Efficacy and safety of intravenous iron repletion in patients with heart failure: a systematic review and meta-analysis

Introduction AFFIRM-AHF and IRONMAN demonstrated lower rates of the combined endpoint recurrent heart failure (HF) hospitalizations and cardiovascular death (CVD) using intravenous (IV) ferric carboxymaltose (FCM) and ferric derisomaltose (FDI), respectively in patients with HF and iron deficiency (ID) utilizing prespecified COVID-19 analyses. Material and methods We meta-analyzed efficacy, between trial heterogeneity and data robustness for the primary endpoint and CVD in AFFIRM-AHF and IRONMAN. As sensitivity analysis, we analyzed data from all eligible exploratory trials investigating FCM/FDI in HF. Results FCM/FDI reduced the primary endpoint (RR = 0.81, 95% CI 0.69–0.95, p = 0.01, I2 = 0%), with the number needed to treat (NNT) being 7. Power was 73% and findings were robust with fragility index (FI) of 94 and fragility quotient (FQ) of 0.041. Effects of FCM/FDI were neutral concerning CVD (OR = 0.88, 95% CI 0.71–1.09, p = 0.24, I2 = 0%). Power was 21% while findings were fragile with reverse FI of 14 and reversed FQ of 0.006. The sensitivity analysis from all eligible trials (n = 3258) confirmed positive effects of FCM/FDI on the primary endpoint (RR = 0.77, 95% CI 0.66–0.90, p = 0.0008, I2 = 0%), with NNT being 6. Power was 91% while findings were robust (FI of 147 and FQ of 0.045). Effect on CVD was neutral (RR = 0.87, 95% CI 0.71–1.07, p = 0.18, I2 = 0%). Power was 10% while findings were fragile (reverse FI of 7 and reverse FQ of 0.002). Rate of infections (OR = 0.85, 95% CI 0.71–1.02, p = 0.09, I2 = 0%), vascular disorder (OR = 0.84, 95% CI 0.57–1.25, p = 0.34, I2 = 0%) and general or injection-site related disorders (OR = 1.39, 95% CI 0.88–1.29, p = 0.16, I2 = 30%) were comparable between groups. There was no relevant heterogeneity (I2 > 50%) between the trials for any of the analyzed outcomes. Conclusions Use of FCM/FDI is safe and reduces the composite of recurrent HF hospitalizations and CVD, while effects on CVD alone are based on available level of data indeterminate. Findings concerning composite outcomes exhibit a high level of robustness without heterogeneity between trials with FCM and FDI. Graphical Abstract


Introduction
Iron deficiency (ID) frequently occurs in patients with heart failure (HF) being prevalent in up to 50% of patients. [1,2] Beyond its connection with frailty, impaired quality of life, exercise and functional capacity [3,4] ID associates with recurrent HF hospitalizations, CVD and allcause mortality [5]. These associations exist independently of the presence of anaemia [6][7][8].
Results from randomized trials not powered for clinical end-points have shown that iron supplementation with FCM is safe and improves quality of life, [9,10] symptoms [9] and exercise capacity [9] in ID, being of the very few HF drugs that have shown benefit in this regard in a recent systematic review [11]. The CONFIRM-HF trial [9] with FCM showed a reduced risk of hospitalizations for worsening HF. This trial was underpowered to determine this clinical outcome. In the powered AFFIRM-AHF trial [12], FCM lowered the rate of composite of total HF hospitalizations and CVD, of note, in a pre-specified COVID-19 sensitivity analysis. Similarly, in IRONMAN [13] treatment with IV ferric derisomaltose (FDI) lowered the same primary endpoint, also in a COVID-19 sensitivity analysis.
In the crude analyses of AFFIRM-AHF and IRONMAN trial, there were only trends for benefit regarding all-cause death or CVD. AFFIRM-AHF and IRONMAN were powered, randomized trials with similar design, identical primary endpoints and both hampered by COVID-19 pandemic [14]. Therefore, we meta-analyzed data from both trials (from crude analyses and those reported in the pre-specified COVID-19 sensitivity analyses) as also data from all eligible trials with similar design to explore whether primary endpoint and CVD is heterogenous between the trials, data were robust and whether the trials had sufficient power to detect the observed effect size. This would strengthen the level of evidence that i.v. iron supplementation improves outcome in HF patients and reassure physicians to screen and treat HF patients for ID.

