Lower versus higher hemoglobin threshold for transfusion in ARDS patients with and without ECMO

The acute respiratory distress syndrome (ARDS) is a common cause for hypoxemia in critically ill patients and associated with a high mortality [1, 2]. Despite advances in treatment strategies during recent decades such as the use of low tidal volume and low-pressure ventilation, adequate positive end-expiratory pressure (PEEP), and prone positioning, mortality of ARDS remains high, exceeding 60% for severe forms of ARDS [1, 3].

Besides cardiac output and pulmonary oxygen uptake, the hemoglobin concentration of the blood determines the blood oxygen carrying capacity and secures vital organ supply. To maintain a certain hemoglobin concentration during a patient’s stay on the intensive care unit (ICU), frequently the transfusion of packed red blood cells (RBCs) is required. For each individual blood transfusion, the transfusion-associated risks have to be carefully balanced against potential transfusion-associated benefits. In the recent decades, evidence has accumulated that accepting a lower than normal hemoglobin concentration can be safe in many different disease conditions and patient populations [4]. Therefore, current practice has gradually shifted to a more restrictive transfusion strategy in patients in the ICU [4‐6].

Data for a target hemoglobin concentration and consecutive transfusion thresholds in patients with ARDS are poor. Current national and international guidelines on the management of patients with ARDS address this topic only vaguely or extrapolate recommendations for non-bleeding anemic, critically ill patients to patients with ARDS [6‐9]. For ARDS patients treated with extracorporeal membrane oxygenation (ECMO), evidence is mainly based on case series and a recent large international survey demonstrated a high heterogeneity of the hemoglobin target concentration with most respondents using a higher transfusion threshold for patients on ECMO compared to other critically ill patients [10‐13]. So far, due to the limited data addressing this issue, the efficacy and safety of different hemoglobin threshold for transfusion of RBCs in adults with ARDS are unknown.

Using a large cohort of 1044 patients with ARDS, the objective of this study was to assess the effect of a lower versus a higher transfusion threshold on short-term outcome in patients with ARDS. We hypothesized that using a transfusion threshold of a hemoglobin concentration of 8 g/dl was non-inferior compared to using a transfusion threshold of a hemoglobin concentration of 10 g/dl with regard to mortality and various failure-free days composites during the first 28 days after ARDS onset.

A total of 1044 patients with ARDS were identified. Of those, 153 patients were excluded because they did not receive RBC transfusions, received the first RBC transfusion later than 28 days after ARDS onset, died within 6 h after admission, or were treated with veno-arterial ECMO. For the remaining 891 patients (85.3% [95% CI, 83.0–87.4]), all transfused during the first 28 days after ARDS onset, a total of 54,915 hemoglobin concentrations (median 38 [IQR, 17–83]) were analyzed to calculate the patient-individual transfusion threshold. After exclusion of patients with other transfusion thresholds and patients with incomplete datasets, a total of 368 patients were included into the matching procedure. Applying a 1:1 paired PSM, 96 pairs (75% of the maximum possible pairs) were identified corresponding to a total of 192 matched patients (Fig. 1). In the matched cohort, the prognostic determinants were well balanced between the two hemoglobin threshold groups (PS 0.25 ± 0.16 vs. 0.27 ± 0.16, P = 0.63). The distribution of the PSs before and after matching is presented in Figures S2 and S3 (Additional file 1).

In the matched cohort, most patients had severe ARDS and the distribution of ARDS etiology was consistent with previously published cohorts. The majority of patients received prone positioning and a lung-protective ventilation with a high PEEP, a low tidal volume, and a low driving pressure. Baseline characteristics of the matched groups are shown in Table 1. There were no differences in demographic data, comorbidities, admissions scores, ARDS severity, ARDS etiology, need for ECLS, and ventilation and ECMO parameters.

In this study, patients were grouped according to their individual hemoglobin concentration that reflected the hemoglobin threshold for RBC transfusion applied during the 28-day period after ARDS onset. Our results suggest that in patients with ARDS who received a blood transfusion 28-day mortality was not higher when a hemoglobin concentration of 8 g/dl was used as transfusion threshold compared to a hemoglobin concentration of 10 g/dl. However, a lower transfusion threshold was associated with a lower chance for successful weaning from the ventilator during the first 28 days after ARDS onset.

In recent years, the number of randomized controlled trials (RCTs) comparing restrictive transfusion thresholds with liberal transfusion thresholds has considerably increased with now more than 30 large RCTs [4]. Classically, hemoglobin levels of 7–8 g/dl are considered as restrictive and hemoglobin levels of 9–10 g/d as liberal transfusion thresholds [4]. In summary, data of these RCTs showed that a restrictive transfusion threshold was not inferior compared to a liberal transfusion threshold with regard to mortality and adverse effects [4]. Current evidence-based recommendations propose a transfusion threshold at a hemoglobin concentration of 7 g/dl for hemodynamically stable ICU patients and patients with septic shock [6, 20]. However, for certain subgroups such as patients with hip fractures and cardiovascular diseases higher hemoglobin thresholds are recommended [6, 20]. For patients with ischemic heart disease, or brain injury, data are not sufficient to formulate reliable recommendations. Likewise, although six RCTs examined transfusion strategies in ICU-patients, these results cannot simply be extrapolated to the specific subgroup of patients with ARDS [21‐26]. While blood oxygen carrying capacity is determined by cardiac output, hemoglobin concentration, and pulmonary gas exchange, the latter is severely compromised in patients with ARDS. The blood oxygen carrying capacity in the complex physiology of this patient population needs to be considered when evaluating the optimal hemoglobin concentration to optimize clinical outcome. Conduction of interventional studies in patients with ARDS, often including therapy with ECMO, is challenging because of the complexity of the disease, the frequent inability to obtain timely patient consent for interventions and the concomitant long recruiting phase [27‐29]. Therefore, using a propensity score based matching procedure is particularly helpful to investigate different transfusion thresholds in this challenging patient population, especially if multiple endpoints are addressed.

