Tseng et al.

Crit Care (2020) 24:693

https://doi.org/10.1186/s13054-020-03419-y

RESEARCH Open Access

Resuscitation fluid types in sepsis, surgical,

and trauma patients: a systematic review
and sequential network meta-analyses
Chien-Hua Tseng1,2,3,4, Tzu-Tao Chen4, Mei-Yi Wu5,6, Ming-Cheng Chan7,8, Ming-Chieh Shih1
and Yu-Kang Tu1,9,10*

Abstract
Background: Crystalloids and different component colloids, used for volume resuscitation, are sometimes associated
with various adverse effects. Clinical trial findings for such fluid types in different patients’ conditions are conflicting.
Whether the mortality benefit of balanced crystalloid than saline can be inferred from sepsis to other patient group
is uncertain, and adverse effect profile is not comprehensive. This study aims to compare the survival benefits and
adverse effects of seven fluid types with network meta-analysis in sepsis, surgical, trauma, and traumatic brain injury
patients.
Methods: Searched databases (PubMed, EMBASE, and Cochrane CENTRAL) and reference lists of relevant articles
occurred from inception until January 2020. Studies on critically ill adults requiring fluid resuscitation were included.
Intervention studies reported on balanced crystalloid, saline, iso-oncotic albumin, hyperoncotic albumin, low molecu-
lar weight hydroxyethyl starch (L-HES), high molecular weight HES, and gelatin. Network meta-analyses were con-
ducted using random-effects model to calculate odds ratio (OR) and mean difference. Risk of Bias tool 2.0 was used to
assess bias. Confidence in Network Meta-Analysis (CINeMA) web application was used to rate confidence in synthetic
evidence.
Results: Fifty-eight trials (n = 26,351 patients) were identified. Seven fluid types were evaluated. Among patients with
sepsis and surgery, balanced crystalloids and albumin achieved better survival, fewer acute kidney injury, and smaller
blood transfusion volumes than saline and L-HES. In those with sepsis, balanced crystalloids significantly reduced
mortality more than saline (OR 0.84; 95% CI 0.74–0.95) and L-HES (OR 0.81; 95% CI 0.69–0.95) and reduced acute
kidney injury more than L-HES (OR 0.80; 95% CI 0.65–0.99). However, they required the greatest resuscitation volume
among all fluid types, especially in trauma patients. In patients with traumatic brain injury, saline and L-HES achieved
lower mortality than albumin and balanced crystalloids; especially saline was significantly superior to iso-oncotic
albumin (OR 0.55; 95% CI 0.35–0.87).
Conclusions: Our network meta-analysis found that balanced crystalloids and albumin decreased mortality more
than L-HES and saline in sepsis patients; however, saline or L-HES was better than iso-oncotic albumin or balanced
crystalloids in traumatic brain injury patients.
Trial registration: PROSPERO website, registration number: CRD42018115641).

*Correspondence: [email protected]
1
Institute of Epidemiology and Preventive Medicine, National Taiwan
University, Room 539, No. 17, Xu-Zhou Rd., Taipei 10055, Taiwan
Full list of author information is available at the end of the article

© The Author(s) 2020. Open Access This 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://creativeco
mmons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Tseng et al. Crit Care (2020) 24:693 Page 2 of 12

