Bioimpedance-Guided Fluid Removal in Continuous KRT

Visual Abstract

Abstract

Key Points

• This study, the sole randomized trial of its kind, proposes guidelines for fluid balance management in continuous KRT (CKRT) patients using bioimpedance.

• Despite this, bioimpedance analysis–guided volume management did not influence the proportion of patients achieving estimated euvolemia at 7 days into CKRT.

• Further investigation is needed to assess whether bioimpedance analysis guidance can facilitate rapid fluid removal in the early phase of CKRT for patients with AKI.

Background

Ultrafiltration with continuous KRT (CKRT) can be used to manage fluid balance in critically ill patients with AKI. We aimed to assess whether bioimpedance analysis (BIA)–guided volume management was more efficacious than conventional management for achieving estimated euvolemia (e-euvolemia) in CKRT-treated patients.

Methods

In a multicenter randomized controlled trial from July 2017 to July 2020, the patients with AKI requiring CKRT were eligible if the weight at the start of CKRT had increased by ≥5% compared with the weight at the time of admission or total body water (TBW)/height (H)² ≥13 L/m². We randomly assigned 208 patients to the control (conventional fluid management; n=103) and intervention groups (BIA-guided fluid management; n=105). The primary outcome was the proportion of attaining e-euvolemia 7 days postrandomization. E-euvolemia was defined as the difference between TBW/H² D₇ and D₀ was <−2.1 L/m², or when TBW/H² measured on D₇ was <13 L/m². The 28-, 60-, and 90-day mortality rate were secondary outcomes.

Results

The primary outcome occurred in 34 patients in the intervention group and 27 in the control group (47% versus 41%; P = 0.50). The mean value of TBW/H² measured on D₇ was the same at 13.9 L/m² in both groups. The differences between TBW/H² D₇ and D₀ were −1.13 L/m² in the intervention group and −1.08 L/m² in the control group (P = 0.84). Patients in the intervention group had a significantly higher proportion of reaching e-euvolemia on D₁ than those in the control group (13% versus 4%, P = 0.02). Adverse events did not differ significantly between the groups.

Conclusions

BIA-guided volume management did not affect the proportion of reaching the e-euvolemia at 7 days of the start of CKRT.

Clinical Trial registry name and registration number:

ClinicalTrials.gov, ID: NCT03330626 (Registered on November 6, 2017; seven study participants were retrospectively registered; nonetheless, Institutional Review Board approval of each institution was completed before study participant registration).

Introduction

Although there are advanced techniques for the management of AKI,^1,2 mortality in patients with AKI remains high (40%–50%)^3–6 and is considered to be significantly associated with fluid overload.^7–9 The positive cumulative amount of fluid—one of the variables used in the calculation of fluid overload—is calculated as the difference between the total amount of fluid intake and output, divided by body weight. This method is acceptable, but not exact when detailed fluid monitoring is required as it only reflects total body water (TBW) instead of individual fluid overload in each body compartment.¹⁰

Bioimpedance analysis (BIA) is used to measure nutritional entities of body composition in diverse conditions^11,12 and evaluate body volume status.¹³ In addition, several studies have shown the accuracy and clinical utility of BIA during fluid status assessment among patients on chronic hemodialysis or peritoneal dialysis.^14–18 However, few recent studies have shown the usefulness of BIA in assessing the volume status of critically ill patients.^19,20

Continuous KRT (CKRT) has been established to maintain fluid balance in critically ill patients with AKI.^21–24 However, there are no objective guidelines for fluid removal despite the emphasis on the role of fluid overload in higher mortality in CKRT-treated patients.^7–9

Thus, we performed a multicenter, prospective, randomized controlled trial to assess whether BIA-guided volume management was more suitable for achieving estimated euvolemia (e-euvolemia) 7 days after CKRT initiation in comparison with the conventional method. We also compared the mortality rates between the two approaches among patients with AKI undergoing CKRT.

Methods

Study Design and Participants

The Volume Management under Body Composition Monitoring in Critically Ill Patients on CKRT trial was a prospective, investigator-initiated, multicenter, parallel-group, open-label, randomized controlled study with two fluid management groups, performed in eight tertiary hospitals in Korea from July 2017 to July 2020.²⁵ This study was approved by the institutional review board of each institution, and written informed consent was obtained from all study participants or their legal representatives (Supplemental Methods). The study followed the Consolidated Standards of Reporting Trials reporting guidelines.

