Risk of fractures according to iron parameters and hemochromatosis HFE genotype in 142,146 general population individuals

Abstract

The risk of fractures may be increased in individuals with iron deficiency, iron overload, and/or HFE hemochromatosis. To test this hypothesis, we followed 142,146 individuals from the Danish general population for a median of 11 years (range, 0-41) after study enrollment for hospital and emergency room admissions with fractures. All individuals had blood samples drawn at study enrollment. We measured iron, transferrin saturation, and ferritin in 136,611, 136,555, and 37,990 individuals, respectively, while 132,499 individuals were genotyped for the HFE C282Y and H63D variants. We found a U-shaped relationship between fracture risk and concentration of plasma iron and transferrin saturation when studying all individuals irrespective of genotype. When studied according to plasma ferritin concentration, fracture risk was increased in individuals with low ferritin concentrations, while risk was not increased in individuals with high concentrations. When compared to non-carriers, HFE C282Y homozygotes had an increased risk of any fracture (hazard ratio [HR]=1.38; 95% confidence interval [95% CI]: 1.09-1.75; P=0.008), and risk was increased even in C282Y homozygotes with normal ferritin concentrations (HR=2.89; 95% CI: 1.50-5.56), which is important as these individuals would not usually be recommended for HFE genotyping according to clinical guidelines. When compared to risk in non-carriers, risk of fracture of the hip and femur was increased in C282Y homozygotes (HR=1.78; 95% CI: 1.17-2.70; P=0.007) but surprisingly also in H63D homozygotes (HR=1.21; 95% CI: 1.00-1.47; P=0.04), C282Y heterozygotes (HR=1.10; 95% CI: 1.00-1.21; P=0.04), and C282Y/H63D compound heterozygotes (HR=1.23; 95% CI: 1.00-1.51; P=0.05). The markedly increased fracture risk in C282Y homozygotes with normal ferritin may challenge the presumption that systemic iron accumulation is the primary mechanism causing these individuals increased fracture risk. Further studies are needed to examine whether phlebotomy reduces fracture risk.

Introduction

Iron is a mineral essential for life due to its role in several fundamental processes, including transport of oxygen, DNA synthesis, and ATP generation.¹ Systemic iron regulation is orchestrated by hepcidin, a liver-produced peptide that binds ferroportin, inhibiting iron export from macrophages and hepatocytes and reducing intestinal absorption.² Its expression is primarily stimulated by elevated plasma iron and high transferrin saturation.² In clinical settings, plasma iron and transferrin saturation are often measured to evaluate the amount of iron in the bloodstream, while ferritin is a marker of overall iron storage.³ Notably, both plasma iron and transferrin saturation are very labile and may fluctuate during common clinical conditions such as acute infections, acute or chronic inflammation, liver disease, kidney disease, blood loss, pregnancy, and when taking certain medications.^1,2,4 Hereditary hemochromatosis is one of the most common genetic diseases in Northern Europe and is primarily caused by variants in the HFE gene.³ It is generally considered an iron-overload disorder, characterized by low hepcidin expression, even when iron stores are high, causing high iron uptake with the potential for iron accumulation and progressive organ damage.⁴ In Northern Europe, the allele frequency of the C282Y variant is around 5-10%, and among individuals diagnosed with hereditary hemochromatosis, more than 80% are homozygous for the C282Y variant.^3,5,6 The penetrance of clinically apparent hemochromatosis in C282Y homozygotes varies according to previous studies, depending on the definitions of evident disease, with estimates ranging from more than 80% to less than 1%.^7-10 The H63D variant has a broader distribution globally, with allele frequencies ranging from 10-20%,⁶ but H63D homozygosity and isolated H63D heterozygosity have generally not been found to cause hemochromatosis-related complications,¹¹ while the clinical effects of C282Y/H63D compound heterozygosity remain controversial.

Iron overload has previously been associated with increased risk of osteoporotic bone loss and fractures in some studies;^12-14 however, results have been conflicting.¹⁵ It is therefore unclear whether individuals in the general population with low or high concentrations of plasma iron, transferrin saturation, or ferritin are at increased risk of fractures. Previous population-based studies found an increased risk of fractures^16,17 in individuals with C282Y homozygosity. Although most clinical guidelines are based on the presumption that complications from hereditary hemochromatosis are primarily a result of systemic iron overload, the exact mechanism causing increased risk of fractures in C282Y homozygotes is unknown. Most current guidelines recommend that testing for HFE variants should primarily be performed in individuals with high ferritin and high transferrin saturation (except for testing family members of patients already diagnosed with HFE hemochromatosis).^3,18–21 However, recent studies from our group have found increased risk of infections in C282Y homozygotes despite normal iron, transferrin saturation, or ferritin, and increased risk of diabetes in C282Y homozygotes with normal transferrin saturation and/or ferritin, who are individuals not usually recommended for HFE genotyping according to clinical guidelines.^22,23

