Red blood cell transfusion is associated with increased hemolysis and an acute phase response in a subset of critically ill children

Abstract

In healthy adults, transfusion of older stored red blood cells (RBCs) produces extravascular hemolysis and circulating non–transferrin-bound iron. In a prospective, observational study of critically ill children, we examined the effect of RBC storage duration on the extent of hemolysis by comparing laboratory measurements obtained before, and 4 hr after, RBC transfusion (N = 100) or saline/albumin infusion (N = 20). Transfusion of RBCs stored for longer than 4 weeks significantly increased plasma free hemoglobin (P < 0.05), indirect bilirubin (P < 0.05), serum iron (P < 0.001), and non-transferrin-bound iron (P < 0.01). However, days of storage duration poorly correlated (R²<0.10) with all measured indicators of hemolysis and inflammation. These results suggest that, in critically ill children, most effects of RBC storage duration on post-transfusion hemolysis are overwhelmed by recipient and/or donor factors. Nonetheless, we identified a subset of patients (N = 21) with evidence of considerable extravascular hemolysis (i.e., increased indirect bilirubin ≥0.4 mg/dL). In these patients, transfusion-associated hemolysis was accompanied by increases in circulating non-transferrin-bound iron and free hemoglobin and by an acute phase response, as assessed by an increase in median C-reactive protein levels of 21.2 mg/L (P < 0.05). In summary, RBC transfusions were associated with an acute phase response and both extravascular and intravascular hemolysis, which were independent of RBC storage duration. The 21% of transfusions that were associated with substantial hemolysis conferred an increased risk of inducing an acute phase response.

Introduction

Up to 50% of critically ill children in pediatric ICUs (PICUs) receive at least one red blood cell (RBC) transfusion [1]. Accumulating evidence suggests that transfusions of PICU patients are associated with increased risks of adverse reactions [2]. In the absence of clear benefits of a liberal transfusion strategy [3], a restrictive transfusion strategy is increasingly favored. One hypothesis for the lack of benefit of a liberal transfusion strategy is that transfusions have adverse effects resulting from RBC storage. The FDA allows RBCs to be refrigerator stored for up to 42 days before transfusion and these RBCs become progressively damaged during storage (i.e., the “RBC storage lesion”). The mean storage time of RBCs transfused in the United States is 18 days [4] and a meta-analysis of observational studies suggests that transfusion of RBCs after longer storage durations increases mortality [5], although this remains controversial [6]. Furthermore, several studies identified increased morbidity and mortality in transfused, critically ill, pediatric patients [7–9]. Although observational, these studies raise fundamental questions about the safety of transfusing older, stored RBCs in these clinical settings.

Important physiological differences arising following transfusions of RBCs after shorter or longer storage durations were identified in humans [10–12], dogs [13,14], and mice [15]. Thus, transfusions of fresh RBCs produce no laboratory evidence of hemolysis, whereas transfusions after longer storage durations result in substantial clearance of storage-damaged RBCs, inducing increased levels of circulating bilirubin, iron, and non-transferrin-bound iron [10]. This circulating non-transferrin-bound iron (e.g., iron not bound to transferrin, the physiologic iron transporter) enhances bacterial growth in plasma in vitro [10] and may enhance bacterial infection in vivo [16]. Furthermore, hemolysis is associated with poorer survival in septic patients [17,18]. In addition, in animal models, transfusion-induced hemolysis induces acute phase and pro-inflammatory cytokine responses [13,15], although these were not observed after transfusing healthy human volunteers with single units of older, stored RBCs [10,11]. Nonetheless, ill animals [15] and hospitalized patients [19,20] exhibited increased cytokine responses following transfusion. Thus, one aim of the current study was to determine whether transfusions, particularly of older, stored RBCs, induce acute phase responses in critically ill children at 4 hr after transfusion.