CKRT (continuous veno-venous hemodiafiltration) was considered when the patients had at least one of the following: urine output <100 ml/6 hours unresponsive to fluid resuscitation and diuretics, medically refractory hyperkalemia (serum potassium [K+] >6.5 mmol/L), medically refractory metabolic acidosis (pH <7.2), urea >70 mg/dl, or clinically significant organ edema in the setting of AKI. Among patients initiating CKRT, the inclusion criteria were (1) CKRT expected to be provided for at least 72 hours, (2) provision of informed consent, (3) body mass index between 18 and 35 kg/m², and (4) fluid overload of 5% (had by ≥5% weight increase at the start of CKRT compared with that at admission) or TBW/height (H)² ≥13 L/m².

The exclusion criteria were (1) <18 years of age; (2) imminent death (<24 hours); (3) participants requiring active resuscitation or rescue at enrollment; (4) bearing devices (e.g., a pacemaker) that could affect the results of BIA; (5) cases where BIA measurement cannot be performed, such as in patients with limb amputation; and (6) any other major illness that would substantially increase the risk associated with study participation.

As randomization procedures, an independent statistician generated the random allocation sequences. Randomization was stratified by the study institutions using permuted blocks (block sizes 2, 4, and 6) and conducted using the web-based Research Electronic Data Capture program (Vanderbilt University, Nashville, TN). The study researchers screened and enrolled the participants, and study coordinator randomly assigned eligible participants 1:1 to either the intervention (BIA-guided fluid management) or control group (conventional fluid management) (Figure 1). After study enrollment, the randomization results (group assignment) were disclosed to the researchers and attending physicians.

Figure 1.

Diagram of participant enrollment process. For protocol violations, measurement date violations were the most common, if measurement is not done in D₁ or D₂.

InBody S10 (InBody, Seoul, Korea) was used to measure fluid status at postrandomization days 0, 1, 2, and 7 in both groups. Electrocardiogram leads that may have affected measurement results were removed during measurement. Another independent investigator measured and recorded the InBody S10 measurements; the participants, other researchers, or attending physicians were blinded to the results during the study period. The InBody S10 machines used in this study underwent regular quality control.

Sample Size Calculation

No previous study has evaluated the effect of BIA-guided fluid management on CKRT-treated patients. Among the recent studies on fluid balance, there has been one retrospective study investigating BIA in critically ill patients. In that study, it was estimated that the number of patients who reached euvolemia was approximately 25%, and therefore, it was assumed that the proportion reaching euvolemia in the control group was about 25%. We randomly and equally divided the enrolled participants into two groups. Assuming a 19% difference (control group 25% versus intervention group 44%) in the proportion of achieving euvolemic status between the groups with 0.8 power, two-sided, and 0.05 of α, 97 patients had to be allocated to each group. Estimating a dropout rate of 20%, each group needed 122 patients. When a total of 122 study patients were registered, an interim analysis was performed.

Fluid Removal Targets

The primary outcome was to compare the proportion of achieving e-euvolemia 7 days after study initiation. In a pilot study preceding the main research, 50 CKRT patients were examined across four hospitals, with their TBW/H² ratio measured using BIA. Among surviving patients (n=37), a consistent decrease in TBW/H² during CKRT treatment was observed. Analysis of data from 28 patients who had multiple BIA measurements revealed no correlation between initial TBW/H² values and daily fluid removal rates. On stratification by initial TBW/H², an average of 0.4 L/m² per day was withdrawn, totaling 1.5 L/m² over CKRT. Referencing the study by Garzotto et al.,⁹ their findings underscored that even in critically ill patients devoid of AKI, there existed a volumetric overload of 2.8% vis-à-vis body weight, indicative of a state diverging from true euvolemia. Consequently, it was deduced that to achieve genuine euvolemia, an amount approximating this volume would necessitate supplementary extraction. Using a baseline adult weight of 70 kg and height of 175 cm, the required additional removal was estimated at 2.1 L/m² or more, considering the withdrawn 1.5 L/m² and an extra 0.6 L/m².

In addition, TBW/H² below 13 was included as an outcome based on the study by Rhee et al.,¹⁰ which associated lower mortality rates and better prognostic outcomes with TBW/H² in the first percentile (approximately 13). Consequently, achieving TBW/H² below 13 was designated as the primary outcome. Comprehensively, we defined the primary outcome, estimated euvolemic status, as follows: (1) TBW/H² measured at D₇ is < 13 L/m² or (2) (D₇–D₀) TBW/H² is less than −2.1 L/m². We considered either condition (1) or (2) to be met for defining the primary outcome.

