Preoperative anaemia and red blood cell transfusion in patients with aneurysmal subarachnoid and intracerebral haemorrhage – A multicentre subanalysis of the German PBM Network Registry


 Purpose

Anaemia is common in patients presenting with aneurysmal subarachnoid (aSAH) and intracerebral haemorrhage (ICH). In surgical patients, anaemia was identified as an idenpendent risk factor for postoperative mortality, prolonged hospital length of stay (LOS) and increased risk of red blood cell (RBC) transfusion. This multicentre cohort observation study describes the incidence and effects of preoperative anaemia in this critical patient collective for a 10-year period.
Methods

This multicentre observational study included adult in-hospital surgical patients diagnosed with aSAH or ICH of 21 German hospitals (discharged from 1 January 2010 until 30 September 2020). Descriptive, univariate and multivariate analyses were performed to investigate the incidence and association of preoperative anaemia with RBC transfusion, in-hospital mortality and postoperative complications in patients with aSAH and ICH.
Results

A total of n = 9,081 patients were analysed (aSAH n = 5,008; ICH n = 4,073). Preoperative anaemia was present at 28.3% in aSAH and 40.9% in ICH. RBC transfusion rates were 29.9% in aSAH and 29.3% in ICH. Multivariate analysis revealed that preoperative anaemia is associated with a higher risk for RBC transfusion (OR= 3.25 in aSAH, OR = 4.16 in ICH, p < 0.001), for in-hospital mortality (OR= 1.48 in aSAH, OR= 1.53 in ICH, p < 0.001) and for several postoperative complications.
Conclusions

Preoperative anaemia is associated with increased RBC transfusion rates, in-hospital mortality and postoperative complications in patients with aSAH and ICH.
Trial registration:
 ClinicalTrials.gov, NCT02147795, https://clinicaltrials.gov/ct2/show/NCT02147795


Conclusions
Preoperative anaemia is associated with increased RBC transfusion rates, in-hospital mortality and postoperative complications in patients with aSAH and ICH.
Trial registration: ClinicalTrials.gov, NCT02147795, https://clinicaltrials.gov/ct2/show/NCT02147795 Background Anaemia is an independent risk factor for postoperative complications, mortality, prolonged hospital length of stay (LOS) and increased risk of red blood cell (RBC) transfusion [24]. The prevalence of anaemia is reported to be 22.8% globally and 26.5%-31.5% in patients undergoing surgery [10,2], Preoperative anaemia was reported for 5.5% of patients suffering from aneurysmal subarachnoid haemorrhage (aSAH) [8] and for 24.1-25.8% of patients suffering from intracerebral haemorrhage (ICH) [17,16]. For patients with ICH, anaemia has also been shown to be an independent predictor for unfavourable long-term outcomes a decade ago [17]. Kumar et al. demonstrated that anaemia is common in acute ICH patients and that its presence on admission is an independent predictor of increased ICH volume; contrary to pathophysiological considerations, however, in 2009, they could not demonstrate an effect on increased mortality [16].
Treatment of anaemia in emergency situations usually involves the administration of allogeneic blood products. The administration of RBC transfusions is known to be associated with multiple risks, such as transfusion-related lung injury, haemolytic reactions and transmission of infectious diseases [12]. RBC transfusions in patients undergoing cranial surgery is also associated with a prolonged LOS, more postoperative complications, a 30-day return to the operating theatre and an increased 30-day mortality rate [4]. In patients with aSAH, RBC transfusions have been shown to result in increased mortality and general worse clinical outcomes [31].
This multicentre cohort study analyses the incidence of preoperative anaemia and its association with RBC transfusion requirements, hospital length of stay (LOS), in-hospital mortality and clinically relevant outcomes in patients with aSAH and ICH.

