Lai et al. Ann.

Intensive Care (2017) 7:38

DOI 10.1186/s13613-017-0265-6

RESEARCH Open Access

Earlier versus later initiation of renal

replacement therapy among critically ill patients
with acute kidney injury: a systematic review
and meta‑analysis of randomized controlled
trials
Tai‑Shuan Lai1,2†, Chih‑Chung Shiao3,4†  , Jian‑Jhong Wang5, Chun‑Te Huang6, Pei‑Chen Wu7, Eric Chueh8,
Shih‑Chieh Jeff Chueh9, Kianoush Kashani10,11* and Vin‑Cent Wu12,13*

Abstract 
Background:  Although the optimal timing of initiation of renal replacement therapy (RRT) in critically ill patients
with acute kidney injury has been extensively studied in the past, it is still unclear.
Methods:  In this systematic review, we searched all related randomized controlled trials (RCTs) that directly com‑
pared earlier and later RRT published prior to June 25, 2016, from PubMed, MEDLINE, and EMBASE. We extracted the
study characteristics and outcomes of all-cause mortality, RRT dependence, and intensive care unit (ICU) and hospital
length of stay (LOS).
Results:  We identified 51 published relevant studies from 13,468 screened abstracts. Nine RCTs with 1627 partici‑
pants were included in this meta-analysis. Earlier RRT was not associated with benefits in terms of mortality [rela‑
tive risk (RR) 0.88, 95% confidence interval (CI) 0.68–1.14, p = 0.33] and RRT dependence (RR 0.81, 95% CI 0.46–1.42,
p = 0.46). There were also no significant differences in the ICU and hospital LOS between patients who underwent
earlier versus later RRT [standard means difference −0.08 (95% CI −0.26 to 0.09) and −0.11 (95% CI −0.37 to 0.16) day,
respectively]. In subgroup analysis, earlier RRT was associated with a reduction in the in-hospital mortality among sur‑
gical patients (RR 0.78, 95% CI 0.64–0.96) and patients who underwent continuous renal replacement therapy (CRRT)
(RR 0.80, 95% CI 0.67–0.96).
Conclusions:  Compared with later RRT, earlier initiation of RRT did not show beneficial impacts on patient out‑
comes. However, a lower rate of death was observed among surgical patients and in those who underwent CRRT. The
included literature is highly heterogeneous and, therefore, potentially subject to bias. Further high-quality RCT studies
are warranted.
Keywords:  Acute kidney injury, Length of stay, Meta-analysis, Mortality, Renal replacement therapy, Timing, CAKS,
NSARF

*Correspondence: [email protected]; [email protected]

†
Tai-Shuan Lai and Chih-Chung Shiao contributed equally to this work
10
Division of Nephrology and Hypertension, Department of Medicine,
Mayo Clinic, No. 200 First St. SW, Rochester, MN 55905, USA
12
Division of Nephrology, Department of Internal Medicine, National
Taiwan University Hospital, No. 7 Chung‑Shan South Road, Zhong‑Zheng
District, Taipei 100, Taiwan
Full list of author information is available at the end of the article

© The Author(s) 2017. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License
(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium,
provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license,
and indicate if changes were made.
Lai et al. Ann. Intensive Care (2017) 7:38 Page 2 of 14

Background opposing results, they added value to the field of critical

Acute kidney injury (AKI) is a common yet potentially care. This systematic review is conducted to include all
fatal complication of illnesses among 1% of the commu- relevant RCTs related to the impact of the timing of RRT
nity-based population, 8–15% of hospitalized patients, initiation among critically ill patients with moderate to
and up to 50% of critically ill patients admitted to the severe AKI.
intensive care unit (ICU) [1–5]. AKI carries increased
risk of morbidity and mortality and adds to the health- Methods
care cost, even in mild temporary form [6–11]. Search strategy and selection criteria
Although renal replacement therapy (RRT) remains In concordance with the Preferred Reporting Items
the primary supportive management strategy for patients for Systematic Reviews and Meta-analyses (PRISMA)
with severe AKI, it could also be associated with com- guidelines, we conducted a systematic review and meta-
plications and adverse events [12–14]. Despite improve- analysis to investigate the effect of earlier initiation of
ments in RRT technology, it is still not clear whether RRT on the outcomes of critically ill patients with AKI
the outcome of patients with AKI who require RRT has who require dialysis [30]. We searched MEDLINE, Pub-
improved over the years [7, 15]. Earlier initiation of RRT Med, and EMBASE databases and identified the relevant
may provide a better control of fluid and electrolyte bal- articles published up to June 25, 2016, using Web of Sci-
ance, superior acid–base homeostasis, removal of ure- ence. We screened references by titles and abstracts and
mic waste, and prevention of subsequent complications included related studies for further analysis. Case reports
attributable to AKI [16]. Furthermore, earlier RRT could or case series, non-English articles, articles not focused
potentially limit the kidney-specific and remote organ on critically ill patients, studies consisting of pediat-
injuries due to fluid overload, electrolyte imbalance, and ric patients, studies that did not present mortality data,
systemic inflammation [17]. However, earlier RRT may and those that did not clearly define the timing of initia-
also expose the patients to increased risks of hemody- tion of RRT were excluded. The keywords used for data-
namic instability, anticoagulation-induced bleeding, base search were provided in Additional file 1: Table S1.
blood-stream infection, and even inflammatory or oxida- We only included studies with randomized controlled
tive stress induced by the bio-incompatibility of the dia- designs in the final meta-analysis. Both abstracts and full
lyzer membranes. In comparison, later initiation of RRT papers were selected for quality assessment and data syn-
may allow more time for hemodynamic optimization theses. We contacted the authors of abstracts for further
prior to RRT, and it may avoid the need for RRT and its details, if available.
associated complications [18].
In recent decades, the timing of RRT initiation has Data extraction and synthesis
been evaluated in different population types (e.g., surgi- We extracted data regarding the year of publication
cal or medical patients). Variability in the definitions of and patient enrollment, leading author, the number
AKI and RRT timing has resulted in contradicting con- of patients, and events from each article. When avail-
clusions among the various studies [19–23]. Similarly, able, odds ratios and 95% confidence intervals (CIs) from
previous systematic analyses regarding the optimal tim- these RCTs were extracted. Other parameters for record
ing of RRT initiation were unable to draw definitive con- included the type of patient setting (surgical/mixed/
clusions owing to the scarcity of large-scale randomized medical), criteria used for AKI diagnosis, cohort size,
controlled trials (RCTs), non-standardized triggers for presence of sepsis, study quality, and the proportions of
RRT initiation, and heterogeneities of population and patients on mechanical ventilation. Two researchers (TSL
study design. In summary, while the observational stud- and CCH) independently extracted the data, and a third
ies tended to show more beneficial effects for earlier investigator (VCW) resolved any disagreements between
RRT, clinical trials were unable to replicate these findings them.
[24–27]. Recently, two large RCTs showed contradictory
results and attracted considerable attention from both Risk of bias assessment
clinicians and researchers. The first was a multicenter We assessed the risk of bias in the included articles
RCT by the AKIKI study group [28], which showed no using structured assessment tools. For RCTs, we use
significant differences in 60-day mortality between early Cochrane review tools to access the risk of bias [31]. We
and delayed RRT groups. Another was the ELAIN trial, evaluated the adequacy of randomization and conceal-
[29] a single-center RCT that showed significant ben- ment, blinding, reporting of outcomes, sample size cal-
efits in terms of 90-day mortality, renal function recov- culation, and disclosure of funding sources. We assessed
ery, and hospital length of stay (LOS) among patients in the overall study quality according to current standards
the early RRT group. Although these two RCTs exhibited [31, 32].
Lai et al. Ann. Intensive Care (2017) 7:38 Page 3 of 14