Keywords: Fluid therapy, Intensive care, Resuscitation, Colloids, Crystalloids

Introduction Methods
Fluid resuscitation is one of the most common and Data sources and searches
important methods in managing critically hypotensive We registered our systematic review process on
patients. Crystalloids, mineral salts, or other water-sol- the PROSPERO website [13] (registration number:
uble molecule solutions have been used for more than CRD42018115641). This NMA followed the preferred
100 years for fluid resuscitation [1, 2]. In the past dec- reporting items for systematic reviews and meta-analyses
ades, several colloids, larger insoluble molecular solu- (PRISMA) extension guideline which incorporated NMA
tions, have been developed to improve intravascular for healthcare interventions (Additional file 1: appendix
volume more effectively. However, since the integrity pp. 8–13) [14]. Electronic databases, including PubMed,
of the endothelial glycocalyx layer might be interrupted EMBASE, and Cochrane CENTRAL, were searched from
under inflammatory conditions, such as sepsis, surgery, their inception until January 2020. The search strategies
trauma, or traumatic brain injury, evaluation of the effi- combined terms for patients’ conditions, clinical out-
cacy and safety of colloids in such patients is challeng- comes, and fluid types (Additional file 1: appendix pp.
ing [3, 4]. 14–15).
Insoluble molecules in colloids include starch,
bovine protein (gelatin), and human protein (albumin).
Hydroxyethyl starch (HES) of higher molecular weight Study selection
has a longer half-life in plasma, but it reduces plasma We included randomized controlled trials (RCTs) on crit-
coagulation factors more than HES of lower molecu- ically ill adult patients (more than 18 years old) who pre-
lar weight [5] and albumin [6]. Starch macromolecule sented with systemic hypoperfusion and required fluid
accumulation also impairs glomerular filtration and is resuscitation. We excluded trials on children with dengue
associated with a higher risk of acute renal failure than fever, those on burn injury patients, or those on mixed
gelatin [7]; however, gelatin is associated with a higher populations without reporting subgroup data (Additional
incidence of anaphylactic shock [8, 9]. Compared to file 1: appendix pp. 17–20).
iso-oncotic albumin, hyperoncotic albumin leads to
a higher osmotic pressure, which may alter intraglo-
merular oncotic force and osmotic nephrosis, and is Data extraction and quality assessment
associated with worse kidney damage [10]. Chemical We divided patients requiring fluid resuscitation into the
components, molecular weights, and colloid concentra- following groups for extraction of data and separate anal-
tion might expose the human body to different levels of yses: sepsis, surgical, trauma, and traumatic brain injury.
hazards [11]. Among crystalloids, saline worsens acido- The 7 interventions included 2 crystalloids [balanced
sis and bleeding tendency compared to balanced crys- crystalloids, including lactated Ringer’s, Ringer acetate
talloids [12]. Consequently, classifying resuscitation or PlasmaLytes and saline (0.9% sodium chloride)], and
fluids into either colloids or crystalloids was no longer 5 colloids [iso-oncotic albumin (4%, 5%); hyperoncotic
enough. albumin (20%, or 25%); HES with molecular weight
From 2012 to 2018, of 15 meta-analyses published on ≦ 130 k (L-HES); HES with molecular weight ≧ 200 k
fluid resuscitation in critically ill patients (Additional (H-HES); and gelatin]. The outcomes included:
file 1: appendix pp. 5–7), 12 (80%) grouped high and
low chloride crystalloids or colloids of different com- (1) All-cause mortality rate. If a study reported out-
ponents into a single type of treatment, and 5 (33.3%) comes at multiple time points, we chose the longest
grouped sepsis, surgical, and trauma patients into one observation.
meta-analysis. Furthermore, no meta-analyses com- (2) Fluid resuscitation volume. The resuscitation target
pared the required fluid volumes for the resuscitation is the reversal of organ hypoperfusion.
target. This study aimed to compare the survival benefit (3) Acute kidney injury, referring to the degree of renal
and any potential adverse effects of seven fluid types dysfunction, based on a 5-level scoring system to
using network meta-analysis (NMA) in sepsis, surgical, evaluate risk, injury, failure, loss, and end-stage
trauma, and traumatic brain injury patients, and inves- renal failure (RIFLE).
tigated the trend in treatment difference using sequen- (4) Transfusion volume.
tial NMA. (5) Allergic reaction rate.
Tseng et al. Crit Care (2020) 24:693 Page 3 of 12

Two authors (CH Tseng and TT Chen) screened the functions [22], the two interventions showed no dif-
studies on RCTs independently, extracted data, and ference in their effects. We used the R software pack-
assessed the risk of bias of studies using the revised age “sequentialnma” to undertake sequential NMA [23].
Cochrane risk of bias tool (RoB 2 tool) at study level [15]. Results from these additional analyses were then com-
A third reviewer (YK Tu) was consulted to resolve any pared to the results from the NMA.
disagreement in data extraction or assessment.
Results
Data synthesis and analysis The literature search identified 18,802 citations, and 377
Transitivity assumption was assessed by checking the full-text articles were assessed for eligibility. Of 58 RCTs
distribution of potential confounding variables across which included 26,351 patients in the analysis, 5 large
studies grouped by interventions. The variables examined RCTs included more than one condition—sepsis, surgery,
included age, male percentage, disease severity scores, trauma, and traumatic brain injury. Thus, we extracted
source of sepsis from the lung, and publication year. We the subgroup data of patients with different conditions.
first used the “network” suite of STATA version 14.0 As a result, 23 RCTs on sepsis patients, 24 on surgi-
[16] (StataCorp, Texas, USA) statistical software, which cal patients, 10 on trauma patients, and 4 on traumatic
implements a frequentist approach to the contrast-based brain injury patients were included for further analysis
model meta-analyses [16], to undertake a random-effect (Fig. 1, Additional file 1: appendix pp. 17–48). We pre-
NMA [17]. We then used network map to illustrate the sent the risk of bias assessment for each included study
distribution of the direct and indirect evidence between in Additional file 1: appendix 7 (appendix pp. 61–70);
all treatment comparisons. The size of the nodes in the eFigure 7.1 shows the overall risk of bias in five domains
map was proportional to the number of patients who in sepsis trials, eFigure 7.2 shows the risk of bias for the
received this intervention in the network, and the width individual studies, and eFigure 7.3 explains the reasons
of the edges was proportional to the number of trials that for upgrading or downgrading in every study (Additional
compared the two treatments. Certainty of the evidence file 1: appendix pp. 60–63, 64–66, 67–69). The reasons
was assessed using CINeMA (Confidence in Network to downgrade are mostly inadequate randomization pro-
Meta-Analysis) web application, which allows for confi- cess, open-labeled design, or no detailed information.
dence in the results to be graded as high, moderate, low, No significant differences in baseline variables between
and very low. This approach was based on a methodology interventions were observed within our NMA (Addi-
developed by the Grading of Recommendations Assess- tional file 1: appendix pp. 49–60).
ment, Development and Evaluation Working Group for
pairwise meta-analyses [18]. Sepsis patients
Surface under the cumulative ranking (SUCRA) prob- Most RCTs used the 2001 International Sepsis Defini-
abilities is the ratio of the area under the cumulative tions Conference sepsis definition [24] and included
ranking curve to the entire area in the plot. The more sepsis patients with shock status or those who had
quickly the cumulative ranking curve approaches one, evidence of tissue or organ hypoperfusion (Additional
the closer to unity this ratio is. SUCRA values may be file 1: eTable 5.2, appendix pp. 23–26). The timing
seen as the percentage of safety a treatment achieves in for fluid resuscitation is when the patient meets the
relation to an imaginary treatment that is always the best enrollment criteria: systemic hypoperfusion defined
without any uncertainty [19]. To adjust for the multiplic- by low blood pressure, low central venous pressure or
ity of statistical testing, we further conducted sequential wedge pressure and elevated lactate level. We com-
NMA, proposed by Nikolakopoulor et al., who extended pared the mean arterial pressure among interventions
the rationales of sequential meta-analyses for defining at baseline (Additional file 1: appendix pp. 56), rang-
sample-path, efficacy boundaries, futility boundaries, ing from 69.0 to 73.9 mmHg, and found no statisti-
and information size in meta-analyses [20]. In sequential cally significant differences among seven fluid types.
NMA, we undertook a series of NMA, providing a path Besides, in Additional file 1: eTable 5.4 (appendix pp.
of estimates for each pairwise comparison, by includ- 29–31), and Additional file 1: eTable 5.6 (appendix
ing studies incrementally into the analysis according to pp. 35–36), we compared resuscitation targets among
their publication years [20]. When the path crossed the included trials. The resuscitation goals are generally to
efficacy boundaries, defined by the α-spending function maintain wedge pressure around 15–18 mmHg or cen-
derived from the O’Brien–Fleming method [21], the dif- tral venous pressure around 8–12 mmHg. The average
ference between the two treatments exceeded the thresh- mean study fluid volume was 2397.4 mL ± 1019.1 mL
old for statistical significance. In contrast, when the path in each arm, and the total resuscitation fluid vol-
fell within the futility area defined by the β-spending ume was 7615.6 mL ± 1729.7 mL (Additional file 1:
Tseng et al. Crit Care (2020) 24:693 Page 4 of 12

