Intravenous or oral iron for treating iron deficiency anaemia during pregnancy: systematic review and meta‐analysis

Anaemia affects about 40% of pregnant women worldwide, with iron deficiency the leading cause.1 Anaemia is associated with an increased risk of adverse perinatal outcomes, including low birthweight and pre‐term birth.2,3 Further, iron deficiency anaemia (IDA) during late pregnancy is associated with adverse effects on the neurodevelopment of the child.4,5

Despite the high incidence and burden of IDA, research findings that could guide clinical management are scarce. While oral iron treatment improves haematological parameters in pregnant women, its benefits for subsequent maternal or neonatal clinical outcomes is not supported by high quality evidence.6 Further, oral iron therapy frequently has adverse gastrointestinal effects that reduce adherence to treatment.6

Intravenous iron is a recognised alternative to oral iron therapy, particularly during late pregnancy, when a rapid treatment response is desirable, or for women who tolerate oral iron poorly.7 Compared with oral iron, intravenous iron significantly increases haemoglobin (mean difference [MD], 0.85 g/dL; 95% confidence interval [CI], 0.31–1.39 g/dL) and ferritin levels (MD, 63.3 μg/L; 95% CI, 39.5–87.2 μg/L), and is associated with fewer adverse gastrointestinal effects than oral iron (risk ratio [RR], 0.50; 95% CI, 0.34–0.73).8 These and similar findings have led to the increased use of intravenous iron therapy during pregnancy.9

Systematic reviews and meta‐analyses have not found that intravenous iron is beneficial with respect to maternal or neonatal clinical outcomes, and it is therefore unclear whether its short term improvements of haematological parameters outweigh the potential harms and greater expense of this approach.6,10 Intravenous iron therapy is actively promoted as a key component of patient blood management strategies.7 However, while antenatal anaemia is a risk factor for needing a blood transfusion post partum, is it is not known whether intravenous iron therapy reduces the risk. The objective of our systematic review and meta‐analysis was therefore to compare maternal and neonatal clinical outcomes at delivery following intravenous and oral iron treatment of IDA during pregnancy.

Methods

Data sources and search strategy

We undertook and report this systematic review in accordance with the Preferred Reporting Items for Systematic reviews and Meta‐Analyses (PRISMA),11 the Grading of Recommendations, Assessment, Development and Evaluation (GRADE),12 and the Cochrane Handbook for Systematic Reviews of Interventions guidelines.13 Our review was registered with PROSPERO (CRD42019120652).

On 20 January 2019, we searched the Ovid MEDLINE, EMBASE, Scopus, Cochrane Central Register of Controlled Trials, and Web of Science databases for publications on randomised controlled trials (RCTs) that compared the effects of intravenous iron and oral iron as first‐line treatment of IDA during pregnancy. A full list of the search terms is included in the Supporting Information, table 1. Earlier reviews, bibliographies of published trials, and cross‐references were also checked for relevant publications. No publication date or language restrictions were applied.

Study selection and data extraction

Eligible studies described women who initially had low haemoglobin levels (< 110 g/L) or were at high risk of developing IDA because their iron reserves were insufficient (serum ferritin level < 30 μg/L). Studies were excluded if they did not report maternal or neonatal haematological parameters at delivery, did not report any perinatal outcomes, provided inadequate outcomes data, or were published in abstract form.

Two reviewers (AQ, LG) independently screened the titles and abstracts of studies. Disagreements were resolved by consensus or consultation with a third reviewer (RG). Two reviewers (AQ, LG) independently extracted the following data on standardised data extraction sheets: study characteristics, patient characteristics, treatment outcome measures.

The primary outcome was need for maternal blood transfusion. Secondary outcomes included maternal (eg, breastfeeding) and neonatal (eg, birthweight) clinical outcomes. Haematological data were also extracted, including maternal and neonatal haemoglobin and ferritin levels at delivery. In one study,14 the standard deviation for mean birthweight was not reported and was imputed from the reported P‐value for the difference between study groups, as outlined in chapter 7.7.3.3 of the Cochrane Handbook.13 In another study,15 the reported median blood loss at delivery was converted to an estimated mean (and standard deviation) with a previously validated method.16,17

Study quality assessment

Two investigators (AQ, LG) independently evaluated the methodologic quality of included studies with the Cochrane risk of bias tool.13 In brief, overall risk of bias was assessed by answering questions regarding the following aspects of studies (responses: “yes” = low risk of bias; “no” = high risk of bias; “unclear” = no information, or uncertainty about potential bias): random sequence generation, allocation concealment, blinding of participants, blinding of outcome assessment, incomplete outcome data, selective reporting, other bias. Any disagreement was resolved by consensus.

