Clinical Outcomes Associated With RBC Transfusions in Critically Ill Children: A 1-Year Prospective Study* Pierre Demaret, MD, MSc1; Marisa Tucci, MD, FRCPC2; Oliver Karam, MD, MSc3; Helen Trottier, PhD4; Thierry Ducruet, MSc5; Jacques Lacroix, MD, FRCPC2
Objective: To identify the potential complications associated with RBC transfusions. Design: Prospective observational study. Setting: PICU in a tertiary children’s hospital. Patients: All children consecutively admitted to our PICU during a 1-year period. Interventions: None. Measurements and Main Results: Data were abstracted from medical charts prospectively. Outcomes possibly attributable to RBC transfusions were looked for daily. In transfused cases, it was considered that an outcome was associated with a transfusion only if it was observed after the first RBC transfusion. During the 1-year study period, 913 consecutive admissions were documented, 842 of which were included. Among them, 144 (17%) were transfused at least once. When comparing transfused cases with nontransfused cases, the odds ratio for new or progressive multiple organ dysfunction syndrome was 5.14 (95% CI, 3.28–8.06; p < 0.001). This association remained statistically significant in the multivari*See also p. 585. 1 Division of Pediatric Critical Care Medicine, Department of Pediatrics, CHC, Liège, Belgium. 2 Division of Pediatric Critical Care Medicine, Department of Pediatrics, Sainte-Justine Hospital and Université de Montréal, Montreal, QC, Canada. 3 Pediatric Intensive Care Unit, Geneva University Hospital, Geneva, Switzerland. 4 Department of Social and Preventive Medicine, Research Center, SainteJustine Hospital and Université de Montréal, Montreal, QC, Canada. 5 Department of Pediatrics, Research Center, Sainte-Justine Hospital, Université de Montréal, Montreal, QC, Canada. This work was performed at PICU, CHU Sainte-Justine, Montréal, QC, Canada. Supported, in part, by Fonds de la Recherche en Santé du Québec (grant #24460). Dr. Trottier consulted and lectured for Merck Frosst and for GSK Belgium. She received support for the development of educational presentations from Merck Frosst and received support for travel from GSK Belgium. Dr. Lacroix’s institution received grant support from Fonds de Recherche en Sante de Quebec (grant #24480). The remaining authors have disclosed that they do not have any potential conflicts of interest. For information regarding this article, E-mail: [email protected] Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies DOI: 10.1097/PCC.0000000000000423
Pediatric Critical Care Medicine
able analysis (odds ratio, 3.85; 95% CI, 2.38–6.24; p < 0.001). Transfused cases were ventilated longer than nontransfused cases (14.1 ± 32.6 vs 4.3 ± 9.6 d, p < 0.001), even after adjustment in a Cox model. The PICU length of stay was significantly increased for transfused cases (12.4 ± 26.2 vs 4.9 ± 10.2 d, p < 0.001), even after controlling for potential confounders. The paired analysis for comparison of pretransfusion and posttransfusion values showed that the arterial partial pressure in oxygen was significantly reduced within the 6 hours after the first RBC transfusion (mean difference, 25.6 torr, 95% CI, 5.7–45.4; p = 0.029). The paired analysis also showed an increased proportion of renal replacement therapy. Conclusions: RBC transfusions in critically ill children were associated with prolonged mechanical ventilation and prolonged PICU stay. The risk of new or progressive multiple organ dysfunction syndrome was also increased in some transfused children. Furthermore, our study questions the ability of stored RBCs to improve oxygenation in critically ill children. Practitioners should take into account these data when prescribing an RBC transfusion to PICU patients. (Pediatr Crit Care Med 2015; 16:505–514) Key Words: child; erythrocyte; intensive care units; outcome; pediatric; transfusion
R
BC transfusion is the quickest way to increase hemoglobin concentration and can be lifesaving for patients with hemorrhagic shock or severe anemia. However, issues with regard to the safety of blood transfusions have emerged in recent years. Advances in donor screening and in blood product management have markedly decreased the risk of pathogen transmission (1). The focus has now shifted to noninfectious serious hazards of transfusion (NISHOT), which are increasingly described in the literature. The most common NISHOT is mistransfusion or transfusing the incorrect blood product to a patient (2). However, there are numerous other NISHOTs, including transfusion-related acute lung injury (TRALI) (3), transfusion-associated circulatory overload (TACO) (4), alloimmunization (2), and transfusionrelated immunomodulation (5). These adverse reactions may www.pccmjournal.org
505
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Demaret et al
exacerbate organ dysfunction and therefore worsen the clinical condition of transfusion recipients. RBC transfusions are frequent in critically ill children. Approximately 15–23% of children admitted to PICUs receive at least one RBC transfusion (6, 7), and nearly one in two children who stay in PICU more than 48 hours will be transfused (8). Stated transfusion practice patterns show significant variation among pediatric intensivists (9). Although there is high-level evidence published that shows it is safe to adopt a restrictive transfusion strategy (i.e., to transfuse RBCs only if the hemoglobin level drops below 70 g/L) in stable noncyanotic critically ill children (10), some data show that it can take several years before new knowledge is applied at the bedside (11). This might be explained by poor physician knowledge regarding the adverse effects of blood products. It is thus important to describe the clinical outcomes associated with RBC transfusions in critically ill children, in order that physicians can better estimate the risk-benefit ratio of transfusing RBCs in PICU. We therefore conducted a 1-year prospective observational cohort study to determine adverse outcomes associated with RBC transfusion in a multidisciplinary PICU population.
MATERIAL AND METHODS Study Site The PICU of Sainte-Justine University Hospital is a multidisciplinary 24-bed pediatric critical care unit, serving both medical and surgical specialties. On average, there are 1,000 admissions per year. Study Population All consecutive admissions to the PICU, from April 21, 2009, to April 20, 2010, were prospectively screened for recruitment. Exclusion criteria were newborn with gestational age less than 40 weeks at the time of PICU admission, age less than 3 days or more than 18 years at PICU admission, pregnancy, or admission just after labor. Cases were defined as transfused cases if at least one RBC transfusion was given during the PICU stay and as nontransfused cases if no RBC transfusion was given. Data Collection and Management Trained research coordinators prospectively collected data daily in a validated case report form. All information was abstracted from patient medical charts. Baseline data collected within 24 hours after PICU entry included age, gender, weight, medical history, and admission diagnoses (more than one diagnosis could be attributed to each patient). A predictive score of mortality, the Pediatric Risk of Mortality (PRISM) score (12), and a descriptive score of severity of multiple organ dysfunction, the Pediatric Logistic Organ Dysfunction (PELOD) score (13) were used to describe severity of illness at entry. This study was observational and provided no transfusion guidelines to medical staff. Institutional review board approval was obtained, waiving the requirement for written informed consent. 506
www.pccmjournal.org
Outcomes The list of outcomes under study was generated a priori. In nontransfused patients, each outcome was documented as a single event that occurred at any moment during the prospective follow-up of a patient’s ICU stay. In transfused patients, a variable was considered an outcome associated with a transfusion only if it was observed after the first RBC transfusion event. Primary Outcome. The primary outcome measure was “new or progressive multiple organ dysfunction syndrome” (NPMODS) (10). This was the proportion of patients with development or progression of multiple organ dysfunction syndrome (MODS), as defined by Goldstein et al (14). In nontransfused patients, new MODS was diagnosed if a patient with no or one organ dysfunction at admission developed two or more organ dysfunctions during the PICU stay, and progressive MODS was considered present if a patient who already had MODS at PICU entry developed dysfunction of at least one other organ during the PICU stay. In transfused patients, new MODS was diagnosed if a patient with no or one organ dysfunction at the time the first RBC transfusion developed two or more organ dysfunctions during the remaining PICU stay after the first RBC transfusion and progressive MODS was diagnosed if a patient who already had MODS when the first RBC transfusion was initiated developed dysfunction of at least one other organ after transfusion and during the remainder of the PICU stay. Patients who died were considered to have the primary outcome. Secondary Outcomes. Nosocomial infections were defined as infections diagnosed at least 2 days after PICU entry. Presence of suspected bacterial or viral infection was extracted from medical records (suspected infection); the infection was considered proven if a bacterial or viral culture in specimens drawn in PICU was positive. We also recorded PICU length of stay, duration of mechanical ventilation, and 28-day mortality. Tertiary Outcomes. Hypotension was defined as a systolic blood pressure below the 5th percentile for age (15). Systemic inflammatory response syndrome (SIRS), sepsis, severe sepsis, and septic shock were defined according to the definitions from the international pediatric sepsis consensus conference (14). Acute respiratory distress syndrome (ARDS) was defined according to the following criteria: acute onset, Pao2/Fio2 ratio up to 200, bilateral infiltrates on chest radiograph, and no evidence of left atrial hypertension (16). Po2 was measured using arterial (Pao2) or capillary (Pcapo2) blood samples. We documented the worst Po2 and the worst oxygen saturation (Spo2) during the 2 hours preceding and during the 6 hours following the first RBC transfusion. Statistical Analysis Results of descriptive statistics were expressed as a fraction of the total population, mean ± sd, and/or median with interquartile range. Continuous variables were compared using Student t test if the variable was normally distributed (as attested by the Shapiro-Wilk test) or Wilcoxon test if not. The chi-square statistic was used for categorical variables. Results were considered statistically significant when the p value was less than 0.05. July 2015 • Volume 16 • Number 6
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Feature Articles
For all dichotomous outcomes, a crude odds ratio (OR) with its 95% CI was calculated using logistic regression. We decided a priori to conduct multivariable analyses for the primary outcome (NPMODS) and the secondary outcomes (nosocomial infection, 28-d mortality, duration of mechanical ventilation, and PICU length of stay). We used multivariable logistic regression for dichotomous variables and Cox regression for time variables. Potential confounders were selected based on the following criteria: bivariate association with RBC transfusion, bivariate association with the primary outcome (p value cutoff, 0.25) (17), clinical relevance (based on previous published literature and on the authors’ expertise), and absence of correlation with another retained confounder. If the occurrence of an outcome was too low to include all the retained confounders in the multivariable model, we adjusted only for the severity of illness, using the PRISM score, because we believed it to be the most pertinent confounder. Continuous variables were categorized if they were not linear in the logit, as tested with the quartile design variable method described by Hosmer et al (18). Interactions were tested one by one (between the variable of interest [transfusion] and age, cardiopathy, and PRISM), using likelihood ratio. Fitting of the final logistic regression models was tested using the Hosmer-Lemeshow statistic. For multivariable survival analyzes, extended Cox models were used with transfusion as a time-dependent variable. Survival curves adjusted to the average value of age and admission PRISM score (for several levels of exposure to RBCs) were obtained from Cox regression. In transfused cases, we compared pretransfusion and posttransfusion values of several variables, using McNemar chi-square for comparison of proportions and Wilcoxon signed-rank test for comparison of means. All statistical analyses were done by P.D. and T.D., using SPSS statistical software (SPSS, release 17.0.1; SPSS, Chicago, IL).