Definition of earlier versus later RRT initiation patients with AKI. These nine RCTs were included in
The definition of earlier initiation of RRT varied substan- the meta-analysis (Table 1). The trials were conducted in
tially among included RCTs. In four of the included tri- Europe, Asia, and Canada between 2002 and 2016.
als, initiation of RRT immediately after randomization A total of 1627 critically ill patients who underwent
was defined as “early” [28, 33–35]. In two studies, early acute dialysis were enrolled in the final analysis. Seven
initiation was considered when RRT started in less than of the nine studies provided quantifiable results for
12 h of admission to ICU, while in another study, authors RRT dependence during the follow-up period. Four
used the serum BUN > 70 mg/dL or creatinine >7 mg/dL trials recruited surgical patients only, and two of them
and the defining criteria for RRT initiation [36, 37]. One enrolled patients undergoing coronary bypass sur-
study compared prophylactic hemodialysis before sur- gery. The remaining five trials enrolled patients in the
gery with standard care [38]. In ELLAIN trial, early RRT mixed surgical/medical ICU setting. Continuous renal
was defined as initiation within 8 h of diagnosis of stage 2 replacement therapy (CRRT) was used as the modality
AKI using the KDIGO classification [29]. We included all of choice in five trials, and two trials used intermittent
definition of early dialysis based on each individual study hemodialysis (IHD). The remaining studies utilized a
design in order to evaluate the potential effect of early mixture of the two dialysis modalities. Early high-vol-
dialysis on the primary outcome; obviously, this leads to ume hemofiltration was compared with the standard
increased heterogeneity observed in our analysis. of care in one RCT. The quality of the included RCTs
varied; most of the studies lacked sufficient informa-
Ascertainment of outcomes tion regarding participants or personnel blinding and
The primary outcome of interest was all-cause mortality, concealment processes (Additional file  1: Fig. S1). We
including in-hospital mortality and 30-, 60-, and 90-day divided these studies into high or low quality, with 3 out
mortality. We also evaluated RRT dependence after hos- of 6 domains of bias as the cutoff for the quality assess-
pital discharge. The secondary outcomes were ICU or ment tool. Publication bias was tested, and funnel plot
hospital LOS. was drawn. There was no obvious impact of study sam-
ple size with a p  =  0.846 for Egger’s test (Additional
Statistical analyses file 1: Fig. S2).
Owing to the significant heterogeneity among the
enrolled studies, we used the random effects model. The Primary outcomes
overall summary risk ratios (RRs) and 95% CIs were cal- Among the included trials, the pooled mortality rates
culated using the Mantel–Haenszel method. We charac- were 38.7% (309 of 798) and 42.5% (352 of 829) in the
terized the heterogeneity with the I2 and τ2 statistic. A p groups of patients who underwent earlier and later
value ≤0.05 was considered statistically significant. Sen- RRT, respectively. Pooled estimates of included stud-
sitivity analyses were conducted for variables that could ies indicated no significant survival benefit in patients
modify the effect of initiation time and mortality. In sub- who underwent earlier RRT compared with those who
group analysis, we performed meta-regression to assess underwent later RRT, with an RR of 0.88 (95% CI 0.68–
the effect of interaction between variables and the tim- 1.14, p = 0.33) (Fig. 2). Substantial heterogeneity existed
ing of RRT initiation on mortality and RRT dependence. among studies, with an I2 value of 64.6%, and Chi-square
Funnel plots were drawn to evaluate the distribution of p  =  0.004. In addition, there were no significant differ-
studies. Begg’s rank correlation test and Egger’s linear ences in the 30-, 60-, and 90-day mortality between
regression were used to assess the publication bias. We groups with earlier and later RRT (Fig. 3).
used STATA (version 13, Stata Corp. 2013. College Sta- In the subgroup analyses, there were no differences
tion, TX: Stata Corp LP), and Review Manager (RevMan) between patients who underwent earlier and later RRT in
(version 5.2. Copenhagen: The Nordic Cochrane Cen- the majority of subgroups, with three exceptions. Nota-
tre, The Cochrane Collaboration, 2014) software for the bly, earlier RRT seemed to provide a survival benefit in
meta-analysis. surgical patients (RR 0.78, 95% CI 0.64–0.95), but not in
patients in the mixed ICUs (RR 1.00, 95% CI 0.87–1.16).
Results This is despite the fact that we found no substantial evi-
Study characteristics dence of such differences when trials were stratified
Figure 1 shows the flowchart of the literature search and by the ICU setting (p  =  0.31 for interaction). Besides,
selection process. We screened 13,468 abstracts, of which the survival benefit of earlier RRT initiation was also
174 were eligible for full-text reviews. A total of 51 stud- observed in the patients who started with CRRT (RR
ies including 9 RCTs and 42 cohort studies presented 0.80, 95% CI 0.66–0.95), but not in those who received
data on the timing of RRT initiation among critically ill mixed modalities (RR 1.01, 95% CI 0.86–1.17) or IHD
Lai et al. Ann. Intensive Care (2017) 7:38 Page 4 of 14

