Zero-Balance Ultrafiltration during Cardiopulmonary Bypass Is Associated with Decreased Urine Output

Abstract:

Zero-balance ultrafiltration (ZBUF) during cardiopulmonary bypass (CPB) has been purported to reduce pro-inflammatory mediators during cardiac surgery. However, its clinical benefit is equivocal and its effect on renal function unknown. The purpose of this study was to examine the effect of ZBUF on urine output in adult patients undergoing CPB. Following institutional review board approval, 98,953 records from a national registry of adult patients at 215 U.S. hospitals between January 2016 and September 2019 were reviewed. Groups were stratified according to ZBUF use. Anuric patients were excluded from the study as they were patients with missing data on urine output, ultrafiltration use, or ZBUF volume. The primary endpoint was intraoperative urine output normalized to body weight and procedure duration (total operative time). Final analysis of this endpoint was carried out using a linear mixed-effects regression model adjusting for patient and procedural characteristics, as well as practice patterns associated with surgeons and perfusionists. There was a significant 16.1% reduction in median urine output for ZBUF patients (.94 [.54, 1.47] mL/kg/h) vs. the non-ZBUF group (1.12 [.70,−1.73] mL/kg/h), p < .001. After statistically adjusting for patient and procedural characteristics, each liter of ZBUF volume was associated with an estimated change in intraoperative urine output of −.03 mL/kg/h (95% CI: [−.04 to −.02], p < .001). The median ZBUF volume was 1,550 [1,000, 2,600] mL, and when ZBUF was used, conventional ultrafiltration (CUF) was more likely to be used as well (88.4% vs. 44.8%, p < .001). ZBUF patients had median asanguineous volume and crystalloid cardioplegia nearly two times more than non-ZBUF patients, and had slightly higher red blood cell transfusions (17.6% vs. 16.3%, p < .05). The application of ZBUF during CPB was associated with patients having lower urine output and significantly higher use of CUF. Further research is required to determine if these results are reproducible in prospective clinical studies.

Managing a patient’s fluid balance during cardiac surgery is a complex process that involves multiple factors that affect both hemodynamics and hemostasis. The steps involved with fluid maintenance represent a complex interaction of multiple factors that include solute delivery and waste product removal, vascular integrity, molecular and cellular function, and the influence of numerous, and often competing, medications. The administration of asanguineous fluids to maintain hemodynamic stability results in perturbations of circulating blood volumes that alters the concentration of formed and non-formed elements of blood. Removing excessive plasma volume mitigates hemodilution and results in a concentrating response. The main processes involved in hemoconcentration during cardiac surgery with cardiopulmonary bypass (CPB) are renal function and the use of extracorporeal ultrafiltration (UF) devices.

In cardiac surgery with CPB, UF has been in use for many years and is often described in terms of its application (1,2). The most common UF modality is described as conventional UF (CUF), whereas the other two common applications are referred to as modified UF (MUF) and zero-balance UF (ZBUF) with the possibility of all three techniques being performed on a single patient (3–7). Whereas MUF is used most often in pediatric applications to concentrate the patient’s blood volume once separated from CPB, ZBUF is used during CPB not as a hemoconcentrating measure but as a means to reduce circulating elements generated most often through inflammatory processes or electrolyte imbalances created by cardioplegic solutions (4,8). Recently, the use of CUF has come under scrutiny as it may reduce urine output and possibly lead to an increased risk for acute kidney injury (AKI) (9–11). The purpose of this study was to evaluate the influence of ZBUF on urine output of adult patients undergoing cardiac surgery during CPB.