RESULTS Over the 1-year study period, there were 913 consecutive PICU admissions involving 802 patients (Fig. 1); 71 cases were excluded. A total of 842 admissions (cases) were retained for analysis. At least one RBC transfusion was given in 144 cases (17.1%). Data at First Day in PICU Data collected within 24 hours after PICU entry are reported in Table 1. Transfused cases were younger, more severely ill (higher PRISM and PELOD scores), and more likely to have congenital heart disease compared with nontransfused cases (34.7% vs 17.3%; p < 0.05). The hemoglobin concentration at admission was significantly lower in transfused cases (92 ± 29.6 g/L vs 115.7 ± 21.4 g/L; p < 0.05). The most frequent admission diagnoses were respiratory disease, bacterial or viral infection, and elective surgery (cardiac and noncardiac). Some admission diagnoses were significantly more prevalent in nontransfused cases (respiratory disease, viral infection, noncardiac surgery, and seizures), whereas others were significantly more prevalent in transfused cases (cardiac surgery and shock). Data on RBC Transfusions Data on transfusions are reported in Table 2. Five-hundred seventy-eight transfusions were given during the 1-year study period. The mean hemoglobin level before the first transfusion was 77.3 ± 27.2 g/L. Of the 144 first transfusion events, 110 (76.4%) were prescribed during the first 2 days in PICU.
Outcomes and Their Association With RBC Transfusion Primary Outcome. NPMODS was observed in 99 of 842 patients (Table 3). The crude OR for development of NPMODS after the first transfusion event was 5.14 (95% CI, 3.28–8.06; p < 0.001). When controlling for potential confounders (admission PRISM score, age, and presence of a congenital heart disease), the association between RBC transfusion and NPMODS remained statistically significant (adjusted OR, 3.85; 95% CI, 2.38–6.24; p < 0.001) (Table 4). Secondary Outcome. Nosocomial infections were increased in transfused cases (crude OR, 4.79; 95% CI, 2.54–9.03; p < 0.001), even after adjustment for the severity of illness at PICU entry (adjusted OR, 3.31; 95% CI, 1.67–6.56; p = 0.001). The association Figure 1. Flow chart of study patients. MODS = multiple organ dysfunction syndrome. between RBC transfusion and Pediatric Critical Care Medicine
www.pccmjournal.org
507
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Demaret et al
Table 1.
Patient Characteristics at the Day of Admission to PICU
Patient Characteristic
All Cases (n = 842)
Transfused Cases (n = 144)
Nontransfused Cases (n = 698)
Gender (male), n (%)
434 (51.5)
74 (51.4)
361 (51.7)
57.7 ± 74
74.9 ± 71.3a
9.6 ± 7
5.2 ± 5.3a
9.1 ± 8.5
3.9 ± 6a
Age (mo), mean ± sd Pediatric Risk of Mortality score, mean ± sd
71.9 ± 72 6 ± 5.8
Pediatric Logistic Organ Dysfunction score, mean ± sd
4.8 ± 6.8
Congenital heart disease, n (%)
171 (20.3)
50 (34.7)
121 (17.3)a
93 (11)
37 (25.7)
56 (8)a
Cyanotic congenital heart disease, n (%) Worst Po2 (torr) Worst Pao2, mean ± sd (n = 278) Worst Pcapo2, mean ± sd (n = 467) Lowest hemoglobin (g/L), mean ± sd (n = 776)
127.1 ± 63.5
129.6 ± 83.6
126.2 ± 53.7
72.6 ± 45.9
72.3 ± 84
72.6 ± 40.1
111.5 ± 24.7
92 ± 29.6
115.7 ± 21.4a
Worst Scvo2 (%), mean ± sd Cyanotic congenital heart disease (n = 52)
64.7 ± 15.5
62.8 ± 13.6
66.9 ± 17.5
Noncyanotic congenital heart disease (n = 104)
72.8 ± 17.1
69.8 ± 18.7
74.2 ± 16.3
3.2 ± 3.1
4.9 ± 5.1
2.8 ± 2.2a
Highest serum lactate level (mmol/L), mean ± sd (n = 543) Admission diagnosis, n (%)b Respiratory disease
298 (35.4)
37 (25.7)
261 (37.4)a
Bacterial infection
237 (28.2)
42 (29.2)
195 (28)
Viral infection
203 (24.1)
22 (15.3)
181 (26)a
Noncardiac surgery
204 (24.2)
25 (17.4)
179 (25.8)a
Cardiac surgery
108 (12.8)
34 (23.6)
74 (10.6)a
Seizures
62 (7.4)
2 (1.4)
60 (8.6)a
Any shock
49 (5.8)
23 (16)
26 (3.7)a
Hypovolemic shock
19 (2.3)
5 (3.5)
Septic shock
27 (3.2)
13 (9)
14 (2) 14 (2)a
Hemorrhagic shock
6 (0.7)
4 (2.8)
2 (0.3)a
Cardiogenic shock
15 (1.7)
8 (5.6)
7 (1)a
Polytrauma without head trauma
18 (2.1)
3 (2.1)
15 (2.2)
Severe head trauma
11 (1.3)
3 (2.1)
8 (1.1)
272 (32.3)
57 (39.6)
235 (33.7)
Otherc
Pcapo2 = capillary partial pressure in oxygen. a Significant difference (p < 0.05) between transfused cases and nontransfused cases. b More than one diagnosis can be attributed to one patient. c Intoxication (30), arrhythmia (21), intracranial hemorrhage (16), diabetic ketoacidosis (14), brain death (6), severe burn (4), cardiorespiratory arrest (3), etc.
28-day mortality (crude OR, 9.49; 95% CI, 4.41–20.43; p < 0.001) was reduced but still statistically significant after controlling for severity of illness at PICU entry (adjusted OR, 5.12; 95% CI, 2.18–12.02; p < 0.001) (Tables 3 and 4). 508
www.pccmjournal.org
Mechanically ventilated transfused cases were intubated for longer periods of time than nontransfused cases (14.1 ± 32.6 d and 4.3 ± 9.6 d, respectively; p < 0.001). This association remained statistically significant after multivariable adjustment (hazard ratio July 2015 • Volume 16 • Number 6
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Feature Articles
Table 2.
Data on RBC Transfusions in Patients Transfused in PICU
Number of RBC transfusions
578 144 (17.1)
Patients who received at least one transfusion, n (%) Incidence density (transfusion events/100 patient-days)
11.2
Volume of blood during first transfusion, mL/kg
11.1 (7.3–15)
Hemoglobin level within 24 hr before first transfusion, g/L
77.5 (63.8–94)
Time from PICU admission to first transfusion, d
1 (0–1)
Pediatric Risk of Mortality score within 24 hr before first transfusion
7 (3–12)
Pediatric Logistic Organ Dysfunction score within 24 hr before first transfusion
10 (1–12)
Length of storage of packed RBC units, d
13 (9–22)
a
Data available for 354 transfusion events. Continuous variables are described with median (interquartile range).
a
[HR] of extubation [reflecting the rate of extubation for transfused cases vs nontransfused cases], 0.59; 95% CI, 0.45–0.79; p < 0.001) (Table 4). We observed an adjusted dose-effect relationship between RBC transfusion and length of mechanical ventilation (Fig. 2A). PICU length of stay was significantly increased in transfused cases (12.4 ± 26.2 d vs 4.9 ± 10.2 d; p < 0.001). The adjusted HR of PICU discharge for transfused cases versus nontransfused cases was 0.7 (95% CI, 0.57–0.85; p < 0.001) (Table 4). When considering different levels of exposure to RBCs, we observed an adjusted dose-effect relationship (Fig. 2B). Tertiary Outcomes. The worst Pao2 was lower in transfused cases than in nontransfused cases (61.1 ± 26.5 and 83.5 ± 28.1 torr, respectively; p < 0.001), while it did not differ between the two groups at PICU admission (Table 1). This difference remained statistically significant even when children with and without a cyanotic heart disease were analyzed separately (Table 3). Furthermore, the worst Pao2 significantly decreased during the 6 hours following the first RBC transfusion (mean difference, 25.6 torr; p = 0.029) (Table 5). Pulse oximetry analysis also showed a small decrease between pretransfusion and posttransfusion worst oxygen saturation (Spo2), statistically but not clinically significant (95.9% ± 7.1% vs 95% ± 7%; p < 0.001) (Table 5). The proportion of ARDS was similar in transfused and nontransfused cases (5.3% vs 4.2%; p = 0.54) (Table 3). Arterial hypotension was observed more frequently in transfused cases than in nontransfused cases (50.4% vs 28.3%; p < 0.001), but pre transfusion and posttransfusion prevalence of this complication did not differ (p = 1) (Table 5). Renal replacement therapy was more frequent in transfused than in nontransfused cases (8.4% vs 0.6%; p < 0.001) (Table 3). It was also more prevalent after than before the first RBC transfusion (8.4% vs 2.3%; p = 0.008) (Table 5). Finally, more transfused cases developed SIRS, sepsis, severe sepsis, or septic shock, but pretransfusion and posttransfusion prevalence of all these did not statistically differ (Table 5).
DISCUSSION This prospective observational study describes several associations between RBC transfusions and clinical outcomes in Pediatric Critical Care Medicine
critically ill children. We report an independent association and a dose-effect relationship between RBC transfusion and increased morbidity, with an increased risk of NPMODS, a longer PICU stay, and an increased duration of mechanical ventilation. RBC Transfusions and Worse Outcomes in Critically Ill Patients A multicenter study conducted by Bateman et al (8) reported that after correction for age and for admission PRISM III score, transfused patients had an increased risk of death, a higher rate of nosocomial infections, more cardiac or respiratory dysfunction, a longer duration of mechanical ventilation, and a longer PICU length of stay than nontransfused patients. Other studies showing worse outcomes associated with RBC transfusions have focused on specific PICU populations (e.g., severely burned children [19] and postoperative children after an open heart surgery [1]) or on a more restricted number of adverse events associated with transfusions in a general PICU population (e.g., bloodstream infections [20] and increased use of some intensive care resources [oxygen, mechanical ventilation, and vasoactive agents infusion] [21]). Most critically ill patients have systemic inflammation in association with the illness that led to ICU admission (first hit). These patients are deemed more vulnerable to any sequential insult (second hit), which may amplify the already primed inflammatory cascade (22). This two-hit model explains many cases of multiple organ failure observed in the intensive care (23). RBC transfusion can be a potential second hit (22). Based on this concept, RBC transfusion may have a more significant impact in the critically ill as compared with less ill patients (24). Our data confirm that RBC transfusions are associated with worse outcomes in critically ill children and are of importance because of the large number of potential complications looked for, the case-mix which included a large multidisciplinary PICU population, the attention paid to the temporal relationship between transfusions and the potential outcomes, and the prospective manner in which data were collected. www.pccmjournal.org
509
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Demaret et al
Table 3.
