Ultrafiltration in cardiac surgery: Results of a systematic review and meta-analysis

Introduction

Given that cardiac surgery traditionally utilizes approximately 10–15% of the nation’s blood supply, it is critically important to examine ways to optimize transfusion practice. Patient blood management (PBM) programs involve the systematic, evidence-based approach to allocating blood products by promoting proven blood conservation techniques. ¹ Cardiopulmonary bypass (CPB) is necessary for most cardiac surgical procedures, but may contribute to coagulopathy, hemodilution, and activation of systemic inflammatory mediators, all of which can potentiate the likelihood of blood product administration. ² In response, PBM programs have adopted a variety of techniques which may be applied to CPB to mitigate these effects, including the use of ultrafiltration.

Ultrafiltration involves the filtration of blood through a semipermeable membrane in an attempt to both remove fluid and concentrate certain blood constituents, including hemoglobin (i.e., hemoconcentration) and platelets. ³ Approaches include conventional ultrafiltration (CUF), which occurs during CPB with the hemofilter maintained in parallel with the bypass circuit, or lesser employed modified ultrafiltration (MUF), which occurs after the patient has separated from CPB. With MUF, blood is removed from the patient through the aortic cannula, hemofiltered, and returned via a separate venous cannula. ⁴ In either case, part of the intent is to increase hemoglobin concentration and potentially reduce transfusion.

Over the past 25 years, there have been numerous trials examining the association between CUF and MUF and perioperative blood transfusions. Some studies have also demonstrated risks of end organ injury, specifically acute kidney injury (AKI), when large volumes of ultrafiltrate are removed.^4–6 Whereas ultrafiltration has been included among a number of potential blood conservation strategies in various consensus statements, grading has been variable and the level of recommendation is potentially subject to revision given the publication of several recent studies.^7–9 To that end, we have performed a systematic review and meta-analysis to determine the association between CUF or MUF and the primary outcome of intraoperative red cell transfusion. We hypothesized that CUF and MUF would be associated with reduced intraoperative red cell transfusions.

Methods

Results

Literature identification

The initial search resulted in 3251 abstracts (820 from PubMed, 1278 from Embase, 914 from Web of Science, and 234 from Cochrane Library) (Figure 1). After removing 1324 duplicate studies, 1927 potentially relevant articles were screened on basis of the abstract. After removing 1862 abstracts that did not meet inclusion criteria, the remaining 65 full text articles were thoroughly reviewed. An additional 55 articles were excluded on the following basis: 1) no data on the primary outcome of interest, 2) transfusion outcome with incomparable units, 3) pediatric patient population, 4) intervention other than MUF or CUF, 5) meta-analysis or systematic review, 6) combined use of MUF and CUF in study group, and 7) lack of non-ultrafiltration control group. Nine studies^{3,5,10,13–19} (n = 47,985 patients) were included in the final analysis. There were 7 RCTs (n = 928)^3,10,13–17 and 2 observational studies (n = 47,007)^5,18 included in the analysis. One observational study was removed during analysis because it only reported raw data and we analyzed adjusted odds ratios with standard errors. ¹⁹

OPEN IN VIEWER

Figure 1.

Outline of literature search strategy yielding 10 publications. Seven studies were randomized controlled trials (n = 928 patients) comparing modified ultrafiltration to control. Three studies were observational studies (n = 47,175 patients) comparing conventional ultrafiltration to control.

Study characteristics

Table 1 summarizes the study characteristics including patient numbers, patient demographics, ultrafiltration type and volume, surgery type, and other blood conservation interventions mentioned in methods. Among the 9 included studies, there is large variability in ultrafiltration volume. Of note, 6 included studies did not report mean weight or body surface area of study groups.^{5,10,13,15–17} Supplemental Figure 2 illustrates the Cochrane Risk of Bias 2.0 reporting for all included RCT studies. Two of 7 studies demonstrated an overall high risk of bias due to randomization process and deviations from intended interventions. ¹³ One study had uncertain risk of bias due to uncertain randomization process and how missing outcome data was addressed. ¹⁰ Supplementary Figure 3 summarizes the Risk of Bias Assessment tool for Non-randomized Studies (RoBANS) for all included observational studies. One of 3 studies had high risk of bias for selection of participants since it was a before/after study. ¹⁹ The same study had high risk of bias for confounding (major confounding variables were not adequately considered during the design and analysis phases) and thus, was excluded from final analysis. ¹⁹

OPEN IN VIEWER

Table 1.

