Summary of main results

We included 12 RCTs (1880 participants) that evaluated preoperative rHuEPO with iron to correct preoperative anaemia in non‐cardiac surgery. Preoperative anaemia treatment was mainly investigated in orthopaedic surgery (five studies: Bernabeu-Wittel 2016; Feagan 2000; Moonen 2008; Na 2011; Weber 2005), and gastrointestinal surgery (five studies: Christodoulakis 2005; Heiss 1996; Kettelhack 1998; Kosmadakis 2003; Qvist 1999). Two trials were performed in women scheduled for hysterectomy (Dousias 2003; Larson 2001). rHuEPO with iron was mainly compared to placebo, whilst four studies investigated rHuEPO + iron versus control treatment with oral iron only (Christodoulakis 2005; Dousias 2003; Larson 2001; Weber 2005). Two trials compared rHuEPO with iron to no treatment (Moonen 2008), or to control treatment with iron supplementation (Qvist 1999).

All included studies reported the proportion of participants requiring RBC transfusion. The risk of transfusion is probably reduced up to five days postoperatively when rHuEPO + iron is given to preoperatively anaemic adults (RR 0.55, 95% CI 0.38 to 0.80; participants = 1880; studies = 12; I² = 84%; moderate‐quality evidence due to inconsistency).

A high dose of rHuEPO + iron may result in increased pre‐surgery haemoglobin concentrations (g/dL) (MD 1.87, 95% CI 1.26 to 2.49; participants = 852; studies = 3; I² = 89%; low‐quality evidence due to inconsistency and risk of bias). In the low‐dose rHuEPO + iron regimen, pre‐surgery haemoglobin concentration (g/dL) may not differ between groups (MD 0.11, 95% CI −0.46 to 0.69; participants = 334; studies = 4; I² = 69%; low‐quality evidence due to inconsistency and risk of bias). 

There is probably no difference in the number of allogeneic RBC units transfused between rHuEPO + iron and control groups (MD −0.09, 95% CI −0.23 to 0.05; participants = 1420; studies = 6; I² = 2%; moderate‐quality evidence due to different counting system of the units).

Mortality within 30 days of surgery is probably not affected when rHuEPO + iron is administered to preoperatively anaemic adults before non‐cardiac surgery compared to control (RR 1.19, 95% CI 0.39 to 3.63; participants = 230; studies = 2; I² = 0%; moderate‐quality evidence due to imprecision).

There is probably no difference in adverse events between rHuEPO + iron and control (RR 0.93, 95% CI 0.68 to 1.28; participants = 1722; studies = 10; I² = 0%; moderate‐quality evidence due to imprecision).

We are uncertain if the administration of rHuEPO + iron compared to control influences length of hospital stay (days) (MD −1.07, 95% CI −4.12 to 1.98; participants = 293; studies = 3; I² = 87%; low‐quality evidence due to inconsistency and imprecision). 

Overall completeness and applicability of evidence

The evidence that rHuEPO + iron reduces the need for allogeneic RBC transfusion in preoperatively anaemic patients remains limited. The generalizability of the results to all kind of operations in non‐cardiac surgery is limited because most included trials were in orthopaedic, gastrointestinal, and gynaecological surgery. We are uncertain if the results apply to other surgical fields. Potential differences in patient characteristics in terms of age, gender, comorbidity or malignancies in different areas than non‐cardiac surgery may alter the generalizability. 

In general, participants in the included studies had mild to moderate preoperative anaemia (haemoglobin from 10 to 12 g/dL). We thus cannot be certain that rHuEPO + iron would be effective in avoiding transfusion in patients with a haemoglobin concentration < 10 g/dL, or whether it would be even more effective in those patients.