METHODS

Consecutive cardiac surgical patients where CPB was used from the SpecialtyCare Operative Procedure rEgistry (SCOPE^™), between January 2016 and September 2019, at 215 hospitals throughout the United States were reviewed. The SCOPE registry was established in 2011 as a national quality control process for systematically collecting intraoperative data from cardiac surgical procedures, and has been described elsewhere (10,11). It serves a multifunctional purpose focused on quality enhancement and performance improvement. It was designed to achieve the following goals: standardization of electronic data recording of specific perioperative quality indicators, creation of reporting tools including dashboards and written reports, and benchmarking of performance at multiple levels including the clinician, the hospital, and geographical region. National policies and procedures were established using a best-available evidence approach and distributed to all facilities, which were used to guide the conduct of CPB. The system uses a proprietary software application¹ that records demographic and operative data for every procedure. Institutional ethics review board approval² was obtained for the use of data from the SCOPE registry.

Study Design

All surgical patients older than 18 years who underwent a procedure requiring CPB were included. Patients were excluded from the analysis if they met any of the following criteria: anuric, dialysis dependent, had MUF performed at the end of the procedure, did not have all required quality indicators recorded, or had missing data. All UF devices were primed according to the manufacturer’s instructions for use. The attending perfusionist decided the make and model of ultrafiltrator that was used, and this decision was made without any outside influence. The decision to use any type of UF process was made locally by cardiac team members on a case-by-case basis. The decision to use ZBUF was similarly determined by the cardiac team. Two broad groups of patients were established based on the use of ZBUF during CPB and identified as either ZBUF or non-ZBUF. Patients who had concomitant use of CUF were included in either group.

Endpoints

The primary endpoint for the study was intraoperative urine output which was normalized to body weight and procedure duration. Secondary endpoints included the use of asanguineous volumes added by both anesthesia and perfusion personnel.

Statistical Analysis

Descriptive statistics were calculated as count and percentage for categorical variables, and mean and SD for continuous variables, with median and interquartile range used for heavily skewed continuous variables. Unadjusted group differences were assessed using chi-squared tests for categorical data, Welch’s analysis of variance for continuous and approximately normally distributed variables, and the Kruskal–Wallis rank sum test for heavily skewed continuous variables. Descriptive analysis was performed using cases with complete covariate data.

An inferential statistical model was estimated to assess the potential effect of ZBUF, the primary independent variable, on intraoperative urine output while controlling for patient and procedural characteristics, as well as any effects arising from the unique practice patterns of individual surgeons and/or perfusionists. Specifically, we used a linear mixed-effects regression method as implemented in the package “lme4” within the R statistical computing environment (12,13).

To minimize any potential bias attributable to missing covariate data (14–16), we used a multiple imputation method before statistical modeling. In the present analysis, we used the R package “Amelia” which relies on the expectation maximization with bootstrapping algorithm to generate five imputed datasets for modeling (17). Our statistical model was estimated on each of these five datasets with final results reported here combined across models according to “Rubin’s rules” (18) using the R package “merTools” (19). This approach to dealing with missing data has been shown to greatly reduce bias relative to the widespread practice of case-wise deletion, and allows the analyst to preserve important information extant in partial data, while also reflecting uncertainty (through model error estimates) about missing values.

Our final statistical model controlled for the following confounding variables identified elsewhere in the cardiac surgery literature: age, height, weight, patient gender, net priming volume in the extracorporeal circuit, first hematocrit (Hct) upon arrival to the operating room (OR), whether retrograde autologous CPB circuit priming was used, procedure type, asanguineous volume added by perfusion while on CPB, asanguineous volume added by anesthesia, crystalloid cardioplegia volume, ultrafiltrate volume removed intraoperatively, presence or absence of allogenic red blood cell (RBC) transfusion, and CPB duration.³

RESULTS

Descriptive statistics on demographics and procedure type for 73,782 patients with complete data are shown in Table 1 (Figure 1). There were fewer females seen in the ZBUF group. Most of the cases in either group were isolated first-time coronary artery bypass graft (CABG) procedures. Retrograde autologous prime was used more often in non-ZBUF patients who also had slightly higher net prime volumes.

Table 1.

Preoperative and intraoperative characteristics among patients undergoing ZBUF.