Characteristics of included studies.

Author, year [Ref]	UF type	UF (n)	Control (n)	UF age (years)	Control age (years)	UF female (%)	Control female (%)	UF weight (kg)/BSA (m2)	Control weight (kg) BSA (m2)	UF volume (mL)	UF CPB time (min)	Control CPB time (min)	Surgery type	Other interventions
Boodhwani, 2010 [4] a	MUF	29	36	66.3 ± 13.1	71.1 ± 10.1	65.5	77.8	2.21 ± 2.08	2.47 ± 4.19	1000 ± 21	88 ± 29	95 ± 33	Elective cardiac surgeries	TXA, antiplatelet/antithrombotics stopped
Grunefelder, 2000 [12]a,b	MUF	30	20	62.1 c	62.1 c	*	*	*	*	*	77 ± 17 d	77 ± 17 d	Elective CABG	Normothermic (37°C) bypass
Grunefelder, 2000 [12]a,b	MUF	30	17	58.7 c	58.7 c	*	*	*	*	*	98 ± 23 d	98 ± 23 d	Elective CABG	Hypothermic (26-28°C) bypass
Luciani, 2001 [13] a	MUF	284	289	63.8 ± 10.6	62.6 ± 13.5	34.2	28.7	72.9 ± 12.1	72.5 ± 12.0	—	111.3 ± 39.2	108.7 ± 40.6	CABG,AVR,MVR, AVR + MVR,AVR + CABG, MVR + CABG, aortic root, ascending aorta, heart txplant	Bypass circuit primed with RBCs to maintain Hct >20; cell salvage
Mongero, 2019 [17]	CUF	20,325	20,325	67.0 (58.0, 74.0) e	66.1 (59.0,74.0) e	31.0	31.4	2.0 (1.8,2.1) e	2.0 (1.8,2.1) e	1300 (800,2000) e	120.7 ± 56.7	115.6 ± 53.3	CABG, aortic surgery, AVR + CABG,AVR + MVR, AVR,MVR,MVR + CABG	Retrograde autologous priming
Paugh, 2015 [18]	CUF	1,362	5,045	* f	* f	24.7	24.7	* g	* g	1365 ± 1202	96.1 ± 42.0	96.0 ± 39.0	Isolated CABG	*
Raman, 2003 [19] h	CUF	61	57	68.4 ± 11.8	65.3 ± 12.9	34.4	22.8	*	*	3400 ± 1921	149.0 ± 71.1	142.3 ± 52.3	High risk surgery i	*
Torina, 2010 [14] a	MUF	20	17	54.9 ± 7.4	56.1 ± 9.8	20.0	0	*	*	1000 ± 155.6	52 ± 10	58 ± 10	Elective CABG	*
Torina, 2012 [15] a	MUF	30	30	55.2 ± 7.3	55.6 ± 9.3	20.0	7.0	*	*	995 ± 160	51 ± 10	56 ± 8	Elective CABG	Hb trigger 9.0 g/dL
Zahoor, 2007 [9] a	MUF	50	46	50.7 ± 14.7	48.7 ± 11.2	30.0	13.0	*	*	*	95.42 ± 27	78.0 ± 21	CABG, AVR, MVR, AVR/MVR.redo MVR, CABG/AVR;congenital	Mannitol (2 mL/kg)

UF: Ultrafiltration; CPB: Cardiopulmonary bypass; MUF: Modified ultrafilration; CUF: Conventional ultrafiltration; TXA: Tranexamic acid; CABG: Coronary artery bypass grafting; AVR: Aortic valve replacement; MVR: Mitral valve repair/replacement; heart txplant: Heart transplant; RBCs: Red blood cells; Hb: Hemoglobin. Other interventions column represents interventions for blood conservation specifically mentioned in methodology section.

^aRandomized controlled trials;

^b2 different interventional arms of same randomized trials (normothermic 37°C MUF compared to controls and hypothermic 26–28°C compared to controls);

^cMean age of normothermic/hypothermic MUF interventional arm and control arm with no standard deviation reported;

^dMean CPB time of normothermic/hypothermic MUF interventional arm and control arm;

^eMedian, interquartile range reported;

^fAge for intervention and control arms reported in sub-categories: <60 yrs, 60–69 yrs, and 70 + years, p = 0.48 between groups at each age sub-category;

^gBody surface area for intervention and control arms reported in sub-categories: <1.60, 1.60–1.79, 1.80–1.99, ≥2.00, p = 0.18 between groups at each age sub-category;

^hUsing Parsonnet’s scoring, the mortality risk for intervention and control >20%;

ⁱRevision CABG, emergency CABG, CABG + valve, urgent valvular, infective endocarditis/valvular replacement, Ross procedure, left ventricular reconstruction.