Fig. 1 Summary of evidence search and selection

appendix pp. 22–31, 61–64). In Additional file 1: eTa- Sepsis patients—mortality

ble 5.3 (appendix pp. 26–27), we presented the base- Between 1983 and 2018, 23 RCTs with 14,659 par-
line characteristics, including age, severity of illness, ticipants presented with usable results on mortality. In
mean arterial pressure, and lactate level. Additional file 1: appendix eTable 5.1 (Additional file 1:
Tseng et al. Crit Care (2020) 24:693 Page 5 of 12

appendix pp. 21), we provided the details of mortal- the comparison between balanced crystalloids and gela-
ity outcome used in our analysis, including in-hospital tin (Fig. 3). According to SUCRA, balanced crystalloid
mortality, 30 day-mortality, and 90-day mortality. If appeared to be the best option; however, saline, L-HES,
multiple time points were reported in a study, we chose and H-HES were not favored (Fig. 4).
the longest observation period for mortality analy-
sis. Balanced crystalloids reduced mortality more than
saline and L-HES with odds ratios (OR) of 0.84 (95% Sepsis patients—fluid resuscitation volume
CI 0.74–0.95) and 0.81 (95% CI 0.69–0.95), respectively Thirteen trials with 10,970 participants reported usable
(Fig. 2a). Sequential NMA further supported the differ- results for fluid resuscitation volume in sepsis patients.
ence in mortality rate between balanced crystalloids ver- Balanced crystalloids and saline required more fluid
sus saline and L-HES by demonstrating that the trend in volume than iso-oncotic albumin with mean differ-
cumulative evidence exceeded the efficacy boundary. The ences (MD) of 2122 mL (95% CI − 300 to 4544 mL) and
cumulative evidence exceeded the futility boundary in 1964 mL (95% CI 89–3840 mL), respectively (Fig. 2b).
the comparison between balanced crystalloids and albu- SUCRA revealed that the colloids were associated with
min, but fell between efficacy and futility boundary in less resuscitation fluid volume than crystalloids (Fig. 4).

Fig. 2 Network geometry and forest plot in sepsis patients with four outcomes. a Mortality, b fluid resuscitation amount, c acute kidney injury, d
transfusion amount. The difference among each comparison is visualized with forest plot, and the effect size and evidence rating are labeled on the
right-hand side. The bold characters are to emphasize significant contrasts. The 95% confidence intervals in the forest plot are clipped to arrows,
when they exceed the limit of x-axis. Abbreviations: OR odds ratio; *p < 0.0.5; **p < 0.01; H high confidence rating, M moderate confident rating,
L low confidence rating, VL very low confidence rating, BC balanced crystalloids, Iso-albumin iso-oncotic albumin, Hyper-albumin hyperoncotic
albumin, L-HES low molecular weight hydroxyethyl starch, H-HES high molecular weight hydroxyethyl starch
Tseng et al. Crit Care (2020) 24:693 Page 6 of 12