	Non-ZBUF	ZBUF	p Value
Number (count [%])	70,309 (95.3)	3,473 (4.7)
Mean weight (SD)	87.8 (19.4)	87.8 (19.4)	.099
Mean height (SD)	1.72 (.10)	1.72 (.10)	<.001
Mean age (SD) (years)	65.6 (11.1)	65.1 (11.1)	.005
Gender: female (%)	20,686 (29.4)	827 (23.8)	<.001
Procedure type (%)
Isolated CABG (count [%])	41,694 (59.3)	1,783 (51.3)	<.001
Aortic surgery (count [%])	1,434 (2.0)	105 (3.0)
AV Surgery + CABG (count [%])	4,988 (7.1)	369 (10.6)
CABG Reoperation (count [%])	689 (1.0)	34 (1.0)
Combined AV/MV surgery (count [%])	1,076 (1.5)	82 (2.4)
Isolated AV surgery (count [%])	7,924 (11.3)	417 (12.0)
Isolated MV surgery (count [%])	4,546 (6.5)	249 (7.2)
MV surgery + CABG (count [%])	1,654 (2.4)	116 (3.3)
Other (count [%])	6,248 (8.9)	318 (9.2)

Figure 1.

Patient flow diagram outlining recruitment and final enrolment.

The use of ZBUF occurred in 4.7% of cases with a median (interquartile range [IQR]) exchange volume of 1,500 (1,000, 2,500) (Table 2), with a range of 0.1–15 L, and 8.5% (297/3,473) of ZBUF patients with exchange volumes >5 L. The use of CUF was almost twice as high in the ZBUF group as compared with non-ZBUF with a UF volume nearly 2.5 times greater as well. In the ZBUF group, all CPB-added asanguineous volumes were nearly double that of the non-ZBUF group, whereas anesthesia volumes trended in the opposite direction. Intraoperative data are shown in Table 2. Both CPB and total intraoperative urine output volumes were higher in the non-ZBUF group with higher urine volumes made by body weight and OR time (Figure 2). Figure 3 gives the distribution across hospitals of the percent of cases within hospitals each that used ZBUF. Approximately 7% of hospitals used ZBUF on 10% or more of their patients, with the remainder of hospitals using ZBUF more sparingly.

Table 2.

Intraoperative characteristics among patients undergoing ZBUF.

	Non-ZBUF	ZBUF	p Value
n	70,309	3,473
Mean first Hct in OR (SD)	36.1 (5.6)	36.4 (5.8)	.010
Retrograde autologous prime used (count [%])	59,970 (85.3)	2,841 (81.8)	<.001
Mean total CPB time (SD)	110.1 (54.1)	122.3 (61.1)	<.001
RBC transfusion (count [%])	12,256 (17.4)	661 (19.0)	.016
Median net prime volume [IQR]	715.0 [600.0, 860.0]	700.0 [575.0, 850.0]	<.001
Median CPB asang. volume [IQR]	259.0 [65.0, 630.0]	650.0 [228.5, 1,271.5]	<.001
Median cardioplegia asang. volume [IQR]	374.0 [58.0, 750.0]	770.0 [200.0, 1,280.0]	<.001
Median anesthesia asang. volume [IQR]	1,500.0 [1,000.0, 2000.0]	1,200.0 [1,000.0, 1800.0]	<.001
Median ZBUF volume [IQR]	.0 [.0, .0]	1,500.0 [1,000.0, 2,500.0]	<.001
Ultrafiltration used (count [%])	31,987 (45.5)	3,086 (88.9)	<.001
Median ultrafiltration volume [IQR]	.0 [.0, 14,000.0]	2,000.0 [1,000.0, 3,000.0]	<.001
Median CPB urine volume [IQR])	220.0 [125.0, 375.0]	210.0 [110.0, 350.0]	.001
Median total urine volume [IQR]	530.0 [350.0, 800.0]	500.0 [300.0, 800.0]	<.001
Median urine volume (mL/kg/h OR time [IQR])	1.12 [.70, 1.73]	.94 [.54, 1.47]	<.001

Asang., asanguineous; Hct, hematocrit; All volumes are reported in mL.