Primary outcomes

Based on the pooled effects of included trials, MUF was associated with a reduced mean intraoperative red cell transfusion per patient (Figure 2; number of comparisons reporting outcome [n = 7]; MD −0.73; 95% CI −1.12 to −0.35 p = 0.04; p for heterogeneity = 0.0001, I ² = 55%) compared to controls. One RCT gave results in proportion of patients transfused and therefore was excluded from the primary analysis. ¹⁷ Based on the pooled effect of observational studies, CUF was not associated with a significant difference in incidence of intraoperative red cell transfusions (Figure 3; number of comparisons reporting outcome [n = 2]; OR 3.09; 95% CI 0.26–36.59; p = 0.37; p for heterogeneity = 0.94, I ² = 0%. The RCTs examining CUF had different comparisons (CUF vs MUF with no control group or combined CUF and MUF vs control) or had disparate units for intraoperative transfusions (proportion transfused vs mean units per patient). There was no evidence of significant publication bias in our analysis according to the symmetry of the funnel plots noted on Egger’s test (Bias = 2.57, p = 0.272) and Begg’s test (p = 0.09; Supplemental Figures 4a and 4b, respectively).

OPEN IN VIEWER

Figure 2.

The weighted (pooled) estimate showed modified ultrafiltration reduced mean intraoperative allogeneic red cell transfusions (MD −0.73; 95% CI −1.12 to −0.35 p = 0.04) compared to controls.

OPEN IN VIEWER

Figure 3.

The weighted (pooled) estimate showed no effect of conventional ultrafiltration on incidence of intraoperative allogeneic red cell transfusion (OR 3.09; 95% CI 0.26–36.59; p = 0.37) compared to controls.

Discussion

Several methods are employed through the cardiac surgical encounter to reduce excessive total body water and prevent hemodilution associated with CPB. These most notably include CPB prime volume reduction techniques such as low volume bypass circuits, retrograde and anterograde autologous blood priming, micro-cardioplegia, autotransfusion and ultrafiltration. ²⁰ Ultrafiltration is commonly been utilized to reverse excessive hemodilution and remove certain inflammatory mediators associated with CPB. ²¹

In the 1990s, Naik et al. ²² innovated and described a modified technique, MUF, for pediatric cardiac surgical patients, which has remained a popular approach in that population. Our results suggest that MUF may represent a useful strategy to reduce intraoperative transfusion. According to a 2016 survey from the U.S., Canada, and non-North American centers, 79% of pediatric cardiac surgical respondents reported use of MUF, ²³ but similar surveys of perfusion practices in adults are lacking.^24,25 MUF is performed after the patient has separated from CPB, thus providers may be reluctant to use the technique both due to the time it entails as well as the potential for relative hemodynamic instability during a phase when a patient’s clinical status can otherwise be quite dynamic. Unfortunately, this review was unable to characterize the association between MUF and organ injury as so few of the studies included AKI as one of their outcomes. This represents an opportunity for future investigation.

The recent STS/SCA/AmSECT/SABM update to the clinical practice guidelines for PBM includes a Class IIB recommendation for the use of MUF for blood conservation in adult cardiac surgical patients. ⁸ Prior guidelines have recommended the use of ultrafiltration in order to hemoconcentrate blood volume with similar Class IIB grade and contrasting level of evidence (LOE) A.^7,9 Unfortunately, these recommendations have largely been based on prior meta-analyses with significant limitations. Boodhwani et al. ²⁶ included studies that combined CUF and MUF and was comprised of studies from the early 2000s, which are unlikely to adequately reflect present day practice. Low et al. ²⁷ similarly included studies that combined CUF and MUF in the intervention arm, and employed dated methodological assessment of quality. In contrast to these prior publications, this present meta-analysis is the largest to date specifically designed to assess the individual methods of ultrafiltration in isolation. As our results suggest, the specific form of ultrafiltration employed is not only likely to impact the frequency of transfusion but may also impact the risk of postoperative organ injury in vulnerable populations as well.

