Abstract
Objectives
To test whether haemochromatosis HFE C282Y homozygotes have increased risk of diabetes, liver disease, and heart disease even when they have normal plasma iron, transferrin saturation, or ferritin concentrations and to test whether C282Y homozygotes with diabetes, liver disease, or heart disease have increased mortality compared with non-carriers with these diseases.
Design
Prospective cohort study.
Setting
Three Danish general population cohorts: the Copenhagen City Heart Study, the Copenhagen General Population Study, and the Danish General Suburban Population Study.
Participants
132 542 individuals genotyped for the HFE C282Y and H63D variants, 422 of whom were C282Y homozygotes, followed prospectively for up to 27 years after study enrolment.
Main outcome measure
Hospital contacts and deaths, retrieved from national registers, covering all hospitals and deaths in Denmark.
Results
Comparing C282Y homozygotes with non-carriers, hazard ratios were 1.72 (95% confidence interval (CI) 1.24 to 2.39) for diabetes, 2.22 (1.40 to 3.54) for liver disease, and 1.01 (0.78 to 1.31) for heart disease. Depending on age group, the absolute five year risk of diabetes was 0.54-4.3% in C282Y homozygous women, 0.37-3.0% in non-carrier women, 0.86-6.8% in C282Y homozygous men, and 0.60-4.80% in non-carrier men. When studied according to levels of iron, transferrin saturation, and ferritin in a single blood sample obtained at study enrolment, risk of diabetes was increased in C282Y homozygotes with normal transferrin saturation (hazard ratio 2.00, 95% CI 1.04 to 3.84) or ferritin (3.76, 1.41 to 10.05) and in C282Y homozygotes with normal levels of both ferritin and transferrin saturation (6.49, 2.09 to 20.18). C282Y homozygotes with diabetes had a higher risk of death from any cause than did non-carriers with diabetes (hazard ratio 1.94, 95% CI 1.19 to 3.18), but mortality was not increased in C282Y homozygotes without diabetes. The percentage of all deaths among C282Y homozygotes that could theoretically be prevented if excess deaths in individuals with a specific disease were eliminated (the population attributable fraction) was 27.3% (95% CI 12.4% to 39.7%) for diabetes and 14.4% (3.1% to 24.3%) for liver disease. Risk of diabetes or liver disease was not increased in H63D heterozygotes, H63D homozygotes, C282Y heterozygotes, or C282Y/H63D compound heterozygotes.
Conclusions
Haemochromatosis C282Y homozygotes with normal transferrin saturation and/or ferritin, not recommended for HFE genotyping according to most guidelines, had increased risk of diabetes. Furthermore, C282Y homozygotes with diabetes had higher mortality than non-carriers with diabetes, and 27.3% of all deaths among C282Y homozygotes were potentially attributable to diabetes. These results indicate that prioritising detection and treatment of diabetes in C282Y homozygotes may be relevant.
Introduction
Iron homoeostasis is controlled by the master hormone hepcidin, a peptide synthesised in the liver and regulated through the homoeostatic iron regulator (HFE) protein. Hereditary haemochromatosis is caused by germline genetic variants, most commonly in the HFE gene, causing increased intestinal iron uptake that can lead to progressive iron accumulation in the body.1 2 3 Approximately 80-90% of patients with a diagnosis of hereditary haemochromatosis in northern Europe are homozygous for the C282Y (rs1800562) variant in the HFE gene, whereas less than 10% are compound heterozygous for the H63D (rs1799945) and C282Y variants.1 4 5 The C282Y variant is common in people of northern European descent, with homozygosity reported in 0.25-1% of the general population.5 However, estimates of penetrance vary markedly between studies, with clinically apparent haemochromatosis reported in between <1% and 60% of C282Y homozygotes.3 6 7 8
Previous studies have found increased risk of liver cirrhosis, liver cancer, and diabetes mellitus in people with C282Y homozygosity, but studies on risk of heart disease have shown conflicting results.9 10 11 Most clinical guidelines on hereditary haemochromatosis are based on the assumption that the increased risk of liver disease and diabetes is mechanistically caused by accumulation of iron in hepatocytes and pancreatic islets, reflected in high levels of transferrin saturation and ferritin. Hence, guidelines from several countries, including Denmark and the UK, recommend genetic testing for the two HFE variants C282Y and H63D only in individuals with a high transferrin saturation and a high ferritin concentration (except for genotyping close relatives of affected patients).1 3 5 12 13 On the basis of observational studies, therapeutic phlebotomy has been the cornerstone of treatment for hereditary haemochromatosis for many years. The effect of therapeutic phlebotomy in individual patients is typically assessed by measuring whether transferrin saturation and ferritin concentrations decrease sufficiently. However, randomised trials on risk of complications and death in people with hereditary haemochromatosis treated with phlebotomy have not been conducted, and whether individuals with hereditary haemochromatosis with normal levels of plasma iron, transferrin saturation, and/or ferritin are at increased risk of diabetes, liver disease, heart disease, and/or death is unknown.
Therefore, we tested the hypothesis that C282Y homozygotes have increased risk of diabetes, liver disease, and heart disease even when they have normal levels of plasma iron, transferrin saturation, or ferritin. Furthermore, we tested the hypothesis that C282Y homozygotes with diabetes, liver disease, or heart disease have higher mortality than do HFE non-carriers with diabetes, liver disease, or heart disease. For this purpose, we studied 132 542 individuals from three Danish general population studies. All individuals were genotyped for the HFE variants C282Y and H63D and were followed for up to 27 years after study enrolment.
Methods
We studied 132 542 consecutive individuals from three Danish cohort studies of the general population: the Copenhagen City Heart Study (9174 individuals examined 1991-94),14 the Copenhagen General Population Study (103 276 individuals examined 2003-14),15 and the Danish General Suburban Population Study (20 092 individuals examined 2010-13).16 Among those invited, 61% participated in the Copenhagen City Heart Study, 43% in the Copenhagen General Population Study, and 43% in the Danish General Suburban Population Study. On the day of enrolment, all individuals underwent a physical examination, had blood samples drawn, and completed a questionnaire on health and lifestyle. Age at study enrolment was 20-100 years. Study procedures were similar for the three cohort studies. No individuals overlapped between the three cohort studies, and no individuals were lost to follow-up. Danish ethics committees approved the studies, and each participant gave written, informed consent. See the supplementary methods for details about study procedures.
HFE genotype
All individuals were genotyped for the two HFE variants C282Y and H63D by using DNA extracted from peripheral blood leukocytes.17 18 Details on genotyping are described in the supplementary methods. Individuals were not informed about their C282Y and H63D genotypes as the ethics committee approvals authorised the disclosure of such genetic findings only when a clear, substantial, and well documented health consequence existed; thus, participating in the study did not influence whether individuals were treated for hereditary haemochromatosis.
Covariates
Information on age and sex came from the Danish Civil Registration System. For the multivariable adjusted supplementary analyses on risk of diabetes and risk of death from any cause, we retrieved information on other covariates reported to be associated with diabetes or all cause mortality on the basis of a priori literature reviews conducted before we did the analyses (see supplementary methods).
Plasma iron, transferrin saturation, and ferritin
We measured plasma iron, transferrin saturation, and ferritin in blood samples from 130 947, 130 902, and 33 428 individuals, respectively. Blood samples used for measuring iron, transferrin saturation, and ferritin were drawn on the day of study enrolment, except for 4791 individuals from the Copenhagen City Heart Study (examined 1991-94) who had ferritin measured on blood samples drawn at an earlier examination in 1981-83. See the supplementary methods for information on sample storage, laboratory assays, calibration, and quality control.
Disease endpoints and vital status
The Danish National Patient Register contains information on all inpatient hospital admissions in Denmark since 1977 and all emergency department and outpatient visits since 1994.19 Using this register, we retrieved information on inpatient hospital admissions from 1 January 1977 until 31 December 2018 and emergency department and outpatient visits from 1 January 1994 until 31 December 2018. Diagnoses were classified according to ICD-8 (international classification of diseases, 8th revision) until 31 December 1993, when it was superseded by ICD-10. ICD codes for diabetes, liver disease, heart disease, and subcategories of these conditions are shown in supplementary table S1. In previous studies, when the validity of diabetes diagnoses in the Danish National Patient Register was ascertained by comparison with detailed hospital chart review conducted by a physician in 50 individuals registered as having diabetes in the Danish National Patient Register, the hospital charts documented a diagnosis of diabetes in 48 (96%).20 Corresponding validity numbers were 100/100 (100%) individuals for liver disease and 99/100 (99%) for heart disease.20
According to Danish national guidelines, patients with hereditary haemochromatosis should be treated exclusively in public hospital departments.21 22 To do stratified analyses according to whether or not C282Y homozygotes had a diagnosis of haemochromatosis, we therefore retrieved information on hospital contacts with haemochromatosis from 1 January 1977 until 31 December 2018, from the Danish National Patient Register (ICD-8 code for haemochromatosis: 27329; ICD-10: E831A). For the analysis on risk of non-alcoholic fatty liver disease, we used the fatty liver index based on body mass index, waist circumference, plasma triglycerides, and plasma γ-glutamyl transferase concentrations at study enrolment.23 24 We defined fatty liver disease cross sectionally as having a fatty liver index ≥60 at study enrolment in individuals without heavy alcohol intake (with heavy defined as >168 g/week for men and >84 g/week for women according to recommendations by Danish health authorities). Information about vital status and emigration until 31 December 2018 came from the national Danish Civil Registration System.25
Statistical analysis
We used Stata 18.0 for all statistical analyses except for the sensitivity analyses using Cox regression with Firth’s penalised maximum likelihood bias reduction method,26 27 for which we used R version 4.3.2. All statistical tests were two sided. Relative risks of diabetes, liver disease, heart disease, and death were modelled by Cox proportional hazards regression, using left truncated age as the timescale, thus adjusting for age. To account for potential differences between the three study cohorts, all main analyses on relative risk used shared frailty Cox models,28 29 with frailties shared for individuals within each of the three study cohorts. However, results were similar to those presented if we adjusted for study cohort as a categorical variable in the Cox models instead of using shared frailty Cox models (data not shown). When estimating absolute five year risk of diabetes, liver disease, heart disease, and death, we modelled each endpoint separately using Poisson regression adjusted for categories of age and sex. As genotype is constant during life and not affected by lifestyle, the primary analyses were adjusted only for age and sex to avoid adjusting for potential mediators that may be part of the causal pathway between HFE genotype and risk of disease. However, we also did supplementary analyses using multivariable adjusted models which gave similar results, as described in the results section and in the supplementary methods.
