Hepcidin, incident heart failure, and cardiac dysfunction in older adults: the ARIC study

Abstract

Aims

Hepcidin regulates plasma and tissue iron levels. We studied the association of hepcidin levels with the risk of incident heart failure (HF) and cardiac dysfunction in older adults.

Methods and results

We included adults from the ongoing, longitudinal Atherosclerosis Risk in Communities (ARIC) study, who were free from prevalent anaemia and HF at Visit 5 (2011–13) and had available hepcidin and covariate data. Associations of plasma hepcidin levels with overall adjudicated incident HF, HF with reduced ejection fraction (HFrEF), and HF with preserved ejection fraction (HFpEF) were assessed using multivariable Cox proportional hazards regression models. Cross-sectional associations of hepcidin with echocardiographic measures of cardiac structure and function were estimated using multivariable linear regression models. The mean age was 75 ± 5 years old, and 56% were women. In fully adjusted models, lower hepcidin levels were associated with a higher risk of overall incident HF [hazard ratio [HR] [95% confidence interval (CI) per 50% lower hepcidin]: 1.15 [1.05–1.26]; P = 0.003] and HFpEF [HR (95% CI): 1.25 (1.10–1.42); P = 0.001]. Plasma hepcidin level was not significantly associated with the risk of incident HFrEF [HR (95% CI): 1.08 (0.94–1.24); P = 0.30]. Lower hepcidin levels were associated with higher E wave (P = 0.046), higher E/e′ ratio (P = 0.002), higher left atrial volume index (P = 0.005), and higher pulmonary artery systolic pressure (P = 0.02).

Conclusion

In community-dwelling older adults without anaemia, lower plasma hepcidin levels associate with a higher risk of incident HF (particularly HFpEF) and diastolic dysfunction.

Graphical Abstract

Open in new tabDownload slide

Introduction

Iron deficiency contributes to the development and progression of heart failure (HF).^1–4 Disordered cardiomyocyte iron handling impairs mitochondrial function, which leads to systolic and diastolic dysfunction.^5–8 Iron deficiency in extra-cardiac tissues also may promote cardiac dysfunction.^9–11 The associations of iron regulatory proteins, such as hepcidin, with incident HF risk and cardiac structure and function remain incompletely understood.

Plasma and tissue iron levels are regulated by hepcidin, which degrades the iron exporter ferroportin.¹² Hepatic hepcidin production decreases in states of iron deficiency to promote absorption and mobilization of iron stores.¹² Increases in hepcidin protect against iron overload.¹² Cardiomyocyte-derived hepcidin regulates cardiac iron levels independently from systemic hepcidin.^13,14 Hepcidin may also play an important role in response to pharmacologic iron repletion.^15–18 Dietary iron intake, which is controlled by hepcidin levels, has been associated with the risk of myocardial infarction and stroke.¹⁹ Studying the relationships of hepcidin, incident HF, and cardiac structure and function may have direct clinical implications given the development of pharmacologic hepcidin therapies and the role of hepcidin in regulating the response to iron repletion.¹²

The aim of this study is to examine the associations between plasma hepcidin levels, and incident HF overall, incident HF with reduced ejection fraction (HFrEF) and incident HF with preserved ejection fraction (HFpEF), and cardiac structure and function in older (mean age 75 years) adults.

Methods

Study cohort

The Atherosclerosis Risk in Communities study (ARIC) is an ongoing, prospective, longitudinal cohort study that enrolled 15 792 adults in 1987–89 from Forsyth County, NC; Washington County, MD; suburban Minneapolis, MN, and Jackson, MS.²⁰ We included Visit 5 (2011–13) attendees with no history of HF or anaemia (haemoglobin <13 mg/dL for males and <12 mg/dL for females)²¹ who had available plasma hepcidin and covariate measurements.

The study protocol was reviewed and approved by the Institutional Review Board of each participating institution. All study participants provided written informed consent.