Bivariate Association Between RBC Transfusions and Outcome Nontransfused Cases (NTD+/NTD– or Mean ± sd)
Transfused Cases (TD+/TD– or Mean ± sd)
New or progressive MODS
55/698
44/144
5.14 (3.28, 8.06)
< 0.001
New MODS
37/698
23/130
3.84 (2.2, 6.72)
< 0.001
Progressive MODS
45/662
43/132
6.62 (4.13, 10.64)
< 0.001
Death during PICU stay
11/698
17/144
8.36 (3.83, 18.27)
< 0.001
24/696
19/130
4.79 (2.54, 9.03)
< 0.001
Clinical Outcome
Bivariate OR (95% CI)
p
Primary outcome
Secondary outcomes Nosocomial infection Pneumonia
5/692
7/129
7.88 (2.46, 25.24)
< 0.001
Tracheitis
4/692
3/129
4.1 (0.91, 18.52)
0.047
Bacteremia
1/692
7/129
39.65 (4.84, 325.1)
< 0.001
Urinary tract infection
8/693
6/129
4.18 (1.42, 12.25)
0.005
Other
8/692
8/129
5.65 (2.08, 15.35)
< 0.001
Endotracheal intubation
269/698
84/129
2.98 (2.01, 4.41)
< 0.001
Length of endotracheal intubation (d)
4.3 ± 9.6
14.1 ± 32.6
< 0.001
PICU length of stay
4.9 ± 10.2
12.4 ± 26.2
< 0.001
11/698
19/144
9.49 (4.41, 20.43)
Acute respiratory distress syndrome
29/698
7/131
1.3 (0.56, 3.01)
Worst Po2
68.5 ± 31
62.3 ± 26.9
0.003
Worst Pcapo2
60.5 ± 29.6
58.1 ± 28.1
0.13
Cyanotic heart disease
50.4 ± 16.5
53.4 ± 18.5
0.674
No cyanotic heart disease
61.1 ± 30.1
58.8 ± 29.5
0.093
Worst Parto2
83.5 ± 28.1
61.1 ± 26.5
< 0.001
Cyanotic heart disease
69.6 ± 27.5
55.3 ± 27
No cyanotic heart disease
85.6 ± 27.6
68.9 ± 25.2
< 0.001
91.1 ± 12
85.2 ± 14.4
< 0.001
Cyanotic heart disease
86.3 ± 12.8
78.2 ± 16.8
0.017
No cyanotic heart disease
91.5 ± 11.9
87.7 ± 12.6
< 0.001
2/698
7/130
19.81 (4.07, 96.45)
< 0.001
169/698
66/131
3.18 (2.17, 4.67)
< 0.001
4/698
11/131
15.9 (5, 50.8)
< 0.001
Systemic inflammatory response syndrome
364/696
87/131
1.8 (1.22, 2.67)
0.003
Sepsis
207/696
55/131
1.71 (1.17, 2.51)
0.006
Severe sepsis
39/698
22/131
3.41 (1.95, 5.97)
< 0.001
Septic shock
12/698
15/131
7.39 (3.38, 16.19)
< 0.001
28-d mortality
< 0.001
Tertiary outcomes
a
b
Worst Spo2c
Cardiac arrest Arterial hypotension Renal replacement therapy
0.540
0.035
NTD+ = nontransfused cases with determinant, NTD– = nontransfused cases without determinants, TD+ = transfused cases with determinant, TD– = transfused cases without determinant, OR = odds ratio, MODS = multiple organ dysfunction syndrome, Pcapo2 = capillary partial pressure in oxygen, Parto2 = arterial partial pressure in oxygen, Spo2 = pulsed oxygen saturation. a Information available for 395 nontransfused cases (NTCs) and for 36 transfused cases (TCs). b Information available for 192 NTCs and for 88 TCs. c Information available for 696 NTCs and for 130 TCs.
510
www.pccmjournal.org
July 2015 • Volume 16 • Number 6
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Feature Articles
Table 4. Adjusted Association Between RBC Transfusion and Clinical Outcomes (Primary and Secondary Outcomes) Clinical Outcome
p
Multivariable Association
Multivariable logistic regression
Odds ratio (95% CI)
New or progressive multiple organ dysfunction syndrome
3.85 (2.38, 6.24)
< 0.001
Nosocomial infectionb
3.31 (1.67, 6.56)
0.001
28-d mortalityb
5.12 (2.18, 12.02)
a
Cox regression
Hazard ratio (95% CI)
Duration of mechanical ventilation PICU length of stay
< 0.001
a,c
a,d
0.59 (0.45, 0.79)
< 0.001
0.7 (0.57, 0.85)
< 0.001
Adjustment made for age, admission Pediatric Risk of Mortality (PRISM), and diagnosis of congenital heart disease. b Adjustment made for admission PRISM only (because of the small number of outcome events). c Only mechanically ventilated children were included in the model (n = 353); the model gives the hazard ratio of extubation: a hazard ratio significantly < 1 indicates a lower risk of extubation at any time for transfused children as compared with nontransfused ones. d The model gives the hazard ratio of PICU discharge: a hazard ratio significantly < 1 indicates a lower risk of PICU discharge at any time for transfused children as compared with nontransfused ones. Transfused children are compared with nontransfused. The goodness-of-fit of the logistic regression models was evaluated using the Hosmer-Lemeshow statistic, showing there was not sufficient evidence to conclude a bad fit. No significant interaction was found in any of these multivariable models. a
Protopathic Bias or the Importance of the Temporal Relationship Reverse causality occurs when the outcome precedes and increases the risk for the putative exposure (25). Protopathic bias is a specific type of reverse causality: it occurs when an outcome is associated with an exposure that actually results from early signs and symptoms of the outcome under study (26). In other
words, it is not sufficient to observe development of a complication or condition after transfusion; one must be certain the complication or condition was not present before the transfusion. Rigorous attention was paid during our study in order to avoid the presence of protopathic bias. When analyses were undertaken to compare pretransfusion and posttransfusion values of the variables studied, several associations between RBC transfusions
Figure 2. Adjusted survival curves showing the time from PICU admission to PICU discharge (A) and to extubation for ventilated children (B), stratified according to different levels of RBC transfusion. Adjustment was made for age and for admission Pediatric Risk of Mortality score. A, The corresponding hazard ratios are significantly higher for each of the two smallest categories (i.e., 0 and 1–20 cc/kg of RBCs during PICU stay) as compared with the reference group (> 80 cc/kg) (p < 0.05). In this analysis, a higher hazard ratio can be interpreted as a higher risk of PICU discharge and thus a shorter length of stay in PICU. B, The corresponding hazard ratios are significantly higher for each of the three smallest categories (i.e., 0, 1–20, and 21–40 cc/ kg of RBCs during PICU stay) as compared with the reference group (> 80 cc/kg) (p < 0.05). In this analysis, a higher hazard ratio can be interpreted as a higher risk of extubation during PICU stay and thus a shorter duration of mechanical ventilation.
Pediatric Critical Care Medicine
www.pccmjournal.org
511
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Demaret et al
Table 5.
Paired Analysis for Comparison of Pretransfusion and Posttransfusion Values Pretransfusion
Posttransfusion
pa
10.6 ± 8.4
11.4 ± 9.8
0.29
70/131
85/131
0.014
Cardiovascular dysfunction
7/131
5/131
0.625
Hematological dysfunction
49/131
56/131
0.248
Neurological dysfunction
90/131
83/131
0.281
Hepatic dysfunction
30/131
39/131
0.093
Renal dysfunction
12/131
18/131
0.031
Clinical Outcome
Primary outcome: organ dysfunctions Worst daily Pediatric Logistic Organ Dysfunction score Respiratory dysfunction
Tertiary outcomes Acute respiratory distress syndrome
8/131
7/131
1
Worst Pao2b,c
154.9 ± 106
129.4 ± 81.9
0.029
Worst Spo2b,d
95.9 ± 7.1
95 ± 6.8
Arterial hypotension
67/131
66/131
1
Renal replacement therapy
3/131
11/131
0.008
Systemic inflammatory response syndrome
75/131
87/131
0.088
Sepsis
46/131
55/131
0.093
Severe sepsis
18/131
22/131
0.289
Septic shock
11/131
15/131
0.219
< 0.001
Spo2 = pulsed oxygen saturation. a McNemar chi-square for comparison of proportions and Wilcoxon signed-rank test for comparison of means. b Lowest value during the 2 hr before and during the 6 hr after the first transfusion. c Pretransfusion and posttransfusion values are available for 67 transfused patients. d Pretransfusion and posttransfusion values are available for 129 transfused patients.