Follow-up for the analyses on relative risk of diabetes, liver disease, and heart disease began at age 20 years or 1 January 1977, whichever came last, and follow-up ended on date of first hospital contact with the disease being studied in the specific analysis (diabetes, liver disease, or heart disease), emigration (384 individuals), death from any cause, or 31 December 2018, whichever came first (supplementary figure S1). For the analyses on absolute five year risk of diabetes, liver disease, and heart disease, follow-up began at study enrolment and ended on date of first hospital contact with the disease being studied in the specific analysis, emigration, death from any cause, or 31 December 2018, whichever came first. For the analyses on relative and absolute risk of death in all individuals irrespective of disease, follow-up began at the day of study enrolment and ended on date of death from any cause, emigration, or 31 December 2018, whichever came first. For the analyses on relative and absolute risk of death from any cause in individuals with diabetes, liver disease, or heart disease, follow-up began at the day of study enrolment or the day of first hospital contact with the disease being studied in the specific analysis, whichever came last, and follow-up ended on date of death from any cause, emigration, or 31 December 2018, whichever came first. We observed no major violations of the proportional hazards assumption when we tested this by first visually plotting –ln(–ln(survival)) against ln(analysis time) and further testing on the basis of Schoenfeld residuals if non-parallel curves were suspected from the visual inspection. Further information on statistical methods is provided in the supplementary methods.
Patient and public involvement
Patients were not directly involved in the planning, design, or conduct of this study. Although patient and public involvement is increasingly recognised as important, the general population studies that formed the foundation of this research were started before formal patient and public involvement practices were common in Denmark. The specific focus of this study was inspired by encounters with patients presenting with classical hereditary haemochromatosis, particularly those with complications of diabetes and other associated health problems. However, no resources were allocated to formal patient involvement before the study’s analysis plan was finalised. During the manuscript review process, one of the reviewers was a patient with hereditary haemochromatosis, and their insights significantly enhanced the quality and relevance of the final manuscript. Although direct patient and public involvement was not formally integrated into the study, public representation was ensured during the ethical review process. The Danish ethical committees, which reviewed and approved the general population studies, include lay members, ensuring that public perspectives were considered during ethical oversight. Furthermore, informal conversations with patients provided additional context and helped to refine our understanding of the condition, even though these discussions were not part of the formal study protocol.
Results
Baseline characteristics according to HFE genotype for the 132 542 individuals from the general population are presented in table 1 and described in the supplementary results. As expected, C282Y homozygotes (C282Y/C282Y) had higher levels of iron, transferrin saturation and ferritin than did individuals who were non-carriers for both C282Y and H63D (non-carrier/non-carrier).
Table 1.
Baseline characteristics of 132 542 individuals from the general population according to haemochromatosis genotypes C282Y and H63D. Values are numbers (percentages) unless stated otherwise
| Characteristics | Haemochromatosis genotype | P for difference between C282Y/C282Y and non-carrier/non-carrier individuals | |||||
|---|---|---|---|---|---|---|---|
| Non-carrier/non-carrier (n=87 313) | H63D/non-carrier (n=27 820) | H63D/H63D (n=2358) | C282Y/non-carrier (n=12 514) | C282Y/H63D (n=2115) | C282Y/C282Y (n=422) | ||
| Median (IQR) plasma iron, µmol/L | 13 (11-16) | 15 (12-18) | 17 (13-20) | 15 (12-18) | 18 (15-22) | 25 (19-31) | <0.001 |
| Median (IQR) transferrin saturation, % | 22 (17-27) | 24 (19-30) | 28 (23-35) | 26 (20-32) | 33 (27-41) | 57 (43-75) | <0.001 |
| Median (IQR) plasma ferritin, µg/L | 98 (49-165) | 103 (52-176) | 121 (57-202) | 106 (52-184) | 125 (58-241) | 312 (94-606) | <0.001 |
| Median (IQR) age, years | 57.7 (47.6-67.2) | 57.6 (47.6-67.0) | 57.9 (47.6-67.3) | 57.8 (47.7-67.0) | 57.2 (47.3-67.3) | 56.5 (47.0-65.5) | 0.02 |
| Median (IQR) year of birth | 1950 (1940-1961) |
1951 (1941-1961) |
1950 (1940-1961) |
1950 (1941-1961) |
1951 (1941-1961) |
1952 (1942-1962) |
0.01 |
| Male sex | 39 386 (45) | 12 483 (45) | 1077 (46) | 5669 (45) | 961 (45) | 194 (46) | 0.72 |
| Premenopausal women (% of women) | 15 941 (33) | 5043 (33) | 415 (32) | 2320 (34) | 392 (34) | 77 (34) | 0.94 |
| Ever smoker | 51 404 (59) | 16 218 (58) | 1413 (60) | 7543 (60) | 1214 (57) | 250 (59) | 0.95 |
| Median (IQR) cumulative smoking, pack years* | 18 (7-30) | 17 (7-30) | 17 (7-30) | 18 (7-31) | 16 (6-32) | 15 (8-34) | 0.71 |
| Alcohol consumption >168/84 g/week† | 31 393 (36) | 10 010 (36) | 832 (35) | 4547 (36) | 769 (36) | 155 (37) | 0.55 |
| Use of antihypertensive drugs‡ | 16 797 (19) | 5573 (20) | 530 (22) | 2507 (20) | 456 (22) | 99 (23) | 0.03 |
| Median (IQR) body mass index | 25.6 (23.2-28.5) | 25.6 (23.2-28.5) | 25.7 (23.3-28.5) | 25.6 (23.2-28.5) | 25.4 (23.0-28.4) | 26.0 (23.3-28.9) | 0.13 |
| Comorbidity (excluding diabetes, liver disease, or heart disease)§ | 15 114 (17) | 4726 (17) | 415 (18) | 2226 (18) | 377 (18) | 76 (18) | 0.70 |
| Median (IQR) plasma non-HDL cholesterol, mmol/L | 3.9 (3.2-4.7) | 3.9 (3.2-4.7) | 3.9 (3.2-4.7) | 3.9 (3.2-4.7) | 3.9 (3.2-4.6) | 3.7 (3.1-4.5) | <0.001 |
| Median (IQR) plasma C reactive protein, mg/L | 1.4 (0.9-2.4) | 1.4 (0.9-2.4) | 1.4 (1.0-2.6) | 1.4 (0.9-2.4) | 1.4 (1.0-2.4) | 1.7 (1.1-3.0) | 0.05 |
| Diagnosis of hemochromatosis before study enrolment | 0 (0) | <5 | <5 | <5 | <5 | 11 (3) | <0.001 |
| Cohort: | 0.84 | ||||||
| Copenhagen City Heart Study | 6135 (7) | 1883 (7) | 158 (7) | 844 (7) | 131 (6) | 23 (5) | |
| Copenhagen General Population Study | 67 944 (78) | 21 682 (78) | 1846 (78) | 9800 (78) | 1667 (79) | 337 (80) | |
| Danish General Suburban Population Study | 13 234 (15) | 4255 (15) | 354 (15) | 1870 (15) | 317 (15) | 62 (15) | |
P for difference between C282Y homozygotes and non-carrier individuals for binary variables was calculated using logistic regression. For cohorts, P for difference between C282Y homozygotes and non-carrier individuals was calculated using Kruskal-Wallis test. For continuous variables, P for difference between C282Y homozygotes and non-carrier individuals was calculated using linear regression. Owing to Danish regulations on data privacy, fewer than five individuals in a column is reported as “<5.”
Non-carrier/non-carrier=non-carrier for both C282Y and H63D; H63D/non-carrier=heterozygous for H63D variant; H63D/H63D=homozygous for H63D variant; C282Y/non-carrier=heterozygous for C282Y variant; C282Y/H63D=compound heterozygous for C282Y and H63D variants; C282Y/C282Y=homozygous for C282Y variant.
HDL=high density lipoprotein; IQR=interquartile range.
Calculated for ever smokers only.
>168 g/week for men and >84 g/week for women.
Use of antihypertensive drugs was self-reported at study enrolment on basis of questionnaire.
Comorbidity at study enrolment (using modified Charlson comorbidity index disregarding diabetes, liver disease, and heart disease).