Fig. 1  Flowchart of study selection for meta-analysis. RCT randomized controlled trials

(RR 1.15, 95% CI 0.66–2.01) (p  =  0.43 for interaction) Secondary outcomes
(Fig. 4; Additional file 1: Fig. S3). The mean weighted ICU LOS was 12.5  days (n  =  604)
Seven of the nine included RCTs reported information in the earlier RRT group and 13.0  days (n  =  614) in
about RRT dependence. There was no statistically signifi- the later RRT group. The mean weighted hospital LOS
cant difference in the risk of RRT dependence between was 29.4  days (n  =  604) in the earlier RRT group and
patients with earlier and later initiation of RRT, with a 31.3 days (n = 614) in the later RRT group. Pooled analy-
pooled RR of 0.81 (95% CI 0.46–1.42) (Fig. 5). There was sis demonstrated no significant differences in the ICU
no evidence for heterogeneity with an I2 value = 0% and LOS and hospital LOS between the two groups, with a
Chi-square p  =  0.748). In the subgroup analysis, there standard difference in the means of −0.08  day (95% CI
was no statistically significant difference in RRT depend- −0.26 to 0.09) and −0.11  day (95% CI −0.37 to 0.16),
ence across different subgroups (Additional file  1: Fig. respectively (Additional file 1: Figs. S5, S6).
S4).
Table 1  Summary of included randomized controlled trials
References Population Study Nation Sites1 Inclusion criteria Exclusion criteria n Male Mean Mode Sepsis Endpoint Counts
setting period Age

1 Bouman [53] Mixed Not speci‑ Netherland M Oliguria <30 cc/h, shock eGFR <30; cirrhosis 71 42 (59%) 68.4 CRRT n/a In-hospital I: 18/35
fied with vasopressor obstructive AKI mortality C: 14/36
support
Lai et al. Ann. Intensive Care (2017) 7:38

In-ICU mor‑ I: 13/35

tality C: 10/36
28-day mor‑ I: 11/35
tality C: 9/36
2 Durmaz [38] Surgical n/a Turkey S n/a n/a 44 n/a n/a IHD 2 (4.5%) In-hospital I: 1/20
mortality C: 7/16
3 Sugahara [33] Surgical 1995–1997 Japan S Urine output < 30 cc/h Pregnant 28 18 (64%) 64.5 CRRT n/a 14-day mor‑ I: 2/14
or increase in Severe hepatic dysfunction tality C: 12/14
SCr > 0.5 mg/dL/day (serum bilirubin level of
5.0 mg/dL or higher)
Mental disorders Cancers
4 Koo [34] Mixed Not speci‑ Korea S Severe sepsis or septic Not specified 102 62 (61%) 62 CRRT 102 In-hospital I: 12/43
fied shock (100%) mortality C: 30/59
5 Jamale [37] Mixed 2011–2012 India S Severe AKI with increas‑ Needing urgent dialysis; 208 141 42.4 IHD 44 (21%) In-hospital I: 21//102
ing BUN and SCr levels Previous dialysis history (68%) mortality C: 13/106
Recovering AKI
6 Combes [35] Surgical 2009–2012 France M Postoperative shock, <18 years old; pregnant; 224 177 59.4 CRRT n/a In-hospital I: 50/112
needing high dose Enrolled into current or (79%) mortality C: 44/112
vasopressors or ECMO other trials previous RRT
Weight > 120 kg In-ICU mor‑ I: 50/112
SAPSII > 90 tality C: 44/112
30-day mor‑ I: 40/112
tality C: 40/112
60-day mor‑ I: 48/112
tality C: 42/112
90-day mor‑ I: 51/112
tality C: 43/112
Page 5 of 14
Table 1  continued
References Population Study Nation Sites1 Inclusion criteria Exclusion criteria n Male Mean Mode Sepsis Endpoint Counts
setting period Age

7 Wald [36] Mixed 2013–2013 Canada M KDIGO AKI Stage 3, Withdrawal of life support 100 72 (72%) 63.1 Mixed 56 (56%) In-hospital I: 16/48
oliguria > 12 h or Intoxication mortality C: 19/52
NGAL > 400 ng/ml RRT within the previous
Lai et al. Ann. Intensive Care (2017) 7:38

2 months; renal obstruc‑ In-ICU mor‑ I: 13/48

tion, rapidly progressive tality C: 16/52
glomerulonephritis, 90-day mor‑ I: 18/48
or interstitial nephritis; tality
eGFR < 30 ml/min per C: 19/52
1.73 m2 the passage of
48 h since doubling of
baseline serum creatinine
8 Zarbock [29] Surgical 2013–2015 Germany S KDIGO AKI stage 2 eGFR < 30 ml/ 231 146 67.0 CRRT 75 (32%) 30-day mor‑ I: 34/112
min/1.732 m2 previous (63%) tality C: 48/119
RRT
AKI caused by permanent 60-day mor‑ I: 43/112
occlusion of renal artery tality C: 60/119
or surgery 90-day mor‑ I: 44/112
Glomerular nephritis, inter‑ tality
stitial nephritis, hemo‑ C: 65/119
lytic uremic syndrome
Pregnancy
AIDS
Heptorenal syndrome
9 Gaudry [28] Mixed 2013–2016 France M Ischemic or toxic AKI BUN > 112 mg/dL 619 405 66.2 Mixed 483 (78%) 30-day mor‑ I: 129/311
and receiving invasive Blood K > 6.0 mmol/L (65% tality C: 134/308
mechanical ventilation, A pH < 7.15
catecholamine infusion Acute pulmonary edema 60-day mor‑ I: 150/311
or both, and AKIN tality C: 153/308
Stage 3
I Intervention group, C control group, RCT randomized controlled trial, S single-center study, M multicenter study, n/a not available, BUN blood urea nitrogen, SCr serum creatinine, ECMO extracorporeal membrane
oxygenator, AKI acute kidney injury
Page 6 of 14
Lai et al. Ann. Intensive Care (2017) 7:38 Page 7 of 14

