Ultrafiltration reduces blood transfusions following cardiac surgery: a meta-analysis

Abstract

Background: Although used routinely in pediatric patients, ultrafiltration techniques that reverse hemodilution are infrequently used in adults. Data from small, unblinded clinical trials suggest that the use of ultrafiltration can reduce inflammatory mediators, improve cardiac function, and reduce hemodilution. We conducted a meta-analysis of randomized trials to evaluate the effects of ultrafiltration on blood transfusions and blood loss following adult cardiac surgery. Methods: Medline, EMBASE, and Cochrane databases were searched and randomized controlled trials evaluating modified and/or conventional ultrafiltration, meeting pre-determined selection criteria, were obtained. Quality evaluation and data extraction were performed by two independent observers blinded to study source. Random effects models were used to determine pooled effect estimates and sources of heterogeneity were explored using meta-regression. Results: One hundred and thirty two studies were screened and 10 randomized trials evaluating 1004 patients (control, n = 495; ultrafiltration, n = 509) were identified of which only two were double-blinded. The use of ultrafiltration was associated with a reduction in postoperative blood transfusions (weighted mean difference [95% CI] of −0.73 units [−1.16, −0.31]; p = 0.001). This reduction was greater in studies evaluating modified ultrafiltration. Use of ultrafiltration was also associated with reduced postoperative bleeding (−70 ml, [−118, −21]; p = 0.005), which was driven primarily by trials evaluating modified rather than conventional ultrafiltration. Conclusions: Use of ultrafiltration is associated with a significant reduction in postoperative blood transfusions as well as reduced bleeding in adults undergoing cardiac surgery. The efficacy and cost-effectiveness of ultrafiltration as a blood conservations strategy should be evaluated in a large, randomized, double-blinded study.

1 Introduction

Cardiac surgery remains one of the greatest consumers of blood products. Blood transfusion following surgery has been linked not only to viral infections such as hepatitis and AIDS, but also to immunosuppression and increased mortality [1,2]. Furthermore, in the near-term future, the demand for blood products is expected to outstrip the supply by up to 20% in certain developed countries [3]. Lastly, blood product use contributes to the economic burden of cardiac surgery. These are compelling reasons to aggressively pursue blood conservation strategies in this patient population.

While numerous factors contribute to bleeding and the need for transfusion after cardiac surgery, one of the key contributors is the hemodilution resulting from the use of cardiopulmonary bypass (CPB). This hemodilution occurs due to the use of 1–2 l of priming solution that fill the circuit before perfusion of the patient begins, and is further exacerbated by the use of cardioplegia solution, which in some cases may add 2–3 l to the pump balance. On discontinuation of CPB, hemodiluted blood remains in the extracorporeal circuit and standard practice involves the slow transfusion of this residual dilute pump blood back to the patient prior to removing the aortic cannula, or its administration intravenously after decannulation.

Hemodilution after CPB can be reversed, or at least reduced, using ultrafiltration (UF) [4]. UF devices can be integrated in parallel into existing CPB circuits primarily in two configurations. Conventional UF (CUF) is run during CPB and cannot be run following its discontinuation. Blood enters the inflow from the venous reservoir and returns to the oxygenator in a concentrated form [5]. While CUF is successful in removing fluid, much of this is usually returned as crystalloid to maintain circulating volume during CPB. In contrast, modified UF (MUF) is initiated after the completion of CPB. The cannulae are left in place and blood is aspirated from a sideport of the aortic cannula into a hemofilter, and then returned to the right atrium in a hemoconcentrated form [4]. External suction is applied on the hemofilter to remove solute solution.

A number of randomized trials have been conducted to evaluate the effects of ultrafiltration on perioperative outcomes following adult cardiac surgery. With the exception of one study, the trials are invariably conducted in small populations, using non-clinical surrogate markers as primary endpoints. The aim of this meta-analysis was to synthesize the existing evidence on the use of ultrafiltration (conventional and modified) as a blood conservation strategy in adult cardiac surgical patients. Specifically, we sought to evaluate the effects of ultrafiltration on postoperative blood product use and perioperative bleeding.

2 Materials and methods

The analyses and reporting were conducted according to the published guidelines in the QUORUM statement [6].

2.1 Data sources and study selection

We searched Medline, Biosis, Embase, CINAHL, HealthStar, and the Cochrane databases from their inception to October 2003. In addition, a hand search of the bibliographies of selected manuscripts was conducted. A broad search strategy was employed, which attempted to identify all randomized controlled trials relevant to modified ultrafiltration, ultrafiltration, or hemofiltration, utilized in cardiac surgical procedures with cardiopulmonary bypass in adults. In order to be included, clinical trials had to be randomized, have a control group, utilize either conventional and/or modified ultrafiltration techniques with the explicitly stated goal of hemoconcentration, and report blood product use and/or postoperative bleeding.