Relative risk of diabetes
For the analyses on relative risk of diabetes, liver disease, and heart disease, follow-up began at age 20 years or 1 January 1977 (whichever came last), giving a median follow-up of 41 (range 0-42) years. During follow-up, 7702 of 132 542 individuals had a hospital contact with diabetes. Compared with non-carriers (non-carrier/non-carrier), C282Y homozygotes (C282Y/C282Y) had a higher overall risk of diabetes (age and sex adjusted hazard ratio 1.72, 95% confidence interval (CI) 1.24 to 2.39) (fig 1) and a higher risk of diabetes with complications (2.03, 1.22 to 3.38) (fig 2). Results were similar when we additionally adjusted for potential risk factors for diabetes (body mass index, alcohol intake, and C reactive protein concentration as a marker of inflammation), with C282Y homozygotes having a multivariable adjusted hazard ratio for diabetes of 1.53 (95% CI 1.10 to 2.12) (supplementary figure S2). Age at diagnosis of diabetes was similar in C282Y homozygotes and non-carriers (mean age at diagnosis 64.4 v 64.5 years) (fig 1).
Fig 1.
Relative risk of diabetes according to haemochromatosis genotypes C282Y and H63D, adjusted for age and sex. Diabetes was defined as any hospital contact with diabetes (inpatient admission, emergency department visit, or outpatient visit) at any time before or after study enrolment. Non-carrier/non-carrier=non-carrier for both C282Y and H63D; H63D/non-carrier=heterozygous for H63D variant; H63D/H63D=homozygous for H63D variant; C282Y/non-carrier=heterozygous for C282Y variant; C282Y/H63D=compound heterozygous for C282Y and H63D variants; C282Y/C282Y=homozygous for C282Y variant. CI=confidence interval
Fig 2.
Relative risk of diabetes, liver disease, and heart disease and subcategories of these conditions for C282Y homozygotes (C282Y/C282Y) compared with non-carriers of haemochromatosis variants C282Y and H63D (non-carrier/non-carrier), adjusted for age and sex. Any diabetes was defined as any hospital contact with diabetes; subcategory “diabetes with complication” was defined as any hospital contact with diabetes with complications (eye complications, nephropathy, angiopathy, gangrene, foot ulcer, neurological complication, or other complications due to diabetes). Any liver disease was defined as any hospital contact with liver disease; subcategory “liver cirrhosis” was defined as any hospital contact with liver cirrhosis. Any heart disease was defined as any hospital contact with heart disease; subcategory “heart failure” was defined as any hospital contact with heart failure. For all diagnoses, all types of hospital contacts (inpatient admission, emergency department visits, and outpatient visits) at any time before or after study enrolment were included. CI=confidence interval
We also analysed risk of diabetes cross sectionally, whereby instead of defining diabetes by hospital diagnoses, we defined it simply as having a non-fasting blood glucose >11 mmol/L (>198 mg/dL) at study enrolment. When we used this alternative definition, risk of diabetes in C282Y homozygotes was also increased (odds ratio for diabetes 2.57 (95% CI 1.21 to 5.47) for C282Y homozygotes compared with non-carriers) (supplementary figure S3). For H63D heterozygotes (H63D/non-carrier), H63D homozygotes (H63D/H63D), C282Y heterozygotes (C282Y/non-carrier), and compound heterozygotes (C282Y/H63D), risk of diabetes was not increased regardless of how diabetes was defined (fig 1; supplementary figure S3).
To examine whether current clinical guidelines on testing for haemochromatosis adequately identify C282Y homozygotes at increased risk of diabetes by recommending testing only in individuals with high transferrin saturation and/or high ferritin, we examined risk of diabetes stratified by levels of iron, transferrin saturation, and ferritin. For this stratified analysis, we excluded individuals (n=11 C282Y homozygotes) with a diagnosis of haemochromatosis at study enrolment to exclude that levels of iron, transferrin saturation, or ferritin in the blood samples obtained at study enrolment were affected by previous therapeutic phlebotomy. Importantly, we found an increased risk of diabetes even in C282Y homozygotes with normal levels of transferrin saturation (hazard ratio for diabetes 2.00 (95% CI 1.04 to 3.84) when comparing C282Y homozygotes with normal transferrin saturation and non-carriers with normal transferrin saturation), normal concentrations of ferritin (3.76, 1.41 to 10.05), or normal levels of both ferritin and transferrin saturation (6.49, 2.09 to 20.18) (fig 3). For C282Y homozygotes with normal iron, risk of diabetes was less pronounced (hazard ratio 1.38, 0.91 to 2.09). Results were similar to those described above if we excluded not only individuals with a diagnosis of haemochromatosis before study enrolment but instead those (n=78 C282Y homozygotes) with a diagnosis of haemochromatosis at any time before or after study enrolment (data not shown). As diabetes could be diagnosed either before or after blood samples were obtained at study enrolment, figure 4 illustrates when diabetes was diagnosed and the median time between study enrolment and diagnosis of diabetes in C282Y homozygotes and non-carriers with normal levels of iron, transferrin saturation, and ferritin. When we restricted the analysis to C282Y homozygotes, the median time from diabetes to study enrolment for C282Y homozygotes with a diagnosis of diabetes before study enrolment was 4.9 years and the median time from study enrolment to diabetes for C282Y homozygotes with a diagnosis of diabetes after study enrolment was 4.2 years.
Fig 3.
Relative risk of diabetes for C282Y homozygotes (C282Y/C282Y) compared with non-carriers for both C282Y and H63D (non-carrier/non-carrier) stratified according to levels of plasma iron, transferrin saturation, and ferritin at study enrolment, adjusted for age and sex. Diabetes was defined as any hospital contact with diabetes (inpatient admission, emergency department visit, or outpatient visit) at any time before or after study enrolment. Analysis excludes individuals (n=11 C282Y homozygotes) with diagnosis of haemochromatosis at study enrolment to exclude that levels of iron, transferrin saturation, or ferritin in blood samples obtained at study enrolment were affected by previous therapeutic phlebotomy. Normal iron was defined as iron 9-34 µmol/L and high iron as >34 µmol/L. Normal transferrin saturation was defined as 10-45% for women aged ≤50 years and 15-45% for women >50 and men regardless of age. High transferrin saturation was defined as >45% for both sexes. Normal ferritin was defined as ferritin 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. As very few C282Y homozygotes had low levels of iron, transferrin saturation, or ferritin, individuals with low levels of iron, transferrin saturation, or ferritin were excluded from analysis stratified for specific parameter that was low. Owing to Danish regulations on data privacy, fewer than 3 individuals with diabetes in a column is reported as “<3” but risk estimates were calculated using exact numbers. CI=confidence interval
Fig 4.
Graphical curves illustrating risk of diabetes by showing proportion of individuals without diabetes as function of time for C282Y homozygotes (C282Y/C282Y) and non-carriers (non-carrier/non-carrier) with normal levels of iron, transferrin saturation, or ferritin and for C282Y homozygotes and non-carriers with normal levels of both ferritin and transferrin saturation. Curves depict proportion of individuals without diabetes estimated using Cox proportional hazards model on risk of diabetes. Diabetes was defined as any hospital contact with diabetes (inpatient admission, emergency department visit, or outpatient visit) at any time before or after study enrolment. Analysis excludes individuals with diagnosis of haemochromatosis at study enrolment to exclude that levels of iron, transferrin saturation, or ferritin in blood samples obtained at study enrolment were affected by previous therapeutic phlebotomy. Arrows on x axis mark median time from diabetes to study enrolment for individuals with diagnosis of diabetes before study enrolment and median time from study enrolment to diabetes for individuals with diagnosis of diabetes after study enrolment. P values are from log-rank test comparing C282Y homozygotes (C282Y/C282Y) with non-carriers for both C282Y and H63D (non-carrier/non-carrier)
When we restricted the analysis to study only risk of diabetes after study enrolment and exclude individuals with a diagnosis of diabetes before study enrolment, confidence intervals became wider owing to lower statistical power, but risk of diabetes was still increased in C282Y homozygotes with normal concentrations of ferritin (4.44, 95% CI 1.43 to 13.82) or normal levels of both ferritin and transferrin saturation (8.03, 2.00 to 32.28), whereas the overall group of C282Y homozygotes with normal transferrin saturation had a hazard ratio for diabetes of 1.04 (95% CI 0.34 to 3.23) (supplementary figure S4).
To assess whether the increased risk of diabetes was also seen in C282Y homozygotes without other comorbidities potentially caused by haemochromatosis, we did stratified analyses according to presence versus absence of liver disease or heart disease. The increased risk of diabetes was found in C282Y homozygotes without liver disease or heart disease at study enrolment (hazard ratio for diabetes 1.56, 95% CI 1.07 to 2.28) (supplementary figure S5), and results were similar for C282Y homozygotes who did not have a diagnosis of liver disease or heart disease at any time before or after study enrolment (supplementary figures S6 and S7). Compared with a reference group of non-carriers with body mass index 18.5-24.9, risk of diabetes was increased in non-carriers with high body mass index (hazard ratio 2.15 (95% CI 1.98 to 2.33) for non-carriers with body mass index 25-29.9; hazard ratio 5.76 (5.33 to 6.24) for non-carriers with body mass index ≥30) and even higher in C282Y homozygotes with high body mass index (hazard ratio 4.17 (95% CI 2.58 to 6.74) for C282Y homozygotes with body mass index 25-29.9; hazard ratio 7.34 (4.24 to 12.69) for C282Y homozygotes with body mass index ≥30) (supplementary figure S8). However, high body mass index and C282Y homozygosity seemed to be independent risk factors for diabetes, as we observed no interaction between body mass index and HFE genotype on risk of diabetes (P for interaction=0.12). Results were similar when we restricted the analysis on risk of diabetes according to body mass index to study only risk of diabetes after study enrolment when measurements of height and weight were obtained for calculation of body mass index (supplementary figure S8).