Fig. 2  Forest plot for all-cause mortality: all studies. RCT randomized controlled trials, RRT renal replacement therapy

Discussion Subgroup analyses

In the current systematic review of nine RCT includ- We hypothesized that using consensus AKI definitions,
ing 1627 critically ill patients with AKI, who received enrolling sepsis-associated AKI, differences in sample
RRT, we did not find any significant survival benefits in sizes and study qualities had high impacts on patient out-
patients who underwent earlier versus later RRT. A con- comes observed among different investigations. When
siderable heterogeneity across studies was observed. we used different AKI definitions, septic AKI, and study
Furthermore, 30-, 60-, and 90-day mortality, dialysis quality for subgroup analyses, we found no difference
dependence, and LOS in the hospital or ICU were not between earlier versus later RRT initiation time.
lower in patients who underwent earlier RRT in com- We found survival benefit for earlier RRT initiation
parison with those who underwent later RRT. We noticed when CRRT was utilized. Previous studies including one
a lower mortality rate in the earlier RRT group only in meta-analysis showed no difference in mortality or RRT
postsurgical patients and among those who underwent dependence between various dialysis modalities [40, 43],
CRRT. while other meta-analyses showed that the use of CRRT
In recent interventional studies, no survival benefits decreases mortality or RRT dependence [42]. However,
have been observed among different intensities [39] and these findings largely were dependent on data from
modalities of dialysis [40]. The optimal timing of RRT ini- observational trials, which were potentially biased by
tiation still remains debatable owing to the contradictory allocation and the qualities were uncertain. Our analysis
reports in the literature. To our knowledge, the current focused on RCTs, mostly with high qualities and appro-
systematic review is the first to exclusively include most priate randomization, and the results were more reliable.
RCTs to address the issue of the timing of RRT initia- The possible mechanisms of the observed benefits from
tion and evaluate its impact on patient survival and RRT CRRT as the dialysis modality include gentler osmolar
dependence (Table 2) [24–27, 41, 42]. shifts, lower overall cumulative fluid balance, and clear-
ance of inflammatory factors [44]. Our study is not able
Lai et al. Ann. Intensive Care (2017) 7:38 Page 8 of 14

Fig. 3  Forest plots for mortality in a 30 days, b 60 days, and c 90 days

Lai et al. Ann. Intensive Care (2017) 7:38 Page 9 of 14

Fig. 4  Forest plot for all-cause mortality: in subgroups. RCT randomized controlled trials, RRT renal replacement therapy

to identify the reasons behind improved outcomes with undergo elective surgeries have undergone extensive pre-
CRRT, and further studies are warranted. operative evaluation and optimization which contributed
We also reported a survival benefit for postsurgi- to their better outcomes in comparison with those of
cal critically ill patients with AKI when they received medical ICU patients [22].
“earlier” RRT. A meta-analysis showed that early initia- In septic patients, earlier RRT was not found to be
tion of RRT for patients with AKI after cardiac surgery associated with improvement in mortality or RRT
improved mortality [45]. Postoperative fluid overload in dependence. In this subgroup of patients, sepsis-associ-
the surgical ICU is very common [46], and these patients ated AKI due to intrinsic renal lesions is only one part of
may benefit from the earlier removal of excessive fluid the puzzle. Often, mortality in these patients correlates
by RRT [47, 48]. Other supporting evidence came from with various sepsis-induced inflammatory tissue dam-
the observation that compared to patients admitted to ages and multi-organ failure [51]. Therefore, a single
medical ICU, those admitted to the surgical ICU admis- intervention may not be able to show a significant change
sions at a greater risk for aggravation of cardiovascular, in ICU outcomes, such as mortality. Furthermore, the
neurological, and respiratory diseases [49]. The literature possible adverse effects of earlier RRT such as enhanced
review indicated that following initial resuscitation in the clearance of antibiotics, amino acids and nutrients and
postsurgical critical care setting, maintaining appropri- hypothermia may counteract the benefits of a timely
ate fluid balance through earlier RRT is clinically relevant RRT. Moreover, in some studies, earlier initiation of RRT
[50]. Unlike surgical patients who often suffer from single showed deleterious effects on the outcomes of patients
organ failure, the heterogeneity of medical ICU patients with severe sepsis and septic shock; in addition, no differ-
may limit the effect of a single intervention (in this case ences were detected in their plasma cytokine levels [52].
“earlier” RRT). Additionally, many surgical patients who Our results confirmed that earlier RRT initiation had
Lai et al. Ann. Intensive Care (2017) 7:38 Page 10 of 14