Search and selection strategy
We conducted a meta-analysis of the published randomized controlled trials (RCTs) in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) statement [15]. The protocol for this analysis has been submitted to PROSPERO for registration (CRD 400,041). The search has been performed in MEDLINE and Embase via OVID @ using the following keywords and Medical Subject Headings (MeSH) terms: [Heart failure AND iron deficiency OR iron repletion OR intravenous iron OR ferric carboxymaltose OR iron derisomaltose OR iron supplementation OR iron therapy AND randomized controlled trial]. The search was restricted to full-text articles published in English between 2000 and 2022. Furthermore, we screened the reference list of the current guidelines for HF of the European Society of Cardiology [7]. Randomized, placebo-controlled cardiovascular clinical trials that investigated effects of IV iron repletion using ferric carboxymaltose or derisomaltose in patients with HF and ID were considered eligible for inclusion. There was no limit regarding number of patients or duration of follow-up in potentially acceptable studies. Finally, data from non-randomized trials, registries or trials that used oral iron substitution were not considered eligible for inclusion. A reference manager software (Zotero) was used for duplicates removal and data management. Two reviewers (DV and AA) reviewed the full texts and used the same template to extract data relevant to the analysis.

Data extraction and analysis
Two authors (DV, AA) extracted all data of interest according to a previously established pattern and evaluated the risk of bias at the study level according to the Cochrane risk-of-bias tool (RoB 2.0) [16]. A publication bias would have been assessed using Funnel plot, but only in cases where outcomes of interest were reported in at least ten studies. We extracted the following data: i) baseline characteristics (study design, primary outcome, duration of follow-up, sample size, comparator, regimen of iron substitution, included population); ii) number of events and crude point estimates like rate ratios (RRs) or hazard ratios (HRs) with its associated confidence intervals for outcome of interest. We explored following outcomes of interest: (i) recurrent outcomes (total HF hospitalizations and CVD, total HF hospitalizations), time-to-event outcomes (time to first HF hospitalization or CVD, time to cardiovascular death), (iii) dichotomous outcomes concerning efficacy (all-cause mortality and CVD) but also concerning safety (infections, general or injection-site related disorder, vascular disorder). We performed a study-level, pairwise meta-analysis based on the intention-to-treat analysis of the summary data exploring the risk of above identified outcomes of interest between the groups (with iron supplementation vs. usual care). In our main analysis, we pooled the data from the AFFIRM-AHF and IRONMAN trials. We compared categorical data from populations included in these trials in terms of age, symptoms, medical treatment and others by applying Pearson's Chi-squared Test. For the composite endpoint consisting of recurrent HF hospitalizations and CVD we computed corresponding number needed to treat for benefit (NNT) or needed to treat for harm (NNH) as appropriate. We performed the following sensitivity analyses: i) running the analysis by pooling the data from COVID-19 prespecified analyses from AFFIRM-AHF and IRONMAN, (ii) running the analysis by pooling the data from all eligible trials.
For recurrent outcomes, we pooled RRs and for timeto-event outcomes we pooled HRs, which were reported in the original publications. For this purpose inverse variance statistical method was applied. In case RR was not reported, we used number of events to calculate it. For dichotomous outcomes we determined odds ratios (ORs) by applying the Mantel-Haenszel method. The data from each trial were pooled using random-effects (DerSimonian-Laird) model. Heterogeneity between the trials was assessed using Cochran's Q test and I 2 statistic. Relevant statistical heterogeneity was considered in case Cochran's Q-test p < 0.05 and I 2 greater than 50%. Study-specific and summary effect estimate with corresponding 95% confidence intervals (CIs) and p-value were visualized using Forest plots.
We explored the robustness of the meta-analysis findings for the composite endpoint (HF hospitalization and CVD) and CVD by determination of the fragility index (FI) for significant outcomes and reversed fragility index (RFI) for non-significant outcomes. We determined FI and RFI by applying the calculator available online http:// clini calep idemio. fr/ fragi lity_ ma/. [17] FI indicates the number of specific events-status modification (events added or subtracted in the treatment or placebo group) needed to turn the statistically significant to statistically non-significant results. RFI indicate the number of specific events-status modification needed to turn the statistically non-significant to significant results. Furthermore, we calculated the fragility quotient (FQ) and reversed FQ (RFQ) by dividing FI or RFI respectively with the sample size to account for different sample sizes. FQ represent the proportion of events, which need to be moved to change the significance of results. For example, meta-analysis A had FI of 2 and sample size of 500 participants while meta-analysis B had FI of 2 and sample size of 1000. Albeit FI is the same in both analyses, FQ can reveal us which analysis is relatively more fragile. Analysis A had FQ of 0.004 indicating that 4 events per 1000 patients will be needed to change the results significance; while analysis B has FQ of 0.002 indicating that 2 events per 1000 patients will be needed to change the results significance. Accordingly, FQ suggest us that results of trial B are more fragile.
Lower FI or RFI suggests less statistical robustness although there are no standardized cut off values that defines robustness or fragility. For the purpose of this analysis, FI and RFI < 20 was considered as fragile, FI and RFI 20-40 was considered as moderately robust and FI or RFI ≥ 40 as robust findings.
We calculated the power of our meta-analysis for observed risk ratio reduction (RRR) for composite of recurrent HF hospitalizations and CVD and odds/hazard ratio reduction (ORR/HRR) for CVD in the main and both sensitivity analyses [18].
All statistical analyses were performed using RevMan Version 5.4 and GraphPad Prism Version 6. All P values were two-sided, with P less than 0.05 considered as significant. For determination of the power of the analysis we used statistic program R Version 4.2.2.