Among the 422 C282Y homozygotes, 78 had a diagnosis of haemochromatosis made at any time before or after study enrolment, of whom 11 had had a diagnosis of haemochromatosis before study enrolment. When we did stratified analyses according to whether or not C282Y homozygotes had a diagnosis of haemochromatosis, risk estimates for diabetes were similar for C282Y homozygotes with a clinical diagnosis of haemochromatosis at any time before or after study enrolment (hazard ratio 2.41 (95% CI 1.29 to 4.47) compared with non-carriers) and in C282Y homozygotes who never had a diagnosis of haemochromatosis (1.55, 1.05 to 2.27) (fig 5).
Fig 5.
Relative risk of diabetes in C282Y homozygotes (C282Y/C282Y) compared with non-carriers (non-carrier/non-carrier) and risk of death from any cause in C282Y homozygotes and non-carriers with and without diabetes, stratified by whether or not C282Y homozygotes had diagnosis of haemochromatosis at any time before or after study enrolment and, adjusted for age and sex. Diagnoses of diabetes and haemochromatosis were retrieved from Danish National Patient Registry. Diabetes was defined as any hospital contact with diabetes; haemochromatosis was defined as any hospital contact with haemochromatosis. For both diagnoses, all types of hospital contacts (inpatient admission, emergency department visits, and outpatient visits) were included. CI=confidence interval
We observed no convincing difference in relative risk estimates for diabetes between C282Y homozygotes aged 20-65 years and those aged ≥66 years (supplementary figure S9). Relative risk estimates were similar for men and women when we stratified the analysis on risk of diabetes according to sex by comparing C282Y homozygous men with non-carrier men and C282Y homozygous women with non-carrier women (supplementary figure S10).
As plasma iron and transferrin saturation can change over time,2 we examined levels of iron and transferrin saturation in 89 C282Y homozygotes who had plasma iron and transferrin saturation measured first in a blood sample obtained at the day of study enrolment and again in a second blood sample obtained for repeat measurements of plasma iron and transferrin saturation at a median of 10 years later. Of the 89 C282Y homozygotes with repeat samples, 76 had normal iron measured in the first blood sample; of these, 67 (88.2%) still had normal iron measured in the second blood sample. Of the 89 C282Y homozygotes, 24 had normal transferrin saturation measured in the first blood sample; of these, 15 (62.5%) still had normal transferrin saturation measured in the second blood sample (supplementary figure S11). No individuals had a repeat measurement of ferritin. Importantly, all analyses on risk of diabetes and other diseases were based on the first blood sample drawn at the day of study enrolment.
Relative risk of liver disease and heart disease
Compared with non-carriers, C282Y homozygotes had a higher risk of any liver disease (hazard ratio 2.22, 95% CI 1.40 to 3.54) and a higher risk of liver cirrhosis (3.42, 1.41 to 8.27) (fig 2). Likewise, C282Y homozygotes had higher risk of non-alcoholic fatty liver disease at study enrolment (odds ratio 1.63, 95% CI 1.22 to 2.19) when this was defined using the fatty liver index (supplementary figure S12). When stratified according to levels of iron, transferrin saturation, and ferritin, risk estimates for any liver disease were less pronounced for C282Y homozygotes with normal plasma iron (1.73, 95% CI 0.96 to 3.13) than for C282Y homozygotes with high plasma iron (5.18, 2.15 to 12.47; P for difference=0.04) (supplementary figure S13). Risk of liver disease was not convincingly increased in C282Y homozygotes with normal transferrin saturation (1.02, 95% CI 0.26 to 4.09) or ferritin (2.22, 0.31 to 15.85) (supplementary figure S13). Risk of any liver disease and risk of non-alcoholic fatty liver disease were not increased in H63D heterozygotes (H63D/non-carrier), H63D homozygotes (H63D/H63D), C282Y heterozygotes (C282Y/non-carrier), or compound heterozygotes (C282Y/H63D) (supplementary figures S12 and S14).
Risk of heart disease (1.01, 0.78 to 1.31) and risk of heart failure (0.84, 0.50 to 1.43) were not increased in C282Y homozygotes (fig 2). Likewise, risk of heart disease was not increased in C282Y homozygotes when we did stratified analyses according to levels of iron, transferrin saturation, and ferritin (supplementary figure S15).
Absolute risk of diabetes, liver disease, and heart disease
To guide clinical decisions for C282Y homozygous individuals, absolute risk estimates may be more useful than relative risk estimates. Therefore, table 2 shows absolute five year risks of having a diagnosis of diabetes, liver disease, or heart disease according to sex and age categories. Depending on age group, the absolute five year risk of having a diagnosis of diabetes was 0.54-4.3% in C282Y homozygous women, 0.37-3.0% in non-carrier women, 0.86-6.8% in C282Y homozygous men, and 0.60-4.80% in non-carrier men. Corresponding absolute five year risks for liver disease were 0.77-2.3% in C282Y homozygous women, 0.28-0.83% in non-carrier women, 0.87-2.6% in C282Y homozygous men, and 0.32-0.94% in non-carrier men.
Table 2.
Absolute five year risk (%) of having diagnosis of diabetes, liver disease, and heart disease made after study enrolment
| Age group, years | Women | Men | |||
|---|---|---|---|---|---|
| Non-carrier | C282Y/C282Y | Non-carrier | C282Y/C282Y | ||
| Diabetes | |||||
| 20-39 | 0.37 | 0.54 | 0.60 | 0.86 | |
| 40-49 | 0.58 | 0.84 | 0.94 | 1.4 | |
| 50-59 | 1.4 | 2.1 | 2.3 | 3.3 | |
| 60-69 | 2.0 | 2.9 | 3.2 | 4.6 | |
| 70-79 | 2.7 | 3.9 | 4.3 | 6.2 | |
| ≥80 | 3.0 | 4.3 | 4.8 | 6.8 | |
| Liver disease | |||||
| 20-39 | 0.28 | 0.77 | 0.32 | 0.87 | |
| 40-49 | 0.36 | 0.98 | 0.40 | 1.1 | |
| 50-59 | 0.65 | 1.8 | 0.73 | 2.0 | |
| 60-69 | 0.80 | 2.2 | 0.91 | 2.5 | |
| 70-79 | 0.81 | 2.2 | 0.92 | 2.5 | |
| ≥80 | 0.83 | 2.3 | 0.94 | 2.6 | |
| Heart disease | |||||
| 20-39 | 0.61 | 0.54 | 0.98 | 0.88 | |
| 40-49 | 1.4 | 1.2 | 2.2 | 2.0 | |
| 50-59 | 3.4 | 3.1 | 5.6 | 5.0 | |
| 60-69 | 6.0 | 5.4 | 9.6 | 8.6 | |
| 70-79 | 11 | 10 | 18 | 16 | |
| ≥80 | 19 | 18 | 30 | 27 | |
Absolute five year risk estimates in percentages are shown for C282Y homozygotes (C282Y/C282Y) and non-carriers for both C282Y and H63D (non-carrier) and stratified according to sex and age categories. Analysis on risk of diabetes included 85 304 non-carriers and 406 C282Y homozygotes as individuals with diagnosis of diabetes before study enrolment were excluded. Analysis on risk of liver disease included 86 561 non-carriers and 418 C282Y homozygotes as individuals with diagnosis of liver disease before study enrolment were excluded. Analysis on risk of heart disease included 82 010 non-carriers and 401 C282Y homozygotes as individuals with diagnosis of heart disease before study enrolment were excluded.
Death from any cause in C282Y homozygotes with diabetes, liver disease, or heart disease
During follow-up, 17 688 individuals died, of whom 48 were C282Y homozygotes. When we studied all individuals irrespective of disease, risk of death from any cause was not higher in C282Y homozygotes than in non-carriers (hazard ratio 1.16, 95% CI 0.87 to 1.54) (fig 6). Likewise, risk of death from any cause was similar for C282Y homozygotes without diabetes and non-carriers without diabetes. Compared with non-carriers without diabetes, risk of death was higher in non-carriers with diabetes (hazard ratio 2.26, 95% CI 2.15 to 2.38). Importantly, however, C282Y homozygotes with diabetes had an even higher relative risk of death from any cause compared with non-carriers without diabetes (4.39, 95% CI 2.69 to 7.18), meaning that relative risk of death was substantially increased for C282Y homozygotes with diabetes compared with non-carriers with diabetes (hazard ratio 1.94, 95% CI 1.19 to 3.18) (fig 6).
Fig 6.