Guidelines for Volume Management (Intervention)

To the intervention group, the fluid removal amount calculated according to the fluid removal guide was recommended (Supplemental Table 1). On the day randomization was implemented (D₀) and the next day (D₁), the recommended amount of fluid removal was presented as follows: The recommended amount of fluid removal for this patient is total input/output (I/O)−() L/d. On D₂, the amount of fluid to be removed over the next 5 days to achieve e-euvolemia was recommended.

The formula for calculating fluid removal for D₂–D₇ was defined as follows: (1) TBW/H² measured at D₂−13.0; (2) the value obtained by subtracting (TBW/H² measured at D₀−2.1) from TBW/H² measured at D₂; (3) the smaller of the above two cases was divided and removed from D₂ to D₇. If on D₁ or D₂, more than the TBW/H² target value was removed, the target amount was reduced by as much as the removed amount from the recommended removal amount up to D₂ or D₇.

The amount of fluid removed was recommended based on the details presented in Supplemental Table 1: (1) If target was not achieved on D₁ and if there was a reasonable hypotensive event, no additional fluid removal was done on D₂; (2) if the target was not achieved on D₁, additional fluid removal was made on D₂ if there was no reasonable hypotensive events; (3) if more than the target value was achieved in D₁, the target amount was reduced by the amount that was additionally omitted from the recommended removal amount in D₂; (4) if the end point had already been reached at D₁ and if there was evidence of excess body fluid in the patient, additional fluid removal was performed based on −0.25 L/m²; and (5) if the end point had already been reached at D₂ and if there was evidence of excess fluid volume in the patient, additional fluid removal was performed under the subjective judgment of the clinician. If on D₁ or D₂, more than the TBW/H² target value was removed, the target amount was reduced by as much as the removed amount from the recommended removal amount up to D₂ or D₇. Supplemental Methods include detailed information on the volume management in reasonable hypotensive events.

In the conventional method group, fluid balancing was conducted based on clinical information, including body weight, hemodynamic stability, and daily input and output records. The target I/O was by the attending physician's judgment, without the researchers' involvement. For safety reasons, it was recommended to remove <50% of the amount of fluid overload or 1 L/d according to the previous day's intake and output records; the attending physicians could adjust these limits at their discretion. The nephrologists of each institution regularly educated the attending physicians on standard treatment based on general CKRT guidelines; therefore, they administered conventional standardized therapy to patients during the study period.

Even if the patient stopped CKRT or switched to intermittent hemodialysis, the measurements and the fluid control guideline were applied in the same way. For hemodialysis patients, fluid removal was conducted through ultrafiltration (UF) during dialysis sessions. Conversely, for patients whose kidney function had recovered, leading to dialysis cessation, fluid removal was managed in line with guidelines using methods like fluid restriction or diuretics administration and adjustment.

Secondary Outcomes

The secondary outcomes were comparisons of the 28-, 60-, and 90-day mortality rates between the control and intervention groups.

Post Hoc Analysis

We compared the following variables between the control and intervention groups: (1) KRT dependence among survivors at 90 days; (2) duration of CKRT application, intensive care unit (ICU), and hospital stays; and (3) cause of 28-day mortality. Supplemental Methods include detailed information on the post hoc analysis.

Adverse Events

Information on the number of hypotensive episodes and the use and type of vasopressors was also collected to assess adverse events. A hypotensive event was defined as a mean BP decrease of <65 mm Hg, when vasopressors were increased or when fluids were given to increase the patient's BP.

Statistical Analyses

Statistical analyses were performed for both per protocol and intention to treat.²⁵ Kaplan–Meier curve analysis and Cox regression analysis were used to analyze the time to mortality and mortality hazard ratio (HR), and logistic regression models were used to obtain the odds ratio and 95% confidence intervals (CIs) for the primary outcome. A competing risk analysis was performed in which expiration was set as a competing risk event, using R version 4.1.3 (cmprsk library). We also compared the TBW/H² measured at each time point between the two groups through the repeated measures analysis of variance. Subgroups were divided into those with a median TBW/H² at D₀ of <14.63 L/m² and those with TBW/H² >14.63 L/m² along with the change in TBW/H² values by the period between each group were examined. A P < 0.05 was considered statistically significant. All analyses were performed using SPSS Statistics software version 22.0 (IBM Corporation, Armonk, NY).

Results

Study Participants

Data regarding the fluid status of 621 patients on CKRT were obtained through InBody measurement (Figure 1). Among them, 208 patients were randomized and included in the intention-to-treat analysis. The dropout rate was lower than anticipated, with only 14 of 208 randomized patients withdrawing. Consequently, 194 patients were eligible for the per-protocol analysis. This comprised 96 patients from the intervention group and 98 from the control group.

