Background
Red blood cell (RBC) transfusions are frequently administered in critically ill patients [
1,
2], although probably less so in recent years than in the past. RBC transfusion may have beneficial effects by increasing oxygen delivery to the tissues and improving the oxygen demand/supply balance [
3], but may also have some harmful effects. In addition to possible type and cross-matching issues, transfusions may be associated with transfusion-related acute lung injury (TRALI), transfusion-associated circulatory overload (TACO) and immunomodulating effects resulting in a higher incidence of nosocomial infections [
4].
Prospective, randomized, controlled trials (RCTs) have suggested that restrictive RBC transfusion strategies are safe, but this may apply more to patients who are less severely ill. Indeed, recent meta-analyses have suggested caution in patients with cardiovascular disease [
5], and that restrictive strategies should be applied with particular care in high-risk patients undergoing major surgery [
6]. The uncertainty concerning the risk/benefit profile of RBC transfusion in critically ill patients has fueled ongoing debate on the subject over the last two decades. Moreover, the heterogeneity of critical illness and the absence of universally accepted guidelines may influence RBC transfusion practice in intensive care units (ICUs) worldwide. Several regional observational studies [
1,
7] have reported on patterns of RBC transfusion in ICU patients, but global differences in transfusion practice have not been addressed. Such data would be useful from an epidemiological point of view, and also to improve our understanding of the factors that may influence transfusion practice.
We used the worldwide Intensive Care Over Nations (ICON) database, to investigate current transfusion practices worldwide and the possible effect of transfusion on outcomes.
Methods
This was a planned post-hoc analysis of the ICON database. ICON was a multicenter, worldwide audit endorsed by the World Federation of Societies of Intensive Care and Critical Care Medicine (WFSICCM). Institutional recruitment for participation was by open invitation and was voluntary, with no financial incentive. Institutional review board approval was obtained by the participating institutions according to local ethical regulations. Informed consent was not required due to the observational and anonymous nature of data collection in this audit. The ICON audit collected data on 10,069 adult (> 16 years of age) patients admitted to the participating centers (Additional file
1) between 8 May and 18 May 2012. The 9553 patients for whom data related to transfusion had been completed were included in the present analysis.
Data collection and management
Details about the conduct of the audit, the data collected and the definitions used are provided elsewhere [
8]. Data were collected prospectively using preprinted case report forms (CRFs) and were electronically introduced by the investigators using an Internet-based website. Data collection on admission included demographic data and data on comorbid diseases (chronic obstructive pulmonary disease (COPD), cancer, metastatic cancer, hematologic cancer, insulin-dependent diabetes mellitus, heart failure (New York Heart Association (NYHA) class III-IV), chronic renal failure, human immunodeficiency virus (HIV) infection, cirrhosis, immunosuppression, steroid therapy or chemotherapy). Clinical data and laboratory data were collected daily for a maximum of 28 days in the ICU, and included respiratory, hemodynamic and hematological parameters, and fluid administration. For the purposes of this study, the minimum hemoglobin concentration reported on the day of the transfusion was considered as the “transfusion trigger”. ICU and hospital length of stay and outcomes were recorded. Clinical and laboratory data for the Simplified Acute Physiology Score (SAPS) II [
9] were reported as the worst value within 24 h after admission. A daily evaluation of organ function according to the sequential organ failure assessment (SOFA) score [
10] was performed and organ failure was defined as a SOFA score >2 for the organ in question. Surgical admissions were identified as patients who had had surgery in the 4 weeks preceding admission.
Data were processed and analyzed in the Department of Intensive Care of the University of Brussels. The study coordinators processed all queries during data collection. Detailed instructions, explaining the aim of the study, instructions for data collection, and definitions for various items were available through a secured website for all participants. Validity checks were made concurrent with data entry by the electronic case record form, including plausibility checks within each variable and between variables. Data were further reviewed by the coordinating center for plausibility and availability of the outcome parameter, and doubts were clarified with the corresponding ICU. We did not perform any other supplementary quality control measures.
Subgroup analyses
Subgroups defined a priori included male versus female patients, medical versus surgical patients, sepsis versus no sepsis, and subgroups according to quartiles of SAPS II and SOFA scores on admission to the ICU and age categories (18–50, 51–65, 66–75 or > 75 years).
Statistical analysis
Data are summarized using means with standard deviation (SD), medians and interquartile ranges (IQ) or numbers and percentages. For the purposes of this study, hemoglobin concentrations were categorized according to cutoff values of < 7, 7–8.9, 9–10.9, 11–13, and > 13 g/dL. Differences between groups were tested using analysis of variance (ANOVA), the Kruskal-Wallis test, Student’s t test, the Mann-Whitney test, the chi-square test or Fisher’s exact test as appropriate.