Two authors (LG, AQ) independently evaluated the quality of evidence of studies contributing data to each outcome according to the five GRADE categories (study limitations/risk of bias, consistency of effect, imprecision, indirectness, publication bias).12,13 Disagreements were resolved by consensus or discussion with a third author (RG).

Data synthesis

Meta‐analyses were performed with the Cochrane review manager software (RevMan 5.3; The Nordic Cochrane Centre) and the R 3.4.3 packages meta and metafor. For rare dichotomous outcomes (frequency lower than 5% in each treatment arm), we calculated Peto odds ratios (ORs) in a fixed effects model;13 for frequent dichotomous outcomes, we calculated pooled relative risks (RRs) in a random effects model. For continuous outcomes, we calculated mean differences (MDs) in a random effects model.

Between‐study heterogeneity was assessed in χ² tests (threshold P = 0.10) and quantified as I². We planned to investigate reporting bias in funnel plots if more than ten studies reported data on the same outcome.

Subgroup and sensitivity analyses

We assessed the robustness of results for the primary outcome (blood transfusion) in sensitivity analyses using a range of meta‐analytic methods.18 We undertook pre‐specified stratified analyses to determine whether treatment effects differed by study setting (low or middle income v high income countries), pre‐treatment haemoglobin level (< 90 v ≥ 90 g/L), or gestational age at treatment (≤ 28 v > 28 weeks’ gestation). We also stratified outcomes according to study quality (high v low risk of bias) and iron formulation (iron polymaltose or ferric carboxymaltose v iron sucrose).

Results

Study selection

Of the 681 publications identified after removing duplicates, 38 full‐text articles were screened for eligibility; 25 were excluded as ineligible. An additional three publications19,20,21 were identified by bibliographic search, yielding a total of 15 eligible studies (16 publications) and data for 1938 participants (Box 1).14,15,19,20,21,22,23,24,25,26,27,28,29,30,31,32

Study characteristics

The total number of women (intervention and control arms) in the included studies ranged from 50 to 252. Most studies were undertaken in India (eight); two were undertaken in Australia, and one each in Singapore, France, Thailand, Turkey, and multiple countries. Baseline mean haemoglobin levels (range, 60–109 g/L) and mean gestation at enrolment (range, 22–33.3 weeks) differed between studies. The most frequently used intravenous iron preparation was iron sucrose (12 studies); iron polymaltose was used in two and ferric carboxymaltose in one. Twelve studies employed variable doses according to baseline haemoglobin level and (in eleven studies) weight, with the target haemoglobin levels ranging from 110 to 150 g/L. Two studies used fixed intravenous iron doses (400 or 500 mg iron sucrose) (Box 2).

Primary outcome: blood transfusion

Nine RCTs (555 intervention, 535 control participants)14,15,20,21,23,25,26,27,32 reported the risk of maternal blood transfusion. Compared with oral iron, intravenous iron therapy was associated with reduced risk of women requiring blood transfusions (Peto OR, 0.19; 95% CI, 0.05–0.78; low quality evidence). This corresponded to a number needed to treat with intravenous iron (v oral iron) of 95 women to avoid one blood transfusion (95% CI, 81–348 women). Pooled effect estimates and forest plots for the primary outcome are included in Box 3 and Supporting Information, figure 1A.

Secondary outcomes: maternal and neonatal clinical outcomes at delivery

Compared with oral iron, intravenous iron therapy was associated with greater infant birthweight (MD, 58 g; 95% CI, 4–112 g; 8 RCTs, low quality evidence).14,20,21,26,27,29,30,31 There were no significant differences in gestation at birth, caesarean delivery, hypertensive disorders of pregnancy, or pre‐term birth (Box 3, Supporting Information, figure 1B–F).

Single studies reported data for other secondary outcomes. Compared with oral iron, intravenous iron did not appreciably influence blood loss at delivery (MD, –9.0 mL; 95% CI, –96.3 to 78.3 mL; low quality evidence),15 post partum haemorrhage (Peta OR, 0.14; 95% CI, 0–6.82; very low quality evidence),27 birth length (MD, 0.1 cm; 95% CI, –0.6 to 0.8 cm; low quality evidence),30 birthweight‐to‐length ratio (MD, 0; low quality evidence),30 or stillbirth (Peto OR, 0.14; 95% CI, 0.00–6.82; very low quality evidence).14 One study found that the rate of breastfeeding cessation within 24 months of delivery was lower for women who received intravenous iron (hazard ratio, 0.70; 95% CI, 0.50–0.99; very low quality evidence)22 (Supporting Information, table 2).