Exposure variable

The primary exposure was plasma hepcidin levels, which were measured using a modified aptamer assay (SomaLogic, Boulder, CO, USA) from frozen samples collected at ARIC Visit 5.^22,23 Specificity of the aptamer targeting hepcidin is supported by the identification of two cis-protein quantitative trait loci for this aptamer (see Supplementary material online, Table S1).^24,25

Heart failure outcomes

The primary outcome was incident HF, defined as the first HF hospitalization or HF death after Visit 5. Follow-up telephone calls, surveillance of local hospital discharge records, local health district death certificates, and the National Death Index were reviewed to identify potential HF events.²⁶ When relevant HF diagnosis codes were present, two physicians adjudicated all HF events following pre-defined criteria.²⁶ Incident HF events were classified as HFpEF [left ventricular ejection fraction (LVEF) ≥50%], HFrEF (LVEF < 50%), and unknown LVEF using LVEF measurements recorded during the HF hospitalization. Follow-up ended on the date of the incident HF event, loss to follow-up, 31 December 2019 (Forsyth County, suburban Minneapolis, and Hagerstown Field Centers), or 31 December 2017 (Jackson Field Center, due to delays in acquiring more recent information on HF hospitalizations).

Cardiac structure and function outcomes

Transthoracic echocardiography was performed at Visit 5, as previously described.²⁷ Briefly, trained and certified sonographers captured all echocardiographic images using the same ultrasound machines across all field centres. The Brigham and Women’s Hospital Echocardiography Laboratory blindly performed all quantitative analyses following the American Society of Echocardiography recommendations.²⁷

Covariates

Date of birth, biological sex, and race were self-reported by participants at enrolment. At Visit 5, smoking status defined as current, former, never smoker, or unknown was self-reported. Hypertension was defined as study visit blood pressure ≥140/90 mm Hg or self-reported use of an anti-hypertensive medication. Diabetes mellitus was defined as a study visit fasting glucose level of at least 126 mg/dL, a study visit non-fasting glucose level of at least 200 mg/dL, self-report of a physician diagnosis of diabetes mellitus, or self-reported glucose-lowering medication use. Prevalent atrial fibrillation was defined as hospitalization with a discharge International Classification of Diseases (ICD)-9 or ICD-10 diagnostic code of atrial fibrillation in the absence of concurrent cardiac surgery. Prevalent coronary heart disease included self-reported myocardial infarction before Visit 1 and, afterwards, definite or probable adjudicated cases of non-fatal myocardial infarction or electrocardiogram evidence of a prior myocardial infarction. Measurements of haemoglobin and other red blood cell indices, creatinine, cystatin C, C-reactive protein, aspartate and alanine aminotransferase, and LDL cholesterol have been described previously.^28–35

Estimated glomerular filtration rate (eGFR) was calculated using the creatinine- and cystatin C-based 2012 Chronic Kidney Disease Epidemiology Collaboration equation.³⁶ Iron deficiency or anaemia treatments (oral or IV iron supplements and erythropoietin-stimulating agents), aspirin use, and oral anticoagulant use were recorded by trained staff from the medications brought to the study visit by participants. Elevated liver enzyme levels [measured at Visit 4 only (1996–98)] were defined as at least one of aspartate aminotransferase or alanine aminotransferase above the sex-specific upper limit of normal for this assay [men: >41 units/L for alanine aminotransferase and >37 for aspartate aminotransferase; women: >31 units/L for alanine (ALT) and aspartate aminotransferase (AST)].³⁴ A sensitivity analysis was performed using Visit 5 alanine aminotransferase and aspartate aminotransferase levels measured using the SomaScan assay.^22,23 Current cancer was defined as a self-reported cancer diagnosis collected during follow-up interviews between June 2011 and the date of Visit 5. Self-reported alcohol drinking status was defined as current, former, and never drinker.

Statistical analysis

Categorical variables were summarized as frequencies and percentages. Continuous variables with an approximately normal distribution were summarized as means and standard deviations, while non-normally distributed variables were summarized as medians and (25–75th percentiles). Study participant characteristics were compared across hepcidin quartiles using the Cuzick test for trend and linear or logistic regression. Normalization and quality control procedures for aptamer-based measurements in ARIC have been previously reported.^25,37,38 SomaScan reports hepcidin measurements as immunofluorescent units on a relative scale instead of absolute concentrations. Therefore, hepcidin levels were log₂-transformed and then multiplied by −1 before the analysis, so that higher estimated coefficients reflect an exposure of 50% lower hepcidin level, instead of 1 relative immunofluorescent unit higher.