It may appear contradictory that rHuEPO reduces the need for allogeneic RBC transfusion, but there is no indication of a difference between groups in the number of RBC units administered. It is possible that despite their higher starting haemoglobin concentration, there may have been some unknown clinical reason for some individuals to receive more units when transfusion was required in the rHuEPO + iron group, but the amount of RBC units per participant in total remained unchanged. Alternatively, the result could be explained by the 'in practice' administration of complete units of RBCs rather than the exact volume of RBCs each individual needed. The available information in these reports on the transfusion protocols used was insufficient to support or refute this supposition.

Two different rHuEPO treatment regimens were investigated in the studies. For low‐dose rHuEPO, no changes in presurgical haemoglobin concentration (g/dL) were observed. Nevertheless, data were insufficient to conclude that low‐dose rHuEPO is less efficient compared to the high‐dose regimen. One reason might be that transfusion, as well as haemoglobin concentration (g/dL), are direct and clinically relevant parameters for the efficiency of rHuEPO, but they did not directly reflect the pharmacological‐kinetic stimulation of erythropoiesis. However, both outcomes are of high clinical relevance and are often used as primary and secondary outcomes in clinical trials, especially in anaemic patients in the perioperative setting.

The duration of preoperative rHuEPO treatment was inconsistent across the trials. In four studies, rHuEPO was given either once a week, Dousias 2003; Moonen 2008; Weber 2005, or twice a week, Larson 2001, whereas in all the other trials, rHuEPO was given daily over a 5‐to‐10‐day period. In one trial, preoperative rHuEPO was given on the morning of surgery and for five days postoperatively (Na 2011). As yet no conclusion can be drawn either in favour of a weekly or daily dose regimen of preoperative rHuEPO or a defined continuous administration over the first postoperative days.

Concomitant iron supplementation was mainly given orally (200 mg/d); intravenous iron was administered in three trials (Christodoulakis 2005; Kettelhack 1998; Kosmadakis 2003). We were unable to determine whether iron should be given intravenously or orally for the best effect, or if it is necessary at all as long as iron deficiency is not proven preoperatively.

Furthermore, the applicability of the results is mainly limited because the cause of anaemia was not analysed preoperatively. Besides the reticulocyte counts, parameters of iron metabolism such as transferrin, ferritin, etc. were defined as secondary outcomes in six trials (Christodoulakis 2005; Heiss 1996; Kosmadakis 2003; Larson 2001; Na 2011; Qvist 1999). None of the studies stratified their intervention related to the cause of anaemia. 

According to the current meta‐analysis of all trials, there are still insufficient primary and secondary outcome data to stratify the analysis of these outcomes for gender, different participant populations, or age (e.g. elderly patients aged > 65 years).

Quality of the evidence

We assessed the quality of the evidence for all primary and secondary outcomes using the GRADE approach (Guyatt 2008; Guyatt 2011; Schünemann 2009). This assessment was based on evaluation of the risk of bias (study limitations), publication bias, imprecision, inconsistency, and indirectness of the evidence (summary of findings Table 1).

For the primary outcome need for allogeneic RBC transfusion, the overall quality of evidence was moderate, meaning there might be an important bias in our estimate of the effect of rHuEPO + iron on this outcome. The risk of publication bias remained unclear, but not serious. We found no pre‐published study protocols without consecutive publication. Due to missing study protocols of already published trials, we cannot rule out selective outcome reporting (Figure 4, see funnel plot for publication bias in primary outcome). Based on small confidence intervals in the trials, and an adequate number of events as well as participants, the clinical decision‐making thresholds did not cross the line between in favour of treatment or control treatment, therefore we did not downgrade the quality of evidence for imprecision. Inconsistency of effect estimates, or heterogeneity of evidence, was illustrated by a wide variance of point estimates and large I² values.

The high I² value for the primary outcome may represent substantial heterogeneity. Assuming that the intervention effects would differ between the included studies, we used the random‐effects model and subgroup analysis to explore heterogeneity. Due to the minimal numeric and clinically important differences between high and low dose, we did not present the subgroups of low and high dose for the primary outcome.

The quality of evidence for presurgical haemoglobin concentration was low due to the findings mentioned above regarding evidence quality and to study limitations in sensitivity analysis. 