In accordance with the results of previous RCTs on transfusion thresholds in various patient populations, the data of this study demonstrate that a lower (restrictive) transfusion threshold compared to a higher (liberal) transfusion threshold was not associated with an increase in 28-day mortality in ICU-patients with ARDS who received a blood transfusion [4, 22, 24, 26, 30]. In addition, applying a higher transfusion threshold was not associated with an increase in the chances for a successful weaning from ECMO within 28 days after onset of ARDS. Therefore, these data argue against the need of a higher transfusion threshold in ARDS patients treated with veno-venous ECMO [6, 10]. In particular, in cases without evidence for anemic hypoxia as reflected by lactate levels and ECMO blood flow settings, a lower transfusion threshold can be safe. Nevertheless, comparing a restrictive versus a liberal transfusion strategy in patients treated with ECMO remains a research priority for the upcoming years [6]. In contrast to the weaning from ECMO, chances for a successful extubation or weaning from the ventilator after tracheostomy within 28 days after onset of ARDS were lower when a lower transfusion threshold was applied. Apart from gas transport and gas exchange disorders, low hemoglobin concentrations can be an important determinant for weaning failure because anemia leads to compensatory stress of the cardio-respiratory system. In patients with a chronic impairment of the respiratory system, RBC transfusion leads to a significant reduction of minute ventilation and work of breathing [31, 32]. In a retrospective analysis of difficult-to-wean patients, patients with a hemoglobin concentration of 8–10 g/dl were more often successfully weaned from mechanical ventilation compared to patients whose hemoglobin concentration was below 8 g/dl [33]. So far, in the absence of any prospective studies investigating effects and complications of different transfusion thresholds in patients undergoing a prolonged pulmonary weaning, our data suggest that a higher hemoglobin threshold in ARDS patients who receive a blood transfusion might facilitate a successful extubation or weaning from mechanical ventilation.

This study has several limitations. Grouping a cohort of ARDS patients according to different hemoglobin thresholds introduced a selection bias with respect to prognostic determinants and study periods between the threshold groups. Furthermore, the studied cohort included mainly patients admitted with a severe ARDS and a concomitant high rate of patients receiving therapy with veno-venous ECMO [1, 28]. Although we cannot rule out that further unknown confounders have affected the results, the selection bias of prognostic determinants in the matched cohort could be reduced to a minimum as indicated by the low standardized mean differences for the majority of variables. Furthermore, we could demonstrate that the primary endpoint was not systematically affected by the change of the transfusion practice during the study period in this patient cohort. Due to the retrospective study design, the individual hemoglobin threshold had to be calculated for each individual patient, allowing less precision than in a prospective study. However, the low coefficient of variation in each transfusion threshold group confirmed a low intra-patient variability of the individual hemoglobin thresholds. Furthermore, the calculation of the individual hemoglobin threshold resulted in the exclusion of 138 patients (13%) that were not transfused. Due to the retrospective study design, transfusion thresholds could only be determined in patients that had received a blood transfusion during the observation period. Patients, who did not receive any RBC transfusion because a lower transfusion threshold was applied, were therefore missed in the analysis. Due to the retrospective study design, evaluation whether a lower transfusion threshold resulted in a lower transfusion rate in the restrictive group was not possible. Despite an appropriate PSM, the current study cannot mimic a RCT, and therefore, comparison of the results of the current study with results of studies that investigated different transfusion thresholds in a prospective setting has to be interpreted with caution. The hemoglobin increment after RBC transfusion is inversely proportional to the pre-transfusion hemoglobin level [34]. This might explain hemoglobin levels around 9 g/dl for patients transfused at a hemoglobin threshold between 7 and 8 g/dl. Although the analyses of this study have to be interpreted as non-confirmatory, Kaplan–Meier curves of the primary endpoint were nearly identical. Therefore, it is unlikely that the loss of statistical power, e.g., due to incomplete datasets and excluded patients, debilitated the validity to detect a significant difference in 28-day mortality between patients in the lower and the higher transfusion threshold groups.

This is the first study exclusively comparing two hemoglobin thresholds for RBC transfusion in patients with ARDS who received a blood transfusion. This single-center cohort study carefully indicates that transfusion at a hemoglobin threshold of 8 g/dl, as compared with a hemoglobin threshold of 10 g/dl, was not associated with an increase in 28-day mortality in adults with ARDS who received a blood transfusion. Whether a higher hemoglobin concentration facilitates successful weaning from the ventilator should be subject of further research.