The search and study selection is depicted in Fig. 1 . In total, 132 titles and abstracts from various sources were reviewed. Of these, 28 manuscripts meeting the inclusion criteria were retrieved. Seventeen studies were further excluded because they included duplicate publications, lacked a control group or were non-randomized. Finally, one more study was excluded because it did not contain sufficient information on the outcomes of interest. In total, 10 studies were used for data synthesis. One study was a 3-arm study and was used for two separate comparisons [7]. Another utilized a factorial design and also provided data for two separate comparisons [8]. Attempts were made to contact the corresponding authors of selected studies, which did not report certain data on outcomes of interest in the manuscript. Out of a total of 16 authors contact by email and surface mail on three separate occasions, 2 responded to the data request.

2.2 Quality evaluation and data extraction

Quality evaluation and data extraction were performed by individuals blinded to the source (authors, institution, and journal) of the articles. Information on randomization, blinding, and patient follow-up was used to calculate the study quality score (also termed the Jadad score, range: 0–5) as described by Jadad et al. [9]. Data extraction was performed independently by two individuals (G.G. and N.S.) using a standardized data collection form. All discrepancies were resolved by consensus. In addition to the endpoints of blood transfusions and postoperative bleeding, data collected from each manuscript included: lead author, year of publication, type of surgery, type of cardioplegia, study intervention, amount of fluid removed during intervention, hematocrit (before and after intervention), and perioperative outcomes (mortality, ICU, and in-hospital length of stay) where available. Data on blood transfusions, reported using differing measures in different studies, were converted to number of units per patient for all studies.

2.3 Statistical analysis

Weighted mean difference (WMD) was used as the primary effect estimate. Due to significant between-study heterogeneity, DerSimonian and Laird random effects models were used to calculate the summary effect estimates and 95% confidence intervals. Sources of heterogeneity were examined using stratified analyses and meta-regression techniques. Sensitivity analyses were conducted and publication bias was assessed using a Begg’s funnel plot. All statistical analyses were performed using STATA version 9 (STATA corporation, College Station, TX, USA).

3 Results

In total, 10 randomized clinical trials were utilized for the analysis, which evaluated a total of 1004 patients (control – 495 patients, treated – 509 patients). Of these, only two were described as blinded. Only three of the ten studies had a study quality score greater than one, indicating generally poor methodology and methods reporting (Table 1 ).

In the treatment group, there was significant variability among studies with respect to the amount of fluid removed. The average amount of fluid removed was 1765 (range – 906 vs 4970) ml/patient and the amount of fluid removed tended to be greater in trials utilizing MUF (1941 ml vs 1472 ml in MUF vs non-MUF studies).

3.1 Effects of ultrafiltration on blood product use

The majority of studies did not individually show a significant beneficial effect of ultrafiltration on blood product use (Fig. 2 ). The pooled estimate, however, demonstrated a significant reduction in blood product use in patients treated with ultrafiltration (WMD [95% CI] of −0.73 units [−1.16, −0.31]; p = 0.001). However, there was significant study heterogeneity (p = 0.042). Stratified analysis by type of intervention indicated a slightly greater beneficial effect in trials that evaluated MUF (−0.71 units [−1.08, −0.35]; p ≪ 0.001) compared with trials that did not use MUF (−0.61 units [−1.65, 0.43]; p = 0.25).

3.2 Ultrafiltration and postoperative bleeding

The pooled effect estimate demonstrated a significant reduction in postoperative bleeding by 70 ml (p = 0.005, 95% CI – [−118, −21]) in patients treated with ultrafiltration (Fig. 3 ). This difference was primarily driven by trials utilizing MUF (−104 ml [−161, −46]; p ≪ 0.001) compared to non-MUF studies (−10 ml [−68, 47]; p = 0.73).

3.3 Stratified analyses and assessment of publication bias

In order to ensure that the pooled effect estimates were stable and not overly influenced by the largest study (n = 578 patients) or the study demonstrating the largest beneficial effect, pooled estimates were calculated after excluding these trials. The beneficial effect on blood transfusions remained significant even after removal of the largest study (−0.76 [−1.27, −0.25]; p = 0.004) as well as the study with the largest beneficial effect (−0.55 [−0.85, −0.26]).

Meta-regression was performed to evaluate sources of between-study heterogeneity. In a multivariate analysis, the study intervention (CUF vs MUF), amount of fluid removed, and type of cardioplegic solution were not significant but study quality score was a significant predictor of the variability between studies (p = 0.006). Studies with higher study quality scores demonstrated lower estimates of treatment effects. A Begg’s funnel plot did not indicate significant publication bias (Begg’s p = 0.94).

4 Discussion

Although many studies have reported benefits of ultrafiltration, this technique is infrequently used in adult cardiac surgical patients. Over 1000 patients have been studied as part of randomized clinical studies evaluating the effects of ultrafiltration. In this meta-analysis, we found that the use of ultrafiltration in adult patients undergoing cardiac surgery is associated with decreased blood product use by approximately 0.7 units/patient. This was further supported by a reduction in postoperative bleeding by approximately 70 ml/patient in those treated with ultrafiltration. These differences in blood transfusions and bleeding were greater in trials evaluating modified ultrafiltration rather than conventional ultrafiltration. Lastly, there was significant between-study variability in effect estimates, which was explained, in part, by variations in study quality with higher quality studies demonstrating lower effect estimates.