Cox proportional hazards regression models were used to estimate cause-specific hazard ratios for the associations between log₂-transformed plasma hepcidin levels with incident HF overall, HFpEF (censoring HFrEF), and HFrEF (censoring HFpEF). Model 1 included age, sex, and a combined field centre and race variable. Model 2 included Model 1 covariates plus diabetes mellitus, hypertension, coronary heart disease, atrial fibrillation, smoking status, body mass index, log₁₀-transformed C-reactive protein, and eGFR. Effect modification by age categories (above and below the median), sex, and eGFR (less than vs. greater than or equal to 60 mL/min per 1.73 m²) was assessed using multiplicative interaction terms in the Cox regression models.

Linear regression was used to model the associations of log₂-transformed plasma hepcidin levels and cardiac structure and function measures. Linear regression models were adjusted for Model 1 and Model 2 covariates in addition to Visit 5 systolic blood pressure and heart rate.

Several sensitivity analyses were performed. Cox proportional hazards models were adjusted further for Model 1 and Model 2 covariates plus haemoglobin levels and confounders that were self-reported (current cancer diagnosis, alcohol drinking history, oral anticoagulant use, aspirin use, and iron deficiency treatment), reported by earlier publications as potential confounders (stroke history and LDL cholesterol), or were measured at Visit 4 only (ALT and AST levels). Model 4 was adjusted for Model 2 covariates plus Visit 5 systolic blood pressure. Model 5 was adjusted for Model 3 covariates including continuous liver enzymes (ALT and AST) measured using SomaLogic proteomic assays instead of elevated levels at Visit 4. The analysis was repeated after reassigning all incident HF cases without LVEF measurements to either HFrEF or HFpEF.

All statistical analyses were conducted using Stata version 17.0 (StataCorp LLC, College Station, TX, USA). A P-value <0.05 was considered statistically significant.

Results

Participants characteristics

The final cohort for this analysis included 3472 participants with a mean age of 75 ± 5 years (56% women and 14% Black; Figure 1). Participants with lower hepcidin levels were more likely to be women with a higher prevalence of coronary heart disease and atrial fibrillation as well as lower C-reactive protein levels (Table 1). eGFR was lower for participants with hepcidin levels in Quartiles 3 and 4 (68 ± 15 and 66 ± 17 mL/min per 1.73 m²) compared with Quartiles 1 and 2 (70 ± 16 and 70 ± 17 mL/min per 1.73 m²).

Associations of plasma hepcidin levels with red blood cell indices

Lower plasma hepcidin levels were associated with lower haemoglobin levels, smaller mean red cell volume, lower mean corpuscular haemoglobin, and lower mean corpuscular haemoglobin concentration (Table 2). Lower plasma hepcidin levels were associated with higher red cell distribution width. Iron deficiency and anaemia treatments were more common among participants with lower hepcidin levels, but this difference was not statistically significant (4% for Quartile 1, 3% for Quartile 2, 2% for Quartile 3, and 3% for Quartile 4; P = 0.51; Table 2).

Associations of plasma hepcidin levels with incident heart failure and heart failure sub-types

Over a median follow-up period of 7.2 years, 293 incident HF events occurred {incidence rate [95% confidence interval (CI) per 1000 person-years]: 12.6 [11.3–14.2]}, including 131 HFrEF events, 131 HFpEF events, and 31 events with unknown LVEF. After adjusting for demographics and HF risk factors (Model 2), a 50% lower plasma hepcidin level was associated with a 1.15-time higher risk of incident HF overall [hazard ratio (HR) (95% CI): 1.15 (1.05–1.26); P = 0.003] and a 1.25-time higher risk of incident HFpEF [HR (95% CI): 1.25 (1.10–1.42); P = 0.001; Figure 2]. Plasma hepcidin level was not significantly associated with the risk of incident HFrEF [HR (95% CI): 1.08 (0.94–1.24); P = 0.30; Figure 2].