Study design and objectives
The current study is a subanalysis of the ongoing prospective multicentre observational study ´Safety and effectiveness of a Patient Blood Management (PBM) programme in surgical patients´ (ClinicalTrials.gov, NCT02147795) [22]. The period analysed covered 1 January 2010 to 30 September 2020. Data from 23 hospitals was screened. The study was approved by the Ethics Committee of the University Hospital Frankfurt, Goethe University ( rst vote ref. The primary objective of the study was to assess the prevalence of preoperative anaemia and its association with RBC transfusion in aSAH or ICH patients. The secondary objective was to investigate the association of potential risk factors (such as preoperative anaemia, RBC transfusions and other factors related with the type of neurosurgical intervention, additional diagnoses and patient characteristics) with common clinical outcomes (including mortality, typical postoperative complications and LOS) in patients with aSAH and ICH (Online Resource 1).

Patient enrolment and inclusion criteria
The underlying PBM database contained by design adult (≥ 18 years) in-hospital patients, who underwent surgery or a procedure (classi ed according to the Operation and Procedure Classi cation System (OPS) code (Online Resource 1)) during their hospital stay. Patients from the PBM database with a diagnosis of aSAH or ICH, de ned by the International Classi cation of Disease (ICD-10) codes and discharged from hospital within the time period from 1 January 2010 to 30 September 2020, were included (Online Resource 1 and 2). Exclusion criteria were diagnoses of additional traumatic SAH and intracranial neoplasm (Online Resource 1). Patients were assigned to either the aSAH (patients diagnosed with aSAH with/without additional ICH) or the ICH group (patients diagnosed with ICH only) ( Figure 1). This classi cation was chosen because an additional ICH can occur after the aetiological event of an aSAH, even though the patients were originally diagnosed only with aSAH. In the group of patients diagnosed with ICH only, ICH was the primary diagnosis and cause of hospitalisation.
Preoperative anaemia status was based on the rst available preoperative Hb value, and postoperative anaemia status was based on the last available Hb value before hospital discharge. Diagnostic criteria were de ned by the relevant ICD-10 codes (Online Resource S1). Vasospasm was de ned by ICD code I67.80. Interventions were de ned by the relevant Operation and Procedure Classi cation (OPS) codes (Online Resource S1). Mortality was de ned by the discharge code. Hospital LOS was de ned by the given admission and discharge dates.
Data collectionThe underlying data source was anonymous routine data from hospital information systems (e.g. Agfa Orbis, Nexus, iMedOne, SAP) and additional data from individual blood bank and pharmacy software systems of the corresponding hospitals participating in the epidemiological research and quality management study of the German Patient Blood Management Network [22]. The data transferred to the PBM Network Coordination Centre did not contain any personal information. A data protection vote from the Hessian data protection o cer was obtained (ref: 43.60; 60.01.21-ga from 24 October 2018). The biostatistician in charge subsequently evaluated the data for completeness and correctness and extracted the cases that ful lled the inclusion criteria for this study before performing the nal analysis.

Statistical analysis
Descriptive analysis was used to determine patient characteristics, the prevalence of pre-and postoperative anaemia, RBC transfusions, surgical interventions, and postoperative outcomes. The results of the descriptive analysis are presented as means (± standard errors), medians (with rst and third quartiles) and rates (with 95% CI).
Multivariate mixed effect regression analysis was performed to identify independent predictors of RBC transfusion and various postoperative outcomes. The multivariate mixed effect regression models included the hospitals as random effects (to account for hospital individual effects) and other potentially relevant factors (such as age, gender, surgical interventions, preoperative anaemia, RBC consumption and vasospasm) as xed effects.
Univariate non-parametric analysis (chi-square tests for binary endpoints and Wilcoxon-Mann-Whitney tests for continuous endpoints) was performed a-priori to assess the correlation of the individual factors where appropriate. To account for the heterogeneity of the aSAH and ICH groups, all analyses (univariate and multivariate) were performed separately by group.All analyses were performed using the free software R (Version 3.6.3).