Relative risk of death from any cause for C282Y homozygotes (C282Y/C282Y) and non-carriers for both C282Y and H63D (non-carrier/non-carrier), adjusted for age and sex, including all individuals irrespective of disease with further stratification according to whether or not individuals had diagnosis of diabetes, liver disease, or heart disease at any time before or after study enrolment. Diabetes was defined as any hospital contact with diabetes, liver disease was defined as any hospital contact with any liver disease, and heart disease was defined as any hospital contact with any heart disease. For all three diagnosis categories, all types of hospital contacts (inpatient admission, emergency department visits, and outpatient visits) at any time before or after study enrolment were included. CI=confidence interval; PAF=population attributable fraction for each disease (diabetes, liver disease, or heart disease) on death from any cause. *Age and sex adjusted hazard ratio (95% CI) for death from any cause comparing C282Y homozygotes with diabetes and non-carriers with diabetes. †Age and sex adjusted hazard ratio (95% CI) for death from any cause comparing C282Y homozygotes with liver disease and non-carriers with liver disease. ‡Age and sex adjusted hazard ratio (95% CI) for death from any cause comparing C282Y homozygotes with heart disease and non-carriers with heart disease
When we included only individuals with normal levels of iron or transferrin saturation at study enrolment and excluded those with a diagnosis of haemochromatosis at study enrolment, results on risk of death in C282Y homozygotes with diabetes were similar to those presented above. However, statistical power was very limited for the stratified analyses on risk of death according to ferritin, as only four C282Y homozygotes with diabetes had a normal ferritin concentration (of six C282Y homozygotes with diabetes who had ferritin measured) (supplementary figure S16).
Results on risk of death in C282Y homozygotes with diabetes were similar to those shown in figure 6 when we excluded individuals with liver or heart disease at study enrolment (supplementary figure S17) and when we did multivariable adjusted analysis including several covariates reported to be associated with mortality (supplementary figure S18). Likewise, risk estimates for death were similar for C282Y homozygotes with diabetes with a diagnosis of haemochromatosis at any time before or after study enrolment (hazard ratio 3.31 (95% CI 1.07 to 10.26) compared with non-carriers without diabetes) and in C282Y homozygotes with diabetes with no diagnosis of haemochromatosis (4.75, 2.76 to 8.19) (fig 5). The number of deaths in C282Y homozygotes with diabetes (n=16) was too low to enable analyses on specific causes of death in C282Y homozygotes with diabetes.
Non-carriers with any liver disease had a higher risk of death from any cause (4.06, 95% CI 3.76 to 4.39) than did non-carriers without liver disease (fig 6). C282Y homozygotes with liver disease had a higher risk of death than did non-carriers without liver disease (6.71, 95% CI 3.35 to 13.43) but not a convincingly higher risk of death than non-carriers with liver disease (1.65, 0.82 to 3.32) (fig 6). Likewise, non-carriers with any heart disease had a higher risk of death from any cause (2.07, 95% CI 1.99 to 2.15) than did non-carriers without heart disease (fig 6). However, although C282Y homozygotes with heart disease had a higher risk of death than did non-carriers without heart disease (2.74, 95% CI 1.74 to 4.29), risk of death was not convincingly higher in C282Y homozygotes with heart disease (1.32, 0.84 to 2.08) than in non-carriers with heart disease (fig 6).
To estimate the proportion of all deaths that could hypothetically be prevented if interventions were able to eliminate the higher risk of death in individuals with diabetes, we calculated the population attributable fraction for diabetes on death from any cause, defined as the proportion of deaths from any cause that could theoretically be prevented at the population level if individuals with diabetes had the same mortality as those without diabetes. The population attributable fraction for diabetes on death from any cause was 7.8% (95% CI 7.2% to 8.4%) among non-carriers and 27.3% (12.4% to 39.7%) among C282Y homozygotes (fig 6), meaning that 27.3% of all deaths among C282Y homozygotes could hypothetically be prevented if C282Y homozygotes with diabetes had the same risk of death as C282Y homozygotes without diabetes.
The population attributable fraction for liver disease on death from any cause was 4.4% (95% CI 4.0% to 4.9%) among non-carriers and 14.4% (3.1% to 24.3%) among C282Y homozygotes, whereas the population attributable fraction for heart disease on death from any cause was 22.2% (21.1% to 23.4%) among non-carriers and 22.8% (4.3% to 37.8%) among C282Y homozygotes (fig 6). To examine the impact of HFE genotype on absolute risk of death from any cause, table 3 and table 4 show five year absolute risk estimates for death from any cause according to age and sex for C282Y homozygotes and non-carriers, showing risk estimates for all individuals irrespective of disease (table 3) and specifically for individuals with and without diabetes, liver disease, or heart disease (table 4).
Table 3.
Absolute five year risk of death (%) from any cause for all individuals irrespective of disease
| Age group, years | Women | Men | |||
|---|---|---|---|---|---|
| Non-carrier | C282Y/C282Y | Non-carrier | C282Y/C282Y | ||
| 20-39 | 0.39 | 0.39 | 0.56 | 0.56 | |
| 40-49 | 0.83 | 0.83 | 1.2 | 1.2 | |
| 50-59 | 2.6 | 2.6 | 3.7 | 3.7 | |
| 60-69 | 6.3 | 6.3 | 8.9 | 8.9 | |
| 70-79 | 15 | 15 | 21 | 21 | |
| ≥80 | 31 | 31 | 41 | 41 | |
Absolute five year risk estimates in percentages are shown for C282Y homozygotes (C282Y/C282Y) and non-carriers for both C282Y and H63D (non-carrier) and stratified according to sex and age categories. Analyses included all 87 313 non-carriers and all 422 C282Y homozygotes. Risk estimates were highly similar for non-carrier individuals and C282Y homozygotes; most five year risks estimates were different only at the second or third decimal place, meaning that risk estimates after rounding of numbers appear identical for non-carriers and C282Y homozygotes.
Table 4.
Absolute five year risk of death (%) from any cause in individuals with and without diabetes, liver disease, or heart disease
| Age group, years | Women | Men | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Non-carrier | C282Y/C282Y | Non-carrier | C282Y/C282Y | ||||||||
| Without disease | With disease | Without disease | With disease | Without disease | With disease | Without disease | With disease | ||||
| Diabetes | |||||||||||
| 20-39 | 0.36 | 0.78 | 0.29 | 1.4 | 0.51 | 1.1 | 0.41 | 1.9 | |||
| 40-49 | 0.81 | 1.7 | 0.65 | 3.0 | 1.1 | 2.4 | 0.92 | 4.2 | |||
| 50-59 | 2.5 | 5.2 | 2.0 | 8.9 | 3.4 | 7.3 | 2.8 | 12 | |||
| 60-69 | 5.9 | 13 | 4.9 | 21 | 8.2 | 17 | 6.7 | 28 | |||
| 70-79 | 14 | 29 | 12 | 44 | 20 | 38 | 16 | 56 | |||
| ≥80 | 30 | 54 | 25 | 74 | 40 | 67 | 34 | 85 | |||
| Liver disease | |||||||||||
| 20-39 | 0.36 | 1.3 | 0.33 | 1.8 | 0.51 | 1.9 | 0.47 | 2.5 | |||
| 40-49 | 0.76 | 2.8 | 0.70 | 3.7 | 1.1 | 4.0 | 1.0 | 5.3 | |||
| 50-59 | 2.5 | 8.8 | 2.3 | 12 | 3.5 | 12 | 3.2 | 16 | |||
| 60-69 | 6.0 | 20 | 5.5 | 26 | 8.5 | 28 | 7.8 | 36 | |||
| 70-79 | 15 | 44 | 13 | 55 | 20 | 57 | 19 | 68 | |||
| ≥80 | 30 | 73 | 28 | 83 | 40 | 85 | 37 | 92 | |||
| Heart disease | |||||||||||
| 20-39 | 0.38 | 0.83 | 0.37 | 0.95 | 0.51 | 1.1 | 0.50 | 1.3 | |||
| 40-49 | 0.78 | 1.7 | 0.77 | 1.9 | 1.0 | 2.3 | 1.0 | 2.6 | |||
| 50-59 | 2.3 | 5.0 | 2.3 | 5.7 | 3.1 | 6.6 | 3.0 | 7.5 | |||
| 60-69 | 5.5 | 12 | 5.4 | 13 | 7.2 | 15 | 7.1 | 17 | |||
| 70-79 | 13 | 26 | 13 | 29 | 17 | 33 | 16 | 36 | |||
| ≥80 | 28 | 52 | 28 | 56 | 36 | 62 | 35 | 67 | |||
Absolute five year risk estimates in percentages are shown for C282Y homozygotes (C282Y/C282Y) and non-carriers for both C282Y and H63D (non-carrier) and stratified according to sex and age categories. Analyses included all 87 313 non-carriers and all 422 C282Y homozygotes.
Sensitivity analyses
As some subgroups in the study had sparse number of events, we did additional sensitivity analyses in which all main analyses on relative risk (fig 1, fig 2, fig 3, fig 5, and fig 6) were instead done using Cox regression with Firth’s penalised maximum likelihood bias reduction method to reduce small sample bias,26 27 which gave results similar to those presented in the main figures (supplementary figures S19-S23). When examining heterogeneity, we found no indication of heterogeneity between the three general population cohorts for the analyses on risk of diabetes (P for heterogeneity=0.61), liver disease (P for heterogeneity=0.12), or heart disease (P for heterogeneity=0.41) (supplementary figure S24).