To determine the adjusted relative risk of hospital death after RBC transfusion during the ICU stay, we developed an extended multivariable Cox model in the overall population and in the subgroups defined a priori. We considered discharge as a competing event. Other confounding variables considered for the Cox analysis included age, sex, SAPS II score and SOFA sub-scores on admission to the ICU, type of admission, the use of mechanical ventilation or renal replacement therapy on ICU admission, comorbidities, presence of sepsis and hemoglobin categories on admission to the ICU. We also adjusted for ICU and hospital-related organizational factors including type of hospital, hospital bed capacity, total number of ICU patients in 2011, number of staffed ICU beds, number of nurses and the presence of 24 h/day in-house intensivist coverage. The geographic region (North America, South America, Western Europe, Eastern Europe, Middle East, South Asia, East and South-East Asia, Oceania and Africa) was considered in these multivariable models. An extended Cox model was constructed, adding interaction terms that involve time, i.e., time-varying fixed effects, computed as the by-product of time and individual covariates in the model (time × covariate). Individual time-varying fixed covariates were introduced one by one and in combinations in the extended model, none of which was found to be significant (Wald chi-square statistics). Colinearity between variables was checked before modeling and all variables were included in the models thereafter, irrespective of the level of significance on univariate basis. Transfusion during the ICU stay was introduced into the model as a time-dependent variable using the time to the first erythrocyte unit given.
Data were analyzed using IBM® SPSS® Statistics software, version 22 for windows. All reported p values are two-sided. A p value of less than 0.05 was considered to indicate statistical significance.
Discussion
The present study provides interesting information about the epidemiology and impact of RBC transfusions in ICUs around the globe. Despite the fact that the degree of anemia in our study was comparable to that reported in previous observational studies [
1], the proportion of patients who received a RBC transfusion (26%) is lower than that reported in older studies: for example, 37% in the European ABC study [
1] and 44% in the North American CRIT study [
7], which were comparable studies conducted about 15 years ago, and 33% in the European SOAP study conducted a few years later [
2]. The transfusion triggers in our study were slightly lower in patients admitted to ICUs from Western Europe (8.1 g/dL) and North America (7.9 g/dL) than those reported in the older ABC (8.4 g/dL) and CRIT (8.6 g/dL) studies.
We observed considerable variation in transfusion practice among geographic regions, and marked differences in pre-transfusion hemoglobin concentrations according to age, type of admission, the presence of sepsis, severity scores and ICU length of stay. There are surprisingly few data on ICU transfusion practice in areas of the world other than Europe, Oceania and North America.
As expected, transfusions were associated with higher severity scores, longer lengths of stay and higher mortality rates. The important finding was that in multivariable analysis, RBC transfusion was associated with greater benefit in sicker patients and with worse outcomes in patients with lower severity scores. In the study by Hebert et al. [
11], mortality rates were also higher in the liberal-strategy group than in the restrictive-strategy group in patients with lower Acute Physiology and Chronic Health Evaluation (APACHE) II scores.
A particular strength of this study is the large size of the database, with participants from all geographical regions; nevertheless, we acknowledge that contributions from some areas were limited and may not be representative of the area in general. We also recognize other limitations. First, contribution to the study was voluntary and inclusion may have been subject to selection bias, particularly in the less-represented regions. Moreover, we did not adjust for repeated measures within ICUs or regions, so that small numbers of clinicians within a few large ICUs contributing a large amount of data may have influenced the estimates, especially in smaller regions. Second, although this was a predefined sub-study, the original ICON study was not specifically designed to investigate this topic and some potentially relevant aspects were not available. For example, the indication for red blood cell transfusion, especially acute bleeding, was not recorded in our database, so that a possible confounding effect cannot be excluded. In addition, availability of blood for transfusion at the regional level and differences in local transfusion guidelines may influence decision making in this context. Nonetheless, we did adjust for several factors related to the severity of illness and the underlying disease process. As an observational study, we cannot control for unmeasured confounding as can RCTs, and can therefore only demonstrate an association and not prove causality; our findings can therefore only be considered as hypothesis-generating. Nevertheless, unlike RCTs in which only a fraction of potentially eligible patients are included, observational studies include all patients thus providing data that are more pragmatic and generalizable to the whole ICU population. Third, although transfusion was associated with a lower hazard of mortality in sicker patients, we are unable to comment on optimum thresholds or targets of transfusion. Fourth, we lack information on complications associated with transfusions, such as TRALI, TACO, transfusion reactions and development of nosocomial infections. We also have no information on whether or not transfused units were leukocyte-depleted.