To this end, we performed a prospective, observational study in 100 critically ill children in two PICUs to compare the effects of transfusions of packed RBCs, after various storage durations, on indicators of hemolysis, iron metabolism, and inflammation. This population was chosen as the effects of RBC transfusions in critically ill children have not been widely explored, and because they are frequently exposed to an RBC dose derived from a single donor per transfusion event, thereby simplifying study design and data analysis. Transfusion effects were assessed by comparing laboratory measurements obtained 4-hr after transfusion to those obtained pre-transfusion. Furthermore, 20 critically ill children who were treated for hypovolemia or hypoalbuminemia with saline or albumin infusions were studied as controls. Finally, a sub-study analysis was performed in the 13 children who received both a saline or albumin infusion and an RBC transfusion at different times during their PICU stay.

Methods

Study and data collection

This prospective, observational study was conducted in the two PICUs of New York-Presbyterian Hospital: at the Children’s Hospital of New York on the Columbia campus and at the Komansky Center on the Weill Cornell campus. Inclusion criteria were weight >5 kg, age <21 years old, and transfusion of a dose of leukoreduced packed RBCs or infusion of a saline/albumin solution. Patients were excluded if they were transfused with platelets and/or plasma within 4 hr of the test RBC transfusion or saline/albumin infusion, or if they received RBC products from more than one donor during the test transfusion. Participation in the study did not affect the transfusion practice in the PICUs or the age of RBC products released from the respective blood banks. Both Institutional Review Boards approved the protocol, which was also registered on ClinicalTrials.gov (NCT02087553).

Blood sampling

A single blood sample was collected into a serum separator tube prior to transfusion/infusion and 2–6 hr after transfusion/infusion, with a target of 4 hr post-transfusion. This time-point was chosen based on our prior results in which circulating non-transferrin-bound iron levels peaked at 4 hr after transfusion in healthy adult volunteers [10]. The sample was allowed to clot for 30 minutes and then centrifuged at 2800 × g for 10 minutes. Sera were aliquoted and frozen at −80°C until testing.

Laboratory testing

All laboratory testing for routine clinical parameters was performed in the New York-Presbyterian Hospital Clinical Laboratories. Circulating non-transferrin-bound iron, measured using an ultrafiltration assay [15], was performed in the Iron Reference Laboratory at the Columbia University Medical Center. Cytokines were measured using a multiplex cytometric bead array (BD Biosciences), following the manufacturer’s instructions. Hepcidin was measured using a commercial ELISA (Bachem), following the manufacturer’s instructions.

Statistical analysis

This prospective, observational study was designed to obtain a pre-transfusion and a 4-hr post-transfusion blood sample from each patient to study the effects of RBC transfusions on indicators of hemolysis, iron metabolism, and inflammation. It was powered (90% power) to detect a partial correlation coefficient of 0.30 for the change in non-transferrin-bound iron predicted by transfusion unit storage duration, after adjusting for a priori defined potential confounders (i.e., RBC dose (ml/kg), RBC unit irradiation (yes/no), sepsis at transfusion (yes/no), age (years), race, and ABO blood type). Correlations and the effect of covariates were determined by linear regression using SAS^™ version 9.4 (SAS Institute, Inc.). Normality of data was assessed using the Kolmogorov-Smirnov Test (with Dallal-Wilkinson-Lillie for P-value). Differences between means were assessed by one-way ANOVA with Tukey’s Multiple Comparison Test or Kruskall-Wallis with Dunn’s Multiple Comparison Test, as appropriate. A sub-study was performed on the 13 transfused patients who also received an infusion of saline/albumin at a different time during their PICU stay. Differences between means in this paired sub-study were assessed using a paired t-test or Wilcoxon matched-pairs signed rank test, as appropriate. Graphs and descriptive statistics were performed using Prism version 5.0 (GraphPad Software, Inc.).

Results

Demographics and clinical characteristics

From December 2013 to October 2014, we recruited 100 critically ill children transfused with RBCs and 20 critically ill children infused with saline (N = 9) or albumin (N = 11). Table I compares transfused patients with saline/albumin infused controls. The groups were well matched with regard to gender, age, and diagnosis. The indication for transfusion was predominantly for anemia, whereas for saline/albumin infusion it was primarily hypovolemia (P <0.001). Furthermore, saline/albumin infusions were completed more rapidly than RBC transfusions [1 ± 0.7 vs. 3 ± 0.7 hr, respectively; (P <0.05)] and were associated with higher pre-transfusion hemoglobin levels (9 ± 1.7 vs. 7.9 ± 1.7 g/dL, respectively; P <0.05).