In this study, we therefore tested the hypotheses that risk of fractures is increased in individuals with low or high plasma iron, transferrin saturation, or ferritin. Furthermore, we tested whether risk of fractures is increased in hemochromatosis HFE C282Y homozygotes overall and specifically in C282Y homozygotes with normal concentrations of iron, transferrin saturation, or ferritin. To do so, we studied 142,146 individuals from the Danish general population who had blood samples drawn at study enrollment, after which they were followed prospectively for a median of 11 years.

Methods

Participants

We studied 142,146 individuals aged 20-100 years from three general population cohort studies: (i) 13,663 individuals from the Copenhagen City Heart Study,^22,24 examined between 1981-1983 or 1991-1994; (ii) 108,051 individuals from the Copenhagen General Population Study,²⁵ examined between 2003-2015; and (iii) 20,432 individuals from the Danish General Suburban Population Study,²⁶ examined between 2010-2013. At study enrollment, all individuals attended a health examination during which they filled out a questionnaire regarding lifestyle and health, had a physical examination, and blood samples drawn. Due to the completeness of the Danish registries, no individuals were lost to follow-up. From the Copenhagen City Heart Study, we included individuals at two different examinations: 9,002 individuals with ferritin measurements from the 1981-1983 examination and 9,727 individuals with iron, transferrin saturation, or HFE genotype data from the 1991-1994 examination. In the Copenhagen City Heart Study 5,063 individuals attended both examinations and had ferritin measured in the blood samples obtained in 1981-1983, while their plasma iron, transferrin saturation, and/or HFE genotype were measured using the blood samples obtained in the 1991-1994 examination. For further details, see the Online Supplementary Methods.

The studies received approval from Danish ethics committees and all individuals supplied written informed consent.

Covariates and comorbidities

All covariates for the multivariable adjusted models were chosen a priori based on previous reports of an association with iron, transferrin saturation, ferritin, and/or risk of fractures.^{27-2 9} For details, see the Online Supplementary Methods.

Plasma iron, transferrin saturation, and ferritin

Blood samples used for measuring plasma iron, transferrin saturation, and ferritin were collected at study enrollment. In total, 136,611 individuals had plasma iron measurements, 136,555 had transferrin saturation measurements, and 37,990 individuals had ferritin measurements. For details, see the Online Supplementary Methods.

Genotypes

The HFE gene was genotyped for the C282Y and H63D variants in 132,499 individuals using blood samples collected at study enrollment. Details on genotyping are described in our previous publications.^22,23 For other details, see the Online Supplementary Methods.

Fractures, vital status, and diagnoses of hemochromatosis

We used the national Danish Patient Register³⁰ to obtain information on all inpatient admissions from January 1^st, 1977, until December 31^st, 2021, and information on all outpatient and emergency room visits from January 1^st, 1994, until December 31^st, 2021. We defined fracture events as any inpatient or emergency room admissions due to fractures, categorized using World Health Organization’s International Classification of Diseases (ICD)-8 and ICD-10 codes (Online Supplementary Table S1).³¹ Information on emigration and vital status until December 31^st, 2021 was obtained from the Danish Civil Registration system,³² with 100% complete information. To perform stratified analyses on whether or not C282Y homozygotes were diagnosed with hereditary hemochromatosis and had therefore potentially been treated with therapeutic phlebotomy, we obtained information on hospital contacts due to hereditary hemochromatosis (ICD-8 code 27329 and ICD-10 code E831A) from January 1^st, 1977 until December 31^st, 2021. For details, see the Online Supplementary Methods.

Statistical analyses

We used Stata version 18.0, and all statistical tests were two-tailed. Risks of fractures were modeled using Cox proportional hazards regression, adjusting for age by using left-truncated age as the timescale. We chose this approach, which is often recommended for general population studies, as the date of study enrollment is chosen at random and, therefore, age is generally a stronger predictor for risk of future disease than time since study enrollment. Furthermore, when using this type of age-adjustment, the impact of age on fracture risk is not assumed to follow a specific pattern, meaning that the method efficiently adjusts for age at any given age during follow-up without underlying assumptions regarding linearity or other specific patterns of the relationship between age and risk of fractures. To estimate the cumulative incidence of any fracture, we used the competing risk model by Fine and Gray,³³ taking competing risk of death from any cause into consideration.

