Abstract
Objective:
Ultrafiltration (UF) during cardiopulmonary bypass (CPB) is a well-accepted method for hemoconcentration to reduce excess fluid and increase hematocrit, platelet count and plasma constituents. The efficacy of this technique may confer specific benefit to certain patients presenting with acquired cardiac defects. The purpose of this study was to retrospectively evaluate the effect of UF on end-CPB hematocrit by cardiac surgical procedure type.
Methods:
A review of 73,506 cardiac procedures from a national registry (SCOPE) was conducted between April 2012 and October 2016 at 197 institutions. Cases included in this analysis were those completed without intraoperative red blood cell transfusion and where zero-balance UF was not used. The primary end point was the last hematocrit reading taken before the end of CPB, with a secondary end point of urine output during CPB. In order to isolate the effect of the UF volume removed, we controlled for a number of confounding factors, including: first hematocrit on CPB, total asanguineous volume, estimated circulating blood volume, CPB urine output, total volume of crystalloid cardioplegia, total volume of other asanguineous fluids administered by both perfusion and anesthesia, type of cardiac procedure, acuity, gender, age and total time on CPB. Descriptive statistics were calculated among five subgroups according to the UF volume removed: no volume removed and quartiles across the range of UF volume removed. The effect of UF volume on primary and secondary end points was modeled using ordinary least squares and restricted cubic splines in order to assess possible non-linearity in the effect of the UF volume while controlling for the above-named confounding factors. An interaction term was included in each model to account for possible differences by procedure type.
Results:
The study found a statistically significant non-linear pattern in the relationship between the UF volume removed and the last hematocrit on bypass (X2 = 172.5, df=24, p<0.001). For most procedure types, UF was most effective at increasing the last hematocrit on CPB, from 1 mL to approximately 2.5 L, with continued improvements in hematocrit coming more slowly as the UF volume was increased above 2.5 L. There were statistically significant interactions between UF and procedure type (X2 = 78.5, df=24, p<0.0001) as well as UF and starting hematocrit on CPB (X2 = 234.0, df=4, p<0.0001). In a secondary end-point model, there was a statistically significant relationship between the ultrafiltration volume removed and urine output on bypass (X2 = 598.9, df=28, p<0.001).
Conclusion:
The use of UF during CPB resulted in significant increases in end-hematocrit, with the greatest benefit shown when volumes were under 2.5 L. We saw a positive linear benefit up to 2.5 L removed and, thereafter, in most procedures, the benefit leveled off. However, of note is markedly decreased urine output on bypass as the ultrafiltration volumes increase.
Introduction
Excessive hemodilution during cardiopulmonary bypass (CPB) procedures due to various component priming and exposure of blood cells to the circuit begins an inflammatory response, characterized by capillary leak and increased total body water, that has been known to cause postoperative organ dysfunction. Furthermore, it provokes the inflammatory responses associated with ischemia-reperfusion injury from cardioplegia solutions delivered directly to the heart. Cytokines and other inflammatory mediators released during CPB contribute to the development of reperfusion injury and myocardial dysfunction and may lead to acute respiratory distress syndrome and multi-organ failure. In some pathologic states, including acute myocardial infarction or reperfusion during CPB, there is an accumulation of plasma mediators that are capable of activating polymorphonuclear neutrophils (PMNs) and endothelial cells. 1 Hemodilutional anemia is a consequence of CPB, due to the asanguineous prime of the circuit. The degree of hemodilutional anemia is related to the patient’s initial hematocrit (Hct) plus the priming volume of the circuit. Variability among clinicians as to what is the optimum Hct during CPB should be and is often weighed against postoperative morbidity and mortality. 2
Many techniques have been developed that boast reduction of hemodilution and its untoward effects, such as: off-pump procedures, minimally invasive approaches requiring shorter CPB time and utilizing mini-extracorporeal circulation (MECC). Other similar techniques have been used to reduce the crystalloid load of patients undergoing CPB and include retrograde autologous priming (RAP), intraoperative conventional ultrafiltration (CUF), during CPB hemoconcentrating volume with zero-balance ultrafiltration (ZBUF) and the post-CPB use of ultrafiltration (UF), either of the CPB circuit contents or by modified ultrafiltration (MUF). 3 The purpose of this study was to evaluate the effect of UF on end-CPB hematocrit by cardiac surgical procedure type.