and clinical outcomes lost their significance, which highlights the possible importance of protopathic bias in this type of research. RBC Transfusions and Oxygenation Oxygenation markers did not improve during the 6 hours following the first RBC transfusion; we rather observed a trend toward worsening of Po2 and Spo2 (Table 5). The posttransfusion decrease of Po2 is particularly questioning, even if we did not expect to a Po2 increase considering a mean pretransfusion value of 154.9 ± 106 mm Hg. Obviously, several factors not related to RBC transfusions may interfere with these oxygenation variables. However, several explanations related to RBC transfusions might also explain this observation, including transfusion-associated respiratory complications such as TRALI (3) and TACO (4). RBCs are frequently prescribed to improve oxygen delivery (27). An RBC transfusion increases arterial oxygen content (Cao2) by increasing the hemoglobin level. It thereby should increase global oxygen delivery (Do2). However, it is not always the case. Indeed, an RBC transfusion increases blood viscosity, which can decrease cardiac output in normovolemic patients whose cardiac function is not impaired by myocardial ischemia (28) and therefore nullify the global Do2 increase that occurs with a higher hemoglobin level. Furthermore, respiratory complications associated with RBC transfusion may cause a decrease in oxygen saturation 512
www.pccmjournal.org
that might also counteract the positive effect of an increased hemoglobin level on Cao2. In addition, RBC transfusions may also exert several effects that reduce tissular Do2 and oxygen consumption. Although several studies have reported that RBC transfusions increase tissue oxygenation (29, 30), others have led to conflicting results (31, 32). Actually, a growing body of evidence suggests that stored RBCs may have adverse effects on microcirculatory flow and oxygen use (33). Many changes occur during RBCs storage, including diminished adenosine triphosphate and 2,3-diphosphoglycerate levels (34), decreased RBC deformability over time (35), hemolysis resulting in generation of proinflammatory microparticles (36), and altered vasoregulation attributable to certain effects stored RBCs have on nitric oxide metabolism (37). Several of these changes may disturb oxygen capture in the lungs and may impact the ability of RBC to increase oxygen delivery and consumption in certain vascular territories, which might contribute to organ dysfunction in critically ill patients. This could explain the greater risk of NPMODS we observed in some transfused children. Limitations and Strengths Our study has several limitations. It was conducted in a single center, which limits its external validity; however, our critical care unit is comparable to most multidisciplinary July 2015 • Volume 16 • Number 6
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Feature Articles
university-affiliated North-American PICUs with regard to case-mix and severity of illness (8, 10, 38). Second, we did not collect data on the length of time between the first RBC transfusion and the adverse outcomes. It is therefore possible that an outcome was associated with RBC transfusion even though it actually occurred a few days after the transfusion. This may have led to an overestimation of the association between RBC transfusions and some of the outcomes assessed. Third, we did not consider RBC transfusions that occurred before PICU admission. Ninety-one children had been transfused within the 24 hours prior to PICU admission and were not considered as transfused children in our study. Fourth, our study was not able to highlight a cause-effect relationship between RBC transfusions and outcomes in critically ill children: only a randomized-controlled trial could establish such a causal link. Caution is thus warranted in the interpretation of our results. Fifth, transfused and nontransfused children are different. The need to receive an RBC transfusion selects a group of subjects at greater risk of adverse outcomes (39). Clinical research in transfusion medicine is subject to confounding by indication, and our study does not escape to this pitfall. This bias occurs when studying the effect of a treatment (e.g., RBC transfusion) while the indication for the treatment causes the outcome (40). The sicker a child is, the more transfusions will be indicated. Also, the sicker a child is, the higher the risk that the child will have a poor outcome (40). Even if we adjusted for several confounders, it is possible that our results are still influenced by residual confounding and that we overestimated the association between adverse outcomes and RBC transfusions (39). Sixth, we did not consider plasma and platelet transfusions in our analyses. Number of children received at least one platelet and/or one plasma transfusion in addition to their RBC transfusion. It is possible that the associations we observed are partially explained by plasma and/or platelet transfusions. Seventh, patients readmitted to PICU more than 24 hours after discharge were considered as new cases. It can be argued that this approach leads to biased results because of correlation between cases. However, we undertook a sensitivity model using a generalized estimating equation approach (41) and showed that considering these patients as one case would have resulted in findings similar to those of our multivariable models. Finally, because of the observational nature of our study, we were not allowed to ask the physicians to prescribe laboratory analyses for the study, which resulted in some missing values for some variables (e.g., Pao2). Our study also has several strengths. This is the first study to prospectively consider such a high number of outcomes related to RBC transfusions in a general PICU population. The list of outcomes possibly associated with RBC transfusion was decided upon before the study was initiated. The study included all consecutive PICU admissions over a 1-year period, which resulted in a case-mix with a limited risk of selection bias and no influence due to seasonal variation. We paid a careful attention to the temporal relationship between the first RBC transfusion in PICU and outcomes. Our paired analysis for the comparison of pretransfusion and posttransfusion values highlights the risk of protopathic Pediatric Critical Care Medicine
bias in this type of research. Finally, the prospective nature of our study was a major asset to minimize information bias.
CONCLUSIONS This observational prospective study suggests that RBC transfusions are associated with worse outcomes in critically ill children. We found that transfusions were associated with a greater risk of developing NPMODS, with prolonged mechanical ventilation and prolonged PICU stay. Furthermore, several of our findings raise questions with regard to the ability of stored RBCs to improve oxygenation in critically ill children and highlight the importance of temporal relationship when assessing the clinical effects of transfusions. Despite some limitations, findings from this study may help caregivers to figure out what can be the risk-benefit ratio of RBC transfusion.
REFERENCES
1. Dodd RY: Current risk for transfusion transmitted infections. Curr Opin Hematol 2007; 14:671–676 2. Hendrickson JE, Hillyer CD: Noninfectious serious hazards of transfusion. Anesth Analg 2009; 108:759–769 3. Toy P, Gajic O, Bacchetti P, et al; TRALI Study Group: Transfusionrelated acute lung injury: Incidence and risk factors. Blood 2012; 119:1757–1767 4. Li G, Rachmale S, Kojicic M, et al: Incidence and transfusion risk factors for transfusion-associated circulatory overload among medical intensive care unit patients. Transfusion 2011; 51:338–343 5. Vamvakas EC, Blajchman MA: Transfusion-related immunomodulation (TRIM): An update. Blood Rev 2007; 21:327–348 6. Armano R, Gauvin F, Ducruet T, et al: Determinants of red blood cell transfusions in a pediatric critical care unit: A prospective, descriptive epidemiological study. Crit Care Med 2005; 33:2637–2644 7. Kneyber MC, Hersi MI, Twisk JW, et al: Red blood cell transfusion in critically ill children is independently associated with increased mortality. Intensive Care Med 2007; 33:1414–1422 8. Bateman ST, Lacroix J, Boven K, et al; Pediatric Acute Lung Injury and Sepsis Investigators Network: Anemia, blood loss, and blood transfusions in North American children in the intensive care unit. Am J Respir Crit Care Med 2008; 178:26–33 9. Laverdière C, Gauvin F, Hébert PC, et al; Canadian Critical Care Trials Group: Survey on transfusion practices of pediatric intensivists. Pediatr Crit Care Med 2002; 3:335–340 10. Lacroix J, Hébert PC, Hutchison JS, et al; TRIPICU Investigators; Canadian Critical Care Trials Group; Pediatric Acute Lung Injury and Sepsis Investigators Network: Transfusion strategies for patients in pediatric intensive care units. N Engl J Med 2007; 356:1609–1619 11. Grol R, Grimshaw J: From best evidence to best practice: Effective implementation of change in patients’ care. Lancet 2003; 362:1225–1230 12. Pollack MM, Ruttimann UE, Getson PR: Pediatric risk of mortality (PRISM) score. Crit Care Med 1988; 16:1110–1116 13. Leteurtre S, Martinot A, Duhamel A, et al: Validation of the Paediatric Logistic Organ Dysfunction (PELOD) score: Prospective, observational, multicentre study. Lancet 2003; 362:192–197 14. Goldstein B, Giroir B, Randolph A; International Consensus Conference on Pediatric Sepsis: International pediatric sepsis consensus conference: Definitions for sepsis and organ dysfunction in pediatrics. Pediatr Crit Care Med 2005; 6:2–8 15. Goldstein B, Giroir B, Randolph A: Values for systolic blood pressure. Pediatr Crit Care Med 2005; 6:500–501 16. Bernard GR, Artigas A, Brigham KL, et al: The American-European Consensus Conference on ARDS. Definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994; 149:818–824 www.pccmjournal.org
513
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Demaret et al 17. Bursac Z, Gauss CH, Williams DK, et al: Purposeful selection of variables in logistic regression. Source Code Biol Med 2008; 3:17 18. Hosmer DW, Lemeshow S, Sturdivant RX: Applied Logistic Regression, Third Edition. Hoboken, John Wiley & Sons, Inc., 2013 19. Jeschke MG, Chinkes DL, Finnerty CC, et al: Blood transfusions are associated with increased risk for development of sepsis in severely burned pediatric patients. Crit Care Med 2007; 35:579–583 20. Elward AM, Fraser VJ: Risk factors for nosocomial primary bloodstream infection in pediatric intensive care unit patients: A 2-year prospective cohort study. Infect Control Hosp Epidemiol 2006; 27:553–560 21. Goodman AM, Pollack MM, Patel KM, et al: Pediatric red blood cell transfusions increase resource use. J Pediatr 2003; 142:123–127 22. McIntyre LA, Hebert PC: Can we safely restrict transfusion in trauma patients? Curr Opin Crit Care 2006; 12:575–583 23. Moore FA, Sauaia A, Moore EE, et al: Postinjury multiple organ failure: A bimodal phenomenon. J Trauma 1996; 40:501–510 24. Spinella PC, Doctor A, Blumberg N, et al: Does the storage duration of blood products affect outcomes in critically ill patients? Transfusion 2011; 51:1644–1650 25. Murk W, Risnes KR, Bracken MB: Prenatal or early-life exposure to antibiotics and risk of childhood asthma: A systematic review. Pediatrics 2011; 127:1125–1138 26. Tamim H, Monfared AA, LeLorier J: Application of lag-time into exposure definitions to control for protopathic bias. Pharmacoepidemiol Drug Saf 2007; 16:250–258 27. Demaret P, Tucci M, Ducruet T, et al: Red blood cell transfusion in critically ill children (CME). Transfusion 2014; 54:365–375 28. Roberson RS, Bennett-Guerrero E: Impact of red blood cell transfusion on global and regional measures of oxygenation. Mt Sinai J Med 2012; 79:66–74 29. van Hoften JC, Verhagen EA, Keating P, et al: Cerebral tissue oxygen saturation and extraction in preterm infants before and
514
www.pccmjournal.org
after blood transfusion. Arch Dis Child Fetal Neonatal Ed 2010; 95:F352–F358 30. Figaji AA, Zwane E, Kogels M, et al: The effect of blood transfusion on brain oxygenation in children with severe traumatic brain injury. Pediatr Crit Care Med 2010; 11:325–331 31. Creteur J, Neves AP, Vincent JL: Near-infrared spectroscopy technique to evaluate the effects of red blood cell transfusion on tissue oxygenation. Crit Care 2009; 13(Suppl 5):S11 32. Kiraly LN, Underwood S, Differding JA, et al: Transfusion of aged packed red blood cells results in decreased tissue oxygenation in critically injured trauma patients. J Trauma 2009; 67:29–32 33. Weinberg JA, MacLennan PA, Vandromme-Cusick MJ, et al: The deleterious effect of red blood cell storage on microvascular response to transfusion. J Trauma Acute Care Surg 2013; 75:807–812 34. Karam O, Tucci M, Toledano BJ, et al: Length of storage and in vitro immunomodulation induced by prestorage leukoreduced red blood cells. Transfusion 2009; 49:2326–2334 35. Bennett-Guerrero E, Veldman TH, Doctor A, et al: Evolution of adverse changes in stored RBCs. Proc Natl Acad Sci U S A 2007; 104:17063–17068 36. Kim-Shapiro DB, Lee J, Gladwin MT: Storage lesion: Role of red blood cell breakdown. Transfusion 2011; 51:844–851 37. Doctor A, Spinella P: Effect of processing and storage on red blood cell function in vivo. Semin Perinatol 2012; 36:248–259 38. Pollack MM, Patel KM, Ruttimann UE: PRISM III: An updated Pediatric Risk of Mortality score. Crit Care Med 1996; 24:743–752 39. van de Watering L; Biomedical Excellence for Safer Transfusion C: Pitfalls in the current published observational literature on the effects of red blood cell storage. Transfusion 2011; 51:1847–1854 40. Middelburg RA, van de Watering LM, van der Bom JG: Blood transfusions: Good or bad? Confounding by indication, an underestimated problem in clinical transfusion research. Transfusion 2010; 50:1181–1183 41. Zeger SL, Liang KY: Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986; 42:121–130