Results
A total of n = 1,325,438 patients from 23 hospitals were screened. Two hospitals in the network did not treat patients with neurosurgical diagnoses. Overall, n = 9,081 eligible patients from 21 hospitals were included and analysed in this study. The aSAH group included n = 5,008 patients and the ICH group included n = 4,073 patients ( Figure 1). The incidence of eligible cases within the entire database of n = 1,325,438 was 0.4% for aSAH and 0.3% for ICH. Most patients received a neurosurgical OPS (84.9% in aSAH and 76.9% in ICH). The remaining OPS are distributed across several specialties (visceral and endocrine surgery accounts for the highest proportion with 5.0%, followed by 3.5% with otorhinolaryngology). The distribution of other surgical OPS can be found in Online Resource 2). Demographic and intervention data are shown in Table 1.
Anaemia -The median preoperative Hb level was 13.2 g/dl in aSAH patients and 12.8 g/dl in ICH patients. Severe, moderate and mild preoperative anaemia was present in aSAH patients at rates of 1.0%, 10.7% and 16.6%, respectively, and in ICH patients at rates of 2.7%, 17.6% and 20.6%, respectively (Table 1).
Descriptive and univariate analysis for postoperative outcomes according to preoperative anaemia are listed for both pathologies in Tables 1/2 and Online  Resource Tables 3/4. Mortality was signi cantly higher in the presence of preoperative anaemia (22.2% versus 13.3%, p < 0.001 in aSAH and 31.5% versus 17.9%, p < 0.001 in ICH) ( Table 2). Figure 2 demonstrates that an increase in the preoperative Hb values corresponds to a decrease in the mortality rate.