Fig. 3 Sequential network meta-analyses (SNMA) over sepsis patient mortality analysis among a balanced crystalloids versus saline, b balanced
crystalloids versus low molecular weight hydroxyethyl starch (L-HES), c balanced crystalloids versus albumin, and d balanced crystalloids versus
gelatin. Y-axis represent the z scores for effect sizes, and green dots (trials) and green line along the X-axis show the trend of cumulating evidence
toward achieving maximal information. The blue line represents the SNMA efficacy boundary, and orange line represents the futility boundary.
The green dots and green line start in the middle; when they pass the blue line, this indicates that a significant difference in the outcome between
the two treatments has been attained. When they pass the orange line, this suggests no difference in the outcome between the two treatments. I
iso-oncotic albumin, H hyperoncotic albumin

Sepsis patients—acute kidney injury evidence certainty in mortality revealed a moderate-to-

Eleven trials with 10,569 participants reported usable high evidence confidence in comparison, including bal-
results for acute kidney injury. Balanced crystalloids sig- anced crystalloids, saline, and L-HES; low-to-moderate in
nificantly reduced a greater risk of acute kidney injury iso-oncotic albumin and hyperoncotic albumin; and very
than L-HES (OR, 0.80; 95% CI 0.65–0.99) and H-HES low in gelatin and H-HES (Additional file 1: appendix pp.
(OR, 0.54; 95% CI 0.37–0.84) (Fig. 2c). SUCRA ranking 139–142). Results of sensitivity analyses with the exclu-
revealed that gelatin, balanced crystalloid, saline, and iso- sion of the largest SMART trials [12] or the inclusion of
oncotic albumin had a lower risk of acute kidney injury the pilot SALT trial [25] in Additional file 1: appendix 14
than L-HES and H-HES (Fig. 4). show no substantial differences from the main analysis.

Sepsis patients—red blood cell transfusion volume Surgical patients

Ten trials with 11,979 participants reported usable From 1979 to 2020, 8 (34.80%), 6 (26.00%), 6 (26.00%),
results for the packed red blood cell transfusion vol- and 3RCTs compared different resuscitation fluids in
ume. Balanced crystalloids required less volume of red patients receiving cardiac surgery, aortic surgery, major
blood cell transfusion than hyperoncotic albumin (MD, abdominal surgery, and hip arthroplasty and cystectomy,
274 mL; 95% CI 5–548 mL), L-HES (MD, 232 mL; 95% respectively (Additional file 1: appendix pp. 32–36). Fluid
CI 35–430 mL), and H-HES (MD, 497 mL; 95% CI 141– resuscitation was provided during surgical procedures to
854 mL). (Fig. 2d). SUCRA revealed that the crystalloids maintain hemodynamic parameters in most trials, and
and iso-oncotic albumin were associated with less trans- the mean resuscitated fluid of interest was 3327.5 mL
fusion volume than other colloids (Fig. 4). (Additional file 1: appendix 65–67).
The funnel plot and Egger’s test did not detect any sig-
nificant publication bias (Additional file 1: appendix pp. Surgical patients—mortality
114–116). Loop inconsistency and design inconsistency Twenty-three trials with 4646 participants had valid
were also not detected (Additional file 1: appendix pp. results on mortality. There were no significant differ-
124–129). The meta-regression did not change the rank- ences in mortality between 7 interventions (Fig. 5);
ing order (Additional file 1: appendix pp. 138–139). The SUCRA showed that hyperoncotic albumin and balanced
Tseng et al. Crit Care (2020) 24:693 Page 7 of 12

Fig. 4 Surface under the cumulative ranking curve area (SUCRA) for mortality, fluid resuscitation volume, acute kidney injury, and blood transfusion
volume among sepsis, surgical, trauma, and traumatic brain injury patients. Dark color bar represents significantly better or worse interventions, and
the differences between fluid types are shown above the bars

crystalloid were associated with less mortality than gela- were associated with less acute kidney injury than HES
tin, HES, and saline (Fig. 4). and gelatin.

Surgical patients—fluid resuscitation volume

Surgical patients—red blood cell transfusion volume
Twenty trials with 4512 participants provided data on
Sixteen trials with 2818 participants presented usable
resuscitation fluid volume. Balanced crystalloids group
results for red blood cell transfusion volume. Rank-
required significantly more fluid resuscitation vol-
ing probabilities showed that albumin, L-HES, and then
ume than iso-tonic albumin (MD, 2612 mL; 95% CI
gelatin were associated with less transfusion volume than
1416–3800), hypertonic albumin (MD, 2852 mL; 95%
H-HES and crystalloids (Fig. 5).
CI 742–4962), L-HES (MD 1494 mL; 95% CI 345–2644),
Publication bias and inconsistency were not significant
H-HES (MD, 1462 mL; 95% CI 418–2505), and gelatin
(Additional file 1: appendix pp. 118–121). The confidence
(MD, 1154 mL; 95% CI 64–2240) (Fig. 5). SUCRA rank-
ratings were low to very low among all comparisons in
ing showed that colloids (albumin, HES, and then gelatin)
surgical trials (Additional file 1: appendix pp. 143–146).
were associated with less fluid resuscitation volume than
crystalloids (Fig. 4).
Trauma and traumatic brain injury patients
Surgical patients—acute kidney injury From 1977 to 2018, 10 RCTs compared different resus-
Fourteen trials with 4248 participants reported results citation fluids in trauma patients who required fluid
for acute kidney injury. The ORs between seven treat- resuscitations, and 4 RCTs in traumatic brain injury
ments were not statistically significant (Fig. 5). SUCRA patients. (Additional file 1: appendix pp. 37–39).
showed iso-oncotic albumin, and balanced crystalloids Patients’ mean age was 48.6 years, predominantly
Tseng et al. Crit Care (2020) 24:693 Page 8 of 12