All age- and sex-adjusted models were also adjusted for study cohort. We decided a priori to study risk of fractures according to HFE genotype using the age- and sex-adjusted model, and to study risk of fractures according to concentrations of iron, transferrin saturation, and ferritin, using a multivariable adjusted model additionally adjusting for alcohol consumption, body mass index, smoking status, cumulative smoking in pack-years, menopausal status (women only), plasma C-reactive protein level, and Charlson comorbidity index. As a sensitivity analysis, we also studied risk of fractures according to HFE genotype using the multivariable adjusted model. Further details are available in the Online Supplementary Methods.

Results

The baseline characteristics for 142,146 individuals according to study cohort are presented in Table 1, while the baseline characteristics according to HFE genotype are presented in Online Supplementary Table S2.

Plasma iron, transferrin saturation, and ferritin and risk of fractures

When studying risk of fractures according to plasma iron, transferrin saturation, and ferritin levels in all individuals irrespective of HFE genotype, follow-up began at the time of study enrollment, giving a median follow-up of 11 years (range, 0-41), during which 11,549 individuals were hospitalized due to any fracture. For individuals diagnosed with fractures during follow-up, the median time from study enrollment to fracture diagnosis was 6 years for those with plasma iron ≤5^th percentile, 6 years for individuals with plasma iron in the 26-74^th percentile, and 7 years for individuals with plasma iron ≥95^th percentile. Similarly, the median time to fracture diagnosis was 6 years for individuals with transferrin saturation ≤5^th, in the 26-74^th, and ≥95^th percentiles, and 7, 7, and 8 years for individuals with ferritin ≤5^th, in the 26-74^th, and ≥95^th percentiles, respectively. When performing univariate analyses, each of the variables included in the multivariable adjusted analyses (age, sex, menopausal status, smoking status, alcohol consumption, body mass index, C-reactive protein level, and comorbid disease) were univariately associated with risk of fractures (Online Supplementary Figure S1).

Risk of any fracture was increased in individuals with low plasma iron ≤5^th percentile and high plasma iron ≥95^th percentile (hazard ratios are listed below the splines in Figure 1). Similarly, risk of any fracture was increased in individuals with low transferrin saturation ≤5^th percentile, high transferrin saturation ≥95^th percentile, and low ferritin ≤5^th percentile, whereas we found no association between high ferritin ≥95^th percentile and risk of any fracture (Figure 1). Results on plasma iron and transferrin saturation were similar to those presented in Figure 1 when only individuals who were non-carriers of C282Y and H63D were investigated, while risk of any fracture was not increased in non-carrier individuals with low ferritin or high ferritin (Online Supplementary Figure S2). As anemia and comorbid diseases may affect risk of fractures, we also performed analyses excluding all individuals with anemia (Online Supplementary Figure S3) or excluding individuals with any comorbid disease at study enrollment (Online Supplementary Figure S4), giving results similar to those presented in Figure 1. Fractures of the hip and femur were the most common type of fractures, and also represent clinically important fracture types, as 1-year mortality after hip fractures is often reported to be 15-25%.³⁴,³⁵ Therefore, risk of fracture of the hip and femur is presented in Figure 2, showing increased risk in individuals with low iron ≤5^th percentile, high iron ≥95^th percentile, low transferrin saturation ≤5^th percentile, high transferrin saturation ≥95^th percentile, or low ferritin ≤5^th percentile. When examining other specific fractures, risk of fracture of the axial skeleton was increased in individuals with high iron (HR=1.22; 95% CI: 1.02-1.46), low transferrin saturation (HR=1.23; 95% CI: 1.02-1.49), and high transferrin saturation (HR=1.27; 95% CI: 1.08-1.51) (Online Supplementary Figure S5). Furthermore, risk of fracture of the humerus was increased in individuals with low iron (HR=1.24; 95% CI: 1.02-1.50), low transferrin saturation (HR=1.25; 95% CI: 1.03-1.53), high transferrin saturation (HR=1.35; 95% CI: 1.10-1.66), and high ferritin (HR=1.53; 95% CI: 1.10-2.13), while levels of iron, transferrin saturation, or ferritin were not associated with risk of fracture of the wrist (Online Supplementary Figure S5). When stratifying by sex, relative risk estimates for any fracture according to levels of iron, transferrin saturation, and ferritin were similar for women and men (Online Supplementary Figure S6).