The baseline characteristics of the two groups were comparable (Table 1). At the time of randomization, TBW/H² of the intervention (14.9 L/m²) and control groups (15.0 L/m²) showed no statistically significant difference. However, serum creatinine at the time of randomization was 3.2 mg/dl in the intervention group, which was significantly higher than that of the control group. In addition, baseline characteristics between the two groups were similar in the per-protocol analysis (Supplemental Table 2).

Table 1.

Baseline characteristics of intention-to-treat population

Characteristic	Intervention (n=105)	Control (n=103)
Male sex, No. (%)	78 (74)	75 (73)
Age, yr, mean (SD)	67 (14)	67 (14)
Body mass index, kg/m2, mean (SD)	25.6 (4.0)	24.6 (3.5)
Body weight, kg, mean (SD)	70.5 (12.1)	68.4 (10.8)
Preexisting conditions, No. (%)
CKD	46 (56)	38 (47)
Hypertension	51 (49)	42 (41)
Diabetes mellitus	50 (48)	40 (39)
Coronary artery disease	14 (13)	12 (12)
Heart failure	14 (13)	13 (13)
Liver disease	19 (18)	20 (19)
Cancer	22 (21)	32 (31)
Age-modified CCI, mean (SD)	3.3 (2.2)	3.3 (2.5)
Contributing factors for AKI, No. (%)
Sepsis	52 (50)	52 (51)
Ischemia	16 (15)	20 (19)
Major surgery	9 (9)	9 (9)
Others (nephrotoxic drugs)	4 (4)	7 (7)
Clinical condition at randomization
APACHE II score, mean (SD)	29.1 (9.7)	29.4 (10.3)
SOFA score, mean (SD)	10.5 (3.7)	10.6 (3.5)
Mechanical ventilation, No. (%)	80 (76)	72 (70)
Vasopressor support, No. (%)	78 (74)	78 (76)
Days from ICU admission to CKRT start, mean (SD)	2.1 (3.6)	1.9 (4.3)
Days from CKRT start to randomization, mean (SD)	1.0 (2.6)	1.0 (1.2)
24-h U/O before CKRT, ml/kg per hour, mean (SD)	0.5 (0.6)	0.4 (0.4)
6-h U/O before CKRT, ml/kg per hour, mean (SD)	0.5 (0.5)	0.4 (0.4)
Laboratory findings at randomization, mean (SD)
WBC, ×103/μl	13.4 (9.5)	14.3 (12.7)
CRP, mg/L	27.3 (63.8)	18.4 (28.3)
Lactate, mmol/L	6.6 (14.5)	5.2 (4.7)
Sodium, mEq/L	137.6 (5.5)	136.3 (7.1)
Potassium, mEq/L	4.3 (0.8)	4.3 (1.0)
Albumin, mg/dl	3.0 (2.5)	2.7 (0.5)
AST, U/L	648 (1992)	620 (1573)
ALT, U/L	256 (633)	284 (650)
Serum creatinine, mg/dl	3.2 (1.9)	2.6 (1.3)
Prothrombin time, INR	1.8 (0.8)	1.6 (0.6)
CKRT setting
Prescribed dose, ml/kg per hour, mean (SD)	34.5 (12.1)	36.5 (12.0)
Delivered dose, ml/kg per hour, mean (SD)	30.9 (7.4)	31.7 (7.4)
Anticoagulation, No. (%)	60 (57)	53 (52)
TBW/H2 at randomization (D0), L/m2, mean (SD)	14.9 (1.3)	15.0 (1.6)

Continuous variables are presented as mean (SD) and categorical variables are presented as frequencies (percentages).

ALT, alanine aminotransferase; AST, aspartate aminotransferase; CCI, Charlson Comorbidity Index; CKRT, continuous KRT; ICU, intensive care unit; INR, international normalized ratio; TBW/H², total body water/height²; U/O, urine output.

Primary Outcome

On D₇ postrandomization, 34 (47%) and 27 (41%) patients in the intervention and control groups, respectively, achieved e-euvolemia; there was no significant difference between the two groups (Figure 2A and Table 2). The results of the repeated measures analysis of variance indicated that there was no statistically significant difference in TBW/H² between the groups (F=0.052, P = 0.820). In addition, the test for interaction effects between group and measurement time also showed no significant interaction effect (F=0.251, P = 0.859).Per-protocol analysis revealed similar results (Figure 2B and Supplemental Table 3). Even considering patients whose e-euvolemia status was not assessed due to death, there was no difference in the achievement of e-euvolemia between the two groups (Supplemental Figures 1 and 2).