Maternal quality of life

One study evaluated quality of life with the 36‐item short form (SF‐36) health survey; the authors found that intravenous iron was associated with greater improvements in vitality (MD, 5.90; 95% CI, 0.75–11.0) and social functioning scores (MD, 5.77; 95% CI, 0.01–11.5) at the time of delivery than oral iron (both very low quality evidence).30 In another study, the authors found no differences in quality of life at delivery or 6–8 weeks’ post partum on any component of an 11‐item modified version of the SF‐36 questionnaire (data not reported in publication).22

Maternal and neonatal hematologic parameters

Compared with oral iron, intravenous iron therapy led to higher maternal haemoglobin (MD, 7.4 g/L; 95% CI, 3.9–11 g/L; nine RCTs, low quality evidence)14,19,20,21,24,26,27,28,29 and ferritin levels (MD, 21.2 μg/L; 95% CI, 6.5–36.0 μg/L; three RCTs, low quality evidence)21,26,29 at the time of delivery, but did not influence neonatal haemoglobin (MD, –1.0 g/L; 95% CI, –4.7 to 2.8 g/L; six RCTs, low quality evidence)14,20,26,27,29,30 or ferritin levels (MD, 11.2; 95% CI, –1.6 to 24.1 μg/L; six RCTs, low quality evidence)14,20,26,27,29,30 (Box 3, Supporting Information, figure 1G–J).

Study quality assessment

All 15 studies were at high risk of bias because of the lack of blinding of participants and study personnel. Ten studies were at high risk of bias in at least one other domain, including five studies with incomplete outcomes data15,22,24,30,31 and five studies20,23,26,27,32 in which oral iron was administered to women in the control arm for only four weeks (Supporting Information, table 3).

The quality of evidence, as determined by the GRADE approach, was rated low to very low for all outcomes, indicating limited confidence or uncertainty about the relative effects of intravenous and oral iron (Supporting Information, table 4).

Sensitivity and subgroup analyses

Sensitivity analyses indicated that the pooled estimate for the primary outcome (post partum blood transfusion) was reasonably consistent across a range of alternative meta‐analytic approaches (Supporting Information, table 5).

In stratified analyses, outcomes were similarly consistent regardless of baseline haemoglobin level, gestational age at treatment, and study setting (Supporting Information, table 6). The treatment effect of intravenous iron on birthweight was greater in studies in which the mean baseline haemoglobin level was less than 90 g/L, gestation at treatment was less than 28 weeks, or intravenous iron sucrose was used, and also in studies at high risk of bias. The treatment effects of intravenous iron on neonatal ferritin levels were greater in studies with a mean baseline haemoglobin level exceeding 90 g/L, undertaken in high income countries, or employing intravenous iron polymaltose or ferric carboxymaltose.

Discussion

Main findings

Our systematic review of the treatment of pregnant women with IDA indicates that high quality evidence for intravenous iron therapy achieving better maternal or neonatal clinical outcomes at delivery than oral iron has not been published. There is low quality evidence that intravenous iron therapy is more effective in reducing the risk of blood transfusion during delivery and increasing neonatal birthweight, but the treatment effects were small and the relevant studies subject to high risk of bias. Further, while intravenous iron was more effective than oral iron in improving maternal haematological parameters at delivery, neonatal haematological outcomes were similar with both approaches.

Interpretation

All studies included in our systematic review were identified as being at high risk of bias with respect to the maternal and neonatal clinical outcomes evaluated. The absence of treatment blinding is particularly important, as knowledge of treatment allocation can bias outcome evaluation and later treatments, especially as clinical thresholds for blood transfusions were not pre‐defined in any study. Evaluating subjective outcomes, including quality of life, is particularly susceptible to knowledge of treatment. Blinding of treatment allocation was successfully employed in a multicentre study of intravenous iron therapy in intensive care units,33 but it is unclear whether blinding would be feasible in a an antenatal study.