As serum iron, transferrin saturation, and ferritin may change over time, we did a sensitivity analysis examining risk of any fracture separately in the first 5 years after blood sampling, and from 5 years after blood sampling and onwards. Generally, we found similar risk estimates for fractures according to concentrations of iron, transferrin saturation, and ferritin during the first 5 years after blood sampling as in the time period from 5 years after blood sampling and onwards (Online Supplementary Figures S7-9).

Table 1.Baseline characteristics of 142,146 individuals of the Danish general population divided by study cohort.

Hemochromatosis HFE genotypes and risk of fractures

Overall, 132,499 individuals were genotyped for the HFE C282Y and H63D variants and 422 were homozygous for the C282Y variant. Since HFE genotype is constant throughout life, follow-up for the analyses on risk of fractures according to HFE genotype began at the time of creation of the Danish National Patient Register in 1977 or each individual’s 20^th birthday, whichever came last. During a median follow-up of 42 years (range, 0-45), 17,474 of the genotyped individuals were hospitalized because of any fracture. C282Y homozygotes had increased risks of any fracture (age- and sex-adjusted HR=1.38; 95% CI: 1.09-1.75; P=0.008 when compared to non-carriers for C282Y and H63D) and fracture of the hip and femur (HR=1.78; 95% CI: 1.17-2.70; P=0.007) (Figure 3). Notably, C282Y homozygotes had fractures at an earlier age than non-carrier individuals (mean age at any fracture: 60.6 years in C282Y homozygotes vs. 64.9 years in non-carriers; P=0.03) (Figure 3). When examining risk of other specific fractures in C282Y homozygotes, we did not find convincingly increased risks of fractures of the wrist (HR=1.35; 95% CI: 0.95-1.91), axial skeleton (HR=1.17; 95% CI: 0.69-1.97), or humerus (HR=1.29; 95% CI: 0.75-2.22) (Online Supplementary Figure S10). Risk of any fracture was not increased in H63D heterozygotes, H63D homozygotes, C282Y heterozygotes, or C282Y/H63D compound heterozygotes. Importantly, however, risk of fracture of the hip and femur was increased to a moderate extent in H63D homozygotes (HR=1.21; 95% CI: 1.00-1.47; P=0.04), C282Y heterozygotes (HR=1.10; 95% CI: 1.00-1.21; P=0.04), and C282Y/H63D compound heterozygotes (HR=1.23; 95% CI: 1.00-1.51; P=0.05) (Figure 3).

Figure 1.Risk of any fracture according to iron, transferrin saturation, and ferritin levels at study enrollment. Based on risk estimates from Cox proportional hazards regression, cubic splines were fitted. Hazard ratios are represented by a solid red line and 95% confidence intervals by dashed black lines. Multivariable adjustment included age, sex, study cohort, alcohol consumption, body mass index, smoking status, cumulative smoking in pack-years, menopausal status (women only), Charlson comorbidity index, and C-reactive protein level. For the categorical analyses and when modeling the splines, all individuals with available measurements of iron, transferrin saturation, or ferritin were included; however, for graphical presentation of the splines, the x-axis was limited to the 99.9^th percentile for iron and transferrin saturation, and the 99^th percentile for ferritin. Reference levels for the graphical splines were set at the median values of iron, transferrin saturation, or ferritin. 95% CI: 95% confidence interval; N: number; sat: saturation; ref: reference group.

Figure 2.Risk of fracture of the hip and femur according to plasma iron, transferrin saturation, and ferritin levels at study enrollment. Based on risk estimates from Cox proportional hazards regression, cubic splines were fitted. Hazard ratios are represented by a solid red line and 95% confidence intervals by dashed black lines. Multivariable adjustment included age, sex, study cohort, alcohol consumption, body mass index, smoking status, cumulative smoking in pack-years, menopausal status (women only), Charlson comorbidity index, and C-reactive protein level. For the categorical analyses and when modeling the splines, all individuals with available measurements of iron, transferrin saturation, or ferritin were included; however, for graphical presentation of the splines, the x-axis was limited to the 99.9^th percentile for iron and transferrin saturation, and the 99^th percentile for ferritin. Reference levels for the graphical splines were set at the median values of iron, transferrin saturation, or ferritin. 95% CI: 95% confidence interval; N: number; sat: saturation; ref: reference group.