Discussion
In this prospective study of 132 542 individuals from the general population, we found that HFE C282Y homozygotes had a higher risk of diabetes and liver disease than did non-carriers, whereas risk of heart disease was not increased. Importantly and a novel finding, C282Y homozygotes had increased risk of diabetes even when they had normal levels of transferrin saturation and/or ferritin in a single blood sample obtained at study enrolment. Furthermore, C282Y homozygotes with diabetes had higher mortality from any cause than did non-carriers with diabetes, and as many as 27.3% of all deaths among C282Y homozygotes were potentially attributable to diabetes.
Strengths and weaknesses of study
Among the strengths of our study is the large general population cohort with 132 542 consecutive genotyped individuals, most of whom also had iron, transferrin saturation, and/or ferritin measured. Our comprehensive data on hospital contacts and deaths from the nationwide Danish registries without any loss to follow-up is also a strength.
Our study is limited by not being able to ascertain with certainty whether it is diabetes itself or hypothetically some other unknown associated condition that causes the increased mortality in C282Y homozygotes with diabetes. Additionally, plasma iron, transferrin saturation, and ferritin were measured only at study enrolment, and we do not have data on the levels at the time of diagnosis of diabetes. Therefore, we cannot exclude the possibility that iron overload occurred during the interval between study enrolment and the onset of diabetes. However, this possibility does not change our finding that even C282Y homozygotes with normal transferrin saturation and/or ferritin concentrations in a single blood sample were at increased risk of diabetes, indicating that the strategy outlined in most current guidelines,3 5 13 recommending genotyping for C282Y only in individuals with increased levels of transferrin saturation and/or ferritin, may fail to detect some C282Y homozygotes at increased risk of diabetes. As only 16 C282Y homozygotes with diabetes died during follow-up, we do not have sufficient statistical power to analyse the risk of specific causes of death. However, when excluding individuals with liver or heart disease, we still found increased mortality in C282Y homozygotes with diabetes, meaning that concomitant liver or heart disease is unlikely to explain the increased mortality in C282Y homozygotes with diabetes.
Strengths and weaknesses in relation to other studies
Our novel finding that C282Y homozygotes had an increased risk of diabetes even when they had normal levels of transferrin saturation and/or ferritin is in contrast to the results of an Australian study comparing 102 C282Y homozygotes, who all had serum ferritin <1000 µg/L, and 291 non-carrier controls, as that study did not find an increased risk of diabetes.30 The different results may be explained by greater statistical power in our study including 422 C282Y homozygotes of whom 36 (8.5%) had diabetes and by the very low prevalence of diabetes in the Australian study, as it found self-reported diabetes in only 1% of C282Y homozygotes. A study of 451 243 individuals from the UK Biobank, including 2890 C282Y homozygotes, found increased risk of diabetes among male but not female C282Y homozygotes.10 By contrast, we found no convincing evidence for a difference in risk estimates between male and female C282Y homozygotes. Importantly, however, as the UK study did not have measurements of iron, transferrin saturation, or ferritin, it could not examine whether increased risk of diabetes was also found in C282Y homozygotes with normal levels of iron, transferrin saturation, or ferritin. In contrast to our finding on increased risk of diabetes in C282Y homozygotes, the baseline examination of the HEIRS haemochromatosis population screening study did not find increased prevalence of self-reported diabetes in 299 C282Y homozygotes at study enrolment, when the median age of the study population was 50 years.31 Likewise, a later study did not find an increased risk of having a diagnosis of diabetes in a 17.7 year follow-up analysis including 93 C282Y homozygotes from the HEIRS cohort in combination with 577 patients referred to the hospital with haemochromatosis due to C282Y homozygosity.32 The different findings may hypothetically be due to the different methods used and the different groups of individuals studied. We studied solely individuals from the general population and obtained information on diabetes diagnoses during follow-up from a nationwide registry covering all hospitals in Denmark. In the HEIRS study, results from the baseline examination were from a cross sectional analysis on prevalent self-reported diabetes at study enrolment, and the HEIRS 17.7 year follow-up study on incident diabetes included a combination of 93 C282Y homozygotes identified by population screening and 577 patients referred to the hospital specifically because of haemochromatosis.32 The Norwegian HUNT study involving 65 238 individuals from the general population found no increase in the prevalence of self-reported diabetes at study enrolment among 297 C282Y homozygotes with a median age of approximately 50 years,33 but this was based on performing selective HFE genotyping only in individuals with increased transferrin saturation in repeat blood samples, meaning that the study solely investigated prevalence of diabetes in C282Y homozygotes with persistently increased transferrin saturation.34
Our findings that C282Y homozygotes with diabetes in the general population had higher mortality than non-carriers with diabetes, and that as many as 27.3% of all deaths among C282Y homozygotes were potentially attributable to diabetes, are potentially clinically important. In a previous study from 2014,35 we studied 84 865 individuals from the Copenhagen City Heart Study and the Copenhagen General Population Study and found a tendency towards increased mortality in C282Y homozygotes with diabetes, but the finding was only borderline statistically significant with a P value of 0.05. In the current study with longer follow-up and more individuals from the general population, the increased mortality is now more robustly confirmed with a P value of 0.008. Our findings are partly supported by a study of patients with diabetes referred to a tertiary diabetes centre in Denmark, which found higher mortality among C282Y homozygotes with diabetes than in non-carrier patients.36
Possible explanations and implications for clinicians and policy makers
Several guidelines on testing for haemochromatosis recommend genotyping patients for the C282Y and H63D variants only when both transferrin saturation and ferritin concentrations are increased (except for genotyping close relatives of affected patients).3 5 13 In contrast to these guidelines, we found that even C282Y homozygotes with normal levels of transferrin saturation and/or ferritin had an increased risk of diabetes. Hence, our findings indicate that the recommended strategy for testing may fail to detect some C282Y homozygotes at increased risk of diabetes. Drawing any conclusions from our observational study about the mechanisms that cause increased risk of diabetes in C282Y homozygotes is not possible, but our findings might question the assumption that progressive systemic iron accumulation is the only mechanism causing increased risk of diabetes in C282Y homozygotes,2 3 5 12 37 as the risk of diabetes was increased in C282Y homozygotes with normal levels of transferrin saturation and/or ferritin. Our findings show some alignment with the HEIRS study, which reported increased insulin resistance in C282Y homozygotes when measured as the homoeostasis model assessment estimated insulin resistance (defined as the fasting serum glucose (mg/dL) multiplied by fasting serum insulin (mU/L) divided by 405). In that study, insulin resistance was not associated with serum ferritin concentrations,38 39 indicating that mechanisms unrelated to iron accumulation may cause insulin resistance. When interpreting the results on insulin resistance, we note that the HEIRS study did not find an increased prevalence of diabetes in C282Y homozygotes, implying that the potential clinical consequences of the higher estimates for insulin resistance could not be determined.31 Hypothetically, the high risk of diabetes may be due to decreased concentrations of hepcidin in C282Y homozygotes. Although primarily produced in the liver, hepcidin is synthesised locally in β cells of the pancreas, where it co-localises with insulin within their granules.40 41 Theoretically, if hepcidin had a non-iron mediated effect on either insulin production or insulin sensitivity, this might explain the increased risk of diabetes among C282Y homozygotes regardless of transferrin saturation and/or ferritin levels. Alternatively, the increased risk of diabetes might be due to a linked variant outside the HFE gene but in the same genomic neighbourhood, as the HFE gene maps within the HLA region encompassing the MHC antigens and tumour necrosis factor α. Specifically, the HFE gene is located in the class I region 6p21.3 in linkage disequilibrium with the HLA A and B loci and encodes a protein that binds β-2 microglobulin, as do other MHC class I proteins, but has lost antigen presenting properties.42 43 However, the HFE protein does inhibit antigen specific CD8+ T cell activation,43 and the C282Y variant may be associated with autoimmune diseases,44 suggesting that this variant may predispose to autoimmune forms of diabetes, including latent autoimmune diabetes in adults. However, these mechanisms and any potential therapeutic consequences are speculative and need further investigation.
Unanswered questions and future research
What causes the higher mortality in C282Y homozygotes with diabetes compared with non-carriers with diabetes is unknown, and further studies are needed to clarify whether mortality can be reduced by intensified treatment of diabetes and/or specific antidiabetic drugs. When considering potential drug candidates for clinical trials in C282Y homozygotes with diabetes, it should be noted that current antidiabetic therapies have different effects on iron metabolism. Whereas insulin stimulates iron uptake,45 metformin stimulates iron export, inhibits iron overload and ferroptosis in liver cells, and prevents non-alcoholic fatty liver disease in rats.46 Interestingly, a register based observational study found that use of glucagon-like peptide (GLP)-1 receptor agonists were associated with lower ferritin concentrations in patients with hereditary haemochromatosis and type 2 diabetes.47 GLP-1 receptor agonists also inhibit β cell iron uptake, improve insulin secretory function,48 and reduce hepatic iron accumulation and ferroptosis in obese animals.49 As an example from a different group of patients with diabetes with increased mortality, a recent randomised trial found that treatment with a GLP-1 receptor agonist reduced mortality from any cause in patients with diabetes and chronic kidney disease.50 Similar clinical trials focusing on C282Y homozygotes with diabetes may be relevant to determine whether specific interventions can reduce mortality in C282Y homozygotes with diabetes.
