Background
Fluid overload (FO) is a common complication of acute illness affecting more than a third of critically ill patients and approximately two-thirds of patients with acute kidney injury (AKI) requiring renal replacement therapy (RRT) [
1,
2]. Several studies have documented that FO is independently associated with more than 50% mortality among patients receiving RRT [
3,
4]. Observational studies suggest that fluid removal using net ultrafiltration (UF
NET) may be associated with improved outcomes [
2], and clinical and consensus guidelines recommend UF
NET for the treatment of FO in patients with oliguric AKI who are resistant to diuretic therapy [
5,
6]. However, the optimal intensity of UF
NET (i.e., rate and volume of net fluid removal) in critically ill patients remains uncertain more than 70 years after the first clinical use of ultrafiltration [
7].
Less intensive UF
NET, characterized by a slower rate or smaller volume of fluid removed, may be associated with prolonged exposure to tissue and organ edema and increased morbidity and mortality [
8,
9]. More intensive UF
NET with a faster rate or larger volume of fluid removal, however, may be associated with increased hemodynamic and cardiovascular stress [
10], leading to ischemic organ injury and mortality in critically ill patients [
11]. Indeed, three observational studies in outpatients with end-stage renal disease suggest that UF
NET intensity > 10 ml/kg/h is associated with increased overall [
12‐
14] and cardiovascular [
12] mortality.
Understanding the relationship between UF
NET intensity and outcome in critically ill patients is essential for two important reasons. First, if more intensive UF
NET is associated with lower mortality, then clinical trials could be designed to reduce the risk of death. Second, understanding the intensity–outcome relationship will aid in standardizing UF
NET intensity and implementing quality measures [
15,
16].
In this observational study involving a large heterogeneous cohort of critically ill patients with ≥ 5% FO and receiving RRT, we examined the association between UFNET intensity and its association with risk-adjusted 1-year mortality. Because the magnitude of FO is independently associated with mortality, we hypothesized that intensive UFNET would be associated with lower mortality. However, our null hypothesis was that there is no difference in mortality for an intensive UFNET group compared with a less intensive UFNET group.
Methods
Data source and study population
We conducted a retrospective study using a large tertiary care academic medical center ICU database: the High-Density Intensive Care dataset, details of which have been published elsewhere (Additional file
1: S1) [
1,
17,
18]. The study population included adults admitted to medical, cardiac, abdominal transplant, cardiothoracic, surgical, neurovascular, neurotrauma and trauma ICUs during July 2000 through October 2008. We included patients with AKI receiving RRT who had a cumulative fluid balance ≥ 5% prior to RRT initiation (Additional file
1: Figure S1). We extracted the daily fluid balance before and for the duration of RRT (Additional file
1: S2), hourly mean arterial pressure (MAP) and vasopressor type and dose (Additional file
1: S3) during RRT. The University of Pittsburgh’s institutional review board approved the study.
Determination of UFNET intensity
For patients receiving continuous renal replacement therapy (CRRT), we first extracted data on the total duration (in hours) of any form of CRRT (i.e., continuous venovenous hemodiafiltration (CVVHDF), continuous venovenous hemofiltration (CVVH), continuous venovenous hemodialysis (CVVHD) and slow continuous ultrafiltration (SCUF)). We then determined the UF volume produced and the amount of substitution fluids given each hour for patients receiving CVVHDF and CVVH. The UF
NET each hour was calculated as the difference between the UF volume and the volume of substitution fluids [
19]. For patients receiving CVVHD and SCUF, UF
NET corresponded to the UF volume removed. We then calculated the total number of days of CRRT for each patient based on the hourly duration of CRRT and the total UF
NET.
For patients receiving intermittent hemodialysis (IHD), we extracted the total number of IHD sessions and the UF volume removed per session from the time of ICU admission to the end of ICU stay. We excluded patients if they received IHD prior to ICU admission. UF
NET corresponded to the volume ultrafiltered during each session. We then expressed the total number of IHD sessions as the number of days for each patient. Subsequently, we estimated the UF
NET intensity using the equation:
$$ {\mathrm{UF}}^{\mathrm{NET}}\mathrm{intensity}\left(\mathrm{ml}/\mathrm{kg}/\mathrm{day}\right)=\kern0.5em \frac{\mathrm{Total}\ {\mathrm{UF}}^{\mathrm{NET}}\mathrm{volume}\ \left(\mathrm{ml}\right)}{\mathrm{Hospital}\ \mathrm{admission}\ \mathrm{weight}\ \left(\mathrm{kg}\right)\ \mathrm{X}\ \mathrm{RRT}\ \mathrm{duration}\ \left(\mathrm{days}\right)}. $$
For instance, if an 80-kg patient is on CVVH with an UF rate of 2000 ml/h and substitution fluid of 1500 ml/h, the total UFNET produced is 500 ml/h (2000 – 1500 = 500 ml) or 500 × 24 = 12,000 ml/day. The total UFNET produced for 5 days is 12,000 × 5 = 60,000 ml. Thus, the total UFNET intensity is [60,000 / (80 × 5)] = 150 ml/kg/day. During CVVHD and IHD, the UF volume is equivalent to UFNET.