Discussion
The data revealed in both, aSAH and ICH patients, that preoperative anaemia is associated with a higher RBC transfusion rate, increased postoperative inhospital mortality and increased complication rates. These ndings align with study results for other patient cohorts. Thus, in neurosurgical patients, preoperative anaemia has been shown to be an independent risk factor for postoperative mortality and increased risk of RBC transfusion [24]. Anaemia is common in aSAH patients [30,14,8,33] and in ICH patients [17,16]. In this study, the prevalence of preoperative anaemia in both groups (aSAH 28.3% and ICH 40.9%) was higher than described in previous publications (aSAH 5.5% and ICH 24.1-25.8%) [8, 17,16]. One explanation for this could be that the database only includes cohorts of patients who underwent surgery or other interventions (e.g., coiling) during a hospital stay, so that selection bias cannot be ruled out.
The physiological and pathophysiological impact of anaemia in patients with aSAH and ICH is multifactorial. The supply of oxygen to the brain depends on several variables. Cerebral oxygen availability (DO 2 ) is the product of cerebral blood ow (CBF) and arterial oxygen content (CaO 2 ).: DO 2 = CBF x CaO2 [19].
The oxygen content (CaO 2 ) itself is represented by the formula CaO 2 = (1.31 × Hb ×SaO 2 × 0.01) + (0.0225 × PaO 2 ) and thus depends on Hb levels, arterial oxygen saturation (SaO 2 ) and arterial oxygen pressure (PaO 2 ) [7]. The formula demonstrates that apart from an increase in SaO 2 , the most signi cant factor for optimising the DO 2 to the target cell is the Hb value; thus, the need arises to consider ways of increasing the Hb value through various measures (such as anaemia management or transfusion in an emergency). In healthy brain, a progressive decrease in Hb is compensated for by vasodilation, resulting in increased CBF and a constant cerebral oxygen supply DO 2 . When Hb falls below 5-6 g/dL, DO 2 decreases and no further vasodilation can occur and maximum CBF levels are reached [19]. We observed that the vasospasm rate was signi cantly lower with preoperative anaemia. It is possible, that the Hb value in uences only patients' outcomes after cerebral vasospasm and not the probability of a cerebral vasospasm event itself. The multivariate analysis also revealed a signi cant association of RBC transfusion (OR= 2.47, p<0.001) with vasospasm. This nding underlines the need for risk assessment prior to transfusion and additional prospective studies on this topic. Scholars have long debated whether elevating the haemoglobin levels in SAH patients with vasospasm and thus avoiding anaemia is bene cial [15,29,18]. In general, based on CONSCIOUS-3, the role of vasospasm on delayed cerebral ischemia should be considered with caution, where clazosentan was shown to signi cantly reduce postaSAH vasospasm, but neither dose improved outcome [20]. Further studies are needed to prove the potential bene cial effects of RBC transfusion on anaemic aSAH patients suffering from cerebral vasospasm. In the eld of critical care, there is a growing evidence that strict transfusion limits remain best practice for the vast majority of cases, due to limited adverse effects, comparable or better clinical outcomes and economic aspects [28]. Thus, a restrictive threshold for RBC transfusions (Hb < 7 g/dl) is still recommended in both critically ill and clinically stable ICU patients [23]. Similar pathophysiological considerations are known in patients with acute myocardial infarction, as a recent study demonstrated that a restrictive transfusion strategy resulted in less major adverse cardiac events after 30 days (11.0% in the restrictive and 14.0% in the liberal group) [6]. In the retrospective study by English et al., only 20% of patients with aSAH received RBC transfusions, mostly in the presence of signi cant anaemia (Hb < 8 g/dl), and this was not associated with worse outcomes [8]. However, Dhar et al. demonstrated that RBC transfusion in aSAH patients improved cerebral oxygenation both globally and particularly in the vulnerable brain regions and thus may potentially minimise the risk for delayed cerebral ischaemia. The study analysed the outcomes over a wide range of haemoglobin levels and suggests that restrictive transfusion practice may not be appropriate in this vulnerable population [5]. Naidech et al. demonstrated no difference in outcomes in SAH for Hb 10.0 versus 11.5 g/dL. Here, however, the difference between the groups is rather minor and well above general limits for transfusions [25]. The answer to the question of the role of treatment of anaemia with red blood cell transfusion could be provided by the still ongoing SAHaRA trial [9]. In our analysis, mortality was increased considerably the more transfusions were given, which is also in line with the results from Ceanga et al. [3].
Although preoperative diagnosis and treatment of anaemia can only be implemented to a limited extent in acute situations of ICH and aSAH, the present data underlines primarily the importance of general anaemia vigilance and treatment (as ICH and aSAH occur sudden and without time frame for treatment), and secondarily, anaemia treatment becomes important in the context of peri-/postoperative care. To identify and manage anaemia at an early stage, a multimodal therapy using patient blood management (PBM) has been developed. PBM is an evidence-based, patient-centred and multidisciplinary approach to minimise anaemia-associated risks, unnecessary blood loss and transfusions in patients undergoing surgery [1]. For this purpose, measures have been implemented to reduce preoperative anaemia, minimise iatrogenic blood loss and optimise patient-speci c anaemia tolerance [11]. If iron de ciency is identi ed in the absence of infection, iron supplementation and erythropoiesis-stimulating agents can be considered [13]. Measures to reduce intraoperative blood loss and optimise coagulopathy should be implemented. This includes the following measures (also in neurosurgery): Treatment of coagulopathy should be based on a xed algorithm. The content of the coagulation algorithm should be the maintenance of basic conditions for haemostasis (body temperature > 36°C, ionised calcium > 1.1 mmol/ L, pH > 7.2) or point-of-care diagnostics. The prevalence of bleeding due to anticoagulation was low in our analysis but point of care technology provides information on coagulation dysfunction and the use of anticoagulation, including NOACs. The use of an anti brinolytic is safe and recommended [32]. Blood sample collections should be reduced to the absolute necessary numbers, blood sample collection tubes should draw as little blood volume as possible, and return systems for blood sample collections should be established. Washed cell salvage-the collection, washing and retransfusion of a patient's own wound blood-can help to reduce the need for blood from other sources [26,21].

