Material and Methods
Study Design
This prospective observational cohort study took place in the PICU of Sainte-Justine Hospital, a tertiary care multidisciplinary university-affiliated pediatric hospital. All children consecutively admitted to the PICU over a 1-year period (April 2009 to April 2010) were eligible. The study was approved by the local institutional review board (Comité d'Éthique de la Recherche du CHU Sainte-Justine, project 2870), waiving the requirement for written informed consent. Exclusion criteria included: 1) gestational age less than 40 weeks; 2) postterm age less than 3 days or more than 18 years; 3) pregnancy; and 4) admission to PICU in the postpartum period.
Population
The study involved patients who received a RBC transfusion while in PICU. Time zero was the time of initiation of the first RBC transfusion in PICU. Any administration of RBCs, regardless of the volume given, was considered a transfusion. Most RBC units were collected from allogeneic whole blood donation except for a few units which were obtained by apheresis. Only prestorage leukocyte-reduced additive solution-3 RBC units were used.
Definition of RDAT
New or progressive RDAT is the primary outcome measure of this study. According to Goldstein et al, a RD is diagnosed when at least one of the following criteria is observed: 1) PaO2/FIO2 ratio less than 300 in the absence of cyanotic heart disease or preexisting lung disease before PICU admission; 2) PaCO2 more than 65 mm Hg or 20 mm Hg over baseline PaCO2; 3) proven need for more than 50% FIO2 to maintain saturation at least 92%; and 4) need for nonelective invasive or noninvasive mechanical ventilation. When arterial blood gases were not available, the criterion SpO2/FIO2 less than 253 was added as an alternative to PaO2/FIO2 less than 300; the diagnostic value of SpO2/FIO2 ratio has been validated in critically ill children. In all instances, the PaO2/FIO2 ratio was considered normal if the SpO2 was at least 99%.
We differentiated between new and progressive RDAT. RD was considered new if no RD was present at time zero and appeared after the first RBC transfusion. RDAT was considered progressive when a RD was present at time zero, but worsened after the first RBC transfusion. The severity of RD was considered to have worsened if the PaO2/FIO2 ratio or the SpO2/FIO2 ratio dropped by at least 20% after the transfusion. This threshold was considered clinically significant by nine pediatric intensivists, using a Delphi method. Transient episodes of SpO2/FIO2 decrease attributable to maneuvers such as endotracheal aspiration were not retained for analysis.
Cases of TRALI were also identified based on the definition proposed in 2004 by a panel of experts. TRALI was considered when a new acute lung injury with hypoxemia occurred within 6 hours of transfusion, with bilateral pulmonary infiltrates, no evidence of left atrial hypertension, and no temporal relationship to an alternative risk factor for acute lung injury. A possible TRALI was considered when the last criterion was missing.
Adjudicating Process
An adjudicating committee ascertained the diagnosis of new/progressive RDAT and TRALI. Two investigators (N.K., E.L.) independently reviewed the patient medical charts and filled out a validated adjudication form. If the two adjudicators did not agree on the diagnosis, they were asked to review their assessment of the data. If consensus could not be reached, a third adjudicator (G.E.) reviewed the case. Consensus was considered to be reached when two of the three adjudicators agreed on the diagnosis.
When RDAT was diagnosed, adjudicators rated their degree of certainty for the diagnosis as well as their degree of certainty that there was a causal link between RBC transfusion and development of RD on a 7-point Likert scale. The lowest score (1) meant that a causal link between RBC transfusion and development of RD was very unlikely given the available data and the definition used, whereas a score of 7 reflected a high likelihood (quasi certainty) that RD was caused by RBC transfusion. No specific criteria were a priori established to define the causality.
Potential Risk Factors and Outcome Variables
To prevent any protopathic bias, only events occurring before time zero were considered as possible risk factors for RDAT; events appearing after time zero were considered as possible adverse events. The following variables were a priori considered as possible risk factors or risk markers for RDAT: age, severity of illness at time zero as evaluated by the Pediatric Risk of Mortality (PRISM) score and the Pediatric Logistic Organ Dysfunction (PELOD) score, prior chronic illness (chronic hematologic, cardiac, or oncologic diseases), presence of multiple organ dysfunction syndrome (MODS) as defined by Goldstein et al at time zero, total number of RBC transfusions, volume (mL/kg) of RBCs received, RBC storage time, and transfusion of other blood products (plasma and platelet concentrates).
Possible complications of RDAT were monitored daily; this included development of sepsis, severe sepsis, septic shock, and MODS as well as daily PELOD scores, mortality, duration of endotracheal intubation, and PICU length of stay. The occurrence of nosocomial infections after the first RBC transfusion was also recorded. The ventilation-free days and PICU-free days at day 28 were calculated as the number of days spent alive and without ventilation (or outside the PICU) by day 28.
Data Management
A validated case report form was used to prospectively collect the following data from admission to PICU discharge: demographic data, medical history, admission diagnosis, clinical data, laboratory results, transfusions, mechanical ventilation, fluid balance, new and/or progressive RD criteria, MODS, PICU length of stay, and mortality. The following data were collected hourly in transfused patients from a time point 2 hours prior to transfusion (start of transfusion = time zero) to 24 hours after time zero and then at least once daily until PICU discharge: worse respiratory rate, ventilatory settings, lowest SpO2, lowest PaO2/FIO2 ratio, and SpO2/FIO2 ratio. Data management fulfilled all requirements of standard good clinical data management practice, and all data were treated anonymously.
Transfusion Policy
In the PICU, no written protocol exists regarding transfusion. In line with the Transfusion Requirements in the Pediatric Intensive Care Unit study, the general attitude during the study period was to avoid transfusion when hemoglobin exceeded 7 g/dL in stabilized patients.
Statistical Analysis
Concordance and κ scores were calculated to assess the reproducibility of the adjudication process. Categorical variables were described by frequency distribution and compared using chi-square or Fisher exact probability test. Continuous variables were reported as mean and SD and/or median and interquartile range and compared across groups using Student t test or nonparametric Wilcoxon tests.
Univariate and multivariate logistic regression analyses were done for potential risk factors by calculating unadjusted and adjusted odds ratios (OR) and their 95% CI. Age, severity of illness (PRISM and PELOD scores), and volume (mL/kg) of RBC units given were categorized by quartiles. Statistically significant determinants after univariate analysis (p < 0.05) were included in the multivariate logistic regression. Age and congenital heart disease diagnosis were also included because they are known determinants involved in the decision to transfuse. PELOD score was a priori excluded from the model because of its collinearity with PRISM; highest lactate level was a posteriori excluded because these data were missing in more than 25% of patients. The association between RDAT and adverse outcomes was analyzed using univariate chi-square or Student t test.
Two-tailed p value of less than 0.05 was considered statistically significant. All analyses were performed by a statistician (T.D.) who used SAS 9.2 (SAS Institute, Cary, NC).