Lower plasma hepcidin levels are associated with a higher risk of incident HF overall among participants 75 years of age or older [HR (95% CI): 1.29 (1.13–1.48)], but not those younger than 75 years [HR (95% CI): 1.06 (0.94–1.20); P-interaction = 0.036; see Supplementary material online, Table S2]. There was no other evidence of effect modification of the association of plasma hepcidin levels with incident HF and its sub-types by age, sex, or eGFR category (see Supplementary material online, Table S2). The overall results did not change after assigning cases with incident HF and unknown LVEF to either HFrEF or HFpEF group (see Supplementary material online, Table S3). Results were consistent in additional analyses that adjusted for self-reported current cancer, stroke history, abnormal liver enzyme levels, LDL cholesterol levels, alcohol use, antithrombotic use and iron deficiency treatment use, systolic blood pressure, or using aptamer-based liver enzyme levels (see Supplementary material online, Table S4).

Associations of plasma hepcidin levels with echocardiographic measures of cardiac structure and function

After adjustment for demographics and HF risk factors, lower hepcidin levels were associated with higher E wave velocity [beta (95% CI): 0.52 (0.01–1.03) cm/s per 50% lower hepcidin; P = 0.046], higher E/e′ ratio [beta (95% CI): 0.18 (0.07–0.30); P = 0.002], higher left atrial volume index [beta (95% CI): 0.33 (0.10–0.56); P = 0.005] mL/m², and higher pulmonary artery systolic pressure [beta (95% CI): 0.25 (0.04–0.45); P = 0.02] mmHg (Table 3). For measures of left ventricular (LV) structure, lower hepcidin levels were associated with higher mean wall thickness [beta (95% CI): 0.004 (0.000–0.007); P = 0.045]. No associations were observed with any measure of LV systolic function (Table 3).

Discussion

In this cohort study of 3472 community-based, older (mean age 75 years) adults without prevalent HF or anaemia, lower hepcidin levels were associated with a higher risk of incident HF overall and incident HFpEF, as well as with worse LV diastolic function. The observed associations were statistically significant after adjusting for a wide range of potential confounders, including haemoglobin levels, and were consistent across age, sex, and eGFR categories. These findings provide insight into the role of hepcidin in the associations between disordered iron metabolism and incident HF in late-life.

Both lower and higher hepcidin levels may occur in people with iron deficiency, albeit through different mechanisms. Inflammation increases hepcidin production, whereas hepcidin levels decrease during iron deficiency, pregnancy, and erythropoiesis.¹² Higher hepcidin levels may contribute to iron deficiency by sequestering iron within macrophages and blocking enteral iron absorption. In contrast, lower hepcidin levels may be a response to low iron stores to facilitate increased iron absorption and utilization. Alternatively, hepcidin may have iron-independent effects on cardiac function. Knockout of hepcidin from bone marrow–derived macrophages promoted the secretion of the anti-inflammatory cytokines IL-4 and IL-13 and improved cardiac healing after acute myocardial infarction.³⁹ In our study, lower hepcidin levels were associated with red blood cell biomarkers of iron deficiency, such as lower haemoglobin levels and higher red cell distribution width, and hepcidin was associated with HF and cardiac dysfunction independent of high-sensitivity C-reactive protein levels. Thus, the observed associations between lower hepcidin levels and higher HF risk may be due to underlying low iron levels, rather than direct effects of lower hepcidin on the heart or mediation of the association between inflammation and HF. Consideration of these mechanisms may have therapeutic implications. Although lower hepcidin was associated with higher incident HF risk, raising hepcidin levels would worsen iron deficiency, which would be expected to increase incident HF risk. In contrast, lowering hepcidin with agents in development would facilitate iron absorption and distribution, which could have the potential to alleviate the adverse cardiac effects of iron deficiency.¹²

The observed associations between hepcidin, incident HF, and cardiac dysfunction were significant in a cohort without baseline anaemia and after adjustment for baseline haemoglobin. While we were unable to adjust for incident anaemia due to a lack of longitudinal haemoglobin levels, our results suggest a minimal role for anaemia and impaired oxygen transportation in the associations between hepcidin and incident HF. Indeed, previous studies in patients and pre-clinical models have identified pathways between iron dysregulation and cardiac dysfunction independent of anaemia.^5,6