Outcomes
The primary outcome was 1-year mortality from the index ICU admission and mortality data were obtained from the Social Security Death Master File [
20]. We chose 1-year mortality because our prior work showed that a positive fluid balance was associated with risk of death at 1 year and use of renal replacement therapy was associated with lower risk of death in patients with a positive fluid balance [
1]. Secondary outcomes included hospital length of stay, hospital mortality and renal recovery. Renal recovery was defined as alive
and independent from RRT at 1 year. Dialysis dependence data were obtained from the US Renal Data System [
21].
Statistical analysis
We stratified UF
NET intensity into three groups because of the nonlinear (i.e., J-shaped) association between UF
NET intensity and hospital mortality (Additional file
1: Figure S2). We defined UF
NET ≤ 20 ml/kg/day as “low” intensity, UF
NET > 20 to ≤ 25 ml/kg/day as “moderate” intensity and UF
NET > 25 ml/kg/day as “high” intensity. Categorical variables were compared using the chi-squared test, and continuous variables using-one way analysis of variance and the Kruskal–Wallis test. We assessed time-to-mortality censored at 1 year using Kaplan–Meier failure plots.
We used three methods to examine the association between UF
NET intensity and mortality. First, we fitted logistic regression and estimated risk-adjusted odds ratios (AORs) for high and moderate intensity, compared with low intensity UF
NET (reference), on 1-year mortality. Second, we fitted Gray’s survival model [
22,
23] to estimate risk-adjusted hazard ratios (AHRs) for time to mortality using four time nodes and five intervals (Additional file
1: S4). We adjusted for differences in age, sex, race, body mass index, history of liver disease and sequela from liver disease, admission for liver transplantation, admission for surgery, baseline glomerular filtration rate, Acute Physiologic and Chronic Health Evaluation (APACHE) III score, presence of sepsis, use of mechanical ventilation, percentage of FO before initiation of RRT, oliguria before initiation of RRT, time to initiation of RRT from ICU admission, MAP on first day of RRT initiation, cumulative vasopressor dose and cumulative fluid balance during RRT, first RRT modality and duration of RRT.
Third, in order to account for indication bias, we conducted a propensity score-matched analysis. Since the mortality associated with moderate (> 20 to ≤ 25 ml/kg/day) vs high (> 25 ml/kg/day) or moderate (> 20 to ≤ 25 ml/kg/day) vs low (≤ 20 ml/kg/day) intensity UF
NET was not different (Table
1), we combined the moderate and low-intensity groups into a single low-intensity group (reference). We then matched the low-intensity UF
NET (≤ 25 ml/kg/day) with the high-intensity UF
NET (> 25 ml/kg/day) using propensity scores on a 1:1 basis without replacement, creating 258 matched pairs (Additional file
1: S5).
Table 1
Baseline characteristics of study population by net ultrafiltration Intensity
Age (years), median (IQR) | 61 (52–69) | 59 (51–71) | 58 (48–70) | 0.16 |
Male sex | 301 (63.4) | 114 (68.7) | 218 (50.2) | < 0.001 |
Race |
Caucasian | 380 (80) | 136 (81.9) | 335 (77.2) | 0.018 |
African-American | 24 (5.1) | 6 (3.6) | 43 (9.9) |
Other | 71 (14.9) | 24 (14.5) | 56 (12.9) |
BMI (kg/m2), median (IQR) | 28.3 (24.2–34.3) | 27.7 (24.2–31.7) | 25.1 (21.9–29.3) | < 0.001 |
Comorbid condition |
Hypertension | 169 (35.6) | 72 (43.4) | 161 (37.1) | 0.19 |
Diabetes | 121 (25.5) | 34 (20.5) | 97 (22.4) | 0.33 |