Figure 2.

Intraoperative urine output for ZBUF groups and the influence of increasing exchange volume.

Figure 3.

Distribution of ZBUF across 215 U.S. hospitals.

After adjusting for patient, procedure characteristics, surgeon, and perfusionist practice patterns, we found a negative association between ZBUF volume processed and intraoperative urine output. Each liter of ZBUF volume processed as associated with an average expected change in urine output of −.03 mL/kg/h (95% CI: [−.04 to −.02], p < .001). Figure 4 provides a graphical depiction of this finding in context with other effects estimated in the model. One finding of note was the strong negative effect of weight upon urine output. Increasing weight by one SD (19.4 kg) from its global mean (87.8 kg) was associated with a change in urine output of −.381 mL/kg/h (95% CI: −.387 to −.375), p < .0001.

Figure 4.

Factors affecting intraoperative urine output. AV, aortic valve; CABG; MV, mitral valve; OR; RBC; ZBUF.

Figure 5 depicts the model estimated change urine output as ZBUF volume increase from zero to 14 L for five groups of patients: those at the mean height and weight across a range of values from minus two SDs of weight to plus two SDs from the mean, with body mass index (BMI) values given for reference. These findings draw attention to the additive effects of obesity and ZBUF where large volumes are processed. Full model results are given in tabular format in Table A1.

Figure 5.

Model estimated change in urine output as ZBUF volume increases from zero to 14 L for three groups of patients with a constant mean height but changing weight and body mass index. BMI.

DISCUSSION

The use of UF for patients undergoing cardiac surgery with CPB is a well-accepted technique for hemoconcentration and correcting electrolyte imbalances, and has been in use for nearly 40 years (1,2,20). The term CUF usually denotes the use of UF during CPB as a hemoconcentrating process without the addition of asanguineous solutions, whereas ZBUF is performed in a euvolemic manner with equal volumes of fluid added and removed. In the early 1990s, its use was MUF and extended in the pediatric population as a technique applied in the immediate post-CPB period (3). The benefits of MUF were not only realized in removing plasma water and decreasing transfusion requirements but also in ameliorating inflammation by lowering CPB generated cytokines and by reducing activation of both complement and coagulation systems (6,21). Measures to reduce the systemic inflammatory response to surgery and CPB have been shown to reduce adverse events, lower mortality, and improve outcomes (22). In 1996, Journois and associates described a technique where large volumes of a balanced saline solution were added to the extracorporeal circuit during rewarming while equal volumes of ultrafiltrate were simultaneously removed (ZBUF) (4). They have demonstrated that in pediatric cardiac surgical patients, the use of ZBUF decreased plasma levels of various cytokines (tumor necrosis factor alpha, interleukin (IL)-10, IL-6, IL-8, C3a, and myeloperoxidase), reductions in postoperative blood loss and intubation time, and enhanced pulmonary recovery. They posited that ZBUF reduced inflammation not only by removing inflammatory mediators in the ultrafiltrate but by some additional mechanism. Tallman and associates measured cytokines in the UF volume of 30 CABG and or valve patients randomized to ZBUF or not, and found detectable levels of all pro-inflammatory mediators, except IL-1 (8). Matata et al. (21) performed a prospective randomized controlled trial on ZBUF with a primary focus on its effect on renal function. Although there were no differences in the primary outcome measure of intensive care unit length of stay and 12-month survivability, there was a slight significant difference favoring the ZBUF group in a composite measure for major adverse events (p < .04). Patients with impaired renal function had better outcomes which the authors believe may have been attributed to the removal of toxins that may target the kidneys protecting them from fluid overload and uremic toxicity. However, none of the benefits seen in the early postoperative phase were sustained to 24 hours or beyond.