Figure 3.Risk of any fracture (upper panel) and fracture of the hip and femur (lower panel) according to HFE genotype. C282Y/C282Y: homozygous for C282Y; C282Y/H63D: compound heterozygous for C282Y and H63D; C282Y/non-carrier: heterozygous for C282Y; H63D/ H63D: homozygous for H63D; H63D/non-carrier: heterozygous for H63D; Non-carrier/non-carrier: non-carrier for both C282Y and H63D; N: number; 95% CI: 95% confidence interval.

Most clinical guidelines recommend testing for the HFE genotype primarily in individuals with high concentrations of ferritin and high transferrin saturation.^3,18-21 To test whether these guidelines are sufficient for identifying C282Y homozygotes at high risk of fractures, we studied risk of any fracture in C282Y homozygotes according to concentrations of iron, transferrin saturation, and ferritin. As therapeutic phlebotomy affects concentrations of iron, transferrin saturation, and ferritin, we excluded individuals diagnosed with hemochromatosis before study enrollment from these stratified analyses as they may have been treated with phlebotomy before blood samples were drawn. Notably, C282Y homozygotes with normal concentrations of ferritin had a high risk of any fracture (HR=2.89; 95% CI: 1.50-5.56, for C282Y homozygotes with normal ferritin vs. non-carriers with normal ferritin). In contrast, risk of any fracture was not increased in C282Y homozygotes with normal plasma iron (HR=1.27; 95% CI: 0.97-1.68) or normal transferrin saturation (HR=0.86; 95% CI: 0.49-1.52) (Figure 4). Results were similar to those presented in Figure 4 when performing sensitivity analyses on risk of any fracture according to concentrations of iron, transferrin saturation, and ferritin using the multivariable adjusted model (Online Supplementary Figure S11). To investigate whether the increased risk of fractures in C282Y homozygotes with normal concentrations of ferritin is due to slow iron accumulation occurring after study enrollment, we performed additional analyses restricted to only study risk of fractures occurring before blood samples were obtained at study enrollment (Figure 5). When doing this, the hazard ratio for any fracture in C282Y homozygotes with normal ferritin was 3.27 (95% CI: 1.36-7.87), similar to the risk estimate in Figure 4 for investigating the total follow-up time. This implies that the increased fracture risk in C282Y homozygotes with normal concentrations of ferritin is not a result of iron accumulation occurring after study enrollment.

When stratifying according to whether or not C282Y homozygotes had been diagnosed with hemochromatosis at any time, risk of any fracture was similar in C282Y homozygotes never diagnosed with hemochromatosis (HR=1.37; 95% CI: 1.05-1.79 when compared to non-carrier individuals) and C282Y homozygotes diagnosed with hemochromatosis (HR=1.42; 95% CI: 0.84-2.40) (Figure 6). Likewise, relative risk estimates for any fracture were similar for C282Y homozygotes never diagnosed with liver disease and/or diabetes (HR=1.33; 95% CI: 1.02-1.73 when compared to non-carriers without liver disease and/or diabetes) and for C282Y homozygotes diagnosed with liver disease and/ or diabetes (HR=1.58; 95% CI: 0.91-2.73 when compared to non-carriers with liver disease and/or diabetes) (Online Supplementary Figure S12). When stratifying by sex, relative risk estimates for any fracture and fracture of the hip and femur were similar for C282Y homozygous men and women when compared to non-carrier men and women, respectively (P values for interaction between genotype and sex were 0.09 and 0.37 for risk of any fracture and fracture of the hip and femur, respectively) (Online Supplementary Figure S13). Likewise, when compared to non-carriers of the same age, relative risk estimates for any fracture and fracture of the hip and femur were similar for C282Y homozygotes <60 years and for C282Y homozygotes ≥60 years of age (Online Supplementary Figure S14).

Figure 4.Risk of any fracture for HFE C282Y homozygotes versus non-carriers for the C282Y and H63D variants stratified by normal versus high concentrations of plasma iron (upper panel), transferrin saturation (middle panel), and ferritin (lower panel) at study enrollment. Normal iron was defined as 9-34 μmol/L and high iron as >34 μmol/L. For transferrin saturation, the reference range varies according to age and sex, and therefore normal transferrin saturation was defined as 10-45% for women ≤50 years of age and 15-45% for women >50 years of age and men of any age. High transferrin saturation was defined as >45% for men and women of any age. Normal ferritin was defined as 12-200 μg/L for women and 12-300 μg/L for men, and high ferritin as >200 μg/L for women and >300 μg/L for men. Follow-up began at the time of creation of the Danish National Patient Register in 1977 or each individual’s 20^th birthday, whichever came last. As therapeutic phlebotomy can affect plasma iron, transferrin saturation, or ferritin levels, we excluded 11 C282Y homozygous individuals because they had been diagnosed with hemochromatosis before the day of study enrollment when blood samples were obtained. C282Y/C282Y: homozygous for C282Y; Non-carrier/ non-carrier: non-carrier for both C282Y and H63D; N: number; 95% CI: 95% confidence interval.