Detection and treatment of diabetes are not included in most clinical guidelines on hereditary haemochromatosis3 5 12 37; however, our finding that as many as 27.3% of all deaths among C282Y homozygotes were potentially attributable to diabetes and 14.4% were attributable to liver disease suggests that recommending screening for diabetes in C282Y homozygotes and giving greater priority to treatment of diabetes in future clinical guidelines may be relevant. Although our absolute risk estimates can give an indication of the extent of increased morbidity and mortality related to diabetes, our observational study is not able to test whether earlier detection of C282Y homozygosity—for example, through genetic population screening—would be beneficial. Additionally, the role of plasma iron, transferrin saturation, and ferritin in predicting which C282Y homozygotes may be at increased risk of other haemochromatosis associated conditions, such as osteoarthritis and dementia,10 11 51 52 remains unclear and requires further research.
Conclusion
Haemochromatosis C282Y homozygotes had an increased risk of diabetes even when they had normal levels of transferrin saturation and/or ferritin. Furthermore, C282Y homozygotes with diabetes had higher mortality than non-carriers with diabetes, and 27.3% of all deaths among C282Y homozygotes were potentially attributable to diabetes.
What is already known on this topic
Most clinical guidelines on hereditary haemochromatosis assume that the increased risk of liver disease and diabetes in C282Y homozygotes is caused by iron accumulation in hepatocytes and pancreatic islets
Therapeutic phlebotomy has been the cornerstone in treating hereditary haemochromatosis, with clinical effect typically assessed by measuring decreases in transferrin saturation and ferritin
Randomised trials on the risk of complications and death in individuals with hereditary haemochromatosis treated with phlebotomy have not been performed
What this study adds
This study found an increased risk of diabetes in C282Y homozygous individuals with normal levels of transferrin saturation and/or ferritin, who are not currently recommended for haemochromatosis genotyping
C282Y homozygotes with diabetes had a higher risk of death from any cause than did non-carriers with diabetes, and 27.3% of all deaths among C282Y homozygotes were potentially attributable to diabetes
Prioritising detection and treatment of diabetes for C282Y homozygotes in future clinical guidelines on hereditary haemochromatosis may be relevant
Acknowledgments
We acknowledge and thank the staff and participants in the three studies of the general population: the Copenhagen City Heart Study, the Copenhagen General Population Study, and the Danish General Suburban Population Study.
Web extra.
Extra material supplied by authors
Web appendix: Supplementary materials
Contributors: MM, JH, and AG conceptualised and designed the study. MM, JH, BGN, CE, and SEB collected data and assembled databases. MM, JH, BGN, CE, TMP, JP, SEB, and AG analysed and interpreted data. MM, JH, and AG wrote the original draft of the manuscript. MM, JH, BGN, CE, TMP, JP, SEB, and AG revised and approved the final manuscript. AG is the guarantor. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.
Funding: Research grants from the Capital Region of Denmark and the Independent Research Fund Denmark made this study possible. Funding from the Danish Heart Foundation and Copenhagen University Hospital - Herlev and Gentofte supports the Copenhagen General Population Study and the Copenhagen City Heart Study. CE is partly funded by the Laboratory Medicine Endowment Fund of Boston Children’s Hospital. Study sponsors had no influence on study design; collection, analysis, or interpretation of data; writing the report; or the decision to submit the paper for publication. The researchers acted independently from the study sponsors in all aspects of this study.
Competing interests: All authors have completed the ICMJE uniform disclosure form at https://www.icmje.org/disclosure-of-interest/ and declare: support from the Capital Region of Denmark and the Independent Research Fund Denmark for the submitted work; AG has received research funding from Novo Nordisk and Sanofi, payment for consultancy work from Pharmacosmos, and payment for consultancy work and advisory board participation from Novo Nordisk without direct relation to the submitted work; JH has received research funding from Sanofi; TMP has stock ownership in Novo Nordisk; no other relationships or activities that could appear to have influenced the submitted work.
Transparency: The lead author (the manuscript’s guarantor) affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.
Dissemination to participants and related patient and public communities: To communicate key findings, we will use social media to disseminate our findings. Additionally, we will engage with patient advocacy groups to ensure effective dissemination to those most affected by haemochromatosis.
Provenance and peer review: Not commissioned; externally peer reviewed.
Ethics statements
Ethical approval
The studies were approved by the Danish ethical committees (KF-01-144/01, H-KF-01-144/01, and SJ-114). Participants gave written informed consent.
Data availability statement
Statistical code or technical details can be made available from the corresponding author at andreas.glenthoej@regionh.dk. To comply with data privacy regulations, access to original data is possible only in case of collaborative agreement.
References
- 1. Zoller H, Schaefer B, Vanclooster A, et al. European Association for the Study of the Liver . EASL Clinical Practice Guidelines on haemochromatosis. J Hepatol 2022;77:479-502. 10.1016/j.jhep.2022.03.033 [DOI] [PubMed] [Google Scholar]
- 2. Adams PC, Jeffrey G, Ryan J. Haemochromatosis. Lancet 2023;401:1811-21. 10.1016/S0140-6736(23)00287-8 [DOI] [PubMed] [Google Scholar]
- 3. Girelli D, Busti F, Brissot P, Cabantchik I, Muckenthaler MU, Porto G. Hemochromatosis classification: update and recommendations by the BIOIRON Society. Blood 2022;139:3018-29. 10.1182/blood.2021011338 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Merryweather-Clarke AT, Pointon JJ, Jouanolle AM, Rochette J, Robson KJ. Geography of HFE C282Y and H63D mutations. Genet Test 2000;4:183-98. 10.1089/10906570050114902 [DOI] [PubMed] [Google Scholar]
- 5. Fitzsimons EJ, Cullis JO, Thomas DW, Tsochatzis E, Griffiths WJH, British Society for Haematology . Diagnosis and therapy of genetic haemochromatosis (review and 2017 update). Br J Haematol 2018;181:293-303. 10.1111/bjh.15164 [DOI] [PubMed] [Google Scholar]
- 6. Zoller H, Henninger B. Pathogenesis, Diagnosis and Treatment of Hemochromatosis. Dig Dis 2016;34:364-73. 10.1159/000444549 [DOI] [PubMed] [Google Scholar]
- 7. Allen KJ, Gurrin LC, Constantine CC, et al. Iron-overload-related disease in HFE hereditary hemochromatosis. N Engl J Med 2008;358:221-30. 10.1056/NEJMoa073286 [DOI] [PubMed] [Google Scholar]
- 8. Grosse SD, Morris JM, Khoury MJ. Disease-related conditions in relatives of patients with hemochromatosis. N Engl J Med 2001;344:1477-8. 10.1056/NEJM200105103441914 [DOI] [PubMed] [Google Scholar]
- 9. Henriksen LF, Petri A-S, Hasselbalch HC, Kanters JK, Ellervik C. Increased iron stores prolong the QT interval - a general population study including 20 261 individuals and meta-analysis of thalassaemia major. Br J Haematol 2016;174:776-85. 10.1111/bjh.14099 [DOI] [PubMed] [Google Scholar]
- 10. Pilling LC, Tamosauskaite J, Jones G, et al. Common conditions associated with hereditary haemochromatosis genetic variants: cohort study in UK Biobank. BMJ 2019;364:k5222. 10.1136/bmj.k5222 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Lucas MR, Atkins JL, Pilling LC, Shearman JD, Melzer D. HFE genotypes, haemochromatosis diagnosis and clinical outcomes at age 80 years: a prospective cohort study in the UK Biobank. BMJ Open 2024;14:e081926. 10.1136/bmjopen-2023-081926 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Crownover BK, Covey CJ. Hereditary hemochromatosis. Am Fam Physician 2013;87:183-90. [PubMed] [Google Scholar]
- 13.Schiødt FV, Junker A, Magnussen K, et al. National guidelines on haemochromatosis from the Danish Society of Gastroenterology and Hepatology (HFE-Hæmokromatose. Udredning, diagnostik og behandling). 2017. https://dsgh.dk/wp-content/uploads/2022/06/HFE-Haemokromatose.pdf.