The use of ZBUF may confer additional benefits beyond inflammatory mediator removal. A recent article conducted during the 2017 sodium bicarbonate shortage in America compared the use of a dialysate solution with normal saline during ZBUF and found no differences in the removal of potassium but a significant decrease in the total sodium bicarbonate requirement between solutions (23). Hyperkalemia is often encountered during CPB as a result of several factors that include the administration of large volumes of potassium-based cardioplegic solution, use of allogeneic RBCs in priming of extracorporeal circuits, extended CPB times, and in patients suffering from preoperative renal insufficiency (24,25). Applying ZBUF to a blood prime in infant cardiac surgical patients has been shown to reduce the metabolic load by significantly reducing sodium, potassium, chloride, glucose, and lactate levels before initialing CPB (26). In a case report in a patient undergoing a prolonged period of circulatory arrest, 10 L of ZBUF was applied over a 60-minute period to reverse a severe metabolic acidosis (27).

How widely the use of UF spread during CPB is difficult to determine. Some describe using UF on all cardiac cases (24), whereas multicenter studies have reported the use around 34% during cardiac surgery (10). In the current study, ZBUF was applied in 4.7% of all cases over a 4-year period. Although we did not assess the reasons for ZBUF use, it is reasonable to expect that individuals used it primarily for its anti-inflammatory benefit and/or for treating hyperkalemia (4,8,22). Approximately 7% of hospitals used ZBUF on 10% or more of their patients, with the remainder of hospitals using ZBUF more sparingly, showing a preferential acceptance by some clinicians. When ZBUF was used, there was a concomitant high use of CUF, with 88.9% of hospitals using it compared with only 45.5% in non-ZBUF facilities. The total ultrafiltrate volume was 2.7 times higher in the ZBUF group, which may indicate either a patient-specific need for hemoconcentration or a more likely employment when an ultrafiltrator was used. This additional volume removed by CUF has been shown to reduce CPB urine output (10,11).

There is no consensus on how any UF process should be conducted during CPB in regard to the rate and volume of removal nor with transmembrane flows or pressure gradients through the filter (28). The rate of ZBUF volume exchange ranges from 1.3 to 10 L/h (23,24,29). In a meta-analysis on the use of ZBUF, the authors describe the UF volume as a function of either weight (10–60 mL/kg), weight per unit time (10 mL/min/kg), or per body surface area (mL/m²) or unit time and body surface area (mL/min/m²) (30). Some have used targeted ZBUF volumes of one to three times the estimated circulating blood volume with the higher volumes in renal compromised patients (31). Using a standard male patient of 80 kg, this would equate to ZBUF volumes between 5 and 15 L. Whereas CPB is most often associated with hydration resulting from the administration of large volumes of asanguineous solutions, the use of UF counteracts this effect by removing plasma water. This has the potential for a dehydrating phenomenon in which the rate of water and solute removal may generate unwanted effects on various organ systems. In an ex vivo model, Fontaine and associates studied the effect of aggressive ZBUF during CPB on brain volume (28). The authors calculated that the osmotic disequilibrium that occurred by solute (urea) removal resulted in an increase in brain size by 59 mL, resulting in cerebral edema, and stated that fluid shifts could result in deleterious effects. We have shown that the use of CUF during CPB results in reduced urine outputs in adult patients undergoing cardiac surgery and there appears to be a procedure-dependent effect with the lowest urine outputs seen when combined coronary and/or valve surgeries were performed (10).

A second study of 40,650 propensity-matched cardiac surgical patients examined the effect of CUF on intraoperative RBC transfusion, and again found urine output to be significantly lower when UF was used (11). Paugh and associates have postulated that in patients at risk of developing renal insufficiency presenting with preoperative creatinine clearances less than 99.6 mL/min, that increasing CUF volumes increased the risk of AKI, which they felt may be related to renal hypoperfusion (9). In their analysis, patients who received ZBUF were excluded from the study. Another plausible explanation is the susceptibility of the kidneys to periods of hypoxia.

