRT in patients with AKI remained uncertain. The initiation of RRT is extremely variable and based primarily on empiricism, local institutional practice, and resources.^23,24 There is also no consensus that “early” RRT has beneficial effect on mortality.

Our meta-analysis combined the results of 15 studies with 2955 patients. We revealed that “early” RRT could reduce mortality of patients with AKI, but heterogeneity across studies was found. Exploring the sources of heterogeneity by the subgroup analysis, we found that the modality influenced the heterogeneity. Therefore, we analyzed the effects in the subgroups of CRRT, IHD, and mixed.

Heterogeneity was not found in CRRT and IHD groups. The results suggested that in the CRRT group, “early” CRRT had beneficial effect on the mortality of patients with AKI (RR 0.69; 95% CI: 0.56, 0.84). We think that “early” CRRT could clear the inflammatory factor away effectively and timely. Inflammatory factors are now thought of as an important factor contributing to AKI. Removal of inflammatory factor may stop the injury of kidney and reduce mortality. “Early” CRRT also could clear wastes of body away which could aggravate the renal function. Similarly, in the IHD group, “early” IHD could also improve the survival (RR 0.26; 95% CI: 0.15, 0.45). However, in the IHD group, there were only two studies. The study by Durmaz et al.¹⁰ did not confirm the conclusion; further investigations are required to explore the effect. Karvellas et al.⁹ and Seabra et al.²⁵ had the same opinion that early institution of RRT in critically ill patients with AKI may have a measurable benefit on survival. However, compared with these meta-analysis,^9,25 we have two strengths. One was that we found the modality was the factor that resulted in heterogeneity, but neither Karvellas et al.⁹ nor Seabra et al.²⁵ could succeed in finding the factors. The other was we have included two new clinical studies^21,22 published recently, which made our results more convincing.

Our meta-analysis had some limitations. First, there were only 3 RCTs, the other 12 studies were non-RCTs, and the quality of many were not very high. Second, because of the small number of studies that met our inclusion criteria, publication bias existed. As a consequence, the beneficial effect of “early” RRT on mortality may be overstated compared with “late” RRT. Third, we did not perform meta-regression to assess the factors that resulted in heterogeneity; we specified subgroup analysis limited to patients with CRRT or IHD according to the opinion of some studies.^6,7

In conclusion, our study suggested that no matter “early” CRRT or “early” IHD both could reduce the mortality of patients with AKI. But due to the limitations of this meta-analysis, more prospective randomized trials are needed to confirm the results.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.