We assessed the overall quality of evidence for the secondary outcome number of transfused allogeneic RBC units as moderate (summary of findings Table 1). Imprecision was present because the width of the confidence interval was consistent with both important benefit and harm. 

The quality of evidence for the outcomes mortality and adverse events was moderate, downgraded because of imprecision due to a small number of events leading to wide confidence intervals.

The quality of the evidence for the outcome length of hospital stay was low. Inconsistency or heterogeneity was supported by a wide variance of point estimates, high I² values, and heterogeneity of effect estimates. The quality of the evidence was also downgraded due to imprecision with a width of confidence interval including both no effect and appreciable harm.

Potential biases in the review process

The strengths of this review lie in the robust and comprehensive methodology employed to search the literature in important databases and assess and analyse all relevant trials. We followed the recommendations specified in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019). Search strategies were generated by the review group and approved by Cochrane Anaesthesia. The review group consists of several experts in the field (CvH, LK) who are in contact with those continuing clinical research in the field. Screening of the literature search was performed independently by at least two review authors. The meta‐analysis was done following the best clinical and systematic review methodology according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019). A clinician and methodologist of the review group were involved during all stages of the review process.

Furthermore, when there was insufficient reporting of data to allow inclusion and assessment of trials in different languages, we contacted authors directly to obtain the necessary data or translated the study review into English (e.g. Gerasimov 2012).

Agreements and disagreements with other studies or reviews

The results of this systematic review are in line with recent recommendations and findings of national and international guidelines in preoperative anaemia (German S3 Guideline 2018; Goodnough 2011; Kozek‐Langenecker 2017). Regardless of the type of surgery, the German GRADE methodology‐based guideline identified 14 RCTs for the outcome haemoglobin concentration, and 13 RCTs for the outcome transfusion requirements after treatment of preoperative anaemia with rHuEPO (German S3 Guideline 2018). Similar to our findings, the authors of that guideline reported a reduced risk (odds ratio 0.32, 95% CI 0.20 to 0.50; P < 0.001) for anaemic patients undergoing cardiac‐ or non‐cardiac surgery receiving RBC transfusion when rHuEPO with iron supplementation was given preoperatively (German S3 Guideline 2018). Additionally, an MD of 0.96 (95% CI 0.55 to 1.37) for presurgical haemoglobin concentration after rHuEPO treatment with iron was identified in the meta‐analysis of anaemic patients before cardiac‐ or non‐cardiac surgery (German S3 Guideline 2018). However, similar to our findings, the German guideline group suggested the treatment of preoperative anaemia with rHuEPO only after the diagnostics have identified the cause of anaemia (e.g. renal insufficiency) due to the quality of the evidence (German S3 Guideline 2018). If concomitant iron deficiency was present, treatment with rHuEPO in combination with iron was recommended (German S3 Guideline 2018). In the recommendations for the "management of severe perioperative bleeding", the European Society of Anaesthesiology confirms a diagnosis of the cause of anaemia and appropriate therapy (Kozek‐Langenecker 2017).

When participants in cardiac‐ and non‐cardiac surgery were merged, two meta‐analyses identified a reduced risk in RBC transfusion after preoperative administration of rHuEPO in combination with iron compared to placebo with or without iron (Cho 2019; Kei 2019): the RR of RBC transfusion was 0.57 (95% CI 0.46 to 0.71; participants = 4179; RCTs = 25) without increase in  mortality (RR 1.31, 95% CI 0.8 to 2.16) or incidence of severe adverse events (e.g. deep vein thrombosis) (RR 1.28, 95% CI 0.95 to 2.31) (Kei 2019); and (RR 0.59, 95% CI 0.47 to 0.73; participants = 4750; studies = 28) without any difference in thromboembolic events (RR 1.02, 95% CI 0.78 to 1.33; studies = 28) (Cho 2019). However, both meta‐analyses included studies of participants who were not preoperatively anaemic, but who received rHuEPO and iron (Cho 2019; Kei 2019).