Limitations
Although studies with routine data have several important advantages over traditional clinical trials (such as a larger number of cases with fewer personnel, time and cost requirements) and are therefore becoming increasingly popular as an alternative in the age of advancing digitalisation, they naturally also have some disadvantages. This study is based on routine data of hospital information systems. Data quantity and quality varied between hospitals. In addition, routine data may have some other limitations, in general, such as missing data or incorrect coding techniques. Since ICD and OPS codes are billing-related, they may be biased.
Furthermore, there is no information on the exact time of occurrence and duration of complications and perioperative interventions (including blood transfusions), so an association does not necessarily indicate causality, nor is it possible to show the direction of causality. For this reason, we report associations rather than causalities of the factors.
Missing information on neurological status, resuscitation and intercurrent diseases cannot be obtained from the register, so that a limitation in the analysis of associations with anaemia and transfusion is possible here. The analysis could not consider the in uence of a potential -and already locally availableanaemia therapy. The Hunt and Hess scale for aSAH, which measures the severity of the aSAH, is not documented in ICD-10 codes; therefore, a severityadapted evaluation was not possible. In neurosurgical therapy, patients are often transferred to a rehabilitation intensive care unit or back to the referring intensive care unit shortly after treatment; this leads to a possible bias in endpoints (e.g., especially for LOS). This is a retrospective analysis of prospectively collected registry data; limitations of a retrospective analysis cannot be avoided.

Conclusions
Preoperative anaemia is associated with increased RBC transfusion rates, in-hospital mortality, and postoperative complications in patients with aSAH and ICH. Prospective multicentre studies with tailored data on the therapy of anaemia, the optimal haemoglobin value and transfusion strategy, both for aSAH and ICH patients, are urgently needed. We also wish to thank the IT staff of the participating centres for providing routine and additional data. In addition, we would like to thank the medical staff of each centre for their efforts in the practical implementation of PBM. We thank Dr. Christoph Füllenbach for his previous general support for the PBM project.
Ethics approval and consent to participate

Consent for publication
Not applicable

Availability of data and materials
The data that support the ndings of this study are available from the corresponding author upon reasonable request.  Tables   Table 1) Table 1 shows the patient characteristics, interventions and anaemia prevalence. Anaemia rates are calculated only from the subset of patients, of whom the required pre-and/or postoperative Hb values were available. The exact numbers are given per individual line. All values are represented as mean (± SE), median (IQR) or as rate (95% CI) and total number.   In preoperative non-anaemic patients 1.5 % (1.1 % -1.9 %); n= 45/3,086 2.9 % (2.2 % -3.7 %); n= 59/2,053 Table 2 shows the postoperative outcomes depending on anaemia status. All values are represented as mean (± SE), median (IQR) or as rate (95% CI) and total number.  Table 3 shows the transfusion rates and number of RBC unitis per 1000 patients. Subanalyses were performed according to anaemia status, bleeding due to anticoagulation, factor de ciency or pro-bleeding interventions.
NOACs: Novel oral anticoagulants. All values are represented as mean (± SE) or as rate (95% CI) and total number.  Table 5 shows the results of the multivariate regression analysis to investigate risk factors for postoperative outcomes in aSAH patients. The vertical column lists the risk factors and the horizontal column lists the corresponding outcomes. All values are represented either as Odds ratio (with 95% CI) for binary endpoints or as difference in mean ± standard error of mean (SE) for continuous endpoints. Non-signi cant results and p-values >0.10 are marked with n.s..  Table 6 shows the results of the multivariate regression analysis to investigate risk factors for postoperative outcomes in ICH patients. The vertical column lists the risk factors and the horizontal column lists the corresponding outcomes. All values are represented either as Odds ratio (with 95% CI) for binary endpoints or as difference in mean ± standard error of mean (SE) for continuous endpoints. Non-signi cant results and p-values >0.10 are marked with n.s.. Figure 1 illustrates the inclusion and exclusion criteria among patients analysed Figure 2 illustrates the mortality rate dependent on the preoperative Hb values for a) Aneurysmal subarachnoid haemorrhage (aSAH) and b) Intracerebral haemorrhage (ICH). 95% con dence intervals (error bars) are shown. Figure 3 illustrates the RBC transfusion rate dependent on the preoperative Hb values for a) Aneurysmal subarachnoid haemorrhage (aSAH) and b) Intracerebral haemorrhage (ICH). 95% con dence intervals (error bars) are shown.