July 2015 • Volume 16 • Number 6
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Objective: To identify the potential complications associated with RBC transfusions. Design: Prospective observational study. Setting: PICU in a tertiary children’s hospital. Patients: All children consecutively admitted to our PICU during a 1-year period. Interventions: None. Measurements and Main Results: Data were abstracted from medical charts prospectively. Outcomes possibly attributable to RBC transfusions were looked for daily. In transfused cases, it was considered that an outcome was associated with a transfusion only if it was observed after the first RBC transfusion. During the 1-year study period, 913 consecutive admissions were documented, 842 of which were included. Among them, 144 (17%) were transfused at least once. When comparing transfused cases with nontransfused cases, the odds ratio for new or progressive multiple organ dysfunction syndrome was 5.14 (95% CI, 3.28–8.06; p < 0.001). This association remained statistically significant in the multivari*See also p. 585. 1 Division of Pediatric Critical Care Medicine, Department of Pediatrics, CHC, Liège, Belgium. 2 Division of Pediatric Critical Care Medicine, Department of Pediatrics, Sainte-Justine Hospital and Université de Montréal, Montreal, QC, Canada. 3 Pediatric Intensive Care Unit, Geneva University Hospital, Geneva, Switzerland. 4 Department of Social and Preventive Medicine, Research Center, SainteJustine Hospital and Université de Montréal, Montreal, QC, Canada. 5 Department of Pediatrics, Research Center, Sainte-Justine Hospital, Université de Montréal, Montreal, QC, Canada. This work was performed at PICU, CHU Sainte-Justine, Montréal, QC, Canada. Supported, in part, by Fonds de la Recherche en Santé du Québec (grant #24460). Dr. Trottier consulted and lectured for Merck Frosst and for GSK Belgium. She received support for the development of educational presentations from Merck Frosst and received support for travel from GSK Belgium. Dr. Lacroix’s institution received grant support from Fonds de Recherche en Sante de Quebec (grant #24480). The remaining authors have disclosed that they do not have any potential conflicts of interest. For information regarding this article, E-mail: [email protected] Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies DOI: 10.1097/PCC.0000000000000423
Pediatric Critical Care Medicine
able analysis (odds ratio, 3.85; 95% CI, 2.38–6.24; p < 0.001). Transfused cases were ventilated longer than nontransfused cases (14.1 ± 32.6 vs 4.3 ± 9.6 d, p < 0.001), even after adjustment in a Cox model. The PICU length of stay was significantly increased for transfused cases (12.4 ± 26.2 vs 4.9 ± 10.2 d, p < 0.001), even after controlling for potential confounders. The paired analysis for comparison of pretransfusion and posttransfusion values showed that the arterial partial pressure in oxygen was significantly reduced within the 6 hours after the first RBC transfusion (mean difference, 25.6 torr, 95% CI, 5.7–45.4; p = 0.029). The paired analysis also showed an increased proportion of renal replacement therapy. Conclusions: RBC transfusions in critically ill children were associated with prolonged mechanical ventilation and prolonged PICU stay. The risk of new or progressive multiple organ dysfunction syndrome was also increased in some transfused children. Furthermore, our study questions the ability of stored RBCs to improve oxygenation in critically ill children. Practitioners should take into account these data when prescribing an RBC transfusion to PICU patients. (Pediatr Crit Care Med 2015; 16:505–514) Key Words: child; erythrocyte; intensive care units; outcome; pediatric; transfusion
R
BC transfusion is the quickest way to increase hemoglobin concentration and can be lifesaving for patients with hemorrhagic shock or severe anemia. However, issues with regard to the safety of blood transfusions have emerged in recent years. Advances in donor screening and in blood product management have markedly decreased the risk of pathogen transmission (1). The focus has now shifted to noninfectious serious hazards of transfusion (NISHOT), which are increasingly described in the literature. The most common NISHOT is mistransfusion or transfusing the incorrect blood product to a patient (2). However, there are numerous other NISHOTs, including transfusion-related acute lung injury (TRALI) (3), transfusion-associated circulatory overload (TACO) (4), alloimmunization (2), and transfusionrelated immunomodulation (5). These adverse reactions may www.pccmjournal.org
505
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Demaret et al
exacerbate organ dysfunction and therefore worsen the clinical condition of transfusion recipients. RBC transfusions are frequent in critically ill children. Approximately 15–23% of children admitted to PICUs receive at least one RBC transfusion (6, 7), and nearly one in two children who stay in PICU more than 48 hours will be transfused (8). Stated transfusion practice patterns show significant variation among pediatric intensivists (9). Although there is high-level evidence published that shows it is safe to adopt a restrictive transfusion strategy (i.e., to transfuse RBCs only if the hemoglobin level drops below 70 g/L) in stable noncyanotic critically ill children (10), some data show that it can take several years before new knowledge is applied at the bedside (11). This might be explained by poor physician knowledge regarding the adverse effects of blood products. It is thus important to describe the clinical outcomes associated with RBC transfusions in critically ill children, in order that physicians can better estimate the risk-benefit ratio of transfusing RBCs in PICU. We therefore conducted a 1-year prospective observational cohort study to determine adverse outcomes associated with RBC transfusion in a multidisciplinary PICU population.
MATERIAL AND METHODS Study Site The PICU of Sainte-Justine University Hospital is a multidisciplinary 24-bed pediatric critical care unit, serving both medical and surgical specialties. On average, there are 1,000 admissions per year. Study Population All consecutive admissions to the PICU, from April 21, 2009, to April 20, 2010, were prospectively screened for recruitment. Exclusion criteria were newborn with gestational age less than 40 weeks at the time of PICU admission, age less than 3 days or more than 18 years at PICU admission, pregnancy, or admission just after labor. Cases were defined as transfused cases if at least one RBC transfusion was given during the PICU stay and as nontransfused cases if no RBC transfusion was given. Data Collection and Management Trained research coordinators prospectively collected data daily in a validated case report form. All information was abstracted from patient medical charts. Baseline data collected within 24 hours after PICU entry included age, gender, weight, medical history, and admission diagnoses (more than one diagnosis could be attributed to each patient). A predictive score of mortality, the Pediatric Risk of Mortality (PRISM) score (12), and a descriptive score of severity of multiple organ dysfunction, the Pediatric Logistic Organ Dysfunction (PELOD) score (13) were used to describe severity of illness at entry. This study was observational and provided no transfusion guidelines to medical staff. Institutional review board approval was obtained, waiving the requirement for written informed consent. 506
www.pccmjournal.org
Outcomes The list of outcomes under study was generated a priori. In nontransfused patients, each outcome was documented as a single event that occurred at any moment during the prospective follow-up of a patient’s ICU stay. In transfused patients, a variable was considered an outcome associated with a transfusion only if it was observed after the first RBC transfusion event. Primary Outcome. The primary outcome measure was “new or progressive multiple organ dysfunction syndrome” (NPMODS) (10). This was the proportion of patients with development or progression of multiple organ dysfunction syndrome (MODS), as defined by Goldstein et al (14). In nontransfused patients, new MODS was diagnosed if a patient with no or one organ dysfunction at admission developed two or more organ dysfunctions during the PICU stay, and progressive MODS was considered present if a patient who already had MODS at PICU entry developed dysfunction of at least one other organ during the PICU stay. In transfused patients, new MODS was diagnosed if a patient with no or one organ dysfunction at the time the first RBC transfusion developed two or more organ dysfunctions during the remaining PICU stay after the first RBC transfusion and progressive MODS was diagnosed if a patient who already had MODS when the first RBC transfusion was initiated developed dysfunction of at least one other organ after transfusion and during the remainder of the PICU stay. Patients who died were considered to have the primary outcome. Secondary Outcomes. Nosocomial infections were defined as infections diagnosed at least 2 days after PICU entry. Presence of suspected bacterial or viral infection was extracted from medical records (suspected infection); the infection was considered proven if a bacterial or viral culture in specimens drawn in PICU was positive. We also recorded PICU length of stay, duration of mechanical ventilation, and 28-day mortality. Tertiary Outcomes. Hypotension was defined as a systolic blood pressure below the 5th percentile for age (15). Systemic inflammatory response syndrome (SIRS), sepsis, severe sepsis, and septic shock were defined according to the definitions from the international pediatric sepsis consensus conference (14). Acute respiratory distress syndrome (ARDS) was defined according to the following criteria: acute onset, Pao2/Fio2 ratio up to 200, bilateral infiltrates on chest radiograph, and no evidence of left atrial hypertension (16). Po2 was measured using arterial (Pao2) or capillary (Pcapo2) blood samples. We documented the worst Po2 and the worst oxygen saturation (Spo2) during the 2 hours preceding and during the 6 hours following the first RBC transfusion. Statistical Analysis Results of descriptive statistics were expressed as a fraction of the total population, mean ± sd, and/or median with interquartile range. Continuous variables were compared using Student t test if the variable was normally distributed (as attested by the Shapiro-Wilk test) or Wilcoxon test if not. The chi-square statistic was used for categorical variables. Results were considered statistically significant when the p value was less than 0.05. July 2015 • Volume 16 • Number 6
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Feature Articles
For all dichotomous outcomes, a crude odds ratio (OR) with its 95% CI was calculated using logistic regression. We decided a priori to conduct multivariable analyses for the primary outcome (NPMODS) and the secondary outcomes (nosocomial infection, 28-d mortality, duration of mechanical ventilation, and PICU length of stay). We used multivariable logistic regression for dichotomous variables and Cox regression for time variables. Potential confounders were selected based on the following criteria: bivariate association with RBC transfusion, bivariate association with the primary outcome (p value cutoff, 0.25) (17), clinical relevance (based on previous published literature and on the authors’ expertise), and absence of correlation with another retained confounder. If the occurrence of an outcome was too low to include all the retained confounders in the multivariable model, we adjusted only for the severity of illness, using the PRISM score, because we believed it to be the most pertinent confounder. Continuous variables were categorized if they were not linear in the logit, as tested with the quartile design variable method described by Hosmer et al (18). Interactions were tested one by one (between the variable of interest [transfusion] and age, cardiopathy, and PRISM), using likelihood ratio. Fitting of the final logistic regression models was tested using the Hosmer-Lemeshow statistic. For multivariable survival analyzes, extended Cox models were used with transfusion as a time-dependent variable. Survival curves adjusted to the average value of age and admission PRISM score (for several levels of exposure to RBCs) were obtained from Cox regression. In transfused cases, we compared pretransfusion and posttransfusion values of several variables, using McNemar chi-square for comparison of proportions and Wilcoxon signed-rank test for comparison of means. All statistical analyses were done by P.D. and T.D., using SPSS statistical software (SPSS, release 17.0.1; SPSS, Chicago, IL).