TABLE I.

Patient Demographics

Demographic	RBCs (N = 100)	Saline/Albumin (N = 9/11)	P-value
Race			NS
White	44 (44%)	4 (20%)
Black	21 (21%)	9 (45%)
Asian	11 (11%)	3 (15%)
Hispanic	24 (24%)	4 (20%)
Gender (Female)	41 (41%)	7 (35%)	NS
Age (Years)	6 (1–13)	6 (2–11)	NS
Weight (kg)	18 (10–43)	20 (13–42)	NS
Diagnosis			NS
Heart failure	29 (29%)	3 (15%)
Respiratory failure	21 (21%)	2 (10%)
Orthopedic surgery	12 (12%)	1 (5%)
Abdominal surgery	24 (24%)	11 (55%)
Hematologic disease	4 (4%)	0
Neurosurgery	4 (4%)	1 (5%)
Septic shock	6 (6%)	2 (10%)
CVVH/ECMO (Yes)	4/3 (4/3%)	0	NS
Vascular access			NS
Central/Arterial line	88 (88%)	18 (90%)
Peripheral vein	12 (12%)	2 (10%)
Indication for transfusion			<0.001
Anemia	96 (96%)	0
Oxygen delivery	4 (4%)	0
Hypovolemia	0	16 (80%)
Hypoalbuminemia	0	4 (20%)
Infusion volume (mean ± sd; ml/kg)	10 ± 4.6	11.5 ± 6.6	NS
Infusion duration (mean ± sd hours)	3 ± 0.7	1 ± 0.7	<0.001
Hemoglobin pre-transfusion (mean ± sd; g/dL)	7.9 ± 1.7	9 ± 1.7	<0.05

NS = Not significant.

Table II provides the RBC unit characteristics. Half of all transfused units were Group O and 72% were irradiated. Most units (73%) were stored in AS-3 solution (Nutricel, Haemonetics), with 25% in AS-1 solution (Adsol, Fenwal). The median storage age of transfused units was 23 days (Fig. 1).

TABLE II.

RBC Unit Characteristics (N = 100)

Characteristic	N (%)
ABO blood type
A	29 (29%)
B	16 (16%)
AB	5 (5%)
O	50 (50%)
Irradiated	72 (72%)
Storage solution
AS-1	25 (25%)
AS-3	73 (73%)
AS-5	1 (1%)
CPDA-1	1 (1%)

Figure 1.

Histogram of RBC storage duration for the 100 RBC transfusions in the study. The median storage duration is 23 days (range 4–39 days).

Effect of RBC transfusion on markers of hemolysis and inflammation

To assess the effect of transfusion per se on markers of hemolysis and inflammation, a sub-study of the 13 volunteers who received both an infusion of saline/albumin and a transfusion of RBCs at different times during their PICU admission was performed. At baseline (i.e., before the saline/albumin infusion and before the RBC transfusion), liver function and inflammation were similar, as assessed by total bilirubin, albumin, AST, ALT, alkaline phosphatase, and CRP levels (Supporting Information Fig. 1). There was little change in the indicators of intravascular hemolysis (haptoglobin and free hemoglobin) or extravascular hemolysis (indirect bilirubin, serum iron, and non-transferrin-bound iron) following saline/albumin infusion (Fig. 2 and Supporting Information Table I). In contrast, following RBC transfusion, there were significant increases in indirect bilirubin, serum iron, and circulating non-transferrin-bound iron levels as well as a trend towards an increase in free hemoglobin (Fig. 2 and Supporting Information Table I). CRP levels following RBC transfusion were increased as compared to those after saline/albumin infusions (Fig. 2 and Supporting Information Table I). No significant differences in cytokine or hepcidin concentrations were observed (Supporting Information Table I and Supporting Information Fig. 2).

Figure 2.