Most included studies were undertaken in low or middle income countries, and interpreting and generalising their findings to high income countries is problematic. Studies undertaken in low or middle income countries were more frequently identified as being at high risk of bias, included women with more severe anaemia, and employed intravenous iron sucrose, which is not often used in high income countries. Further, the intravenous iron doses in these studies were frequently much lower than usually recommended for clinical practice.7,34 However, there was no evidence for differences in outcomes by study setting. More importantly, we identified only three studies that investigated treatment with intravenous ferric carboxymaltose or iron polymaltose, the usual formulations in Australia.34,35 Notable differences with respect to administration have been described,34,36 but whether the choice of intravenous iron formulation affects treatment outcomes is unknown. The dose of intravenous iron administered may also influence outcomes.37

Our meta‐analysis significantly builds on another recent meta‐analysis10 which led its authors to conclude that the risks of requiring a blood transfusion were similar after intravenous and oral iron therapy (RR, 1.02; 95% CI, 0.99–1.04). Our finding of a significant difference is explained by our identifying three studies15,25,32 not included in the earlier analysis, and by our more appropriate meta‐analytic approach to handling rare events. Other recent reviews have either focused only on haematological parameters33 or did not meta‐analyse clinical outcomes.21

A 2013 meta‐analysis examined the safety and efficacy of intravenous iron therapy in reducing the need for blood transfusions.38 In the 22 RCTs that included data on blood transfusions, intravenous iron administration led to a substantial reduction in risk (RR, 0.74; 95% CI, 0.62–0.88); four studies included pregnant women, but the meta‐analysis results were not stratified by clinical setting. However, from these data it can be calculated that the prevalence of blood transfusions was lower among patients receiving intravenous iron than among those receiving oral iron, both in non‐obstetric (15.7% v 21.8%) and obstetric studies (0.8% v 2.8%). Based on our estimated number needed to treat in our meta‐analysis (95 women to avert one transfusion; 95% CI, 66–283) and the reported cost in Australia of $400 per woman for intravenous iron therapy,39 the amount saved per red blood cell transfusion by using intravenous rather than oral iron therapy would be $38 000 (95% CI, $34 400–$139 200). This estimate is based on using ferric carboxymaltose, but the savings are likely to be similar for iron polymaltose, which requires significantly longer infusion times and therefore has higher administration costs.36

The relative cost‐effectiveness of intravenous iron therapy in averting the need for blood transfusions therefore remains unclear. While blood transfusions are lifesaving interventions for managing acute severe blood loss, concerns about serious adverse events (alloimmunisation, transfusion‐related infections and acute lung injury), costs, and the availability of blood, makes assessing their cost‐effectiveness difficult.38,40 The benefits of intravenous iron administration must also be balanced against potential medication‐related harms; the authors of a recent systematic review reported that the incidence of moderate to severe adverse events requiring intervention was 6–10 events per 1000 pregnant women treated with intravenous iron;36 the prevalence of persistent skin staining after intravenous iron administration is reported to be 1.3%.41 An earlier meta‐analysis (not specific to treatment during pregnancy) found that the risk of infection was higher with intravenous than oral iron therapy (RR, 1.33; 95% CI, 1.10–1.64).38 None of the studies included in our systematic review reported data for infection‐related outcomes, but the possibility of increased risk should generate caution about treating pregnant women with intravenous iron. On the other hand, the risk of anaphylaxis following first exposure to newer non‐dextran intravenous iron formulations (iron sucrose, iron gluconate, and ferumoxytol) is low (24 events per 100 000 persons).42

Strengths and limitations

We identified three studies not included in earlier meta‐analyses, and our systematic review applied a rigorous approach guided by PRISMA recommendations, without limiting our search by publication date or language.

Limitations of our meta‐analysis include the small numbers of participants in individual studies, which were underpowered for detecting clinically relevant differences in outcomes. Further, we were unable to explore publication bias because no outcome had been assessed by at least ten eligible studies. Finally, the daily iron dose in the control arms of the included studies ranged from 80 to 300 mg elemental iron, and this may have affected treatment response in control participants.

Conclusion

Despite the high incidence and burden of disease associated with IDA in pregnant women, evidence regarding differences in the effectiveness of intravenous and oral iron for improving maternal and neonatal outcomes at delivery is at best weak. Most relevant studies have been undertaken in low or middle income countries, and their results do not provide strong support for preferring intravenous iron to oral iron as routine first‐line therapy for pregnant women in clinical practice. Large, high quality studies powered to investigate perinatal clinical outcomes and treatment‐related harms are needed to determine whether intravenous iron is superior to oral iron for treating IDA during pregnancy.

Box 1 – PRISMA flow diagram of studies identified for our systematic review

---

PRISMA = Preferred Reporting Item for Systematic Reviews and Meta‐analyses.11