Figure 2.

Trends of TBW/H² at each time point. (A) Intention-to-treat analysis. (B) Per-protocol analysis. At 7 days postrandomization, the TBW/H² values in both the intervention and control groups were continuously decreased, but no difference was recorded between the two groups at any time point excluding D₁. In the first 24 hours (first day), the TBW/H² value of the control group slightly increased, whereas that of the intervention group rapidly decreased. Blue dot line, intervention group; red dot line, control group. TBW/H², total body water/height².

Table 2.

Primary, secondary, and post hoc outcomes in the intention-to-treat population

Outcomes	Intervention (n=105)	Control (n=103)	P Value
Primary outcome
Reaching e-euvolemia on D7 (n=73, 66)	34 (47)	27 (41)	0.50
ΔTBW/H2 (D7–D0) <−2.1 L/m2	20 (27)	18 (27)	0.99
TBW/H2 on D7 <13 L/m2	25 (34)	19 (29)	0.49
Secondary outcomes
Death from any cause, d
At 28	39 (37)	53 (52)	0.04
At 60	46 (44)	58 (56)	0.07
At 90	47 (45)	59 (57)	0.07
Post hoc outcomes
TBW/H2 on D1 (L/m2) (n=99, 99)	14.7 (1.6)	15.1 (1.8)	0.01
TBW/H2 on D2 (L/m2) (n=91, 88)	14.6 (1.5)	14.8 (1.9)	0.36
TBW/H2 on D7 (L/m2) (n=73, 66)	13.9 (1.8)	13.9 (1.8)	0.86
Fluid status, mean (SD), L/m 2
Total intake-total output (ml) on D1 (n=100, 101)	−143.3 (1443.8)	145.3 (1675.5)	0.19
Total intake-total output (ml) on D2 (n=96, 93)	−300.3 (1397.0)	−35.9 (1489.2)	0.21
Total intake-total output (ml) on D7 (n=74, 70)	315.1 (1072.6)	205.1 (933.4)	0.51
Body weight (kg) on D1 (n=99, 99)	70.1 (12.5)	68.3 (10.8)	0.27
Body weight (kg) on D2 (n=91, 88)	69.8 (12.6)	67.4 (10.7)	0.17
Body weight (kg) on D7 (n=73, 66)	67.6 (11.8)	65.3 (10.3)	0.23
Total UF (ml) on D1	3146.8 (1397.6)	2942.7 (1245.7)	0.27
Total UF (ml) on D2	3069.2 (1298.4)	3029.6 (1213.7)	0.83
Total UF (ml) on D7	2285.3 (1114.9)	2277.5 (1183.8)	0.97
Cumulative UF (ml) from D0 to D1	5392.8 (2162.5)	5329.9 (2318.4)	0.85
Cumulative UF (ml) from D0 to D2	8449.0 (3016.7)	8622.2 (2786.2)	0.69
Cumulative UF (ml) from D0 to D7	20,685.0 (5012.8)	20,981.0 (6221.6)	0.77
Death during hospitalization	48 (46)	59 (57)	0.10
KRT dependence among survivors at 90 d	5/37 (14)	9/32 (28)	0.13
Duration of CKRT, d, mean (SD)	6.8 (4.9)	6.2 (5.8)	0.67
Median length of ICU stay, d, mean (SD)
Survivors	18.5 (23.5)	14.8 (13.4)	0.32
Nonsurvivors	8.4 (7.9)	9.1 (8.3)	0.70
Median length of hospital stay, d, mean (SD)
Survivors	39.3 (29.6)	34.5 (23.1)	0.35
Nonsurvivors	8.4 (7.9)	9.3 (8.6)	0.63
Cause of 28-d mortality			0.07
Sepsis	10 (26)	14 (26)
Respiratory+cardiogenic	11 (28)	25 (47)
Cancer	7 (18)	2 (4)
Others	11 (28)	12 (23)

Continuous variables are presented as mean (SD) and categorical variables are presented as frequencies (percentages).

CKRT, continuous KRT; e-euvolemia, estimated euvolemia; ICU, intensive care unit; TBW/H², total body water/height²; UF, ultrafiltration.