Results
The selection strategy of the statistical analysis is shown in Fig. 1. Finally, 7 studies met the predefined inclusion criteria and were considered eligible for meta-analysis. In our main analysis we explored data from 2,245 participants from the AFFIRM-AHF [12] and IRONMAN [13] trials. For the purpose of sensitivity analysis we explored data from 2171 participants from the COVID-19 sensitivity analyses reported in the AFFIRM-AHF and IRONMAN trials and from 3258 participants by including additional data from the FAIR-HF [9], CONFIRM-HF [10], EFFECT-HF [19] trials and from two small studies (FER-CARS-01 and EFFICACY-HF), whose data were extracted from an article published by Anker et al. [20] as study level data are not published yet. Recurrent and time-to-event outcomes were reported as rates per 100 patient-years in AFFIRM-AHF, IRONMAN and analysis by Anker et al. [20]. Trials FAIR-HF, CONFIRM-HF, EFFECT-HF, AFFIRM-AHF and IRONMAN were regarded as high quality trials (Figure S1, Supplement), for two not published trials (FER-CARS-01 and EFFICACY-HF) evaluations could not be performed for lack of information. Baseline characteristics of the included trials are visualized in Table 1. Evaluation of publication bias was not performed due to futility (less than 10 studies included).

Results from AFFIRM-AHF and IRONMAN trials
Treatment with i.v. iron supplementation compared to placebo reduced the composite endpoint of recurrent HF hospitalizations and CVD (RR = 0.81; 95% CI 0.69-0.95, p = 0.01, I 2 = 0%) ( Fig. 2A) and the risk of recurrent HF hospitalizations (RR = 0.77; 95% CI 0.65-0.91, p = 0.003, I 2 = 0%) (Fig. 2B). NNT for the composite endpoint was 7 and for recurrent HF hospitalizations it was 8 over a weighted mean follow-up of 96 weeks. The meta-analysis had 73% power to detect a 19% RRR in the composite endpoint, while the summary result for the primary endpoint with FI of 94 and FQ of 0.041 was robust without heterogeneity (Table S2, Supplement).

COVID-19 prespecified analysis from IRONMAN and AFFIRM-AHF trial
Treatment with FCM/FDI compared to placebo reduced the composite endpoint of recurrent HF hospitalizations and CVD (RR = 0.75; 95% CI 0.63-0.91, p = 0.003) (Fig. 4-A) and the risk of recurrent HF hospitalizations (RR = 0.72; 95% CI 0.60-0.88, p = 0.001) (Fig. 4-B). There was no heterogeneity (I 2 = 0%) for any outcome between the trials. NNT for the composite outcome was 11 over a weighted mean of follow-up of 96 weeks. The meta-analysis had 85% power to detect a 25% RRR in the composite endpoint, while summary result for primary endpoint with FI of 98 and FQ of 0.045 was robust (Table S2, Supplement).