Fig. 5 Network geometry and forest plot in surgical patients with four outcomes. a Mortality, b fluid resuscitation amount, c acute kidney injury,
d transfusion amount. The difference among each comparison is visualized with forest plot, and the effect size and evidence rating are labeled on
the right-hand side. The bold characters are to emphasize significant contrasts. The 95% confidence intervals in the forest plot are clipped to arrows,
when they exceed the limit of x-axis. OR odds ratio; *p < 0.05; **p < 0.01; H high confidence rating, M moderate confident rating, L low confidence
rating, VL very low confidence rating, BC balanced crystalloids, Iso-albumin iso-oncotic albumin, Hyper-albumin hyperoncotic albumin, L-HES low
molecular weight hydroxyethyl starch, H-HES high molecular weight hydroxyethyl starch

male (69.8%), and mean resuscitation study fluid was Discussion

5481 mL among trauma trials. (Additional file 1: appen- To our knowledge, this analysis is the largest NMA
dix pp. 82–86). in the field of fluid resuscitation, as we considered a
Ten trials with 5076 participants had valid results on larger number of outcomes and undertook separate
mortality in trauma patients, and differences in mortal- analyses for patients with different conditions. In sepsis
ity were not significant between interventions in trauma patients, balanced crystalloids and iso-oncotic albumin
patients. Balanced crystalloid required less volume of were associated with lower mortality rates, lower risks
red blood cell transfusion than saline (MD, 350 mL; 95% of acute kidney injury, and less red blood cell transfu-
CI 160 mL to 540 mL) and L-HES (MD, 964 mL; 95% CI sion volume. In surgical patients, nonsignificant differ-
400 mL to 1527 mL). Four trials with 1970 participants ences in the risks of mortality and acute kidney injury
had valid results on mortality in traumatic brain injury were observed between the seven interventions, but
patients, and saline reduced mortality than albumin with balanced crystalloids required the greatest volume of
OR of 0.55 (95% CI 0.35–0.87) (Additional file 1: appen- fluid resuscitation among all fluid types. In traumatic
dix pp. 103–114). The confidence ratings were low to brain injury trials, iso-oncotic albumin was associated
very low among all comparisons in traumatic and trau- with higher mortality than saline.
matic brain injury trials (Additional file 1: appendix pp.
123–124, 128, 148–150).
Tseng et al. Crit Care (2020) 24:693 Page 9 of 12