Secondary Outcomes

D₂₈ mortality rates of the intervention and control groups were 37% and 52%, respectively (P = 0.04; Table 2), showing an approximately 44% decrease in the intervention group than in the control group (odds ratio, 0.56; 95% CIs, 0.32 to 0.97; Figure 3A). In the per-protocol analysis, the mortality rate on day 28 was lower in the intervention than in the control group (34% and 50%, respectively; P = 0.03; Figure 3B, Supplemental Table 3). In the Cox regression analysis, the findings from the intention-to-treat analysis were as follows: HR 0.78 (95% CI, 0.54 to 1.13), P = 0.19. When adjusting for baseline serum creatinine value, the results were as follows: HR 0.82 (95% CI, 0.57 to 1.19), with a P value of 0.30, similarly showing no statistical difference. Similarly, the per-protocol analysis yielded the following outcomes: HR, 0.75 (95% CI, 0.51 to 1.10), P = 0.14. The 60- and 90-day mortality rates did not differ between the two groups.

Figure 3.

Kaplan–Meier estimates of survival at 28 days. (A) Intention-to-treat analysis. (B) Per-protocol analysis. The mortality rate at 28 days was significantly lower among patients in the intervention group than in the control group.

Subgroup Analyses

In patients with D₀ TBW/H² <14.63 L/m², the value of TBW/H² at each time point showed no difference between both groups; however, in patients with TBW/H² >14.63 L/m², the TBW/H² values continued to decrease from D₀ to D₇ in both groups. Although neither group reached e-euvolemia, the TBW/H² measured on D₁ showed a significant difference between the two groups (Supplemental Figure 3).

Post Hoc Outcome (End Points)

KRT dependence at 90 days, duration of CKRT application, duration of ICU, and hospital stays did not differ between the two groups (Table 2 and Supplemental Table 3).

Change (delta) in TBW/H² by time point and the percentage attaining e-euvolemia were assessed for all patients in the intention-to-treat analysis (Figure 2 and Supplemental Table 4). TBW/H² at each time point continued to decrease in both groups, but there was no statistical difference. The total intake/output measured on each study day did not demonstrate statistically significant differences between the groups. Similarly, the total UF measured on each study day did not exhibit any discernible distinctions between the groups. Cumulative UF over the period of D₀ to D₇ displayed a gradual increase in both cohorts throughout the study duration, yet no statistically significant disparities were noted between them. These outcomes are consistent with the absence of intergroup discrepancies in TBW/H² measurements during the study intervals (D₁, D₂, and D₇). However, a trend was observed where the intake/output measured on D₁ and D₂ were more negative in the intervention group, although this discrepancy did not achieve statistical significance. The change (delta) in TBW/H² between D₀ and D₁ decreased by 0.32 L/m² in the intervention group while an increase of 0.1 L/m² was observed in the control group. Furthermore, the numbers of patients who attained e-euvolemia on D₁ were 13 (13%) and 4 (4%) in the intervention and control groups, respectively (P = 0.02). Per-protocol analysis revealed similar results, although this statistical significance decreased slightly (Supplemental Table 5).

Adverse Events

No difference was observed in the frequency of hypotensive episodes that occurred in the BIA-guided or conventional volume control group at any time point. Furthermore, among patients with hypotensive episodes, no statistically significant difference was observed between the two groups (Table 3). In the per-protocol analysis, adverse events did not differ between the two groups (Supplemental Table 6).

Table 3.

Adverse events in the intention-to-treat population

Adverse Events	Intervention (n=105)	Control (n=103)
Hypotensive episode on D 0	52/105 (50)	51/103 (50)
Fluid loading (target I/O adjustment)	24 (46)	23 (45)
Increase in the dose of vasopressor	28 (54)	28 (55)
Hypotensive episode on D 1	41/99 (41)	43/99 (43)
Fluid loading (target I/O adjustment)	21 (51)	14 (33)
Increase in the dose of vasopressor	20 (49)	29 (67)
Hypotensive episode on D 2	41/91 (45)	41/88 (47)
Fluid loading (target I/O adjustment)	28 (68)	25 (61)
Increase in the dose of vasopressor	13 (32)	16 (39)
Hypotensive episode on D 7	18/73 (25)	19/66 (29)
Fluid loading (target I/O adjustment)	13 (72)	11 (58)
Increase in the dose of vasopressor	5 (28)	8 (42)

Categorical variables are presented as frequencies (percentages).

I/O, input/output.