Fig. 5  Forest plot for RRT dependence: all studies. RCT randomized controlled trials, RRT renal replacement therapy

no beneficial effects on the clinical outcomes of patients Furthermore, observational studies are more subject to
with sepsis-related AKI. the selection bias when compared with RCTs.
There have been previous systematic reviews consist- As we showed, there is a significant heterogeneity
ing of a mixture of non-randomized cohort studies and among the studies related to the timing of RRT initia-
a limited number of RCT regarding the optimal tim- tion which may impact the results we found in this sys-
ing of RRT initiation [26, 27, 41, 42]. We were not able tematic analysis. There are some possible explanations
to confirm these reports in our systematic review. Previ- for the discordance and heterogeneity among differ-
ous studies concluded earlier RRT was associated with ent studies. Using varied definitions of AKI and differ-
decreased mortality and RRT dependence in critically ill ent AKI stage criteria for RRT initiation accounted for
patients with AKI [26]. Contrary to the previous reports, part of the observed heterogeneity. In the majority of
we did not find a significant effect of earlier RRT on previously reported cohort studies, the differences in
either ICU or hospital mortality and LOS, and dialysis pre-intervention study groups contributed to the hetero-
dependence. In our study, only patients who underwent geneity of the results, making the systematic reviews dif-
CRRT or postsurgical patients showed benefits in terms ficult to interpret. Furthermore, “patients without RRT”
of the mortality rate for earlier RRT initiation. were not used as “control” in cohort studies. As illus-
One of the differences between the current study and trated by the AKIKI trial [28], the mortality in patients
previous reports was the inclusion of RCTs only, includ- in the “delayed RRT” arm who never underwent RRT was
ing the two latest published RCT studies [28, 29], which lower than the mortality of patients who actually under-
accounted for the different results of our study from went RRT. Excluding patients who did not undergo RRT
those of the previously studies [24, 26, 27, 41]. Prior resulted in a significant bias. Another explanation is that
meta-analyses that concluded survival benefit attributed compared with RCTs, observational studies (especially
to earlier RRT initiation relied heavily on data from ret- retrospective studies) are more subject to the selection
rospective cohort studies that may possess incomplete bias. This highlights the critical need for a consensus
pre-intervention data, preexisting significant differ- definition of earlier versus later dialysis for the future
ences among groups and heterogeneous study designs. studies and highlights the knowledge gap in the field. The
Table 2  Comparisons of meta-analyses evaluating timing of RRT initiation in AKI and patients outcomes
Population setting Enrolled studies Outcomes Results (benefit of early RRT) Limitations

Current study Mixed patients with AKI (n = 1627) Total nine RCTs. (publication date: In-hospital mortality; RRT depend‑ All: no significant advantage in High heterogeneity among studies,
2002–2016) ence, 30-, 60-, 90-day mortality survival (in-hospital, 30-, 60-, varied definitions of early RRT
after hospital discharge 90-day mortality) and RRT
dependence
Subgroups: Significant survival
benefit in surgical patients and
those started with CRRT
Seabra [41] Mixed patients with AKI (n = 2378) Total 23 studies including 4 RCTs, 1 Mortality RCTs nonsignificant 36% reduction Paucity of RCTs, varied definitions of
quasi-RCTs, 1 prospective study, of mortality risk early RRT, many small sized stud‑
Lai et al. Ann. Intensive Care (2017) 7:38

16 retrospective studies, and 1 Cohort studies significant 28% ies, publications bias
single-arm study (publication reduction of mortality risk (with
date: 1961–2006) significant heterogeneity among
cohort studies)
Subgroups smaller studies
(n < 100) were more likely to
show the benefit of early RRT
Karvellas [26] Mixed patients with AKI (n = 2684) Total 15 studies including 2 RCTs, 28-day mortality All significant 55% reduction of Varied quality and high heterogene‑
4 prospective studies, and 9 mortality risk (with significant ity among studies Some studies
retrospective studies (publica‑ heterogeneity) were of small sample size Diverse
tion period: 1999–2010) Subgroups survival benefit remains definitions of early vs late RRT
in subgroup analysis according
to ICU type, study design, and
illness severity
Wang [42] Mixed patients with AKI (n = 2955) Total 15 studies including 3 RCTs, Mortality All significant 29% reduction of Many studies were of relative low
About 50% studies involved 2 prospective studies, and 10 mortality risk (with high hetero‑ quality, small sample size, diverse
surgical patients retrospective studies (publica‑ geneity) definitions of early vs late RRT
tion period: 1990–2011) Subgroups significant reduction of
mortality risk of 31% in CRRT and
74% in IHD (without evidence of
heterogeneity)
Liu [45] Surgical patients with AKI (after Total 11 studies including 2 RCTs 28-day mortality; ICU LOS Significant 71% reduction of Based on studies with various qual‑
cardiac surgery) (n = 841) and 9 retrospective studies (pub‑ 28-day mortality risk and ity with very high heterogeneity
lication period: 1972–2011) 3.9 days shorter ICU LOS (with of results
high heterogeneity among
studies)
Wierstra [24] Mixed patients with AKI (n = 1042) Total nine high-quality studies 1-month mortality; ICU/hospital All no significant advantage in Statistically significant heterogene‑
including 6 RCTs, 1 prospec‑ LOS survival and ICU/hospital LOS ity among studies Diverse defini‑
tive study, and 2 retrospective Subgroups no advantage in sur‑ tions of early vs late RRT
studies (publication period: vival and ICU/hospital LOS
2002–2015)
Xu [25] Mixed patients with AKI (n = 1257) Total six RCTs (publication period: Mortality, renal recovery, compos‑ No difference in mortality, renal Insufficient number of studies
2002–2016) ite endpoint recovery, composite endpoint included, some RCTs were rela‑
tively small, diverse definitions of
early vs late RRT
AKI acute kidney injury, CRRT continuous renal replacement therapy, ICU intensive care unit; IHD intermittent hemodialysis, LOS length of stay, RCT randomized controlled trial
Page 11 of 14
Lai et al. Ann. Intensive Care (2017) 7:38 Page 12 of 14