- 14. Jørgensen AB, Frikke-Schmidt R, Nordestgaard BG, Tybjærg-Hansen A. Loss-of-function mutations in APOC3 and risk of ischemic vascular disease. N Engl J Med 2014;371:32-41. 10.1056/NEJMoa1308027 [DOI] [PubMed] [Google Scholar]
- 15. Mortensen MB, Nordestgaard BG. Elevated LDL cholesterol and increased risk of myocardial infarction and atherosclerotic cardiovascular disease in individuals aged 70-100 years: a contemporary primary prevention cohort. Lancet 2020;396:1644-52. 10.1016/S0140-6736(20)32233-9 [DOI] [PubMed] [Google Scholar]
- 16. Bergholdt HKM, Bathum L, Kvetny J, et al. Study design, participation and characteristics of the Danish General Suburban Population Study. Dan Med J 2013;60:A4693. [PubMed] [Google Scholar]
- 17. Moen IW, Bergholdt HKM, Mandrup-Poulsen T, Nordestgaard BG, Ellervik C. Increased Plasma Ferritin Concentration and Low-Grade Inflammation-A Mendelian Randomization Study. Clin Chem 2018;64:374-85. 10.1373/clinchem.2017.276055 [DOI] [PubMed] [Google Scholar]
- 18. Mottelson M, Glenthøj A, Nordestgaard BG, et al. Iron, hemochromatosis genotypes, and risk of infections: a cohort study of 142 188 general population individuals. Blood 2024;144:693-707. 10.1182/blood.2023022235 [DOI] [PubMed] [Google Scholar]
- 19. Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient Register. Scand J Public Health 2011;39(Suppl):30-3. 10.1177/1403494811401482 [DOI] [PubMed] [Google Scholar]
- 20. Thygesen SK, Christiansen CF, Christensen S, Lash TL, Sørensen HT. The predictive value of ICD-10 diagnostic coding used to assess Charlson comorbidity index conditions in the population-based Danish National Registry of Patients. BMC Med Res Methodol 2011;11:83. 10.1186/1471-2288-11-83 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Milman NT, Schioedt FV, Junker AE, Magnussen K. Diagnosis and Treatment of Genetic HFE-Hemochromatosis: The Danish Aspect. Gastroenterology Res 2019;12:221-32. 10.14740/gr1206 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Milman NT, Schiødt FV, Junker AE, Magnussen K, Nathan T, Sandahl TD. [Genetic HFE-haemochromatosis]. Ugeskr Laeger 2018;180:V09180619. [PubMed] [Google Scholar]
- 23. Bedogni G, Bellentani S, Miglioli L, et al. The Fatty Liver Index: a simple and accurate predictor of hepatic steatosis in the general population. BMC Gastroenterol 2006;6:33. 10.1186/1471-230X-6-33 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Busquets-Cortés C, Bennasar-Veny M, López-González A-A, Fresneda S, Aguiló A, Yanez A. Fatty liver index and progression to type 2 diabetes: a 5-year longitudinal study in Spanish workers with pre-diabetes. BMJ Open 2021;11:e045498. 10.1136/bmjopen-2020-045498 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Pedersen CB, Gøtzsche H, Møller JO, Mortensen PB. The Danish Civil Registration System. A cohort of eight million persons. Dan Med Bull 2006;53:441-9. [PubMed] [Google Scholar]
- 26. Heinze G, Schemper M. A solution to the problem of monotone likelihood in Cox regression. Biometrics 2001;57:114-9. 10.1111/j.0006-341X.2001.00114.x. [DOI] [PubMed] [Google Scholar]
- 27.Heinze G, Ploner M, Jiricka L, et al. coxphf: Cox Regression with Firth’s Penalized Likelihood. 2023. https://CRAN.R-project.org/package=coxphf.
- 28. Austin PC. A Tutorial on Multilevel Survival Analysis: Methods, Models and Applications. Int Stat Rev 2017;85:185-203. 10.1111/insr.12214 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. de Jong VMT, Moons KGM, Riley RD, et al. Individual participant data meta-analysis of intervention studies with time-to-event outcomes: A review of the methodology and an applied example. Res Synth Methods 2020;11:148-68. 10.1002/jrsm.1384 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Allen KJ, Bertalli NA, Osborne NJ, et al. HealthIron Study Investigators . HFE Cys282Tyr homozygotes with serum ferritin concentrations below 1000 microg/L are at low risk of hemochromatosis. Hepatology 2010;52:925-33. 10.1002/hep.23786 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Adams PC, Reboussin DM, Barton JC, et al. Hemochromatosis and Iron Overload Screening (HEIRS) Study Research Investigators . Hemochromatosis and iron-overload screening in a racially diverse population. N Engl J Med 2005;352:1769-78. 10.1056/NEJMoa041534 [DOI] [PubMed] [Google Scholar]
- 32. Adams PC, Richard L, Weir M, Speechley M. Survival and development of health conditions after iron depletion therapy in C282Y-linked hemochromatosis patients. Can Liver J 2021;4:381-90. 10.3138/canlivj-2021-0016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Åsberg A, Hveem K, Krüger O, Bjerve KS. Persons with screening-detected haemochromatosis: as healthy as the general population? Scand J Gastroenterol 2002;37:719-24. 10.1080/00365520212510 [DOI] [PubMed] [Google Scholar]
- 34. Åsberg A, Hveem K, Thorstensen K, et al. Screening for hemochromatosis: high prevalence and low morbidity in an unselected population of 65,238 persons. Scand J Gastroenterol 2001;36:1108-15. 10.1080/003655201750422747 [DOI] [PubMed] [Google Scholar]
- 35. Ellervik C, Mandrup-Poulsen T, Tybjærg-Hansen A, Nordestgaard BG. Total and cause-specific mortality by elevated transferrin saturation and hemochromatosis genotype in individuals with diabetes: two general population studies. Diabetes Care 2014;37:444-52. 10.2337/dc13-1198 [DOI] [PubMed] [Google Scholar]
- 36. Ellervik C, Andersen HU, Tybjærg-Hansen A, et al. Total mortality by elevated transferrin saturation in patients with diabetes. Diabetes Care 2013;36:2646-54. 10.2337/dc12-2032 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Bacon BR, Adams PC, Kowdley KV, Powell LW, Tavill AS, American Association for the Study of Liver Diseases . Diagnosis and management of hemochromatosis: 2011 practice guideline by the American Association for the Study of Liver Diseases. Hepatology 2011;54:328-43. 10.1002/hep.24330 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Acton RT, Barton JC, Barton JC. Serum ferritin, insulin resistance, and metabolic syndrome: clinical and laboratory associations in 769 non-hispanic whites without diabetes mellitus in the HEIRS study. Metab Syndr Relat Disord 2015;13:57-63. 10.1089/met.2014.0106 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Barton JC, Barton JC, Adams PC, Acton RT. Risk Factors for Insulin Resistance, Metabolic Syndrome, and Diabetes in 248 HFE C282Y Homozygotes Identified by Population Screening in the HEIRS Study. Metab Syndr Relat Disord 2016;14:94-101. 10.1089/met.2015.0123 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Kulaksiz H, Fein E, Redecker P, Stremmel W, Adler G, Cetin Y. Pancreatic beta-cells express hepcidin, an iron-uptake regulatory peptide. J Endocrinol 2008;197:241-9. 10.1677/JOE-07-0528 [DOI] [PubMed] [Google Scholar]
- 41. Barton JC, Acton RT. Diabetes in HFE Hemochromatosis. J Diabetes Res 2017;2017:9826930. 10.1155/2017/9826930 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. de Campos WN, Massaro JD, Cançado ELR, et al. Comprehensive analysis of HFE gene in hereditary hemochromatosis and in diseases associated with acquired iron overload. World J Hepatol 2019;11:186-98. 10.4254/wjh.v11.i2.186 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Reuben A, Phénix M, Santos MM, Lapointe R. The WT hemochromatosis protein HFE inhibits CD8+ T-lymphocyte activation. Eur J Immunol 2014;44:1604-14. 10.1002/eji.201343955 [DOI] [PubMed] [Google Scholar]
- 44. Barton JC, Barton JC. Autoimmune Conditions in 235 Hemochromatosis Probands with HFE C282Y Homozygosity and Their First-Degree Relatives. J Immunol Res 2015;2015:453046. 10.1155/2015/453046 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Biswas S, Tapryal N, Mukherjee R, Kumar R, Mukhopadhyay CK. Insulin promotes iron uptake in human hepatic cell by regulating transferrin receptor-1 transcription mediated by hypoxia inducible factor-1. Biochim Biophys Acta 2013;1832:293-301. 10.1016/j.bbadis.2012.11.003 [DOI] [PubMed] [Google Scholar]
- 46. Yue F, Shi Y, Wu S, et al. Metformin alleviates hepatic iron overload and ferroptosis through AMPK-ferroportin pathway in HFD-induced NAFLD. iScience 2023;26:108560. 10.1016/j.isci.2023.108560 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Bain SC, Carstensen B, Hyveled L, et al. Glucagon-like peptide-1 receptor agonist use is associated with lower blood ferritin levels in people with type 2 diabetes and hemochromatosis: a nationwide register-based study. BMJ Open Diabetes Res Care 2023;11:e003300. 10.1136/bmjdrc-2022-003300 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Danielpur L, Sohn Y-S, Karmi O, et al. GLP-1-RA Corrects Mitochondrial Labile Iron Accumulation and Improves β-Cell Function in Type 2 Wolfram Syndrome. J Clin Endocrinol Metab 2016;101:3592-9. 10.1210/jc.2016-2240 [DOI] [PubMed] [Google Scholar]
- 49. Song J-X, An J-R, Chen Q, et al. Liraglutide attenuates hepatic iron levels and ferroptosis in db/db mice. Bioengineered 2022;13:8334-48. 10.1080/21655979.2022.2051858 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Perkovic V, Tuttle KR, Rossing P, et al. FLOW Trial Committees and Investigators . Effects of Semaglutide on Chronic Kidney Disease in Patients with Type 2 Diabetes. N Engl J Med 2024;391:109-21. 10.1056/NEJMoa2403347 [DOI] [PubMed] [Google Scholar]
- 51. Chehade S, Adams PC. Severe Hemochromatosis Arthropathy in the Absence of Iron Overload. Hepatology 2019;70:1064-5. 10.1002/hep.30469 [DOI] [PubMed] [Google Scholar]
- 52. Atkins JL, Pilling LC, Heales CJ, et al. Hemochromatosis Mutations, Brain Iron Imaging, and Dementia in the UK Biobank Cohort. J Alzheimers Dis 2021;79:1203-11. 10.3233/JAD-201080 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Web appendix: Supplementary materials
Data Availability Statement
Statistical code or technical details can be made available from the corresponding author at andreas.glenthoej@regionh.dk. To comply with data privacy regulations, access to original data is possible only in case of collaborative agreement.