Discussion

Many patients with AKI encounter concomitant fluid overload,²⁶ and approximately 10%–20% of these patients require extracorporeal support, such as CKRT.²⁷ However, the mortality rate in such patients even under CKRT management remains high, and fluid overload has emerged as the main cause of higher mortality in these patients.^7–9Although fluid balance is imperative for critically ill patients, fluid therapy is often dependent on a simple calculation of intake and output. Body composition, such as the distribution of body fluid in each compartment, including TBW, should be considered in the management of an overhydrated patient.¹⁰

Rochwerg et al.²⁸ recently assessed volume status in ICU patients using BIA. Calculation of intake and output may have errors and mistakes in measurement, collection, and calculation, and blood markers such as brain natriuretic peptide are invasive, take time to obtain results and are inconvenient to repeat tests. In addition, while repeated chest x-ray imaging may have risk of radiation exposure, BIA has the advantages of being noninvasive without such risks and discomfort, easy to measure, and able to confirm results by performing tests quickly and repeatedly.

However, no previous study has ever evaluated fluid management in CKRT-treated cases. Thus, we sought a novel way to assess fluid balance using bioimpedance and compared the e-euvolemia–reaching proportion and mortality between the two groups. To determine the ideal volume status more objectively, a TBW/H² value was presented as a marker that could have both an absolute and a delta value. When the correlation with the existing volume status markers was evaluated, there was a significant correlation between TBW/H² and the actual weight at each time point or between the difference of actual weight and the ideal body weight at each time point (Supplemental Table 7). The significant correlation between ΔTBW/H² and other figures related to daily fluid balance (Δ total I/O or Δ body weight) was also investigated (Supplemental Table 8). Considering this, TBW/H² has the potential as an indicator comparable with traditional ones for fluid balance. Further analysis or additional research focusing on extracellular water among our study participants may be necessary.

The proportion of reaching e-euvolemia on D₇ was not higher in the intervention group. There are several potential reasons for this. First, the proportion of reaching e-euvolemia on D₇ may not be an appropriate end point. It may have been more reasonable to assess the degree of reaching e-euvolemia on D₁ or D₂ or how quickly fluid is removed. Even if fluid is similarly removed, the removal speed at the beginning may be more important. In the post hoc analysis, fluid removal in the early phase (D₀–D₁) was significantly higher in the interventional than in the control group (∆TBW/H² between D₁ and D₀; intervention versus control; −0.32 versus 0.10 L/m², P = 0.01). The positive association between early fluid removal on D₁ and reduction of 28-day mortality was also shown. Furthermore, two previous studies have emphasized the importance of early fluid management among patients receiving CKRT and the benefit of rapid fluid removal for the reduction of mortality rate.^9,29 Second, reducing the fluid removal target when hypotensive episodes occurred was considered important owing to safety concerns. No significant difference was found between groups in the proportion of patients achieving the recommended target each study day. Although no statistical difference was observed in hypotensive episodes among patients not reaching the target, a slightly higher proportion was noted in the intervention group. Furthermore, initial CKRT prescription may have been changed because of decreased BP and the subjective opinions of the attending physicians. However, because such decisions are made in real time without traces in medical records, it has been impossible to investigate them fully. Third, relying on measurements up to D₂ to determine the achievement of the primary outcome on D₇ has its limitations. Fluid removal was not precisely performed through daily monitoring and measurement from D₃ to D₆. If the fluid status was measured every day from D₃ to D₆ and the fluid target was set based on this, the e-euvolemia reaching proportion might have changed.

Nevertheless, we consider our guideline of using BIA a step toward the continuous exploration of a novel method to manage fluid balance in critically ill patients undergoing CKRT. Lumlertgul et al.³⁰ showed significant variation in clinical UF practice among different physicians, suggesting that decisions dependent on physician experiences may not be sufficient for managing fluid balance in critically ill patients receiving CKRT. However, it is not easy to use the results predicted by BIA measurements in clinical decision making. This is due to the complexity of the causal relationship between fluid overload and survival, as well as the intricate nature of fluid management decisions. While accurately assessing a patient's fluid status may potentially facilitate recovery, in reality, patient recovery may be hindered by more aggressive fluid removal. Further research is needed to establish the clinical utility and significance of BIA results in actual clinical decision making. In addition, research and analysis are required to validate the validity of the outcomes we arbitrarily selected.