RESULTS Over the 1-year study period, there were 913 consecutive PICU admissions involving 802 patients (Fig. 1); 71 cases were excluded. A total of 842 admissions (cases) were retained for analysis. At least one RBC transfusion was given in 144 cases (17.1%). Data at First Day in PICU Data collected within 24 hours after PICU entry are reported in Table 1. Transfused cases were younger, more severely ill (higher PRISM and PELOD scores), and more likely to have congenital heart disease compared with nontransfused cases (34.7% vs 17.3%; p < 0.05). The hemoglobin concentration at admission was significantly lower in transfused cases (92 ± 29.6 g/L vs 115.7 ± 21.4 g/L; p < 0.05). The most frequent admission diagnoses were respiratory disease, bacterial or viral infection, and elective surgery (cardiac and noncardiac). Some admission diagnoses were significantly more prevalent in nontransfused cases (respiratory disease, viral infection, noncardiac surgery, and seizures), whereas others were significantly more prevalent in transfused cases (cardiac surgery and shock). Data on RBC Transfusions Data on transfusions are reported in Table 2. Five-hundred seventy-eight transfusions were given during the 1-year study period. The mean hemoglobin level before the first transfusion was 77.3 ± 27.2 g/L. Of the 144 first transfusion events, 110 (76.4%) were prescribed during the first 2 days in PICU.
Outcomes and Their Association With RBC Transfusion Primary Outcome. NPMODS was observed in 99 of 842 patients (Table 3). The crude OR for development of NPMODS after the first transfusion event was 5.14 (95% CI, 3.28–8.06; p < 0.001). When controlling for potential confounders (admission PRISM score, age, and presence of a congenital heart disease), the association between RBC transfusion and NPMODS remained statistically significant (adjusted OR, 3.85; 95% CI, 2.38–6.24; p < 0.001) (Table 4). Secondary Outcome. Nosocomial infections were increased in transfused cases (crude OR, 4.79; 95% CI, 2.54–9.03; p < 0.001), even after adjustment for the severity of illness at PICU entry (adjusted OR, 3.31; 95% CI, 1.67–6.56; p = 0.001). The association Figure 1. Flow chart of study patients. MODS = multiple organ dysfunction syndrome. between RBC transfusion and Pediatric Critical Care Medicine
www.pccmjournal.org
507
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Demaret et al
Table 1.
Patient Characteristics at the Day of Admission to PICU
Patient Characteristic
All Cases (n = 842)
Transfused Cases (n = 144)
Nontransfused Cases (n = 698)
Gender (male), n (%)
434 (51.5)
74 (51.4)
361 (51.7)
57.7 ± 74
74.9 ± 71.3a
9.6 ± 7
5.2 ± 5.3a
9.1 ± 8.5
3.9 ± 6a
Age (mo), mean ± sd Pediatric Risk of Mortality score, mean ± sd
71.9 ± 72 6 ± 5.8
Pediatric Logistic Organ Dysfunction score, mean ± sd
4.8 ± 6.8
Congenital heart disease, n (%)
171 (20.3)
50 (34.7)
121 (17.3)a
93 (11)
37 (25.7)
56 (8)a
Cyanotic congenital heart disease, n (%) Worst Po2 (torr) Worst Pao2, mean ± sd (n = 278) Worst Pcapo2, mean ± sd (n = 467) Lowest hemoglobin (g/L), mean ± sd (n = 776)
127.1 ± 63.5
129.6 ± 83.6
126.2 ± 53.7
72.6 ± 45.9
72.3 ± 84
72.6 ± 40.1
111.5 ± 24.7
92 ± 29.6
115.7 ± 21.4a
Worst Scvo2 (%), mean ± sd Cyanotic congenital heart disease (n = 52)
64.7 ± 15.5
62.8 ± 13.6
66.9 ± 17.5
Noncyanotic congenital heart disease (n = 104)
72.8 ± 17.1
69.8 ± 18.7
74.2 ± 16.3
3.2 ± 3.1
4.9 ± 5.1
2.8 ± 2.2a
Highest serum lactate level (mmol/L), mean ± sd (n = 543) Admission diagnosis, n (%)b Respiratory disease
298 (35.4)
37 (25.7)
261 (37.4)a
Bacterial infection
237 (28.2)
42 (29.2)
195 (28)
Viral infection
203 (24.1)
22 (15.3)
181 (26)a
Noncardiac surgery
204 (24.2)
25 (17.4)
179 (25.8)a
Cardiac surgery
108 (12.8)
34 (23.6)
74 (10.6)a
Seizures
62 (7.4)
2 (1.4)
60 (8.6)a
Any shock
49 (5.8)
23 (16)
26 (3.7)a
Hypovolemic shock
19 (2.3)
5 (3.5)
Septic shock
27 (3.2)
13 (9)
14 (2) 14 (2)a
Hemorrhagic shock
6 (0.7)
4 (2.8)
2 (0.3)a
Cardiogenic shock
15 (1.7)
8 (5.6)
7 (1)a
Polytrauma without head trauma
18 (2.1)
3 (2.1)
15 (2.2)
Severe head trauma
11 (1.3)
3 (2.1)
8 (1.1)
272 (32.3)
57 (39.6)
235 (33.7)
Otherc
Pcapo2 = capillary partial pressure in oxygen. a Significant difference (p < 0.05) between transfused cases and nontransfused cases. b More than one diagnosis can be attributed to one patient. c Intoxication (30), arrhythmia (21), intracranial hemorrhage (16), diabetic ketoacidosis (14), brain death (6), severe burn (4), cardiorespiratory arrest (3), etc.
28-day mortality (crude OR, 9.49; 95% CI, 4.41–20.43; p < 0.001) was reduced but still statistically significant after controlling for severity of illness at PICU entry (adjusted OR, 5.12; 95% CI, 2.18–12.02; p < 0.001) (Tables 3 and 4). 508
www.pccmjournal.org
Mechanically ventilated transfused cases were intubated for longer periods of time than nontransfused cases (14.1 ± 32.6 d and 4.3 ± 9.6 d, respectively; p < 0.001). This association remained statistically significant after multivariable adjustment (hazard ratio July 2015 • Volume 16 • Number 6
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Feature Articles
Table 2.
Data on RBC Transfusions in Patients Transfused in PICU
Number of RBC transfusions
578 144 (17.1)
Patients who received at least one transfusion, n (%) Incidence density (transfusion events/100 patient-days)
11.2
Volume of blood during first transfusion, mL/kg
11.1 (7.3–15)
Hemoglobin level within 24 hr before first transfusion, g/L
77.5 (63.8–94)
Time from PICU admission to first transfusion, d
1 (0–1)
Pediatric Risk of Mortality score within 24 hr before first transfusion
7 (3–12)
Pediatric Logistic Organ Dysfunction score within 24 hr before first transfusion
10 (1–12)
Length of storage of packed RBC units, d
13 (9–22)
a
Data available for 354 transfusion events. Continuous variables are described with median (interquartile range).
a
[HR] of extubation [reflecting the rate of extubation for transfused cases vs nontransfused cases], 0.59; 95% CI, 0.45–0.79; p < 0.001) (Table 4). We observed an adjusted dose-effect relationship between RBC transfusion and length of mechanical ventilation (Fig. 2A). PICU length of stay was significantly increased in transfused cases (12.4 ± 26.2 d vs 4.9 ± 10.2 d; p < 0.001). The adjusted HR of PICU discharge for transfused cases versus nontransfused cases was 0.7 (95% CI, 0.57–0.85; p < 0.001) (Table 4). When considering different levels of exposure to RBCs, we observed an adjusted dose-effect relationship (Fig. 2B). Tertiary Outcomes. The worst Pao2 was lower in transfused cases than in nontransfused cases (61.1 ± 26.5 and 83.5 ± 28.1 torr, respectively; p < 0.001), while it did not differ between the two groups at PICU admission (Table 1). This difference remained statistically significant even when children with and without a cyanotic heart disease were analyzed separately (Table 3). Furthermore, the worst Pao2 significantly decreased during the 6 hours following the first RBC transfusion (mean difference, 25.6 torr; p = 0.029) (Table 5). Pulse oximetry analysis also showed a small decrease between pretransfusion and posttransfusion worst oxygen saturation (Spo2), statistically but not clinically significant (95.9% ± 7.1% vs 95% ± 7%; p < 0.001) (Table 5). The proportion of ARDS was similar in transfused and nontransfused cases (5.3% vs 4.2%; p = 0.54) (Table 3). Arterial hypotension was observed more frequently in transfused cases than in nontransfused cases (50.4% vs 28.3%; p < 0.001), but pre transfusion and posttransfusion prevalence of this complication did not differ (p = 1) (Table 5). Renal replacement therapy was more frequent in transfused than in nontransfused cases (8.4% vs 0.6%; p < 0.001) (Table 3). It was also more prevalent after than before the first RBC transfusion (8.4% vs 2.3%; p = 0.008) (Table 5). Finally, more transfused cases developed SIRS, sepsis, severe sepsis, or septic shock, but pretransfusion and posttransfusion prevalence of all these did not statistically differ (Table 5).
DISCUSSION This prospective observational study describes several associations between RBC transfusions and clinical outcomes in Pediatric Critical Care Medicine
critically ill children. We report an independent association and a dose-effect relationship between RBC transfusion and increased morbidity, with an increased risk of NPMODS, a longer PICU stay, and an increased duration of mechanical ventilation. RBC Transfusions and Worse Outcomes in Critically Ill Patients A multicenter study conducted by Bateman et al (8) reported that after correction for age and for admission PRISM III score, transfused patients had an increased risk of death, a higher rate of nosocomial infections, more cardiac or respiratory dysfunction, a longer duration of mechanical ventilation, and a longer PICU length of stay than nontransfused patients. Other studies showing worse outcomes associated with RBC transfusions have focused on specific PICU populations (e.g., severely burned children [19] and postoperative children after an open heart surgery [1]) or on a more restricted number of adverse events associated with transfusions in a general PICU population (e.g., bloodstream infections [20] and increased use of some intensive care resources [oxygen, mechanical ventilation, and vasoactive agents infusion] [21]). Most critically ill patients have systemic inflammation in association with the illness that led to ICU admission (first hit). These patients are deemed more vulnerable to any sequential insult (second hit), which may amplify the already primed inflammatory cascade (22). This two-hit model explains many cases of multiple organ failure observed in the intensive care (23). RBC transfusion can be a potential second hit (22). Based on this concept, RBC transfusion may have a more significant impact in the critically ill as compared with less ill patients (24). Our data confirm that RBC transfusions are associated with worse outcomes in critically ill children and are of importance because of the large number of potential complications looked for, the case-mix which included a large multidisciplinary PICU population, the attention paid to the temporal relationship between transfusions and the potential outcomes, and the prospective manner in which data were collected. www.pccmjournal.org
509
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Demaret et al
Table 3.