In patients with HFpEF, iron deficiency is associated with impaired skeletal muscle metabolism and skeletal muscle dysfunction during exercise.^10,11 Cardiac-specific hepcidin controls cardiac iron homeostasis independently of liver-derived hepcidin.^9,14,40 Since we measured hepcidin levels in peripheral venous samples, our measurements likely reflect an association between hepatic hepcidin, which regulates peripheral iron skeletal muscle levels, and incident HF and HFpEF, since cardiac-derived hepcidin likely constitutes a small proportion of systemic circulating levels. Our finding of an association between lower plasma hepcidin levels and higher risk of incident HFpEF and diastolic dysfunction suggests that iron deficiency in peripheral skeletal muscle may help explain the mechanisms underlying impaired cardiorespiratory fitness in adults with HFpEF.⁴¹

Another potential explanation for the differential association of lower hepcidin levels with HFpEF and not HFrEF might be due to the underlying inflammatory process involved in the pathophysiology of HFrEF. A rise of hepcidin levels as an acute phase reactant even in the presence of iron deficiency at baseline might mask the presumed association of iron deficiency with HFrEF. In our analysis, we controlled for the presence of inflammation by adjusting to the C-reactive protein measurements given that other inflammatory biomarkers were not measured at baseline. Future studies with absolute iron and inflammatory markers are warranted to confirm this hypothesis.

A prior study in the PREVEND cohort of middle-aged adults did not find an association between hepcidin levels and higher risk of incident HF overall.² Our study builds upon this prior work by focusing on older adults, who have a higher prevalence of both iron deficiency and HF, including a larger number of HF events with LVEF data, and performing analyses of cardiac structure and function. The aforementioned PREVEND analysis did, however, find an association between higher hepcidin levels and higher risk of HFpEF in women, but not men.² The differences in the PREVEND analysis and our results may be due to differences in the relationship between hepcidin levels and HF risk in mid-life and late-life or differences in statistical power. Moreover, differences in the range of hepcidin levels represented in each sample likely play a role as well.

The strengths of this study include the adjudication of HF events, the availability of a wide range of echocardiographic measures, and adjustment for a range of potential confounders. Our findings are relevant directly to older adults, a population at high risk of incident HF, and iron deficiency.

Limitations

There are limitations to this study. Hepcidin assays have not been standardized, and our analysis reports hepcidin levels on a relative scale. Hepcidin measurements were available only at a single timepoint from stored, frozen samples. The study participants constitute a relatively healthy group, as those with the highest risk of declining heart function may have already experienced HF prior to Visit 5. Other potential confounders, such as vitamin D levels, which also suppress hepcidin production, were not analysed. Heart failure events that occurred outside the catchment area and were not reported on routine telephone follow-up or diagnosed in the outpatient setting were not captured. Residual confounding cannot be excluded.

Conclusions

In older, community-dwelling adults without anaemia or prevalent HF, lower plasma hepcidin levels, reflecting iron deficiency, associate with a higher risk of incident HF (particularly HFpEF) and diastolic dysfunction.

Supplementary material

Supplementary material is available at European Journal of Preventive Cardiology.

Acknowledgements

We thank the staff and participants of the ARIC study for their important contributions. SomaLogic Inc. conducted the SomaScan assays in exchange for use of ARIC data. We are grateful for the insightful discussions on this manuscript with Dr Tomas Ganz.

Author contribution

I.A.F.A. performed the analysis, wrote the first draft, and revised subsequent drafts. B.L.C., V.L., and P.D. contributed to the design, analysis, interpretation, and manuscript revisions. K.M., P.L.L., B.Y., B.W.L., and C.E.N. contributed to the design, interpretation, and manuscript revisions. Y.M.K.F. and A.M.S. contributed to the conception, design, analysis, interpretation, and manuscript revisions. L.F.B. contributed to the conception, design, analysis, interpretation, and manuscript revisions and provided supervision.

Funding

This work was supported in part by National Institutes of Health grant R01 HL134320 and National Institutes of Health grant K23 HL150311. The ARIC study has been funded in whole or in part with Federal funds from the National Heart, Lung, and Blood Institute, National Institutes of Health, and Department of Health and Human Services, under contract nos. (75N92022D00001, 75N92022D00002, 75N92022D00003, 75N92022D00004, and 75N92022D00005).

Data availability

Investigators can request access to ARIC data through the ARIC website (https://aric.cscc.unc.edu/aric9/) or the NHLBI BioLINCC website (https://biolincc.nhlbi.nih.gov/home/).

References

Author notes