Although some studies have used a euvolemic approach to ZBUF by concurrently adding and removing equal volumes of solution by dedicated roller pumps, this is not the usual method used during CPB. More often, the perfusionist is adding asanguineous volume and then removing it through UF at a rate dependent upon both flow and transmembrane pressure across ultrafiltrator. Conducting ZBUF in this manner could result in a temporary decline in oxygen content and delivery by hemodilution and lowered hemoglobin. Such a condition would be volume dependent, and successive reductions in oxygen delivery would occur related to the total exchange volume applied. We have shown using multiple imputation model that patient weight strongly influences urine output independent of confounding factors. Furthermore, if we use height as a constant and increase the patient’s BMI by sequential increases in weight, there is a profound decrease in urine output that is dependent on the total exchange volume of ZBUF as shown in Figure 5. Using this model, a patient with a BMI >36 would have reduced urine outputs <1 mL/kg/h, and ZBUF would further depress urine in a volume-dependent manner. Individuals with larger BMIs who may be more susceptible to AKI would be at greater risk of developing it with the use of ZBUF. There is an inverse relationship between ZBUF volume and urine output, with increasing volumes of ZBUF resulting in lower urine outputs that is seen across all modeled BMI values.

In a prospective randomized study of CUF in adult cardiac surgical patients, Kuntz and associates found no differences in CPB or OR urine volumes, but non-CUF patients had significantly higher 24-hour urine than CUF patients (32). The median volume removed by CUF was 5,560 mL, with a range of 1,900 mL to 13,220 mL, which reflected the historically large volumes of intraoperative asanguineous solutions in use during that time, which were nearly four times greater than that seen in the present study. Although hypovolemia resulting in dehydration is known to alter hemodynamics by reducing circulating blood volume and inducing hypotension (33), in cardiac surgery, this is less likely to occur because of the continuous monitoring of filling and systemic pressure. However, during CPB, the effects of the normal physiologic response to hypovolemia are masked by the conduct of extracorporeal circulation where blood flow is dependent on the perfusionist interventions, and blood pressure is dependent on the flow rate, vascular tone, and the administration of vasoactive substances (34). In renally compromised patients undergoing hemodialysis, the effects of rapid UF (greater than 13 mL/kg/h) have been shown to increase all-cause morbidity including death when compared with patients <10 mL/kg/h (35). Whereas it is tempting to identify a similar effect with the use of intermediate UF rates, we excluded anuric or dialysis-dependent patients from analysis, so any comparison would be inappropriate.

Diagnosing AKI is performed in a number of ways, with the most often techniques measuring creatinine levels, calculating glomerular filtration rates, or by measuring urine output where values < .3 mL/kg/h, or anuria for >12 hours, are confirmatory (36). Although the latter two methods are most often reported, the use of urine output is simple and often more feasible, especially in large multicenter analyses. Urine output in non-anuric patients undergoing cardiac surgery is inversely associated with renal function, which makes its assessment an attractive indicator for the early detection of AKI (37). Early detection of oliguria could be used to use measures shown to reduce the risk of AKI. Increased urine output may not indicate adequate renal function because damage to the tubular structures as a result of inflammation from central diabetes insipidus may also result in high urine volumes (38). Recently, it has been shown that urine output during CPB may be an important predictor of AKI and is especially attractive because of it being easily documented (39,40). In a retrospective study of nearly 700 cardiac surgical patients, Song and associates identified a biphasic relationship of urine output and AKI where non-CPB urine rates of > or <4 mL/kg/h independently predicting AKI (39). A second study in CABG patients evaluating CPB urine output as a predictor of postoperative AKI found a strong correlation with injury when CPB urine outputs dropped <3.7 mL/kg/h (40). Both of these studies are well above an arbitrary threshold of CPB urine production that targets 1–2 mL/kg/h as goals. Both authors stated the higher rates of urine output necessary to reduce the risk of injury taking into account the use of pro-diuretic mechanisms (use of diuretic drugs, elevated blood flows) during CPB.