Previous studies and important differences from this study patients, and greater fluid volume and hypotonic solu-
In many previous meta-analyses on fluid resuscita- tion may further raise intracranial pressure, leading to a
tion, sepsis, surgical, trauma, and traumatic brain injury higher mortality [30].
patients were put together as a single group. In 2013,
Perel et al. published a meta-analysis [26] of critically Hydroxyethyl starch (HES): L-HES and H-HES
ill patients of all causes, and another meta-analysis on HES of higher molecular weight has been retracted from
HES [27], including patients with different causes being the market, but the HES of lower molecular weight is still
grouped together. Our analyses separated patients’ con- in use in daily practice, especially in surgical or trauma
ditions, thereby providing more precise information patients. However, this study found that L-HES was asso-
applied to specific subgroups of patients. Furthermore, ciated with the highest mortality rate in sepsis, surgical,
previous meta-analyses also combined different fluid and trauma patients, and a greater risk of acute kidney
types into a single treatment. Our network meta-analysis injury and greater transfusion volume was required dur-
used a more comprehensive classification of resuscitation ing the resuscitation period. However, for traumatic
fluids according to the current knowledge, yielding more brain injury patients, L-HES and saline, both hyper-
clinically meaningful information. tonic solutions, were associated with better survival than
hypotonic solution, including iso-oncotic albumin and
Crystalloids: balanced crystalloids and saline balanced crystalloid.
Several meta-analyses and current sepsis guideline rec-
ommended that crystalloids are the fluid of choice for Gelatin
resuscitation [26, 28]. The present study found that Many review articles are opposed to gelatin use for fluid
among crystalloids, balanced crystalloids show better resuscitation due to the risk of anaphylaxis and acute
survival benefit and renal function for sepsis and surgi- kidney injury, but those opinions were based on animal
cal patients than saline does, and the reverse was found studies, case series, or RCTs designed for other purposes
in traumatic brain injury patients. Instead of considering [11, 31, 32]. Recent large RCTs reveal conflicting results,
crystalloids as one treatment group, we could be more in that gelatin is associated with a nonsignificant, lower
specific in considering balanced crystalloids for sep- mortality than balanced crystalloids and saline3. Our
sis and surgical patients, and saline for traumatic brain sequential NMA demonstrated that the z-score trend for
injury patients. However, both crystalloids required a the difference between balanced crystalloids and gelatin
higher volume to achieve resuscitation goals. Therefore, has not yet exceeded the efficacy or futility boundary,
in addition to evaluate fluid responsiveness with passive indicating that the evidence was still insufficient (Fig. 2).
leg raising or other static tests continuously, choosing
optimal fluid types could also prevent fluid overload. [29]. Strengths and limitations
The present NMA analyzed all outcomes from previous
Albumin: iso-oncotic and hyperoncotic albumin RCTs, especially on the fluid resuscitation volume, which
The osmotic pressure in iso-oncotic solution was simi- has never been considered in previous meta-analyses.
lar to plasma, and hyperoncotic solution was higher This study also analyzed seven fluid types and patients’
than plasma. Iso-oncotic albumin was designed for fluid conditions separately and demonstrated that the benefit
resuscitation and has volume-sparing effect; hyperon- or harmful effects of the fluid types were largely depend-
cotic albumin was used to maintain target serum albu- ent on patients’ conditions. We present results from
min concentration, which helps to maintain effective NMA followed by those from sequential NMA, in which
volume by recruiting endogenous fluid11. This study the dynamic updates of the effect size help to corrobo-
found that iso-oncotic albumin was associated with bet- rate the NMA results and estimate evidence uncertainty
ter survival benefit in sepsis patients who suffer hypov- by depicting the trend and making allowance for mul-
olemia due to extravascular fluid loss caused by increased tiple testing. Our NMA also has some limitations: first,
vascular permeability. However, hyperoncotic albumin in sepsis trials (sample size [n] = 14,659), the evidence
achieved better survival possibilities in surgical patients, was adequate between balanced crystalloids and saline,
whose blood loss was caused by uncorrected blood loss. L-HES, and albumin, but insufficient between balanced
This indicated that iso-oncotic albumin helps with pro- crystalloids and gelatin. The confidence rating was low
viding more volume for sepsis resuscitation, while hyper- in surgical (n = 3871) and traumatic trials (n = 5076)
oncotic albumin is more beneficial for uncorrected blood because the sample size was insufficient and confidence
loss patients with normal vascular permeability. Besides, intervals were wide. The confidence rating was very low
iso-oncotic albumin in hypotonic solution was associ- for traumatic brain injury trials (n = 1970) because the
ated with higher mortality rate in traumatic brain injury direct and indirect evidence was inconsistent and sample
Tseng et al. Crit Care (2020) 24:693 Page 10 of 12

sizes were insufficient. Secondly, the benefit or harm of physician decision in the included trials. In Additional
gelatin could not be determined from the current evi- file 1: appendix 8.1.5 (appendix pp. 74–75), we listed the
dence. Acute kidney injury was ranked best for gelatin in blood transfusion volume and number of bleeding events
sepsis patients (only one trial) but was worse in surgical requiring transfusion.Finally, the amount of investiga-
patients (only two trials). Survival benefit was also incon- tion fluid was often very limited in many trials, and large
sistent between sepsis and surgical patients (Table 1). As volumes of these resuscitation fluids have not been well
very few trials included gelatin, the evidence on gelatin investigated. Some undetected adverse events may occur
should be interpreted with cautions. Third, blood trans- if larger volumes are used.
fusion thresholds are unclear and largely dependent on

Table 1 Characteristics of the fluids assessed and qualitative summary from this network meta-analysis
Components Plasma Balanced Saline Albumin (Iso-/ L-HES Gelatin
crystalloid Hyperoncotic)

Osmolarity (mOsm/ 291 Hypotonic (254–273) Isotonic (286) Hypotonic (4%, 260; Isotonic to Hyper- Isotonic to Hypertonic
kg) 5%, 250; 20%, 200; tonic (283–304) (274–301)
25%, 250)
Na/Cl (mmol/l) 140/103 130–140/98–111 154/154 130–160/128–130 137–154/110–154 145–154/120–145
K/Ca (mmol/l) 40/4 4–5/2–2.7 0/0 < 2/0 0–4/0–2.5 0–5.1/0–6.25
Conditions Outcome Balanced Saline Albumin L-HES Gelatin
crystalloid

Sepsis NMA results Lowest mortality Higher mortality Lower mortality (Iso- Highest mortality
Lowest acute kidney More fluid volume oncotic) More acute kidney
injury required Least fluid volume injury
Lowest transfusion required More transfusion
volume volume
More fluid volume
required
Comments Fluid of choice for Not favored for sepsis Iso-oncotic albumin Not favored for Require further trials
sepsis for sepsis patients sepsis
with risk of fluid
overload
Surgery NMA results Most fluid volume More fluid volume Lower mortality Highest mortality Less fluid volume
required required (Hyper-oncotic) Less fluid volume required
Lower acute kidney More blood transfu- Less fluid volume required
injury sion volume required
Less acute kidney
injury
Less blood transfu-
sion volume
Comments More favored for Less favored for Favored for surgery Not favored for Require further trials
surgery surgery surgery
Trauma Mortality Lower mortality Lower mortality Higher mortality Higher mortality
Less acute kidney More acute kidney Less acute kidney
injury injury injury
Less transfusion More transfusion Less transfusion
volume volume volume
More fluid volume
required
Comments More favored for Damage control
trauma resuscitation. May
consider blood
products for resus-
citation
Traumatic brain Mortality Higher mortality Lower mortality Highest mortality Lowest mortality
injury (TBI) (Iso-oncotic)
Comments Hypotonic solution Favored for TBI Iso-oncotic albumin May consider for TBI Require further trials
was not suggested with hypotonic
for TBI solution was not
favored for TBI
Tseng et al. Crit Care (2020) 24:693 Page 11 of 12