RBC transfusions result in hemolysis and an acute phase response. The median with interquartile range for the change in serum levels of (A) indirect bilirubin, (B) serum iron, (C) non-transferrin-bound iron, (D) CRP, (E) free hemoglobin, and (F) haptoglobin from 4-hr after transfusion to pre-transfusion in the patients (N = 13) infused with both saline/albumin and transfused with RBCs at different times in their PICU admission. *P <0.05, **P <0.01.

Storage duration is a poor predictor of the extent of intravascular or extravascular hemolysis

To examine the correlation between RBC storage duration and the changes in indicators of hemolysis assessed above, the complete data-set of 100 independent transfusions was explored. When considered as a continuous variable, the storage age of transfused RBCs was not correlated with changes in either haptoglobin or free hemoglobin (Fig. 3). This relationship was unchanged when adjusted for the following potential confounders: RBC dose per kg, RBC unit irradiation, sepsis, age, race, ABO blood type, ABO identical or compatible status, time between transfusion and post-transfusion sampling, method of blood sampling (e.g., arterial/central line versus peripheral vein) and baseline haptoglobin level (data not shown). In contrast, when categorized into consecutive 2-week intervals of RBC storage, RBCs stored for greater than four weeks were associated with an increase in plasma free hemoglobin (Fig. 4 and Supporting Information Table II).

Figure 3.

RBC storage duration is a poor predictor of the extent of intra-vascular and extravascular hemolysis observed at 4-hr post-transfusion. A Spearman correlation was used to determine the relationship between the RBC storage duration and changes in circulating levels of (A) haptoglobin, (B) free hemoglobin, (C) indirect bilirubin, (D) serum iron, (E) non-transferrin-bound iron, and (F) hepcidin observed at 4-hr post-transfusion to pre-transfusion in patients (N= 100) transfused with RBCs. The P-values and Spearman correlation coefficient values are as specified in the Figure.

Figure 4.

Measures of hemolysis are increased following transfusion of older, stored RBCs. The median with interquartile range for serum levels of (A) haptoglobin, (B) free hemoglobin, (C) indirect bilirubin, (D) serum iron, (E) non-transferrin-bound iron, and (F) hepcidin from 4-hr after transfusion to pre-transfusion in the patients infused with saline/albumin (N = 20) or transfused with RBCs stored for 0–14 days (N = 15; 1–2 weeks), 15–28 days (N= 65; 3–4 weeks), or 29–42 days (N= 20; 5–6 weeks). *P <0.05, **P <0.01, ***P <0.001 for the comparison to the saline/albumin control group.

When considered as a continuous variable, the storage age of transfused RBCs correlated poorly with changes in indirect bilirubin, serum iron, circulating non-transferrin-bound iron, and hepcidin (Fig. 3). This relationship did not change significantly when adjusted for the following potential confounders: RBC dose per kg, RBC unit irradiation, sepsis, age, race, ABO blood type, ABO identical or compatible status, time between transfusion and post-transfusion sampling, and baseline liver function as assessed by baseline levels of AST, ALT, total bilirubin, albumin, and alkaline phosphatase (data not shown). However, when categorized into consecutive 2-week intervals of RBC storage, transfusions of RBCs stored for more than two weeks were associated with increased serum iron and transfusions of RBCs stored for more than four weeks were associated with increased indirect bilirubin and non-transferrin-bound iron levels. (Fig. 4 and Supporting Information Table II).

Effect of RBC storage duration on markers of inflammation and the acute phase response

The critically ill children in this study had elevated CRP levels at baseline (51.8 mg/L [10.6–103.8]; median with interquartile range; healthy reference range: <1.8 mg/L). No relationship was found between storage age of transfused RBCs and changes in CRP levels, whether storage age was treated as a continuous or categorical variable (Fig. 5A,B and Supporting Information Table II). By contrast, in a post-hoc analysis of the subset of transfused subjects exhibiting substantial hemolysis (N = 21), as defined by an increase in indirect bilirubin of ≥0.4 mg/dL, the CRP level increased significantly (median change of 21.2 mg/L [0.1–37.7] with greater RBC storage duration; P <0.05; Fig. 5C). In this subset of 21 children, non-transferrin-bound iron and free hemoglobin levels also increased following RBC transfusion as compared to the 79 children with a bilirubin increase of <0.4 mg/dL (0.30 μM [0.10–0.52] versus 0.085 μM [−0.10–0.25], respectively; P <0.01) and (11.6 mg/dL [5.2–25.1] versus 4.4 mg/dL [−4.2–13.5], respectively; P <0.05), respectively. Finally, RBC transfusion was not associated with changes in any inflammatory cytokines tested (i.e., MCP-1, IL-6, IL-8, or TNF-α) in any of the analyses performed above (data not shown).