STARRT-AKI study, a 2015 pilot trial that evaluated the Authors’ contributions
VCW chaired the group, conceived and designed the study, performed statisti‑
feasibility and safety of early versus standard timing for cal analysis, and contributed to data collection, data interpretation, and critical
starting RRT, will provide more evidence about the opti- revision of the manuscript. TSL conducted a literature search and statistical
mal timing of RRT initiation in AKI. analysis and wrote the manuscript. CCS performed a literature search, wrote
the manuscript, and performed a critical revision of the manuscript. JJW, CTH,
There are some limitations associated with our study. and PCW performed literature search and summary. EC, KK, and SJC wrote
In our systematic review, we found no further informa- the manuscript and performed a critical review of the manuscript. All authors
tion regarding the other factors associated with mortal- contributed to subsequent drafts and examined the paper. All authors read
and approved the final manuscript.
ity; therefore, we cannot comment on the differences
in the outcomes on the basis of a single intervention, Author details
i.e., earlier or later RRT initiation. Furthermore, no trial 1
 Division of Nephrology, Department of Internal Medicine, National Taiwan
University Hospital Bei-Hu Branch, No. 87, Neijiang St, Taipei 108, Taiwan.
standardized the dialysis modality or dose delivered dur- 2
 Community and Geriatric Research Center, National Taiwan University Hospi‑
ing RRT. We were not able to access the unpublished tal Bei-Hu Branch, No. 87, Neijiang St, Taipei 108, Taiwan. 3 Division of Nephrol‑
reports, which might have biased our results. Although ogy, Department of Internal Medicine, Saint Marys Hospital Luodong, No. 160,
Zhongheng S. Rd., Luodong, Yilan 26546, Taiwan, ROC. 4 Saint Mary’s Medicine,
our funnel meta-regression analysis showed a limited Nursing and Management College, No. 100, Ln. 265, Sec. 2, Sanxing Rd., Sanx‑
publication bias (Additional file  1: Fig. S2), the bias was ing Township, Yilan County 266, Taiwan, ROC. 5 Division of Nephrology, Depart‑
always difficult to ascertain with a small sample number ment of Internal Medicine, Chi-Mei Medical Center, Liouying. No. 201, Taikang,
Taikang Vil., Liuying Dist.736, Tainan City, Taiwan. 6 Division of Internal and Criti‑
of the included studies. Finally, the definition of “ear- cal Care Medicine, Department of Critical Care Medicine, Taichung Veterans
lier” RRT was variable and may have unduly influenced General Hospital, No. 1650 Taiwan Boulevard Sect. 4, Taichung 40705, Taiwan.
pooled effect estimates. As defined by traditional mark- 7
 Division of Nephrology, Department of Internal Medicine, Mackay Memorial
Hospital, No. 92, Sec. 2, Zhongshan N. Rd., Taipei 10449, Taiwan. 8 Case Western
ers, RRT was initiated relatively late which may influence Reserve University, No. 10900 Euclid Ave., Cleveland, OH 44106, USA. 9 Cleve‑
the effectiveness of the early treatment. Furthermore, in land Clinic Lerner College of Medicine and Glickman Urological and Kidney
the majority of the enrolled studies, clinical patient care Institute, Cleveland Clinic, No. 9980, Carnegie Ave, Cleveland, OH 44195, USA.
10
 Division of Nephrology and Hypertension, Department of Medicine, Mayo
is individualized based on the discretion of the clinician. Clinic, No. 200 First St. SW, Rochester, MN 55905, USA. 11 Division of Pulmonary
This would add to the heterogeneity of the studies and and Critical Care Medicine, Department of Medicine, Mayo Clinic, No. 200
their results. First St. SW, Rochester, MN 55905, USA. 12 Division of Nephrology, Department
of Internal Medicine, National Taiwan University Hospital, No. 7 Chung‑Shan
The strength of our present analysis rested on our South Road, Zhong‑Zheng District, Taipei 100, Taiwan. 13 National Taiwan
extensive literature search on RCTs. We used stand- University Study Group on Acute Renal Failure (NSARF), Taipei, Taiwan.
ard Cochrane protocols and had the largest cumulative
Acknowledgements
RCT study sample size in comparison with the previous The authors thank Prof. Hsien-Ho Lin and Dr. Chien-Chang Lee for providing
reports. We only focused on the RCTs that had a reason- statistical opinions of meta-analysis.
able quality with limited differential dropout based on
Competing interests
the assigned treatment arm. The authors declared that they have no competing interests.

Conclusion Funding
This work was supported by the TR15, CAKS, National Research Program for
Compared to later initiation of RRT, earlier RRT initia- Biopharmaceuticals, ROC, Taiwan.
tion in critically ill patients with AKI does not decrease
mortality and long-term RRT dependence and does not Publisher’s Note
alter the length of hospital stay. Earlier initiation of CRRT Springer Nature remains neutral with regard to jurisdictional claims in pub‑
and earlier RRT in postsurgical patients may be associ- lished maps and institutional affiliations.
ated with improved mortality. Future large-scale, mul- Received: 14 November 2016 Accepted: 28 March 2017
ticenter, prospective interventional trials are needed to
delineate the characteristics of patients who benefit from
earlier initiation of RRT.

Additional file References

1. Abelha FJ, Botelho M, Fernandes V, Barros H. Determinants of postopera‑
tive acute kidney injury. Crit Care. 2009;13:R79.
Additional file 1. Supplementary information. 2. Nisanevich V, Felsenstein I, Almogy G, Weissman C, Einav S, Matot I. Effect
of intraoperative fluid management on outcome after intraabdominal
surgery. Anesthesiology. 2005;103:25–32.
Abbreviations 3. Clark E, Wald R, Walsh M, Bagshaw SM. Timing of initiation of renal
AKI: acute kidney injury; CI: confidence interval; CRRT: continuous renal replacement therapy for acute kidney injury: a survey of nephrologists
replacement therapy; ICU: intensive care unit; IHD: intermittent hemodialysis; and intensivists in Canada. Nephrol Dial Transpl. 2012;27:2761–7.
LOS: length of stay; PRISMA: Preferred Reporting Items for Systematic Reviews 4. Balk RA. Severe sepsis and septic shock. Definitions, epidemiology, and
and Meta-analyses; RCT: randomized controlled trial; RRT: renal replacement clinical manifestations. Crit Care Clin. 2000;16:179–92.
therapy; RR: risk ratio. 5. Stein A, de Souza LV, Belettini CR, Menegazzo WR, Viégas Jú R, Costa
Pereira EM, et al. Fluid overload and changes in serum creatinine after
Lai et al. Ann. Intensive Care (2017) 7:38 Page 13 of 14