Figure 5.Risk of any fracture for HFE C282Y homozygotes versus non-carriers for the C282Y and H63D variants stratified by normal versus high concentrations of plasma iron (upper panel), transferrin saturation (middle panel), and ferritin (lower panel) when restricting the analyses to only study risk of fractures occurring before study enrollment (when blood samples were obtained). Normal iron was defined as 9-34 μmol/L and high iron as >34 μmol/L. For transferrin saturation, the reference range varies according to age and sex, and therefore normal transferrin saturation was defined as 10-45% for women ≤50 years of age and 15-45% for women >50 years of age and men of any age. High transferrin saturation was defined as >45% for men and women of any age. Normal ferritin was defined as 12-200 μg/L for women and 12-300 μg/L for men, and high ferritin as >200 μg/L for women and >300 μg/L for men. Follow-up began at the time of creation of the Danish National Patient Register in 1977 or each individual’s 20^th birthday, whichever came last. As therapeutic phlebotomy can affect plasma iron, transferrin saturation, or ferritin levels, we excluded 11 C282Y homozygous individuals because they had been diagnosed with hemochromatosis before the day of study enrollment when blood samples were obtained. C282Y/C282Y: homozygous for C282Y; Non-carrier/non-carrier: non-carrier for both C282Y and H63D; N: number; 95% CI: 95% confidence interval.

Cumulative incidence of any fracture

Cumulative incidence of any fracture according to age was analyzed using Fine-Gray competing risk models to take into account the competing risk of death from any cause. For the overall group of C282Y homozygotes, the cumulative incidence of any fracture was 8% at the age of 60 years and 24% at the age of 80 years, while the corresponding cumulative incidences were 6% and 18% in non-carriers aged 60 and 80 years, respectively. Notably, C282Y homozygotes with normal ferritin had a substantially higher cumulative incidence of any fracture, as their cumulative incidences were 19% at the age of 60 and 49% at the age of 80 years (Figure 7).

Discussion

In this prospective study of 142,146 individuals from the general population, risk of any fracture and risk of fracture of the hip and femur were increased in individuals with low or high plasma iron, low or high transferrin saturation, or low ferritin when studying all individuals irrespective of HFE genotype. Individuals with HFE C282Y homozygosity had increased risk of any fracture, and fracture risk was markedly increased even in C282Y homozygotes with normal ferritin, which is important as these individuals would not usually be recommended for HFE genotyping according to most guidelines. Risk of fracture of the hip and femur was increased in HFE C282Y homozygotes but surprisingly also in H63D homozygotes, C282Y heterozygotes, and C282Y/ H63D compound heterozygotes, which is a novel finding. Our discovery that risk of any fracture was increased in individuals with low or high plasma iron, low or high transferrin saturation, or low ferritin, irrespective of HFE genotype, indicates that fracture risk may be increased in both iron deficiency and iron overload. These results are to some extent supported by previous studies.^12,13 Kim et al. investigated 14,017 individuals from the Korean National Health and Nutrition Examination Survey and found an inverse association between serum ferritin and bone mineral density scores in women ≥45 years of age, and an increased prevalence of fractures in women in the highest vs. lowest ferritin quartile.¹² No previous studies have investigated fracture risk in individuals with low plasma iron, transferrin saturation, or ferritin, underlining the novelty of these results.

Mechanistically, both low and high iron may hypothetically affect bone remodeling, which could potentially explain our findings. Bones constantly undergo remodeling involving osteoclasts that resorb bone tissue and osteoblasts that actively synthesize new bone.³⁶ Iron overload facilitates bone resorption by promoting osteoclastogenesis and increasing osteoclast activity, both in vitro³⁷ and in vivo.³⁸,³⁹ In addition, high iron suppresses osteoblast function.⁴⁰ Low iron concentrations may also affect bone remodeling, as iron is involved in collagen synthesis,⁴¹ vitamin D hydroxylation,⁴² and as hypoxia, which may be seen in anemia, also stimulates bone resorption.⁴³ However, in our study, the risk of fractures remained high in individuals with low concentrations of plasma iron, transferrin saturation or ferritin after excluding individuals with anemia, indicating that anemia is not a likely explanation for the high risk of fractures.