Cardiac disease | 84 (17.7) | 36 (21.7) | 99 (22.8) | 0.14 |
Heart failure | 70 (14.7) | 30 (18.1) | 86 (19.8) | 0.12 |
Vascular disease | 41 (8.6) | 16 (9.6) | 43 (9.9) | 0.79 |
Liver disease | 164 (34.5) | 47 (28.3) | 107 (24.7) | 0.005 |
Sequela from liver disease | 137 (28.8) | 43 (25.9) | 95 (21.9) | 0.056 |
Malignancy | 23 (4.8) | 4 (2.4) | 14 (3.2) | 0.26 |
Liver transplantation | 43 (9.1) | 13 (7.8) | 42 (9.7) | 0.77 |
Multiple comorbidity | 298 (62.7) | 93 (56) | 252 (58.1) | 0.19 |
Surgical admission | 321 (67.6) | 122 (73.5) | 301 (69.4) | 0.72 |
Medical admission | 131 (27.6) | 37 (22.3) | 112 (25.8) | 0.72 |
Admission for liver transplantation | 102 (21.5) | 31 (18.7) | 53 (12.2) | 0.001 |
Baseline serum creatinine (mg/dl), median (IQR) | 1.029 (0.81–1.27) | 1.035 (0.83–1.3) | 1.032 (0.8–1.3) | 0.89 |
Baseline eGFR (ml/min/1.73 m2) |
> 90 | 107 (22.5) | 27 (16.3) | 91 (20.9) | 0.54 |
60–90 | 235 (49.5) | 97 (58.4) | 212 (48.9) |
30–60 | 89 (18.7) | 30 (18.1) | 92 (21.2) |
15–30 | 34 (7.2) | 8 (4.8) | 31 (7.1) |
< 15 | 10 (2.1) | 4 (2.4) | 8 (1.8) |
APACHE III score, median (IQR)a | 95 (70–118) | 91 (71–116) | 91 (69–112) | 0.27 |
Sepsisa | 128 (26.9) | 39 (23.5) | 138 (31.8) | 0.08 |
Mechanical ventilationa | 353 (74.3) | 129 (77.7) | 329 (75.8) | 0.66 |
Vasopressora | 261 (54.9) | 87 (52.4) | 218 (50.2) | 0.36 |
Oliguria before initiation of RRTb |
Stage 2 | 50 (10.5) | 9 (5.4) | 21 (4.8) | 0.017 |
Stage 3 | 406 (85.5) | 154 (92.8) | 402 (92.6) |
MAP during RRT (mmHg), mean (SD)c |
All patients | 75.1 (0.58) | 77.5 (1.19) | 79.4 (0.62) | < 0.001 |
CRRT only (n = 386) | 72.7 (0.70) | 72.4 (1.89) | 77.5 (1.01) | < 0.001 |
IHD only (n = 210) | 85 (1.84) | 84.1 (2.85) | 82.1 (1.27) | 0.77 |
CRRT and IHD (n = 487) | 74.5 (0.91) | 79.1 (1.66) | 79.7 (0.98) | 0.002 |
Vasopressor dose (NE), median (IQR)c,d |
All patients | 0.11 (0.04–0.25) | 0.09 (0.03–0.21) | 0.09 (0.04–0.25) | 0.25 |
Patients on CRRT only | 0.14 (0.05–0.30) | 0.13 (0.03–0.25) | 0.10 (0.03–0.28) | 0.31 |
Patients on IHD only | 0.01 (0.01–0.03) | 0.06 (0.01–0.11) | 0.03 (0.01–0.07) | 0.67 |
Patients on both CRRT and IHD | 0.08 (0.03–0.16) | 0.08 (0.02–0.16) | 0.07 (0.03–0.19) | 0.85 |
We performed five sensitivity analyses and two subgroup analyses. First, we restricted the UF
NET intensity only up to 72 h from initiation of RRT. Second, we used an alternative definition of UF
NET intensity moving the threshold down as follows: low, < 15 ml/kg/day; moderate, 15–20 ml/kg/day; and high, > 20 ml/kg/day. Third, we moved the threshold up: low, < 25 ml/kg/day; moderate, 25–30 ml/kg/day; and high, > 30 ml/kg/day. Fourth, we divided the cohort into tertiles: low, ≤ 16.7 ml/kg/day; moderate, 16.7 to ≤ 27.7 ml/kg/day; and high, > 27.7 ml/kg/day. Fifth, we performed quantitative bias analysis to assess the magnitude of a hypothetical unmeasured confounder that would be necessary to account for the association between UF
NET intensity and risk-adjusted mortality (Additional file
1: S6) [
24,
25].
Sixth, we restricted our analyses only to the subgroup of patients with > 20% FO. Seventh, we confined our analysis of UFNET intensity to the hour (i.e., ml/kg/h) instead of the day among the subgroup of patients who only received CRRT as follows: low, < 0.5 ml/kg/h; moderate, 0.5–1.0 ml/kg/h; and high, > 1 ml/kg/h. Statistical analyses were performed using SAS 9.3 (SAS Institute, Cary, NC, USA), Gray’s model used R 3.2.1, and quantitative bias analysis was performed using STATA 15 (STATCorp., TX, USA). All hypotheses tests were two-sided with a significance level of p < 0.05.