AKI is often assessed by using the RIFLE criteria (risk of renal dysfunction, injury to the kidney, failure of kidney function, loss of kidney function, and end-stage kidney disease), which assesses elevations in serum creatinine (sCr) concentration, reductions in glomerular filtration rate, and urine output of <.3 mL/kg/h, or anuria for >12 hours (36). A recent report from the 20th International Conference of the Acute Disease Quality Initiative identified criteria for defining AKI in cardiac and vascular surgery patients by using both sCr and urine output (41). However, when ZBUF is used, sCr levels in the immediate postoperative period are significantly lowered, but this may not be related to filtration because the molecular weight of creatinine is 113 kDa, which is well above the cutoff value of 30 kDa (21). These authors felt that the reduction in sCr was more than likely because of the surface adsorption through membrane interactions rather than through UF. Although we did not measure chemical markers of renal function, the use of intraoperative urine output would serve as an early marker for risk of developing AKI (37,39,40). After controlling for multiple factors, we found a reduction of .034 mL in urine output for each 1 L of ZBUF exchange. Although this reduced rate is modest, in situations where large volumes of ZBUF are exchanged (greater than 10 L), it is not inconceivable that in high-risk patients, the lowered volume of urine output may increase the risk of AKI.

There were a number of intergroup statistically significant differences in both the use of retrograde autologous priming of the circuit and the administration of asanguineous solutions (Table 2). Although several of these differences may not be clinically significant (use of retrograde autologous prime, net prime volume, and anesthesia asanguineous volume), the volume of asanguineous solutions added during CPB was nearly twice as high as that in the ZBUF group. The CUF volume was more than 2.5 times higher in the ZBUF group, which may indicate a likelihood for higher asanguineous volume administration if the clinician is planning on using UF.

Very rarely is patient urine output reported in publications on ZBUF. In a randomized study in adult cardiac surgical patients comparing subzero-balanced UF, which is described as a combination of both CUF and ZBUF, the authors reported lower volumes of total urine in the UF group although significance was not met (1,465.2 ± 378.9 vs. 865.5 ± 577.2, p < .065) (5). Filtration rates during CPB varied depending on the state of the procedure (during cross clamp 10–20 mL/kg and for rewarming 50–100 mL/kg), with a total CUF average volume of 3.5 L. Although there was significantly lower combined hospital morbidity in the treatment group (p = .025), the average total urine output was 600 mL less than that in the control group (p = .065) (5). There were no differences in postoperative acute renal failure. The randomized trial by Matata et al. (21) reported slightly higher nonsignificant median intraoperative urine volumes in non-ZBUF patients (450 vs. 400 mL, p = .560) and no difference in ICU total urine volumes. In a recent randomized study of 33 CABG patients, Friedrich et al. (42)used ZBUF with a dialysate solution performed during the entire CPB period, and then continued postoperatively for an additional 6 hours via a temporary dialysis catheter. They performed urine output measurements every 20 minutes during the procedure and at 4 hours postoperatively, comparing treated patients with a control group. They reported higher urine outputs during CPB in the treatment group, but not after CPB. They did not report or normalize the volumes by body weight or time on CPB, so it is difficult to relate these outcomes to the current study or to others.

The type of hydration fluid used for the exchange process during ZBUF may confer a benefit by reducing the deleterious effects of normal saline such as hypernatremia and hyperchloremia, which may lead to an iatrogenic metabolic acidosis (43,44). Whereas a balanced electrolyte solution is often used (5,8,29,31), others have used dialysate solutions (23,24,42), or a combination of saline and balanced solutions (26). In the present study, we did not confirm with the sites what type of exchange fluid was used so cannot confirm that the type of exchange solution would confer an effect on urine output.