Figure 6.Risk of any fracture for HFE C282Y homozygotes versus non-carriers for the C282Y and H63D variants stratified according to whether or not C282Y homozygotes had been diagnosed with hemochromatosis at any time. Diagnoses of hemochromatosis were obtained from the Danish National Patient Register, which covers all hospitals in Denmark. C282Y/C282Y: homozygous for C282Y; Non-carrier/non-carrier: non-carrier for both C282Y and H63D; N: number; 95% CI: 95% confidence interval.

Our finding of an overall increased risk of any fracture and fracture of the hip and femur in C282Y homozygotes is partly supported by findings from the UK Biobank.^16,17 Banfield et al. investigated 451,143 individuals from the UK general population and found increased risk of hip and femur fractures in C282Y homozygous men, but not in women, while they did not find an overall increased risk of any fracture in women and men.¹⁷ In contrast, we observed an overall increased risk of any fracture and fracture of the hip and femur in all C282Y homozygous individuals combined, without convincing evidence for interaction between genotype and sex on risk of fractures. Importantly, and a novel finding, we found a markedly increased risk of any fracture even in C282Y homozygotes with normal concentrations of ferritin, which has not been studied in other large general population cohorts such as the UK Biobank, as ferritin was not measured in that cohort. The reliability of our results is further illustrated by the fact that fractures occurred at a younger age in C282Y homozygous individuals compared to non-carrier individuals. Another surprising finding is our novel observation of a moderately increased risk of fracture of the hip and femur in H63D homozygotes, C282Y heterozygotes, and C282Y/H63D compound heterozygotes. Although the magnitude of the risk estimates for fracture of the hip and femur in individuals with the above mentioned genotypes was lower than for C282Y homozygotes, the finding of an increased risk in H63D homozygotes and C282Y heterozygotes is particularly surprising, as previous studies were found that individuals with these genotypes are not at increased risk of other hemochromatosis-related complications such as liver disease, infections, or diabetes.^16,22,23 While this is interesting from a mechanistic point of view, the moderately increased risk may be of limited clinical relevance, and we do not believe our findings should cause changes in the current clinical practice for these intermediate genotypes.

Figure 7.Cumulative incidence of any fracture in HFE C282Y homozygotes and non-carriers of C282Y and H63D variants as a function of age. Fine-Gray competing risk regression was used to calculate cumulative incidence of any fracture, taking death from any cause as a competing event into consideration. The data presented are cumulative incidences of any fracture for all C282Y homozygotes and non-carriers, and specifically for C282Y homozygotes and non-carriers with normal plasma iron, normal transferrin saturation, or normal ferritin at study enrollment. Normal iron concentration was defined as 9-34 μmol/L and high iron concentration as >34 μmol/L. For transferrin saturation, the reference range varies according to age and sex, and therefore normal transferrin saturation was defined as 10-45% for women ≤50 years of age and 15-45% for women >50 years of age and men of any age. High transferrin saturation was defined as >45% for men and women of any age. Normal ferritin concentration was defined as 12-200 μg/L for women and 12-300 μg/L for men, and high ferritin concentration as >200 μg/L for women and >300 μg/L for men. For the analyses on individuals with normal concentrations of plasma iron, transferrin saturation, and ferritin, C282Y homozygous individuals were excluded if they had been diagnosed with hemochromatosis before the day of study enrollment when blood samples were obtained, as therapeutic phlebotomy can affect plasma iron, transferrin saturation, or ferritin levels. The cumulative incidence curves for C282Y homozygotes and non-carriers with normal transferrin saturation overlap in their entire range, and may therefore appear graphically as a single curve, but are indeed two separate curves. Log-rank P values are for the comparison of C282Y homozygotes with non-carrier individuals. C282Y/ C282Y: homozygous for C282Y; Non-carrier/non-carrier, non-carrier: for both C282Y and H63D; N: number.