Discussion
We found that UFNET intensity > 25 ml/kg/day, compared with < 20 ml/kg/day, was independently associated with lower risk-adjusted 1-year mortality in critically ill patients with FO. Using Gray’s model, this survival benefit was greater early on after ICU admission and persisted up to 39 days. In the propensity-matched analysis, UFNET > 25 ml/kg/day, compared with ≤ 25 ml/kg/day, was also associated with lower risk of death. To our knowledge, this is the first study in the literature examining the association between UFNET intensity and long-term mortality.
Our finding is somewhat analogous to the association between intensity of solute control and mortality in critically ill patients receiving RRT in which a threshold intensity of at least 20–25 ml/kg/h of effluent dosing in CRRT or KT/V of 1.2–1.4 per session in patients receiving IHD is associated with improved survival [
26,
27]. However, in contrast to studies on solute control, the optimal “dosing” for UF
NET in critically ill patients with fluid overload is unclear. In our study, we first explored whether there was an association between UF
NET dose and mortality, and then aimed to determine the overall “average dose” that is associated with a long-term mortality benefit. It is important to note that our finding does not suggest that UF
NET should be dosed > 25 ml/kg/day throughout the duration of fluid removal. Day-to-day dosing may vary in patients depending on the severity of fluid overload, patient tolerability and hemodynamics.
In our study only 40% of patients received intensive UFNET, whereas 44% of patients received less intensive UFNET that has implications for care. Unlike a prescription for solute clearance, the concept of a minimum or adequate “dose” for volume clearance is not usually considered in clinical practice. Although patients who received less intensive UFNET were hemodynamically unstable in our study, our findings persisted after accounting for hemodynamics, vasopressor dose and severity of illness, suggesting that less intensive UFNET per se might be associated with mortality. These findings may suggest that failure to tolerate UFNET > 25 ml/kg/day may portend a poor prognosis and, conversely, tolerating UFNET > 25 ml/kg/day may be a predictor of recovery and lower mortality in critically ill patients with fluid overload.
Our study addresses an important knowledge gap not addressed by prior studies. While numerous studies have documented an association between the severity of FO and incremental risk of death [
3,
4], none examined the UF
NET intensity–mortality relationship. Using the Program to Improve Care in Acute Renal Disease (PICARD) study, Bouchard et al. [
4] found that patients in whom FO was corrected during RRT had lower mortality than those who remained fluid overloaded despite RRT. Using the Randomized Evaluation of Normal versus Augmented Level of Renal Replacement Therapy (RENAL RRT) cohort, Bellomo et al. [
2] found that a negative fluid balance during RRT was associated with a mortality benefit. However, we asked a different question: does UF
NET intensity and a threshold “dose” of UF
NET matter in the treatment of FO independent of fluid balance?
There may be several biologic explanations for the association between UF
NET intensity and outcome. First, intensive UF
NET may reduce prolonged exposure to FO and modify host response, and could reduce the incidence of subsequent organ dysfunction [
28]. Second, the salutary effects of intensive UF
NET may be mediated through unknown marker clearance independent of fluid balance since the association persisted despite controlling for cumulative fluid balance. Third, clinicians who decide to initiate intensive UF
NET may select for a unique group of patients to monitor and carefully titrate fluid removal. Fourth, clinicians and nurses may also have a broad variation in how they prescribe and/or practice UF
NET in the real world, which may be associated with differences in outcomes [
29].
The strengths of our study was that it was robust to three different methods of sensitivity analysis. We accounted for confounding due to severity of illness, hemodynamics, vasopressor dose and cumulative fluid balance before and during RRT. Using Gray’s model, we found that high-intensity UFNET was associated with survival only up to 39 days after ICU admission. This finding is in contrast with the logistic model and propensity-matched analyses, which showed mortality benefit up to 1 year. This discordant finding is due to the differences in the models that were used. In Gray’s model, the number of events between high-intensity and low-intensity UFNET groups was not different within the time interval of 39–365 days. Using the logistic regression model, however, a lower odds of cumulative deaths occurred by 1 year in the high-intensity UFNET group compared with the low-intensity UFNET group.
Our study is not without limitations. First, given the observational nature, it is not possible to make causal inferences between UFNET intensity and outcomes. Second, we do not know precisely whether a UFNET threshold > 25 ml/kg/day is associated with better outcomes, although our findings were robust to several sensitivity analyses. Third, our single-center study may not be generalizable to other ICU populations. Nevertheless, our study included patients typical of an academic medical center ICU population. Fourth, we were unable to distinguish whether patients received low-intensity UFNET due to low prescription, failure to remove fluid (e.g., circuit downtime, trip to operating room, etc.) or other variations in practice with respect to fluid removal. Fifth, although the sensitivity analysis indicated that any unmeasured confounder would need to be highly prevalent and have an OR < 0.7 to mask a null association, it is possible that there may be more than one residual confounder and that it may not be a binary variable.