Data from all eligible trials
Analysis of all eligible trials confirmed findings from the main analysis. Treatment with FCM/FDI compared to placebo reduced the composite endpoint of recurrent HF hospitalizations and CVD (RR = 0.77; 95% CI 0.66-0.90, p = 0.0008) (Fig. S3-A, Supplement) and the risk of recurrent HF hospitalizations (RR = 0.72; 95% CI 0.57-0.92, p = 0.007) (Fig. S3-B Figure S4-B, Supplement). There was no heterogeneity (I 2 = 0%) for any outcome between the trials. The meta-analysis had power of 10% to detect 14% ORR in CVD, while summary result of meta-analysis for CVD with RFI of 7 and RFQ of 0.002 was fragile (Table S2, Supplement).

Discussion
The key findings of our analysis are the following: (i) IV iron repletion with FCM or FDI compared to placebo reduced the composite endpoint of recurrent HF hospitalization and CVD in HF patients with ID; (ii) effects of FCM or FDI concerning all-cause death or CVD are based on current level of data indeterminate; (iii) use of FCM and/or FDI was safe. In addition, findings concerning the composite endpoint were very robust while findings concerning CVD were fragile. There was no relevant heterogeneity for any outcome between the trials with FCM and those with FDI. One reason to perform meta-analyses is to increase the power of trials with borderline patient and limited event numbers [22]. It has been recognized that the COVID-19 pandemic had unpredictable and adverse impacts on conduction, results, clinical outcome as recruitment of patients and adequate follow-up [14]. To overcome these obstacles in AFFIRM-AHF and IRONMAN prespecified COVID-19 sensitivity analyses were planned which censored patients at the date when the first COVID-19 patient was reported (AFFIRM-AHF) [12] or at the start of the national lockdown in the United Kingdom (IRONMAN) [13]. COVID-19 pre-specified analyses showed significant reductions in the primary endpoint with FCM/FDI compared to placebo.
Meta-analysis of the data from AFFIRM-AHF [12] and IRONMAN [13] showed that treatment with FCM/FDI reduced the composite endpoint of recurrent HF hospitalizations and CVD, which was mainly driven by reduction of  HF is one of the leading causes for hospitalization being responsible for about 5% emergency hospital admissions worldwide [23]. Median number of HF discharges per million people in Europe amounts 2671 (IQR 1771-4317) with median length of hospital stay of 8.50 days (IQR 7.38-10) [24]. According to the same source median prevalence of HF approximate 17.2 per 1000 people [24]. Taking these figures into account it becomes obvious to what extent HF poses a burden on health-care givers and national economies worldwide. Therefore, reduction in HF rehospitalizations is of paramount importance as one of the mechanisms in improvement of outcome in HF patients.
Our analysis suggests a numerical but no significant effect of i.v. iron repletion on CVD or all-cause death in HF patients. Incident rates of CVD were numerically lower by 2% in patients receiving FCM/FDI compared with usual care group. Of note, both trials were not powered concerning these outcomes. In line, calculated power of meta-analyses for CVD was too low across all explored populations, thus not allowing us to make any reliable conclusion regarding effects of FCM/FDI on this outcome. However, summary findings of meta-analyses for endpoint CVD were fragile (RFI of 14, i.e., 14 events of CVD added to placebo or subtracted from FCM/FDI group of patients needed to render the result positive for CVD, and RFQ of 6, i.e., indicating 6 events per 1000 patients added to the placebo group needed to render the results positive for CVD). Fragility of the meta-analysis summary results for CVD was even more pronounced in analysis where all eligible trials were explored (RFI of 7 and RFQ of 0.002). Furthermore, duration of the follow-up could have been too short in AFFIRM-AHF. Notably, incidence rate reduction was more expressed with iron repletion compared to placebo during longer follow-up as presented in IRONMAN trial. However, this speculate should be regarded as hypothesis generating. Summarized, according to the totality of current available evidence effects of FCM/FDI on CVD are debatable and indeterminate.