Bivariate Association Between RBC Transfusions and Outcome Nontransfused Cases (NTD+/NTD– or Mean ± sd)
Transfused Cases (TD+/TD– or Mean ± sd)
New or progressive MODS
55/698
44/144
5.14 (3.28, 8.06)
< 0.001
New MODS
37/698
23/130
3.84 (2.2, 6.72)
< 0.001
Progressive MODS
45/662
43/132
6.62 (4.13, 10.64)
< 0.001
Death during PICU stay
11/698
17/144
8.36 (3.83, 18.27)
< 0.001
24/696
19/130
4.79 (2.54, 9.03)
< 0.001
Clinical Outcome
Bivariate OR (95% CI)
p
Primary outcome
Secondary outcomes Nosocomial infection Pneumonia
5/692
7/129
7.88 (2.46, 25.24)
< 0.001
Tracheitis
4/692
3/129
4.1 (0.91, 18.52)
0.047
Bacteremia
1/692
7/129
39.65 (4.84, 325.1)
< 0.001
Urinary tract infection
8/693
6/129
4.18 (1.42, 12.25)
0.005
Other
8/692
8/129
5.65 (2.08, 15.35)
< 0.001
Endotracheal intubation
269/698
84/129
2.98 (2.01, 4.41)
< 0.001
Length of endotracheal intubation (d)
4.3 ± 9.6
14.1 ± 32.6
< 0.001
PICU length of stay
4.9 ± 10.2
12.4 ± 26.2
< 0.001
11/698
19/144
9.49 (4.41, 20.43)
Acute respiratory distress syndrome
29/698
7/131
1.3 (0.56, 3.01)
Worst Po2
68.5 ± 31
62.3 ± 26.9
0.003
Worst Pcapo2
60.5 ± 29.6
58.1 ± 28.1
0.13
Cyanotic heart disease
50.4 ± 16.5
53.4 ± 18.5
0.674
No cyanotic heart disease
61.1 ± 30.1
58.8 ± 29.5
0.093
Worst Parto2
83.5 ± 28.1
61.1 ± 26.5
< 0.001
Cyanotic heart disease
69.6 ± 27.5
55.3 ± 27
No cyanotic heart disease
85.6 ± 27.6
68.9 ± 25.2
< 0.001
91.1 ± 12
85.2 ± 14.4
< 0.001
Cyanotic heart disease
86.3 ± 12.8
78.2 ± 16.8
0.017
No cyanotic heart disease
91.5 ± 11.9
87.7 ± 12.6
< 0.001
2/698
7/130
19.81 (4.07, 96.45)
< 0.001
169/698
66/131
3.18 (2.17, 4.67)
< 0.001
4/698
11/131
15.9 (5, 50.8)
< 0.001
Systemic inflammatory response syndrome
364/696
87/131
1.8 (1.22, 2.67)
0.003
Sepsis
207/696
55/131
1.71 (1.17, 2.51)
0.006
Severe sepsis
39/698
22/131
3.41 (1.95, 5.97)
< 0.001
Septic shock
12/698
15/131
7.39 (3.38, 16.19)
< 0.001
28-d mortality
< 0.001
Tertiary outcomes
a
b
Worst Spo2c
Cardiac arrest Arterial hypotension Renal replacement therapy
0.540
0.035
NTD+ = nontransfused cases with determinant, NTD– = nontransfused cases without determinants, TD+ = transfused cases with determinant, TD– = transfused cases without determinant, OR = odds ratio, MODS = multiple organ dysfunction syndrome, Pcapo2 = capillary partial pressure in oxygen, Parto2 = arterial partial pressure in oxygen, Spo2 = pulsed oxygen saturation. a Information available for 395 nontransfused cases (NTCs) and for 36 transfused cases (TCs). b Information available for 192 NTCs and for 88 TCs. c Information available for 696 NTCs and for 130 TCs.
510
www.pccmjournal.org
July 2015 • Volume 16 • Number 6
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Feature Articles
Table 4. Adjusted Association Between RBC Transfusion and Clinical Outcomes (Primary and Secondary Outcomes) Clinical Outcome
p
Multivariable Association
Multivariable logistic regression
Odds ratio (95% CI)
New or progressive multiple organ dysfunction syndrome
3.85 (2.38, 6.24)
< 0.001
Nosocomial infectionb
3.31 (1.67, 6.56)
0.001
28-d mortalityb
5.12 (2.18, 12.02)
a
Cox regression
Hazard ratio (95% CI)
Duration of mechanical ventilation PICU length of stay
< 0.001
a,c
a,d
0.59 (0.45, 0.79)
< 0.001
0.7 (0.57, 0.85)
< 0.001
Adjustment made for age, admission Pediatric Risk of Mortality (PRISM), and diagnosis of congenital heart disease. b Adjustment made for admission PRISM only (because of the small number of outcome events). c Only mechanically ventilated children were included in the model (n = 353); the model gives the hazard ratio of extubation: a hazard ratio significantly < 1 indicates a lower risk of extubation at any time for transfused children as compared with nontransfused ones. d The model gives the hazard ratio of PICU discharge: a hazard ratio significantly < 1 indicates a lower risk of PICU discharge at any time for transfused children as compared with nontransfused ones. Transfused children are compared with nontransfused. The goodness-of-fit of the logistic regression models was evaluated using the Hosmer-Lemeshow statistic, showing there was not sufficient evidence to conclude a bad fit. No significant interaction was found in any of these multivariable models. a
Protopathic Bias or the Importance of the Temporal Relationship Reverse causality occurs when the outcome precedes and increases the risk for the putative exposure (25). Protopathic bias is a specific type of reverse causality: it occurs when an outcome is associated with an exposure that actually results from early signs and symptoms of the outcome under study (26). In other
words, it is not sufficient to observe development of a complication or condition after transfusion; one must be certain the complication or condition was not present before the transfusion. Rigorous attention was paid during our study in order to avoid the presence of protopathic bias. When analyses were undertaken to compare pretransfusion and posttransfusion values of the variables studied, several associations between RBC transfusions
Figure 2. Adjusted survival curves showing the time from PICU admission to PICU discharge (A) and to extubation for ventilated children (B), stratified according to different levels of RBC transfusion. Adjustment was made for age and for admission Pediatric Risk of Mortality score. A, The corresponding hazard ratios are significantly higher for each of the two smallest categories (i.e., 0 and 1–20 cc/kg of RBCs during PICU stay) as compared with the reference group (> 80 cc/kg) (p < 0.05). In this analysis, a higher hazard ratio can be interpreted as a higher risk of PICU discharge and thus a shorter length of stay in PICU. B, The corresponding hazard ratios are significantly higher for each of the three smallest categories (i.e., 0, 1–20, and 21–40 cc/ kg of RBCs during PICU stay) as compared with the reference group (> 80 cc/kg) (p < 0.05). In this analysis, a higher hazard ratio can be interpreted as a higher risk of extubation during PICU stay and thus a shorter duration of mechanical ventilation.
Pediatric Critical Care Medicine
www.pccmjournal.org
511
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Demaret et al
Table 5.
Paired Analysis for Comparison of Pretransfusion and Posttransfusion Values Pretransfusion
Posttransfusion
pa
10.6 ± 8.4
11.4 ± 9.8
0.29
70/131
85/131
0.014
Cardiovascular dysfunction
7/131
5/131
0.625
Hematological dysfunction
49/131
56/131
0.248
Neurological dysfunction
90/131
83/131
0.281
Hepatic dysfunction
30/131
39/131
0.093
Renal dysfunction
12/131
18/131
0.031
Clinical Outcome
Primary outcome: organ dysfunctions Worst daily Pediatric Logistic Organ Dysfunction score Respiratory dysfunction
Tertiary outcomes Acute respiratory distress syndrome
8/131
7/131
1
Worst Pao2b,c
154.9 ± 106
129.4 ± 81.9
0.029
Worst Spo2b,d
95.9 ± 7.1
95 ± 6.8
Arterial hypotension
67/131
66/131
1
Renal replacement therapy
3/131
11/131
0.008
Systemic inflammatory response syndrome
75/131
87/131
0.088
Sepsis
46/131
55/131
0.093
Severe sepsis
18/131
22/131
0.289
Septic shock
11/131
15/131
0.219
< 0.001
Spo2 = pulsed oxygen saturation. a McNemar chi-square for comparison of proportions and Wilcoxon signed-rank test for comparison of means. b Lowest value during the 2 hr before and during the 6 hr after the first transfusion. c Pretransfusion and posttransfusion values are available for 67 transfused patients. d Pretransfusion and posttransfusion values are available for 129 transfused patients.