Figure 5.

RBC transfusion results in an acute phase response, predominantly in recipients exhibiting greater amounts of extravascular hemolysis. (A) A Spearman correlation was used to determine the relationship between the RBC storage duration and the change in CRP from 4-hr post-transfusion to pre-transfusion in patients (N= 100) transfused with RBCs. (B) The median with interquartile range for serum levels of CRP from 4-hr post-transfusion to pre-transfusion in patients infused with saline/albumin (N= 20) or transfused with RBCs stored for 0–14 days (N= 15; 1–2 weeks), 15–28 days (N= 65; 3–4 weeks), or 29–42 days (N= 20; 5–6 weeks). (C) The median with interquartile range for change in serum CRP levels in patients infused with saline/albumin (N= 20) or transfused with RBCs and with a change in indirect bilirubin of either <0.4 mg/dL (N= 79) or ≥0.4 mg/dL (N= 21). The P-values and Spearman correlation coefficient values are as specified in the figure, *P<0.05, **P<0.01.

Discussion

RBCs are damaged during refrigerated storage, and transfusions of these RBCs are associated with a variable extent of hemolysis in human adults [10]. This study shows that variable amounts of extravascular and intravascular hemolysis also occur following transfusion in critically ill children. Nonetheless, in the PICU, RBC storage duration was only weakly associated with the extent of post-transfusion hemolysis, suggesting that other, still undetermined, donor and/or recipient factors are responsible for the variability in the amounts of hemolysis observed. Importantly, the subset of twenty-one patients with the most hemolysis, as assessed by increases in indirect bilirubin of ≥0.4 mg/dL, developed increased concentrations of circulating non-transferrin-bound iron and free hemoglobin together with an enhanced acute phase response, as assessed by a marked increase in median CRP of 21.2 mg/L (P <0.05). Thus, one in five of the transfusions to critically ill children in this study exhibited significant hemolysis, and this conferred an increased risk of exacerbating their pre-existing acute phase response. These increases in CRP were not associated with changes in inflammatory cytokine (MCP-1, IL-6, IL-8, or TNF-α) levels.

The clinical implications of these findings remain to be determined, but add to the emerging evidence that liberal transfusion strategies, by increasing the overall risk of transfusion-induced hemolysis, worsen outcomes. Because hemolysis increases infectious complications [10,15,16,21] and predicts a poor outcome in sepsis [17], this implies that transfusion-induced hemolysis of storage-damaged RBCs provides one plausible mechanism for the lack of benefit associated with liberal transfusion strategies [22–25]. Furthermore, because RBC transfusions producing the most hemolysis were also associated with an enhanced acute phase response, as assessed by increases in CRP, a role in exacerbating the systemic inflammatory response syndrome and sepsis is also possible.

Two recently published randomized trials were designed to test the benefit of fresh RBCs rather than the harm associated with transfusing old RBCs [26,27]. Neither found any meaningful differences in clinical outcomes when comparing transfusion of fresh RBCs to the standard of care (mean ± SD of 6.1 ± 4.9 days versus 22.0 ± 8.4 days [26] and 5.1 ± 2.0 days versus 14.6 ± 8.3 days [27] of storage, respectively, in the two studies). Importantly, these studies did not “address whether the use of RBCs stored for very prolonged periods results in harm [26].” Furthermore, the recently published RECESS study [28] randomized cardiac surgery patients to transfusions of RBCs stored for either ≤10 days or ≥21 days. Of note, 96% of this study population underwent cardiopulmonary bypass. Cardiopulmonary bypass induces RBC damage and hemolysis [29], potentially confounding any additional harmful effect of hemolysis resulting from RBC transfusion. To the extent that hemolysis of transfused, storage-damaged RBCs is responsible for adverse effects, the results of the RECESS study are not generalizable to other patient populations. Finally, in our study of critically ill children, RBC storage duration poorly correlated with indicators of hemolysis and inflammation. Significant increases in recipient circulating free hemoglobin, indirect bilirubin, and non-transferrin-bound iron did not occur until donor RBCs were stored for 5–6 weeks pre-transfusion (Fig. 4 and Supporting Information Fig. 3). Consequently, assuming that hemolysis contributes substantially to the adverse effects of RBC transfusion, then randomized controlled trials comparing the extremes of storage duration are more likely to detect differences in clinical outcome.