Iron represents the essential element for transport and storage of oxygen, especially in cells with high energy demand like skeletal and heart muscle cells [5]. Iron deficiency independently of anemia negatively affects oxidative metabolism, cellular energetic [25] and immune mechanisms which result in decreased oxygen storage in myoglobin and reduced myocardial oxygen capacity leading to mitochondrial and left ventricle dysfunction [5]. Reduced myocardial iron correlates with reduced reactive oxygen species (ROS) protecting enzymes and mitochondrial oxygen consumption [5,25]. These pathological mechanisms contribute to myocardial dysfunction and adverse remodeling which further deteriorate functional capacity of HF patients and promote worsening of HF. Positive findings of pre-specified COVID-19 analyses of AFFIRM-AHF and IRONMAN on outcome of HF patients with ID, strengthen by results of this analysis unequivocally point out that iron substitution with FCM/FDI should be recognized as a life-saving therapy for appropriate patients (those with ID).
Populations enrolled in the AFFIRM-AHF and IRON-MAN differed slightly concerning gender, load of comorbidities, HF symptoms (NYHA class) and background medication. AFFIRM-AHF trial enrolled patients being hospitalized for acute HF while IRONMAN included predominantly (67%) ambulatory HF patients. This makes its comparisons difficult. Nevertheless, composite outcome was similar without statistical heterogeneity (p for Cochran Q = 0.82; I 2 = 0%). Therefore, benefit from i.v. iron repletion remains consistent despite some different presentations of HF patients.
Both substances explored in this analysis (FCM and FDI) are comparable regarding their ability to restore iron stores based on increases in ferritin and transferrin saturation [12,13]. However, there are some discussions about different safety profiles between FCM and FDI related to hypophosphatemia. It has been shown that FCM is associated with higher incidence of hypophosphatemia compared to FDI in patients (mainly women > 90%) with ID anaemia that lasted up to 35 days [26] as in patients with inflammatory bowel disease and ID anaemia [27]. Of note, patients with HF were underrepresented in these randomized trials. In a small (n = 23) single-center study FCM was investigated in HFrEF patients with (CKD +) and without chronic kidney disease (CKD-) [28]. Interestingly, significant serum phosphate decreases were present only in CKD(-) patients, while in both groups transient hypophosphatemia (< 0.8 mmol/l) was observed (9/11 in CKD(-) and 5/12 in CKD( +) patients) in parallel with a decrease in the levels of 1,25-OH vitamin D. Hypophosphatemia following i.v. iron replacement has been induced by increased secretion of fibroblast growth factor (FGF) 23 that leads to increased urinary phosphate excretion and decreased concentration of active vitamin D [28,29]. Dose, repetitive iron infusions, severity of ID, increasing age and vitamin D deficiency among others have been identified as predisposing factors for development of this side-effect [30]. Nevertheless, it remains unclear whether and at what extent this might have negative long-term effects on HF patients. In the AFFIRM-AHF study from week 12 to week 52 level of change regarding serum phosphate was similar (literary one patient in each group experienced hypophosphatemia) between the FCM and placebo group [12] while in IRONMAN study data regarding serum phosphate were not collected [13]. A placebo controlled study is needed to further explore potential hazard.
Administration of FCM and FDI was well tolerated. There was no difference between treatment and control group in terms of infection, general or injection-site disorder or vascular disorder.

Limitations
There are several limitations that need to be acknowledged. This is a post-hoc meta-analysis using the data provided in the official publications and not the individual patient-level data, which does not allow identifying possible covariates that might have impact the final results. Dosing regimen concerning initial dose, maintenance doses, probably total doses as well timing of control of iron values and iron re-administration varied substantially between the trials depending mostly on the duration of the trials which all might have affected results.

Conclusion
In summary, the totality of evidence of ID treatment trials suggests that the use of FCM/FDI in patients with HF and ID is safe and associated with reduced rate of composite of recurrent hospitalizations for heart failure and CVD. The data are robust derived from analyses with sufficient power to detect observed treatment effects and do not show relevant heterogeneity between the trials with the two iron derivates.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.