and clinical outcomes lost their significance, which highlights the possible importance of protopathic bias in this type of research. RBC Transfusions and Oxygenation Oxygenation markers did not improve during the 6 hours following the first RBC transfusion; we rather observed a trend toward worsening of Po2 and Spo2 (Table 5). The posttransfusion decrease of Po2 is particularly questioning, even if we did not expect to a Po2 increase considering a mean pretransfusion value of 154.9 ± 106 mm Hg. Obviously, several factors not related to RBC transfusions may interfere with these oxygenation variables. However, several explanations related to RBC transfusions might also explain this observation, including transfusion-associated respiratory complications such as TRALI (3) and TACO (4). RBCs are frequently prescribed to improve oxygen delivery (27). An RBC transfusion increases arterial oxygen content (Cao2) by increasing the hemoglobin level. It thereby should increase global oxygen delivery (Do2). However, it is not always the case. Indeed, an RBC transfusion increases blood viscosity, which can decrease cardiac output in normovolemic patients whose cardiac function is not impaired by myocardial ischemia (28) and therefore nullify the global Do2 increase that occurs with a higher hemoglobin level. Furthermore, respiratory complications associated with RBC transfusion may cause a decrease in oxygen saturation 512
www.pccmjournal.org
that might also counteract the positive effect of an increased hemoglobin level on Cao2. In addition, RBC transfusions may also exert several effects that reduce tissular Do2 and oxygen consumption. Although several studies have reported that RBC transfusions increase tissue oxygenation (29, 30), others have led to conflicting results (31, 32). Actually, a growing body of evidence suggests that stored RBCs may have adverse effects on microcirculatory flow and oxygen use (33). Many changes occur during RBCs storage, including diminished adenosine triphosphate and 2,3-diphosphoglycerate levels (34), decreased RBC deformability over time (35), hemolysis resulting in generation of proinflammatory microparticles (36), and altered vasoregulation attributable to certain effects stored RBCs have on nitric oxide metabolism (37). Several of these changes may disturb oxygen capture in the lungs and may impact the ability of RBC to increase oxygen delivery and consumption in certain vascular territories, which might contribute to organ dysfunction in critically ill patients. This could explain the greater risk of NPMODS we observed in some transfused children. Limitations and Strengths Our study has several limitations. It was conducted in a single center, which limits its external validity; however, our critical care unit is comparable to most multidisciplinary July 2015 • Volume 16 • Number 6
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Feature Articles
university-affiliated North-American PICUs with regard to case-mix and severity of illness (8, 10, 38). Second, we did not collect data on the length of time between the first RBC transfusion and the adverse outcomes. It is therefore possible that an outcome was associated with RBC transfusion even though it actually occurred a few days after the transfusion. This may have led to an overestimation of the association between RBC transfusions and some of the outcomes assessed. Third, we did not consider RBC transfusions that occurred before PICU admission. Ninety-one children had been transfused within the 24 hours prior to PICU admission and were not considered as transfused children in our study. Fourth, our study was not able to highlight a cause-effect relationship between RBC transfusions and outcomes in critically ill children: only a randomized-controlled trial could establish such a causal link. Caution is thus warranted in the interpretation of our results. Fifth, transfused and nontransfused children are different. The need to receive an RBC transfusion selects a group of subjects at greater risk of adverse outcomes (39). Clinical research in transfusion medicine is subject to confounding by indication, and our study does not escape to this pitfall. This bias occurs when studying the effect of a treatment (e.g., RBC transfusion) while the indication for the treatment causes the outcome (40). The sicker a child is, the more transfusions will be indicated. Also, the sicker a child is, the higher the risk that the child will have a poor outcome (40). Even if we adjusted for several confounders, it is possible that our results are still influenced by residual confounding and that we overestimated the association between adverse outcomes and RBC transfusions (39). Sixth, we did not consider plasma and platelet transfusions in our analyses. Number of children received at least one platelet and/or one plasma transfusion in addition to their RBC transfusion. It is possible that the associations we observed are partially explained by plasma and/or platelet transfusions. Seventh, patients readmitted to PICU more than 24 hours after discharge were considered as new cases. It can be argued that this approach leads to biased results because of correlation between cases. However, we undertook a sensitivity model using a generalized estimating equation approach (41) and showed that considering these patients as one case would have resulted in findings similar to those of our multivariable models. Finally, because of the observational nature of our study, we were not allowed to ask the physicians to prescribe laboratory analyses for the study, which resulted in some missing values for some variables (e.g., Pao2). Our study also has several strengths. This is the first study to prospectively consider such a high number of outcomes related to RBC transfusions in a general PICU population. The list of outcomes possibly associated with RBC transfusion was decided upon before the study was initiated. The study included all consecutive PICU admissions over a 1-year period, which resulted in a case-mix with a limited risk of selection bias and no influence due to seasonal variation. We paid a careful attention to the temporal relationship between the first RBC transfusion in PICU and outcomes. Our paired analysis for the comparison of pretransfusion and posttransfusion values highlights the risk of protopathic Pediatric Critical Care Medicine
bias in this type of research. Finally, the prospective nature of our study was a major asset to minimize information bias.
CONCLUSIONS This observational prospective study suggests that RBC transfusions are associated with worse outcomes in critically ill children. We found that transfusions were associated with a greater risk of developing NPMODS, with prolonged mechanical ventilation and prolonged PICU stay. Furthermore, several of our findings raise questions with regard to the ability of stored RBCs to improve oxygenation in critically ill children and highlight the importance of temporal relationship when assessing the clinical effects of transfusions. Despite some limitations, findings from this study may help caregivers to figure out what can be the risk-benefit ratio of RBC transfusion.
REFERENCES
1. Dodd RY: Current risk for transfusion transmitted infections. Curr Opin Hematol 2007; 14:671–676 2. Hendrickson JE, Hillyer CD: Noninfectious serious hazards of transfusion. Anesth Analg 2009; 108:759–769 3. Toy P, Gajic O, Bacchetti P, et al; TRALI Study Group: Transfusionrelated acute lung injury: Incidence and risk factors. Blood 2012; 119:1757–1767 4. Li G, Rachmale S, Kojicic M, et al: Incidence and transfusion risk factors for transfusion-associated circulatory overload among medical intensive care unit patients. Transfusion 2011; 51:338–343 5. Vamvakas EC, Blajchman MA: Transfusion-related immunomodulation (TRIM): An update. Blood Rev 2007; 21:327–348 6. Armano R, Gauvin F, Ducruet T, et al: Determinants of red blood cell transfusions in a pediatric critical care unit: A prospective, descriptive epidemiological study. Crit Care Med 2005; 33:2637–2644 7. Kneyber MC, Hersi MI, Twisk JW, et al: Red blood cell transfusion in critically ill children is independently associated with increased mortality. Intensive Care Med 2007; 33:1414–1422 8. Bateman ST, Lacroix J, Boven K, et al; Pediatric Acute Lung Injury and Sepsis Investigators Network: Anemia, blood loss, and blood transfusions in North American children in the intensive care unit. Am J Respir Crit Care Med 2008; 178:26–33 9. Laverdière C, Gauvin F, Hébert PC, et al; Canadian Critical Care Trials Group: Survey on transfusion practices of pediatric intensivists. Pediatr Crit Care Med 2002; 3:335–340 10. Lacroix J, Hébert PC, Hutchison JS, et al; TRIPICU Investigators; Canadian Critical Care Trials Group; Pediatric Acute Lung Injury and Sepsis Investigators Network: Transfusion strategies for patients in pediatric intensive care units. N Engl J Med 2007; 356:1609–1619 11. Grol R, Grimshaw J: From best evidence to best practice: Effective implementation of change in patients’ care. Lancet 2003; 362:1225–1230 12. Pollack MM, Ruttimann UE, Getson PR: Pediatric risk of mortality (PRISM) score. Crit Care Med 1988; 16:1110–1116 13. Leteurtre S, Martinot A, Duhamel A, et al: Validation of the Paediatric Logistic Organ Dysfunction (PELOD) score: Prospective, observational, multicentre study. Lancet 2003; 362:192–197 14. Goldstein B, Giroir B, Randolph A; International Consensus Conference on Pediatric Sepsis: International pediatric sepsis consensus conference: Definitions for sepsis and organ dysfunction in pediatrics. Pediatr Crit Care Med 2005; 6:2–8 15. Goldstein B, Giroir B, Randolph A: Values for systolic blood pressure. Pediatr Crit Care Med 2005; 6:500–501 16. Bernard GR, Artigas A, Brigham KL, et al: The American-European Consensus Conference on ARDS. Definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994; 149:818–824 www.pccmjournal.org
513
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
Demaret et al 17. Bursac Z, Gauss CH, Williams DK, et al: Purposeful selection of variables in logistic regression. Source Code Biol Med 2008; 3:17 18. Hosmer DW, Lemeshow S, Sturdivant RX: Applied Logistic Regression, Third Edition. Hoboken, John Wiley & Sons, Inc., 2013 19. Jeschke MG, Chinkes DL, Finnerty CC, et al: Blood transfusions are associated with increased risk for development of sepsis in severely burned pediatric patients. Crit Care Med 2007; 35:579–583 20. Elward AM, Fraser VJ: Risk factors for nosocomial primary bloodstream infection in pediatric intensive care unit patients: A 2-year prospective cohort study. Infect Control Hosp Epidemiol 2006; 27:553–560 21. Goodman AM, Pollack MM, Patel KM, et al: Pediatric red blood cell transfusions increase resource use. J Pediatr 2003; 142:123–127 22. McIntyre LA, Hebert PC: Can we safely restrict transfusion in trauma patients? Curr Opin Crit Care 2006; 12:575–583 23. Moore FA, Sauaia A, Moore EE, et al: Postinjury multiple organ failure: A bimodal phenomenon. J Trauma 1996; 40:501–510 24. Spinella PC, Doctor A, Blumberg N, et al: Does the storage duration of blood products affect outcomes in critically ill patients? Transfusion 2011; 51:1644–1650 25. Murk W, Risnes KR, Bracken MB: Prenatal or early-life exposure to antibiotics and risk of childhood asthma: A systematic review. Pediatrics 2011; 127:1125–1138 26. Tamim H, Monfared AA, LeLorier J: Application of lag-time into exposure definitions to control for protopathic bias. Pharmacoepidemiol Drug Saf 2007; 16:250–258 27. Demaret P, Tucci M, Ducruet T, et al: Red blood cell transfusion in critically ill children (CME). Transfusion 2014; 54:365–375 28. Roberson RS, Bennett-Guerrero E: Impact of red blood cell transfusion on global and regional measures of oxygenation. Mt Sinai J Med 2012; 79:66–74 29. van Hoften JC, Verhagen EA, Keating P, et al: Cerebral tissue oxygen saturation and extraction in preterm infants before and
514
www.pccmjournal.org
after blood transfusion. Arch Dis Child Fetal Neonatal Ed 2010; 95:F352–F358 30. Figaji AA, Zwane E, Kogels M, et al: The effect of blood transfusion on brain oxygenation in children with severe traumatic brain injury. Pediatr Crit Care Med 2010; 11:325–331 31. Creteur J, Neves AP, Vincent JL: Near-infrared spectroscopy technique to evaluate the effects of red blood cell transfusion on tissue oxygenation. Crit Care 2009; 13(Suppl 5):S11 32. Kiraly LN, Underwood S, Differding JA, et al: Transfusion of aged packed red blood cells results in decreased tissue oxygenation in critically injured trauma patients. J Trauma 2009; 67:29–32 33. Weinberg JA, MacLennan PA, Vandromme-Cusick MJ, et al: The deleterious effect of red blood cell storage on microvascular response to transfusion. J Trauma Acute Care Surg 2013; 75:807–812 34. Karam O, Tucci M, Toledano BJ, et al: Length of storage and in vitro immunomodulation induced by prestorage leukoreduced red blood cells. Transfusion 2009; 49:2326–2334 35. Bennett-Guerrero E, Veldman TH, Doctor A, et al: Evolution of adverse changes in stored RBCs. Proc Natl Acad Sci U S A 2007; 104:17063–17068 36. Kim-Shapiro DB, Lee J, Gladwin MT: Storage lesion: Role of red blood cell breakdown. Transfusion 2011; 51:844–851 37. Doctor A, Spinella P: Effect of processing and storage on red blood cell function in vivo. Semin Perinatol 2012; 36:248–259 38. Pollack MM, Patel KM, Ruttimann UE: PRISM III: An updated Pediatric Risk of Mortality score. Crit Care Med 1996; 24:743–752 39. van de Watering L; Biomedical Excellence for Safer Transfusion C: Pitfalls in the current published observational literature on the effects of red blood cell storage. Transfusion 2011; 51:1847–1854 40. Middelburg RA, van de Watering LM, van der Bom JG: Blood transfusions: Good or bad? Confounding by indication, an underestimated problem in clinical transfusion research. Transfusion 2010; 50:1181–1183 41. Zeger SL, Liang KY: Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986; 42:121–130
July 2015 • Volume 16 • Number 6
Copyright © 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies. Unauthorized reproduction of this article is prohibited