Methods
This is a review of 73,506 cardiac procedures from a national registry, the
Descriptive statistics first were calculated among five subgroups according to the UF volume removed: no volume removed, less than 1.0 L removed, between 1.0 and 1.5 L removed, between 1.5 L and 2.4 L removed and 2.4 L or more removed. A second set of descriptive statistics were calculated according to the procedure type of each case. Normally distributed continuous variables are reported as mean and standard deviation, with Welch’s ANOVA used to assess possible group differences. Heavily skewed continuous variables are reported as median and interquartile range, with the Kruskal-Wallis rank sum test to detect group differences. Categorical variables are reported as count and percentage with Chi-squared tests being used to determine the statistical significance of group differences.
The effect of UF volume on the primary end point, the last hematocrit on CPB, was modeled using ordinary least squares and restricted cubic splines in order to assess possible non-linearity in the effect of UF volume while controlling for the above-named confounding factors. Ultrafiltration volume and all other continuous independent variables were modeled using restricted cubic splines with 5 knots. Among a subset of all cases where ultrafiltration was used, a second least squares model was estimated to assess the secondary end point, urine output during CPB, using these same control variables and spline modeling techniques. Models for both the primary and secondary end points included an interaction term to assess potential differences among procedure types in the effects of ultrafiltration.
Cases with missing data were excluded from the regression analysis. Analyses were completed using the R statistical computing environment 4 in conjunction with the tableone 5 , rms 6 , and ggplot2 7 packages.
Results
Tables 1 and 2 give descriptive statistics by ultrafiltration volume removed and procedure type, respectively. Of the 73,506 patients in our sample, 61.9% (45,471) had no ultrafiltration volume removed, 9.3% (6,812) had between 1 mL and 1.0 L removed, 8.0% (5,885) had between 1.0 L and 1.5 L removed, 10.9% (8,039) had between 1.5 L and 2.4 L removed and 9.9% (7,299) had 2.4 L or more removed. The mean final Hct on CPB was highest among those in the 2.4 L or more group (28.9), followed by the 1.5 L to 2.4 L group (28.2), the 1.0 L to 1.5 L group (27.6), the less than 1.0 L group (27.1) and, finally, the no volume removed group (27.0, p<0.001). The assessment of the ultrafiltration volume removed as a function of estimated blood volume yielded a ratio of ultrafiltrate to estimated blood volume. The median of this ratio for patients with non-zero ultrafiltrate volume was 0.296 (interquartile range: 0.176 − 0.458). Group differences in the first Hct on CPB were statistically significant, but very small. Group differences on the median crystalloid cardioplegia solution administered were rather large, ranging from 0.27 L in the 1 mL to 1.0 L removed group to 0.74 L in the more than 2.4 L group (p<0.001), while anesthesia volume administered and volume added on bypass showed rather small, but statistically significant group differences that did not follow the same monotonic trends described above.
Descriptive statistics by ultrafiltration volume removed.
Mean and standard deviation; global test of group differences assessed with Welch’s ANOVA. B Median and inter-quartile range; global test of group differences with Kruskal-Wallis rank sum test. C Count with column-wise percentage in parentheses; global test of group differences assessed with Chi-squared test.
Descriptive statistics by cardiac procedure type.
Count with column-wise percentage in parentheses; global test of group differences assessed with Chi-squared test. B Median and inter-quartile range; global test of group differences with Kruskal-Wallis rank sum test. C Mean and standard deviation; global test of group differences assessed with Welch’s anova.
A total of 65,617 cases had complete data on all variables and were, thus, included in our statistical model for the primary outcome, the last Hct on bypass (Table 3). After controlling for all confounding variables listed in Table 1, we found a statistically significant non-linear pattern in the relationship between UF volume removed and last Hct on bypass (χ2 = 172.5, df=24, p<0.0001). For most procedure types, UF was most effective at increasing the last Hct on CPB from 1 mL to about 2500 mL, with continued improvements in Hct progressing more slowly as the UF volume was increased above 2500 mL. While we did note statistically significant interactions between UF and procedure type (Figure 1, χ2 = 78.5, df=24, p<0.0001) as well as UF and starting Hct on CPB (χ2 = 234.0, df=4, p<0.001), by far the most powerful explanatory variable in this analysis was the starting Hct on bypass (χ2 = 112,572.3, df=5, p<0.001).
Independent predictors of final Hct on CPB.