Conclusions Received: 14 August 2020 Accepted: 30 November 2020

Among sepsis and surgical patients, balanced crystal-
loids and albumin attained lower mortality rates, lower
risks of acute kidney injury, and less red blood transfu-
References
sion volume than did saline and L-HES. Balanced crys-
1. Lee JA. Sydney Ringer (1834–1910) and Alexis Hartmann (1898–1964).
talloids required the greatest fluid resuscitation volume Anaesth. 1981;36(12):1115–21.
than all the other fluid types. In traumatic brain injury 2. Awad S, Allison SP, Lobo DN. The history of 0.9% saline. Clin Nutr.
2008;27(2):179–88.
patients, saline and L-HES showed better mortality
3. Annane D, Siami S, Jaber S, et al. Effects of fluid resuscitation with colloids
rates than hypotonic solutions, including iso-oncotic vs crystalloids on mortality in critically ill patients presenting with hypov-
albumin and balanced crystalloids. olemic shock: the CRISTAL randomized trial. JAMA. 2013;310(17):1809–17.
4. Myburgh JA, Mythen MG. Resuscitation fluids. N Engl J Med.
2013;369(13):1243–51.
Supplementary Information 5. Treib J, Haass A, Pindur G. Coagulation disorders caused by hydroxyethyl
The online version contains supplementary material available at https://doi. starch. Thromb Haemost. 1997;78(3):974–83.
org/10.1186/s13054-020-03419-y. 6. Wilkes MM, Navickis RJ, Sibbald WJ. Albumin versus hydroxyethyl starch in
cardiopulmonary bypass surgery: a meta-analysis of postoperative bleed-
ing. Ann Thorac Surg. 2001;72(2):527–33 discussion 534.
Additional file 1. Appendix. 7. Schortgen F, Lacherade JC, Bruneel F, et al. Effects of hydroxyethylstarch
and gelatin on renal function in severe sepsis: a multicenter randomized
study. Lancet. 2001;357(9260):911–6.
Acknowledgements 8. Farooque S, Kenny M, Marshall SD. Anaphylaxis to intravenous gelatin-
Not applicable. based solutions: a case series examining clinical features and severity.
Anaesth. 2019;74(2):174–9.
Authors’ contributions 9. Ring J, Messmer K. Incidence and severity of anaphylactoid reactions to
CT and TT performed systemic reviews. CT, MY, and MC analyzed data and did colloid volume substitutes. Lancet. 1977;1(8009):466–9.
the sequential network meta-analysis. TT, MY, and MC provided clinical aspects 10. Mendes RS, Oliveira MV, Padilha GA, et al. Effects of crystalloid, hyper-
implications. CH and YK were major contributors in writing the manuscript. All oncotic albumin, and iso-oncotic albumin on lung and kidney damage in
authors read and approved the final manuscript. experimental acute lung injury. Respir Res. 2019;20(1):155.
11. Finfer S, Myburgh J, Bellomo R. Intravenous fluid therapy in critically ill
Funding adults. Nat Rev Nephrol. 2018;14(9):541–57.
This work was partly funded by grants from the Ministry of Science and 12. Matthew WS, Wesley HS, Jonathan PW, et al. Balanced crystalloids versus
Technology in Taiwan (grant no. MOST 106-2314-B-002-098-MY3 & MOST saline in critically ill adults. N Engl J Med. 2018;378(9):829–39.
109-2314-B-002-150-MY3). 13. PROSPERO register. https://www.crdyorkacuk/prospero/.
14. Krajewski ML, Raghunathan K, Paluszkiewicz SM, Schermer CR, Shaw AD.
Availability of data and materials Meta-analysis of high- versus low-chloride content in perioperative and
All data generated or analyzed during this study are included in this published critical care fluid resuscitation. Br J Surg. 2015;102(1):24–36.
article and its supplementary files. 15. Higgins JPT, Sterne JAC, Savović J, et al. A revised tool for assessing risk of
bias in randomized trials. In: Chandler J, McKenzie J, Boutron I, Welch V,
Ethics approval and consent to participate editors. Cochrane methods. Cochrane Database Syst Rev. 2016;10 (Suppl
Not applicable. 1).
16. White IR. Network meta-analysis. Stata J. 2015;15(4):951–85.
Consent for publication 17. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials.
Not applicable. 1986;7(3):177–88.
18. Salanti G, Del Giovane C, Chaimani A, Caldwell DM, Higgins JP. Evaluat-
Competing interests ing the quality of evidence from a network meta-analysis. PLoS ONE.
The authors declare that they have no competing interests. 2014;9(7):e99682.
19. Salanti G, Ades AE, Ioannidis JP. Graphical methods and numerical sum-
Author details maries for presenting results from multiple-treatment meta-analysis: an
1
Institute of Epidemiology and Preventive Medicine, National Taiwan Univer- overview and tutorial. J Clin Epidemiol. 2011;64(2):163–71.
sity, Room 539, No. 17, Xu-Zhou Rd., Taipei 10055, Taiwan. 2 Division of Pulmo- 20. Nikolakopoulou A, Mavridis D, Egger M, Salanti G. Continuously updated
nary Medicine, Department of Internal Medicine, School of Medicine, College network meta-analysis and statistical monitoring for timely decision-
of Medicine, Taipei Medical University, Taipei, Taiwan. 3 Division of Critical Care making. Stat Methods Med Res. 2018;27(5):1312–30.
Medicine, Department of Emergency and Critical Care Medicine, Shuang Ho 21. Lan KKG, DeMets DL. Discrete sequential boundaries for clinical trials.
Hospital, Taipei Medical University, New Taipei City, Taiwan. 4 Division of Pul- Biometrika. 1983;70(3):659–63.
monary Medicine, Department of Internal Medicine, Shuang Ho Hospital, 22. Lachin JM. A review of methods for futility stopping based on conditional
Taipei Medical University, New Taipei City, Taiwan. 5 Division of Nephrology, power. Stat Med. 2005;24(18):2747–64.
Department of Internal Medicine, Shuang Ho Hospital, Taipei Medical Univer- 23. Tsai S-J, Ding Y-W, Shih M-C, et al. Systematic review and sequential net-
sity, New Taipei City, Taiwan. 6 Division of Nephrology, Department of Internal work meta-analysis on the efficacy of periodontal regenerative therapies.
Medicine, School of Medicine, College of Medicine, Taipei Medical University, J Clin Periodontol. 2020. https://doi.org/10.1111/jcpe.13338.
Taipei, Taiwan. 7 Division of Critical Care and Respiratory Therapy, Department 24. Levy MM, Fink MP, Marshall JC, et al. 2001 SCCM/ESICM/ACCP/
of Internal Medicine, Taichung Veterans General Hospital, Taichung, Taiwan. ATS/SIS international sepsis definitions conference. Crit Care Med.
8
College of Science, Tunghai University, Taichung, Taiwan. 9 Department 2003;31(4):1250–6.
of Dentistry, National Taiwan University Hospital, National Taiwan University, 25. Semler MW, Wanderer JP, Ehrenfeld JM, et al. Balanced crystalloids versus
Taipei, Taiwan. 10 Research Center of Big Data and Meta-Analysis, Wan Fang saline in the intensive care unit. The SALT randomized trial. Am J Respir
Hospital, Taipei Medical University, Taipei, Taiwan. Crit Care Med. 2017;195(10):1362–72.
26. Perel P, Roberts I, Ker K. Colloids versus crystalloids for fluid resuscitation
in critically ill patients. Cochrane Database Syst Rev. 2013(2):Cd000567.
Tseng et al. Crit Care (2020) 24:693 Page 12 of 12