Our study has several limitations. First, laboratory parameters were only measured 4-hr after transfusion; thus, important changes occurring at earlier or later times post-transfusion may have been missed. For example, free hemoglobin infused with the transfused unit (i.e., hemolysis “in the bag”) might have been detected in serum at earlier time points. In addition, inflammatory cytokines peaked in dogs at 6 hr after transfusion [13], after the peak observed in mice [15]. Later increases in inflammatory cytokine levels would not have been detected our study. Furthermore, although a single “unit” of storage-aged mouse RBCs did not induce a measurable rise in circulating cytokines, a distinct inflammatory transcriptional program was induced in the spleen [30]. Thus, it is possible that more sensitive measures of inflammation would yield different results. Nonetheless, increases in CRP levels were observed, especially in the 21% of children exhibiting substantially increased hemolysis following transfusion, a finding to be validated as an outcome defined a priori in future studies. Furthermore, because this was an observational study, unmeasured confounding factors may cause the observed differences in hemolysis and the acute phase response. Nevertheless, adjusting our results for multiple a priori identified potential confounders (e.g., RBC dose (mL/kg), RBC unit irradiation (yes/no), sepsis at transfusion (yes/no), age (years), race, ABO blood type), and baseline liver function did not significantly change any of these results (data not shown). We also only measured a focused panel of indicators of hemolysis and inflammation determined a priori, whereas RBC transfusions might cause differences in other, unmeasured parameters. In addition, this study did not assess the effects of large volume RBC transfusions, because only patients receiving RBC transfusions derived from a single donor were studied. Finally, this prospective, observational study assessed a pediatric critically ill patient sample with an underlying inflammatory state, as demonstrated by elevated baseline CRP levels. Thus, the generalizability of these results to other PICU populations remains to be determined.

As in an earlier study of critically ill adults [31], we did not observe a significant change in circulating pro-inflammatory cytokines at 4 hr after RBC transfusion. Nonetheless, in contrast to that study [31], a subset of critically ill pediatric patients did exhibit increased CRP levels at 4 hr post-transfusion, an effect that was more evident after transfusions resulting in greater amounts of extravascular hemolysis. A correlation between hemolysis and the acute phase response in adults was not reported in the earlier study [31]. In addition, in several studies in adults, stored RBC transfusions produced increased circulating free hemoglobin levels [11,12,32]. Our study complements these and suggests that RBC transfusions, particularly of RBCs during the final two weeks of storage, may result in intravascular hemolysis in critically ill pediatric patients.

Current FDA standards allow refrigerated RBC units to be stored for up to 42 days before transfusion, based on various criteria, including that, on average, <25% of the transfused RBCs will be cleared from the circulation of healthy volunteers in the first 24 hr post-transfusion. Although post-transfusion RBC recovery is worse in ill patients as compared to healthy adult volunteers [33], it has not been examined in critically ill pediatric patients. In the current study, variable degrees of extravascular and intravascular hemolysis, as assessed by increased serum indirect bilirubin, serum iron, non-transferrin-bound iron, indirect bilirubin, and free hemoglobin levels, were observed at 4 hr post-transfusion. Because RBC storage duration poorly predicted the extent of hemolysis, it is important to identify other indicators that would better predict the extent of hemolysis induced by RBC transfusion. Finally, to our knowledge, this is the first study to observe an acute phase response associated with RBC transfusion in a subset of patients; one that is even greater after the 21% of transfusions resulting in substantial hemolysis.