Model-adjusted last hematocrit on CPB by procedure type and volume of ultrafiltration removed.
The statistical model for the secondary outcome, urine output on bypass, included a total of 25,041 cases where ultrafiltration was used and where data were complete for all covariates (Table 4). After controlling for all covariates given in Table 1 (save for urine output which was the dependent variable in this model) a statistically significant relationship was revealed between the ultrafiltration volume removed and urine output on bypass (χ2 = 598.9, df=28, p<0.001). Unlike the model for the primary outcome, model terms for the secondary outcome assessing possible non-linearity in the ultrafiltration and urine output relationship were non-significant, as was the interaction term assessing differential effects of ultrafiltration on urine output by procedure type.
Independent predictors of urine output on CPB.
Discussion
The primary principle governing UF is the selective separation of plasma water and low molecular-weight solutes from the intravascular cellular components and plasma proteins of blood, using a semipermeable membrane filter. Contrasting hemodialysis, the UF technique is simplified by eliminating the requirement for a dialysate solution. Rather than solute osmotic pressure, the driving force for UF is provided by the hydrostatic pressure differential occurring across the UF membrane. The application of a negative pressure on the effluent side of the membrane or the use of an increased perfusion pressure applied to the blood side of the membrane results in improved solute and fluid filtration. 8
One obvious reason to utilize UF during CPB is to reduce the need for blood transfusion, which has been reported to increase mortality. Cardiac surgery remains one of the greatest consumers of blood products. Many studies report the benefit of UF; however, this technique is used less frequently in adult cardiac surgery. 9 In this analysis, just 34% of the total patients were ultrafiltered (n=29,277). Currently, most perfusionists incorporate some form of free water removal in their clinical practice to eliminate priming volume and diminish the opportunity for capillary leak and attenuate whole body inflammatory response. UF, RAP, MUF and, even, autologous blood removal are techniques with proposed benefits of: increased Hct, reduction in circulating mediators, improved cardiovascular performance and improved neurologic outcome.3,10
Quite possibly, the use of UF during CPB could be injurious, based on an observation study to investigate UF and the risk of acute kidney injury (AKI), which showed exposure to UF led to a higher adjusted risk of AKI. Paugh et al. proposed that a potential mechanism between UF and AKI would be the excessive amount of fluid removed, resulting in renal hypoperfusion. 11 Renal hypoperfusion is the main factor causing acute renal failure after cardiac surgery. Since urine output is used as an indicator for renal perfusion during CPB and is used as a guide for perfusion management, it is an early indicator of renal injury, in theory, as discussed by Song et al. who showed an independent association between the amount of CPB urine output and postoperative AKI. 11 Some controversy exits between the Paugh study and Kuntz randomized trial which warrants further investigation in this area. Our study looks at a second end point of urine output on CPB and notes markedly decreased urine output on bypass as ultrafiltration volumes increase. It is unknown what effect this may have on AKI.
By virtue of the technique of hemoconcentration of particles during UF, although there exists numerous studies comparing UF in terms of inflammatory mediators, often with equivocal or contradictory results, is it possible that we may be doing more harm than good? Torina, et al. also found an increased inflammatory response with the use of UF in adults undergoing coronary bypass grafting while, in another study, Soliman, et al. found an increase in serum lactate levels after UF during CPB.12,13 Nonetheless. findings in another study by Tamari et al. state that UF also increases other blood components, including hemoglobin and free hemoglobin molecules, which would make it an effective ultrafilter to control hemodilution during CPB. 14 There are many thoughts about the ideal means to manage patients on CPB and the vast majority survive with little incidence of harm. 3 Regardless of the performance of these techniques with respect to outcome, we need to further investigate their use.
One most important focus today during CPB is the desire for keeping the patient Hct level >(21%), ideally (24-26%), since delivery of oxygen (DO2) is diminished as the Hct decreases.3,15,16 According to Ranucci et al., the pump flow should be increased when the Hct falls on CPB to increase oxygen delivery. 15 Oxygen delivery should be higher than 270 to 280 mL/min/m2 and not just the measurement of Hct alone.
Some perfusionists combine more than one blood conservation technique; for example, where the patient is exposed to both RAP and UF during CPB. In doing so, the clinician needs to weigh the safety of maintaining an appropriate blood level in their reservoir vs removing free water to enhance DO2. Low pump flow rates caused by inadequate venous reservoir volume, decreased SvO2 and low mean arterial pressure may ensue if too much volume is removed. In addition, the process of UF may lead to hypovolemia, with increased osmolarity in the intravascular volume causing interstitial fluid to slowly shift into the vascular space. 17 As you continue the process of UF, total body water will be diminished upon entry to the intensive care unit (ICU).