27. Nenna G, Costantino G. 0.9 % Saline vs 6 % HES 130/0.4 for fluid resuscita- 31. Moeller C, Fleischmann C, Thomas-Rueddel D, et al. How safe is gelatin?
tion in critically ill patients. Intern Emerg Med. 2013;8(4):347–8. A systematic review and meta-analysis of gelatin-containing plasma
28. Zarychanski R, Abou-Setta AM, Turgeon AF, et al. Association of hydroxy- expanders vs crystalloids and albumin. J Crit Care. 2016;35:75–83.
ethyl starch administration with mortality and acute kidney injury in 32. Lorenz W, Duda D, Dick W, et al. Incidence and clinical importance of
critically ill patients requiring volume resuscitation: a systematic review perioperative histamine release: randomized study of volume loading
and meta-analysis. JAMA. 2013;309(7):678–88. and antihistamines after induction of anesthesia. Trial Group Mainz/Mar-
29. Bentzer P, Griesdale DE, Boyd J, MacLean K, Sirounis D, Ayas NT. Will this burg Lancet. 1994;343(8903):933–40.
hemodynamically unstable patient respond to a bolus of intravenous
fluids? JAMA. 2016;316(12):1298–309.
30. Cooper DJ, Myburgh J, Heritier S, et al. Albumin resuscitation for trau- Publisher’s Note
matic brain injury: is intracranial hypertension the cause of increased Springer Nature remains neutral with regard to jurisdictional claims in pub-
mortality? J Neurotrauma. 2013;30(7):512–8. lished maps and institutional affiliations.

Ready to submit your research ? Choose BMC and benefit from:

• fast, convenient online submission

• thorough peer review by experienced researchers in your field
• rapid publication on acceptance
• support for research data, including large and complex data types
• gold Open Access which fosters wider collaboration and increased citations
• maximum visibility for your research: over 100M website views per year

At BMC, research is always in progress.

Learn more biomedcentral.com/submissions