Proposed benefits of ultrafiltration include increased hematocrit [4], reduction in circulating inflammatory mediators [10,11], improved cardiovascular performance [6,10,12], and improved neurologic outcome [13]. The benefits of UF, particularly MUF, are so pronounced in pediatric patients that MUF has been adopted as the standard of care in this population [14]. In adults, however, this technique is rarely used [15].

These benefits are weighed against the costs of ultrafiltration, which include the actual cost of the filter and potentially increased operative time with the patient systemically heparinized. Furthermore, as indicated in Table 1, the majority of clinical trials in adults are limited by small sample sizes that preclude an adequately powered assessment of clinically relevant outcomes. Lastly, as indicated in Figs. 2 and 3, most of the studies do not individually show a statistically significant benefit.

The exception to this is the largest study evaluating the effects of MUF on perioperative outcome in adults [13]. In this study, Luciani et al. demonstrated a reduction in blood product use, neurologic, respiratory, and gastrointestinal complications in patients treated with MUF. The most significant limitations of this study were the lack of blinding of treating physicians and the possibility of false positive results due to testing of multiple endpoints in the same population.

In this meta-analysis, we found that elements of study quality (i.e., randomization, blinding, and patient follow-up) were significant predictors of variability in study results. In particular, studies with higher study quality scores demonstrated lower effect estimates. This trend has been previously reported and further underscores the need for more rigorous design and reporting of studies [9]. Some of these methodological improvements include using and reporting appropriate methods of randomization, the use of double-blinding, adequate tracking of study participants, and pre-specification of study endpoints. Although ensuring adequate blinding can sometimes be difficult in trials evaluating surgical techniques, it is important in order to obtain unbiased estimates of treatment effect.

In this study, we combined trials evaluating different methods of ultrafiltration. However, it is important to note that there are key differences between conventional and modified ultrafiltration techniques. The major difference is that CUF can only be carried out while the patient is on cardiopulmonary bypass whereas MUF is performed at the end of cardiopulmonary bypass before reversal of heparin. As a result, MUF can be used to hemoconcentrate the circulating blood volume even as the hemodiluted blood in the pump reservoir is infused into the patient. Therefore, theoretically MUF may be a more effective strategy for achieving hemoconcentration. This idea is supported by the results of this meta-analysis in which the effect of ultrafiltration on blood transfusions and postoperative bleeding was greater in trials evaluating MUF rather than CUF.

Low body weight has repeatedly been shown to be a predictor of postoperative blood transfusions [16,17], due to the relatively increased hemodilution experienced by these patients. Although none of the studies included in this meta-analysis specifically evaluated the benefits of UF in patients with low body weight, it is likely that the benefits of UF are even greater in this patient population. Lastly, only one of the included studies evaluated the cost of MUF and reported an increase of $85 (cost of the ultrafilter) in the treatment group [13]. Compared to the potential cost savings due to blood transfusions alone (approximately $500 per unit), MUF may reduce the economic burden of cardiac surgery.

In summary, this meta-analysis has synthesized evidence from 10 randomized controlled studies evaluating the effects of ultrafiltration in adult cardiac surgical patients. Despite limitations of small size and poor methodologic quality, the balance of evidence seems to indicate a beneficial effect of ultrafiltration on postoperative blood transfusions and blood loss. Using the nomenclature established by the American Heart Association [18], this conforms to a class IIa recommendation and level B evidence for the use of ultrafiltration (and particularly modified ultrafiltration) in adult cardiac surgery patients, i.e., there is some support from small randomized studies which suggest that the benefits of the procedure outweigh the risks, but larger, more focused randomized trials are needed to clearly define the benefits of treatment and to identify subgroups in whom the benefit is greatest.

4.1 Limitations

We employed a broad search strategy, but the possibility that important clinical trials were missed remains. The Begg’s funnel plot, however, did not indicate significant publication bias. To pool treatment effect estimates, we used weighted mean difference which assumes normal distribution of the outcome. While the majority of studies did not specifically state the distribution of outcome data, it was usually reported as mean ± standard deviation and analyzed using Student’s t-tests, suggesting that the data were not severely non-normal.

4.2 Conclusions

The use of ultrafiltration is associated with a reduction in postoperative bleeding and blood transfusions in adult cardiac surgical patients. These effects appear to be greater for modified rather than conventional ultrafiltration. This meta-analysis supports the evaluation of these benefits of ultrafiltration in large, randomized, double-blinded studies and the data from this study may aid in the design of such trials.

Acknowledgments

The authors would like to acknowledge Kathy Momtahan and Judith Van Berkman for their contributions to the literature search, and all authors who responded to questions regarding their published studies.

References