cardiac surgery: predictors of mortality and longer intensive care stay. A critically ill patients with acute kidney injury: a systemic review and meta-
prospective cohort study. Crit Care. 2012;16:R99. analysis. Crit Care. 2011;15:R72.
6. Bagshaw SM, Laupland KB, Doig CJ, Mortis G, Fick GH, Mucenski M, et al. 27. Crews DC, Scialla JJ, Liu J, Guo H, Bandeen-Roche K, Ephraim PL, et al.
Prognosis for long-term survival and renal recovery in critically ill patients Predialysis health, dialysis timing, and outcomes among older United
with severe acute renal failure: a population-based study. Crit Care. States adults. J Am Soc Nephrol. 2014;25:370–9.
2005;9:R700–9. 28. Gaudry S, Hajage D, Schortgen F, Martin-Lefevre L, Pons B, Boulet E, et al.
7. Uchino S, Kellum JA, Bellomo R, Doig GS, Morimatsu H, Morgera S, et al. Initiation strategies for renal-replacement therapy in the intensive care
Acute renal failure in critically ill patients: a multinational, multicenter unit. N Eng J Med. 2016;375:122–33.
study. JAMA. 2005;294:813–8. 29. Zarbock A, Kellum JA, Schmidt C, Van Aken H, Wempe C, Pavenstädt H,
8. Manns B, Doig CJ, Lee H, Dean S, Tonelli M, Johnson D, et al. Cost et al. Effect of early vs delayed initiation of renal replacement therapy
of acute renal failure requiring dialysis in the intensive care unit: on mortality in critically ill patients with acute kidney injury: The ELAIN
clinical and resource implications of renal recovery. Crit Care Med. Randomized Clinical Trial. JAMA. 2016;315:2190–9.
2003;31:449–55. 30. Moher D, Liberati A, Tetzlaff J, Altman DG, Group P. Reprint–preferred
9. Pannu N, James M, Hemmelgarn B, Klarenbach S. Association between reporting items for systematic reviews and meta-analyses: the PRISMA
AKI, recovery of renal function, and long-term outcomes after hospital statement. Phys Ther. 2009;89:873–80.
discharge. Clin J Am Soc Nephrol. 2013;8:194–202. 31. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency
10. Wu VC, Shiao CC, Chang CH, Huang TM, Lai CF, Lin MC, et al. Long-term in meta-analyses. BMJ. 2003;327:557–60.
outcomes after dialysis-requiring acute kidney injury. Biomed Res Int. 32. Juni P, Witschi A, Bloch R, Egger M. The hazards of scoring the quality of
2014;2014:365186. clinical trials for meta-analysis. JAMA. 1999;282:1054–60.
11. Wu VC, Huang TM, Lai CF, Shiao CC, Lin YF, Chu TS, et al. Acute-on-chronic 33. Sugahara S, Suzuki H. Early start on continuous hemodialysis therapy
kidney injury at hospital discharge is associated with long-term dialysis improves survival rate in patients with acute renal failure following coro‑
and mortality. Kidney Int. 2011;80:1222–30. nary bypass surgery. Hemodial Int. 2004;8:320–5.
12. Rondon-Berrios H, Palevsky PM. Treatment of acute kidney injury: an 34. Koo JR, Yoon JW, Oh JE, Lee YK, Kim SG, Seo JW, et al. Prospective evalua‑
update on the management of renal replacement therapy. Curr Opin tion of early continuous venovenous hemofiltration (CVVH) on the out‑
Nephrol Hypertens. 2007;16:64–70. come in patients with severe sepsis or septic shock. J Am Soc Nephrol.
13. Schneider AG, Bellomo R, Bagshaw SM, Glassford NJ, Lo S, Jun M, et al. 2006;17:50A.
Choice of renal replacement therapy modality and dialysis dependence 35. Combes A, Brechot N, Amour J, Cozic N, Lebreton G, Guidon C, et al. Early
after acute kidney injury: a systematic review and meta-analysis. Intensive high-volume hemofiltration versus standard care for post-cardiac surgery
Care Med. 2013;39:987–97. shock. The HEROICS Study. Am J Respir Crit Care Med. 2015;192:1179–90.
14. Akhoundi A, Singh B, Vela M, Chaudhary S, Monaghan M, Wilson GA, 36. Wald R, Adhikari NK, Smith OM, Weir MA, Pope K, Cohen A, et al. Compari‑
et al. Incidence of adverse events during continuous renal replacement son of standard and accelerated initiation of renal replacement therapy
therapy. Blood Purif. 2015;39:333–9. in acute kidney injury. Kidney Int. 2015;88:897–904.
15. Mehta RL, Pascual MT, Soroko S, Savage BR, Himmelfarb J, Ikizler TA, et al. 37. Jamale TE, Hase NK, Kulkarni M, Pradeep KJ, Keskar V, Jawale S, et al.
Spectrum of acute renal failure in the intensive care unit: the PICARD Earlier-start versus usual-start dialysis in patients with community-
experience. Kidney Int. 2004;66:1613–21. acquired acute kidney injury: a randomized controlled trial. Am J Kidney
16. Wald R, Bagshaw SM. The timing of renal replacement therapy Dis. 2013;62:1116–21.
initiation in acute kidney injury: is earlier truly better? Crit Care Med. 38. Durmaz I, Yagdi T, Calkavur T, Mahmudov R, Apaydin AZ, Posacioglu H,
2014;42:1933–4. et al. Prophylactic dialysis in patients with renal dysfunction undergo‑
17. Gibney N, Hoste E, Burdmann EA, Bunchman T, Kher V, Viswanathan ing on-pump coronary artery bypass surgery. Annals Thorac Surg.
R, et al. Timing of initiation and discontinuation of renal replacement 2003;75:859–64.
therapy in AKI: unanswered key questions. Clin J Am Soc Nephrol. 39. VA/NIH ARFT Network, Palevsky PM, Zhang JH, O’Connor TZ, Chertow
2008;3:876–80. GM, Crowley ST, Choudhury D, et al. Intensity of renal support in critically
18. Shingarev R, Wille K, Tolwani A. Management of complications in renal ill patients with acute kidney injury. N Eng J Med. 2008;359:7–20.
replacement therapy. Semin Dial. 2011;24:164–8. 40. Bagshaw SM, Berthiaume LR, Delaney A, Bellomo R. Continuous versus
19. Carl DE, Grossman C, Behnke M, Sessler CN, Gehr TWB. Effect of timing of intermittent renal replacement therapy for critically ill patients with acute
dialysis on mortality in critically ill, septic patients with acute renal failure. kidney injury: a meta-analysis. Crit Care Med. 2008;36:610–7.
Hemodial Int. 2010;14:11–7. 41. Seabra VF, Balk EM, Liangos O, Sosa MA, Cendoroglo M, Jaber BL. Timing
20. Wu VC, Ko WJ, Chang HW, Chen YS, Chen YW, Chen YM, et al. Early renal of renal replacement therapy initiation in acute renal failure: a meta-
replacement therapy in patients with postoperative acute liver failure analysis. Am J Kidney Dis. 2008;52:272–84.
associated with acute renal failure: effect on postoperative outcomes. J 42. Wang X, Jie Yuan W. Timing of initiation of renal replacement therapy
Am Coll Surg. 2007;205:266–76. in acute kidney injury: a systematic review and meta-analysis. Ren Fail.
21. Elahi MM, Lim MY, Joseph RN, Dhannapuneni RRV, Spyt TJ. Early hemo‑ 2012;34:396–402.
filtration improves survival in post-cardiotomy patients with acute renal 43. Mehta RL, McDonald B, Gabbai FB, Pahl M, Pascual MT, Farkas A, Kaplan
failure. Eur J Cardiothorac Surg. 2004;26:1027–31. RM. Collaborative Group for Treatment of ARFitICU. A randomized clinical
22. Shiao CC, Ko WJ, Wu VC, Huang TM, Lai CF, Lin YF, et al. U-curve asso‑ trial of continuous versus intermittent dialysis for acute renal failure.
ciation between timing of renal replacement therapy initiation and Kidney Int. 2001;60:1154–63.
in-hospital mortality in postoperative acute kidney injury. PLoS ONE. 44. Kashani K, Mehta RL. We restrict CRRT to only the most hemodynamically
2012;7:e42952. unstable patients. Semin Dial. 2016;29:268–71.
23. Shiao CC, Wu VC, Li WY, Lin YF, Hu FC, Young GH, et al. Late initiation of 45. Liu Y, Davari-Farid S, Arora P, Porhomayon J, Nader ND. Early versus late
renal replacement therapy is associated with worse outcomes in acute initiation of renal replacement therapy in critically ill patients with acute
kidney injury after major abdominal surgery. Crit Care. 2009;13:R171. kidney injury after cardiac surgery: a systematic review and meta-analysis.
24. Wierstra BT, Kadri S, Alomar S, Burbano X, Barrisford GW, Kao RLC. The J Cardiothorac Vasc Anesth. 2014;28:557–63.
impact of “early” versus “late” initiation of renal replacement therapy in 46. Lowell JA, Schifferdecker C, Driscoll DF, Benotti PN, Bistrian BR. Postopera‑
critical care patients with acute kidney injury: a systematic review and tive fluid overload: not a benign problem. Crit Care Med. 1990;18:728–33.
evidence synthesis. Crit Care. 2016;20:122. 47. Bagshaw SM, Uchino S, Bellomo R, Morimatsu H, Morgera S, Schetz
25. Xu Y, Gao J, Zheng X, Zhong B, Na Y, Wei J (2016) Timing of initiation of M, et al. Timing of renal replacement therapy and clinical outcomes
renal replacement therapy for acute kidney injury: a systematic review in critically ill patients with severe acute kidney injury. J Crit Care.
and meta-analysis of randomized-controlled trials. Clin Exp Nephrol. 2009;24:129–40.
[Epub ahead of print]. 48. Smak Gregoor PJ, van Saase JL, vd Ingh HF, Weimar W, Kramer P. Dissemi‑
26. Karvellas CJ, Farhat MR, Sajjad I, Mogensen SS, Leung AA, Wald R, et al. A nated histoplasmosis in a haemodialysis patient on immunosuppression
comparison of early versus late initiation of renal replacement therapy in after graft failure. Nephrol Dial Transpl. 1996;11:542–4.
Lai et al. Ann. Intensive Care (2017) 7:38 Page 14 of 14