Limitations

The present study has limitations. This study was conducted using a national registry of data collected in a prospective, but non-randomized manner. Registry data do not permit the investigation of certain factors that may be pertinent in determining effects not found with limited variable analysis. We did not measure chemical measures of AKI that may have generated evidence in support of renal dysfunction. Although database analysis is useful for identifying trends, the absence of physiological determinants of AKI precludes a definitive outcome of UF on renal function. Ultrafiltration guidelines were not standardized across and within individual hospitals, so the application of this technology may have been biased by clinical decisions. The retrospective study design is subject to limitations of inherent selection bias, and the reported results are limited to describe observed associations between the implementation of the described protocol and the improved patient outcomes, and do not demonstrate a direct cause-and-effect relationship. All results are limited to short-term intraoperative outcomes, and intermediate or long-term follow-up data were not available. This was not a longitudinal study, so the effect of ZBUF on postoperative AKI cannot be ascertained. And finally, there exists a potential for the miscoding of data, which, despite steps for validation, must be considered in any secondary analysis of registry data.

CONCLUSIONS

The use of registry data from more than 200 hospitals across America has shown that the use of ZBUF is associated with reduced intraoperative urine output in cardiac surgical patients that supports our previous findings when CUF was examined. Although the reductions in urine output were modest, they are related to the total volume of exchange fluid and may affect renal injury when large volumes are used, especially in patients sensitive to acute injury. Further research is necessary to confirm these findings in prospective evaluations.

Appendix

Table A1.

Multiple imputation method prior to statistical modeling.

Model Term	Estimate and 95% CI	p Value
Intercept†	1.343 (1.306–1.38)	p < .0001
ZBUF volume (L)	−.034 (−.043 to −–.024)	p < .0001
Weight*	−.381 (−.387 to −.375)	p < .0001
Height*	.027 (.02–.035)	p < .001
Age*	−.044 (−.05 to −.039)	p < .0001
Female	.097 (.081–.113)	p < .0001
First hematocrit in OR*	.057 (.051–.064)	p < .0001
Aortic surgery	.043 (.01–.076)	p < .05
AV surgery + CABG	.072 (.049–.095)	p < .0001
CABG reoperation	.029 (−.023–.082)	p = .2753
Combined AV/MV surgery	.11 (.064–.156)	p < .0001
Isolated AV surgery	.165 (.145–.184)	p < .0001
Isolated MV surgery	.113 (.088–.139)	p < .0001
MV surgery + CABG	−.024 (−.061–.012)	p = .1902
Other procedure	.065 (.044–.087)	p < .0001
Time on CPB*	−.067 (−.075 to −.06)	p < .0001
Anesthesia asanguineous volume*	.076 (.07–.083)	p < .0001
Crystalloid cardioplegia volume*	.067 (.059–.074)	p < .0001
Net CPB prime volume*	.006 (−.001–.014)	p = .1006
Perfusion asanguineous volume*	.102 (.095–.108)	p < .0001
Retrograde autologous priming used	.064 (.044–.083)	p < .0001
Allogenic RBC transfusion	−.082 (−.099 to −.066)	p < .0001
Ultrafiltration volume (L)	−.104 (−.11 to −.099)	p < .0001
Grouping variable	n	Percent of variation in individual outcome due to factors at this level
Surgeon	936	10.9%
Perfusionist	728	6.6%

* Model terms listed with an asterisk have been centered and scaled before model estimation, and thus should be interpreted as the effect on intraoperative urine output given a one-SD increase in the term of interest from its mean. See Table A2 for these means and SDs.

† The intercept may be interpreted as follows: the average expected intraoperative urine output for a male undergoing first-time CABG surgery without the use of ZBUF, allogenic RBC transfusion, or retrograde autologous priming, and with all other quantitative variables set to their respective means.

Table A2.

Model term and entire sample mean and SD.

Model Term	Entire Sample Mean and SD
Height (m)	1.72 (.10)
Weight (kg)	87.67 (19.40)
Age (years)	65.38 (11.21)
Time on CPB (minutes)	112.91 (56.83)
Perfusion asanguineous volume (mL)	522.12 (654.12)
Ultrafiltration volume (L)	.92 (1.33)
Net CPB prime volume (mL)	749.57 (260.20)
Crystalloid cardioplegia volume (mL)	492.16 (522.48)
Anesthesia asanguineous volume (mL)	1,548.85 (744.15)
First hematocrit in OR (%)	36.12 (5.73)