Sahoo, et al. provide evidence that the practice of cardiac surgery is seeing an increasing number of more elderly and acute patients present for CABG surgery as well as valvular surgeries. 1 These patients are expected to have higher morbidity and mortality due to the higher incidence of perioperative myocardial infarction, acute respiratory distress syndrome and low cardiac output syndrome. The beneficial effects of UF procedures in terms of improved hemodynamics, hemostasis and pulmonary functions have been documented in selected high-risk adult patients. The patients with valvular and combined procedures, (Figure 2) reached a level >2.5 L in this analysis and even after 5 L of fluid was removed, the change in Hct was only 2% higher than the mean for all groups. Additional hemoconcentration and fluid removal without additional change in hematocrit may be explained by a fluid shift from interstitial to vascular space due to increased oncotic pressures. Disease processes cause changes in fluid dynamics and we did notice that those patients who had multiple surgeries and valvular disease had >2.5 L removed with UF, although there was no more positive beneficial effect in rise in Hct. We would suggest that further exploration of ultrafiltration and its effect is warranted.
Model-adjusted urine output on CPB by volume of ultrafiltration removed.
The limitations to this large, multi-institutional, retrospective review are the lack of outcome data from the SCOPE registry. Also, the degree of hydration of these individuals or knowledge of renal status was not evaluated. The observation of UF volume revealed differences with regard to procedure type that cannot easily be explained.
Conclusion
The use of UF during CPB resulted in significant increases in end-hematocrit, with the greatest benefit shown when volumes were under 2.5 L. A positive linear benefit was found up to 2.5 L removed and, thereafter, in most procedures, the benefit leveled off.
Footnotes
Presented at the 38th Annual Seminar of The American Academy of Cardiovascular Perfusion, San Diego, California, 19-22 January 2017.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article:
Discussion: The effect of ultrafiltration on end-cardiopulmonary bypass hematocrit during cardiac surgery
Mr. DENNIS WIEMER (Leesburg, GA): When you refer to a urinary output of 4 mL/kg/hr, is that affected by medication? Some have suggested that you should avoid the use of Lasix or mannitol as much as possible while still maintaining adequate urinary output. By administering that medication, do you think it has anything to do with an increase in acute kidney injury?
MS. LINDA MONGERO (Locust Valley, NY): I am advocating that more clear fluid should pass through the kidneys than to ultrafiltrate it. Years ago, we used to always give Lasix and never did ultrafiltration. The clear fluid was gone by the end of the case. We used autotransfusion. That is how we did it. I do not have any reference to confirm that acute kidney injury then was different from now, but it is getting a lot of attention. This is just something that we found in our research and I thought it was very interesting. Now I am questioning whether or not it is efficacious to use.
MR. JOSHUA WALKER (San Antonio, TX): Anytime you would like to see more urinary output, anesthesia personnel tend to give Lasix and that is probably not the correct approach. If you are having to force those kidneys to produce more urinary output, you ought to increase the pump flow and/or increase the arterial blood pressure. There is definitely a nadir of oxygen delivery, which is around 270 mL/min, as you mentioned. I use an Excel spreadsheet, updated during cases, that I will often times look at quickly. It has two columns: one column has oxygen delivery at that time so you get a blood gas, quickly enter your values and calculate what the oxygen delivery is based on at that one moment. The other column is a minimum value. So, if I chose a minimum of 280 mL/min and the patient’s hemoglobin is 8 mg/dL, then the minimum flow needs to be “Y”. So, anytime the online blood gas monitor (CDI) estimates what the hemoglobin/hematocrit is, I can adjust my pump flow in real time, based on what I am seeing coming back through the CDI.
This discussion is taken from the dialogue that followed the presentation of the previous paper at the 38th Annual Seminar of the American Academy of Cardiovascular Perfusion. Although the paper has been through Perfusion’s stringent peer-review process, the discussion is a transcript of the dialogue, edited for clarity, and the views expressed in the discussion are those of the commentators and do not necessarily represent, and should not be attributed to, the journal Perfusion, the Editors, authors or the Publisher, SAGE.