49. Singbartl K, Kellum JA. AKI in the ICU: definition, epidemiology, risk strati‑ early phase of severe sepsis: a randomized controlled trial. Crit Care Med.
fication, and outcomes. Kidney Int. 2012;81:819–25. 2009;37:803–10.
50. Xu J, Shen B, Fang Y, Liu Z, Zou J, Liu L, et al. Postoperative fluid overload 53. Bouman CSC, Oudemans-Van Straaten HM, Tijssen JGP, Zandstra DF,
is a useful predictor of the short-term outcome of renal replace‑ Kesecioglu J. Effects of early high-volume continuous venovenous
ment therapy for acute kidney injury after cardiac surgery. Medicine. hemofiltration on survival and recovery of renal function in intensive care
2015;94:e1360. patients with acute renal failure: a prospective, randomized trial. Crit Care
51. Bagshaw SM, Uchino S, Bellomo R, Morimatsu H, Morgera S, Schetz M, Med. 2002;30:2205–11.
et al. Septic acute kidney injury in critically ill patients: clinical characteris‑
tics and outcomes. Clin J Am Soc Nephrol. 2007;2:431–9.
52. Payen D, Mateo J, Cavaillon JM, Fraisse F, Floriot C, Vicaut E, et al. Impact
of continuous venovenous hemofiltration on organ failure during the