Interestingly, according to the results of our pooled analysis of large observational studies with low risk of bias, CUF was not associated with a significant reduction in red cell transfusion. This is inconsistent with the promise of the technique, which involves removal of excessive small solutes, along with fluid, to preserve the cellular elements of blood. ¹⁸ This hemoconcentration effect would theoretically reduce the need for transfusion, particularly when teams adhere to a specific hemoglobin transfusion trigger. The often widely employed practice has been the target of increased scrutiny given large observational studies have revealed an association between larger volumes of CUF removal and postoperative AKI.^4,5 These findings were mirrored in the results of this systematic review. One study, authored by Paugh et al., ⁵ examined 6407 patients undergoing isolated CABG and reported that CUF was associated with a higher adjusted risk of AKI (adjusted OR, 1.36; 95% CI, 1.12–1.65; p < 0.05). They also found that patients with preoperative creatinine clearance less than 99.6 mL/min, increasing volume of CUF/kg was associated with a greater odds of AKI. In a separate study not included in this meta-analysis due to methodological limitations, Manning et al. ⁴ analyzed results from a large, single institution cohort and found that a CUF volume of 32.9 mL/kg was the strongest predictor of AKI. The CUF volume ≥32.6 mL/kg (2.2 L for a 70 kg adult) was also associated with severity of AKI by staging, longer hospital LOS and ICU LOS, and higher vasoactive infusion scores in their multivariate regression model. Paradoxically, patients with both AKI and CUF volume removal ≥32.9 mL/kg were transfused with the largest volume of allogeneic blood, perhaps to overcome hemodynamic derangement and reverse organ injury. The hypothesis for the mechanism of AKI is due to renal hypoperfusion when large volumes of ultrafiltrate are removed. Nevertheless, neither study accounted for delivered oxygen on bypass, which has shown to be an independent predictor of AKI.^28,29

There are several important limitations to our review. There is a moderate degree of heterogeneity associated with the results, likely due to variability in ultrafiltrate volume removed, wide range in the type of surgeries included, patient population, and other PBM interventions that may have been employed (though not specifically reported). The evolution of PBM as well as the application of other transfusion prevention strategies is likely to impact the effect size of ultrafiltration alone. Although other blood conservation strategies are listed in Table 1, we presume there to be underreporting of all relevant approaches. The RCTs included were mostly small studies with a median of 60 patients of varying methodological precision. Although we performed sensitivity analysis, there is potential for unmeasured confounders as well as bias in reporting of the individual study results. We also did not also did not account for postoperative red cell transfusions or non-red cell transfusions since most studies did not examine these.

It is important to note the major limitations of evaluating the technique of CUF on the outcome of intraoperative red cell transfusions when only 2 of 3 observational studies could be used in the meta-analysis.^5,18,19 Appropriate meta-analysis methodology uses adjusted OR instead of raw data for observational studies and thus, one study was eliminated. ¹⁹ Observational studies are at a higher risk of confounding bias due to their design. The predominant study had 20,325 patients in each arm, CUF compared to controls, and uses rigorous methodological techniques, such as propensity scoring analyses and mixed-effects binary logistic regression model. ¹⁸ This study was performed retrospectively on a data registry of 195 centers and therefore, there was no standardization of ultrafiltration techniques which could bias the results. There is also concern for unmeasured confounders, especially in observational studies.

Similarly, for the secondary outcome of AKI, only 2 of 3 studies that examined intraoperative red cell transfusion and CUF reported on rates of AKI.^5,19 These 2 observational studies have differences in surgery type (isolated CABG versus high-risk surgery), reported CPB times, and ultrafiltration volumes (1365 ± 1202 mLs versus 3400 ± 1921 mLs) which increase heterogeneity. There is risk of confounding bias in the predominant study (N = 6407 patients) due to its retrospective, observational design and no controlled process of how CUF was performed in 21 medical centers. ⁵ The authors attempted to control for risk of variation between centers by using a generalized linear mixed-effect model.

The results of this systematic review and meta-analysis suggest that whereas MUF is associated with less intraoperative red cell transfusion, CUF confers no similar benefit based on limited evidence. Larger, multicenter, randomized trials should be performed to determine the optimal MUF and CUF volume that corresponds to the greatest benefit in prevention of transfusion for the adult cardiac surgical patient as well as examine key safety endpoints, such as organ injury.

Supplemental Material

Supplemental Material

Supplemental Material

Supplemental Material

Supplemental Material