This study has several limitations. First, we arbitrarily defined e-euvolemia and the fluid removal target. Because there is no consensus definition for euvolemia, we defined it on the basis of the studies by Garzotto et al.⁹ and Rhee et al.¹⁰ However, because this study was conducted only in men, this TBW/H² may have limitations in being applicable to all sexes. However, even when the study patients were divided by sex for analysis, no differences were observed in the primary outcome between the two groups. Second, information regarding UF prescribed by clinicians and actual UF delivered in both treatment groups was unavailable. Furthermore, we did not consider CKRT circuit patency and maintenance of electrolyte and acid–base homeostasis, which may have been influenced by fluid balance management. Third, physicians' opinions regarding controlling the volume status were inevitably reflected during the management of hypotension or fluid removal. Furthermore, because it is a multicenter study, it was not possible to standardize all protocols for monitoring participants, operation, and management of CKRT. However, this may have reflected a real-time clinical setting. Fourth, we did not verify the results of this study with other bioimpedance devices. However, as the results of previous studies using InBody S10 supported the usefulness of this device,^10,19,20 a prospective study was able to be performed. It may be necessary to validate this with other devices. Finally, this study showed that the 28-day mortality rate, one of the secondary outcomes, may be improved with or without BIA intervention. The intervention in the intervention group showed a favorable effect on 28-day mortality but did not affect 90-day mortality. A plausible explanation could be the significant fluid removal in the initial days, with minimal deviation from the control group afterward. This suggests that although initial fluid removal rate may affect short-term mortality, other factors could mitigate its impact on longer-term mortality. Further investigation is needed to clarify these findings.

As the sole randomized trial of the value of bioimpedance in the acute clinical setting, this study supports implementation of a guideline for fluid balance management in patients undergoing CKRT using bioimpedance; however, BIA-guided volume management did not affect the proportion of reaching e-euvolemia at 7 days after the start of CKRT. The patients in the intervention group had a higher proportion of reaching e-euvolemia on D₁ than the control group; thus, further investigations are required on whether the guidelines using bioimpedance in patients with AKI using CKRT will help rapid fluid removal in the early phase of CKRT.

Supplementary Material

cjasn-19-1527-s001.pdf ^{(1.4MB, pdf)}

cjasn-19-1527-s002.pdf ^{(628.7KB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C23.

Funding

S. Kim: Korea Health Industry Development Institute (HI17C1827).

Author Contributions

Conceptualization: Dong Ki Kim, Sejoong Kim, Jung Pyo Lee, Hyung Jung Oh, Harin Rhee, Dong-Ryeol Ryu.

Data curation: Shin Young Ahn, Jung Nam An, Seon Ha Baek, Jang-Hee Cho, Dong Ki Kim, Sejoong Kim, Jung Pyo Lee, Hyung Jung Oh, Harin Rhee, Dong-Ryeol Ryu, Eun Young Seong.

Formal analysis: Soyeon Ahn, Jung Nam An, Sejoong Kim, Hyung Jung Oh, Sohee Oh.

Funding acquisition: Sejoong Kim.

Project administration: Sejoong Kim.

Supervision: Shin Young Ahn, Soyeon Ahn, Seon Ha Baek, Jang-Hee Cho, Dong Ki Kim, Sejoong Kim, Jung Pyo Lee, Sohee Oh, Harin Rhee, Dong-Ryeol Ryu, Eun Young Seong.

Validation: Soyeon Ahn, Sohee Oh.

Visualization: Jung Nam An, Hyung Jung Oh, Sohee Oh.

Writing – original draft: Jung Nam An, Hyung Jung Oh.

Writing – review & editing: Jung Nam An, Sejoong Kim, Hyung Jung Oh, Sohee Oh.

Data Sharing Statement

All data are included in the manuscript and/or supporting information. Study protocol and statistical analysis plan are available for sharing. Individual patient information cannot be shared. Study protocol and statistical analysis plan can be shared by contacting Jung Nam An, Hyung Jung Oh, and Sejoong Kim by email. This is available immediately after the paper is published.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/CJN/C22.

Supplemental Methods

Supplemental Table 1. Guidelines for fluid management in the intervention group.

Supplemental Table 2. Baseline characteristics of the per-protocol population.

Supplemental Table 3. Primary and secondary outcomes in the per-protocol population.

Supplemental Table 4. Post hoc outcomes in the intention-to-treat population.

Supplemental Table 5. Post hoc outcomes in the per-protocol population.

Supplemental Table 6. Adverse events in the per-protocol population.

Supplemental Table 7. Correlation between TBW/H² and the various volume status markers.

Supplemental Table 8. Correlation between TBW/H² and the daily fluid balance markers.

Supplemental Figure 1. Competing risk analysis for the primary outcome in the intention-to-treat population.

Supplemental Figure 2. Competing risk analysis for the primary outcome in the per-protocol population.

Supplemental Figure 3. Trends of TBW/H² at each time point divided by initial TBW/H².