Mechanistically, common signaling pathways involved in bone remodeling and hepcidin expression may potentially explain the increased fracture risk in HFE hemochromatosis, including the increased risk of fractures in C282Y homozygotes with normal ferritin. The expression of hepcidin in hepatocytes is controlled by signaling through the bone morphogenetic protein pathway, which is also a critical regulator of bone metabolism.^44,45 Furthermore, HFE is highly expressed in both osteoblasts and osteoclasts.⁴⁶ HFE is located on chromosome 6 within the extended HLA class I region,⁴⁷ and the HLA locus also contains genes encoding major histocompatibility complex proteins and tumor necrosis factor-alpha (TNF-α).⁴⁷ Notably, TNF-α might be implicated in the risk of fractures by promoting bone resorption through osteoclast activation,⁴⁸ but we cannot make any conclusions on the exact mechanisms underlying our findings due to the observational nature of the study. Our finding of a markedly increased risk of fractures in individuals with C282Y homozygosity and normal concentrations of ferritin may challenge the presumption that systemic iron accumulation is the primary mechanism causing increased fracture risk in C282Y homozygotes.^14,45 Although our observational study cannot determine the potential effect of lowering ferritin through therapeutic phlebotomy, the high risk of any fracture in C282Y homozygotes with normal ferritin questions whether phlebotomy aimed at reducing iron storage is sufficient for reducing fracture risk in C282Y homozygotes, underlining the need for further studies.

Given the high morbidity and mortality associated with hip and femur fractures,³⁴ interventions aimed at reducing fracture risk may possibly improve quality of life while also reducing healthcare spending.⁴⁹ Several prophylactic measures, such as smoking and alcohol cessation, regular weight-bearing exercise, or treatment with antiresorptive agents, including calcium and vitamin D, either alone or in combination with anabolic agents, have been found to reduce fracture risk in other high-risk populations.⁵⁰ Further studies are needed to determine whether similar interventions could effectively mitigate fracture risk in HFE hemochromatosis.

Strengths of this study include the large general population cohort of 142,146 individuals who were invited based on age and geography without taking prior health conditions or healthcare utilization into account. Another strength is the extensive national Danish registries with detailed information on hospitalizations due to fractures, comorbidities, and vital status without any losses to follow-up. Among the limitations of our study is that we do not know whether concentrations of iron, transferrin saturation, or ferritin might have changed between study enrollment and the time of fracture. However, we have previously investigated changes in plasma iron and transferrin saturation in repeat blood samples after a median of 10 years in 89 of the same C282Y homozygotes who were included in our present study. At study enrollment, 76 of the 89 C282Y homozygotes had normal plasma iron, and in 67 of these (88.2%) plasma iron levels were persistently normal after a median of 10 years. Among 24 C282Y homozygotes with normal transferrin saturation at study enrollment, 15 (62.5%) persistently had normal transferrin saturation in the repeat blood samples.^22,23 Another limitation is that more than 99% of individuals in our study were white and of Danish descent, making us unable to examine whether our findings are generalizable to other ethnic groups.

In conclusion, risk of any fracture and fracture of the hip and femur was increased in individuals with low or high iron, low or high transferrin saturation, or low ferritin when studying all individuals irrespective of HFE genotype. HFE C282Y homozygotes had increased risks of any fracture and fracture of the hip and femur. Importantly, even C282Y homozygotes with normal concentrations of ferritin had an increased risk of any fracture, highlighting that this subgroup, typically not recommended for HFE genotyping under current clinical guidelines, remains at considerable fracture risk with a cumulative incidence of any fracture of 49% at age 80 years.

Footnotes

• Received September 25, 2025
• Accepted March 18, 2026

Correspondence

J. Helby jens.helby.petersen.02@regionh.dk

Disclosures

AG has received specific research funding for this study from Novo Nordisk A/S. Outside of the submitted work, AG has received research funding from Agios, Bristol Myers Squibb, Novo Nordisk, and Sanofi, and is an advisory board member for Agios, Disc Medicine, Novo Nordisk, Pharmacosmos, and Vertex Pharmaceuticals. Outside of the submitted work, JH has received research funding, payment for advisory board participation, and a conference travel grant from Sanofi, and payment for advisory board participation from Disc Medicine. CE is partly funded by the Laboratory Endowment Fund at Boston Children’s Hospital, Boston, MA, USA. The remaining authors have no conflicts of interest to disclose.

Contributions

AG, CE, JP, and JH were responsible for the study concept and design. BGN, CE, SEB, and JH collected data and constructed the database. MW and JH performed the statistical analysis. MW and JH wrote the initial drafts of the manuscript. MW, AG, BGN, MM, CE, JP, SEB, and JH interpreted data and results. MW, AG, BGN, MM, CE, JP, SEB, and JH revised and approved the final manuscript.

Funding

This study was supported by research funding from Novo Nordisk A/S given to AG. Novo Nordisk had no role in data collection, data analysis, or writing the manuscript, and the opinions in this manuscript are those of the authors and do not necessarily represent the opinion of Novo Nordisk.

Acknowledgments

The authors would like to thank the staff and participants of the Copenhagen City Heart Study, the Copenhagen General Population Study, and the Danish General Suburban Population Study.