Morbidity, risk of cancer and mortality in 3645 HFE mutations carriers

Mutations in the HFE gene can lead to hereditary haemochromatosis (HH) and have been suggested to increase the risk of extra‐hepatic diseases, especially breast and colorectal cancer. Here we investigated long‐term outcomes of Swedish patients with HFE mutations.


| INTRODUC TI ON
Mutations in the HFE gene are linked to hereditary haemochromatosis (HH) type 1, which is the most common genetic disease in individuals of European descent. 1,2 The two most important HFE mutations are p.Cys282Tyr (p.C282Y) and p.His63Asp (p.H63D), found in the heterozygous form in 4%-12% and 10%-18% respectively, in the European population. 1,3 HFE encodes a cell surface protein that regulates hepcidin expression and iron absorption. 4 Homozygosity for p.C282Y results in disrupted HFE signalling, hepcidin deficiency and increased iron release from enterocytes and macrophages, which may lead to the HH phenotype. 4 Persons that are compound heterozygous for p.C282Y and p.H63D typically have a milder phenotype.
A homozygous p.C282Y mutation in HFE is found in 80%-90% of patients with HH type 1, whereas only about 5% are compound heterozygotes. 2 However, the homozygous p.C282Y genotype is present in approximately 0.5% of the general population with European descent, and only a minority of these develop phenotypic HH. [5][6][7][8][9][10] In different population studies, the penetrance of clinical HH in p.C282Y homozygotes varies from 1% to 24%, with the highest numbers in men. [5][6][7]10,11 Previous cohort studies have linked HH to an increased risk for liver cirrhosis and hepatocellular carcinoma (HCC), osteoarthritis, diabetes mellitus and extrahepatic malignancies, mainly colorectal and breast cancer. 6,[12][13][14][15][16][17][18][19] After the introduction of HFE genotyping in 1996, 20 similar cohort studies have been performed on HFE mutation carriers with or without clinical HH, but the results regarding the risk of extrahepatic malignancies are conflicting, which could possibly be attributed to differences in why the included populations were sampled for HFE (e.g. asymptomatic blood donors vs. targeted testing in symptomatic individuals). 16,17,[21][22][23][24][25][26] Associations of HFE mutations have also been made with non-malignant diseases such as type 2 diabetes (T2D), 8 Parkinson's disease 27,28 and osteoarthritis. 29,30 However, most previous investigations on especially cancer risk in HFE carriers have been conducted in case-control studies, and studies of long-term outcomes associated with mutations in HFE are scarce. Recently, a large population-based study presented data from the UK biobank. 11 The authors found an increased risk for incident liver cirrhosis, HCC, diabetes mellitus and osteoarthritis in a large cohort of individuals homozygous for the p.C282Y mutation during a mean follow-up of 7 years. However, the risk of overall mortality was not increased compared to other participants in the UK biobank, and the risk of extra-hepatic malignancies was not reported. 11 In the present study we aimed to examine long-term outcomes in a large cohort of individuals with homozygous or compound heterozygous HFE mutations in Sweden and compare disease risk to reference individuals from the general population.

| MATERIAL AND ME THODS
This was a multi-centre cohort study with retrospectively assembled data through the Swedish hepatology network (SweHep, www.swehep.se), consisting of clinical researchers from all eight Swedish university hospitals. In 2018, we contacted all laboratory departments reporting to these hospitals and obtained data on all persons who had been tested for HFE mutations and found to have a homozygous HFE p.C282Y mutation or a compound heterozygote mutation in p.C282Y and p.H63D between 1 January 1997 and 31 December 2017. No information on why the HFE-test was performed was available, and we did not obtain data on those that tested negatively for HFE. Study baseline was defined as the date of HFE testing.

| Variables
We recorded the following variables at baseline: Age, sex, type of HFE mutation (homozygous for p.C282Y or compound heterozygote for both p.C282Y and p.H63D), ferritin (measured in μg/L) and alanine aminotransferase (ALT, measured in μkat/L). Data on ALT and ferritin was not automatically reported by the laboratories but was instead recorded from medical charts separately by one of the coauthors where available within 1 year before HFE testing. We did not record ALT and ferritin obtained after HFE diagnosis, as initiated treatment (phlebotomy) might have impacted such results. diabetes, osteoarthritis and death. Excess mortality was only seen in men. No increased risk was seen for colorectal or breast cancer. Liver-related outcomes were rare, with a cumulative incidence of <1%.
Conclusions: Individuals found to be HFE mutation carriers in a university hospital setting had an increased risk for mortality in men, along with increased risks of cirrhosis, HCC, diabetes type 2, and osteoarthritis. In general, the absolute risk for adverse outcomes was low and no increased risk for colon or breast cancer was observed.

| Follow-up
For each person with a HFE mutation, up to 10 reference individuals were obtained from the general population as recorded in the Total Population Register. 31 Reference individuals were matched on sex, age and county of residence at the time of HFE testing. Data on HFE genotype, ALT and ferritin were not available in the reference population. We excluded any person missing a Swedish personal identity number (six persons with HFE mutations and no reference individuals) or who had undergone a liver transplantation before baseline (four persons with HFE mutations and five reference individuals).
Removal of a person with HFE mutations also meant removal of that person's respective reference individuals.
The combined cohort was linked to national, population-based registers. We used outcome data from the National Patient Register,

| Outcomes
We considered the following outcomes: a diagnosis of HH, cirrhosis, HCC, breast cancer, colorectal cancer, type 1 and 2 diabetes, Parkinson's disease, osteoarthritis and hypothyroidism.
Administrative codes used to define outcomes are listed in Table   S1. Cancer outcomes were defined as the first occurrence of breast cancer, colorectal cancer or HCC in either the Patient Register, the Causes of Death register or the Cancer Register. We allowed persons to have multiple outcomes, e.g. a person could first have a breast cancer and later a colorectal cancer. These outcomes were available for the full study duration and included the full cohort.
Parkinson's disease and osteoarthritis were defined as a corresponding ICD-code (Table S1) in the Inpatient or Outpatient Register, or the Causes of Death Register. Cirrhosis was defined as a composite outcome, including the first diagnosis of either cirrhosis or a complication of cirrhosis, including ascites, oesophageal varices, hepatic encephalopathy or hepatorenal syndrome (definitions in Table S1). For these outcomes, we began follow-up at 1 January 2001 and excluded from all analyses study persons whose HFE was performed before that date in order to be able to use the Outpatient Register.
Type 1 and 2 diabetes and hypothyroidism are diseases that do not necessarily require inpatient care, or a specialist visit and are thus not captured by the Inpatient or Outpatients Registers, but most of these patients are treated with specific pharmacotherapies. Therefore, we defined these diseases as the first time a prescribed drug corresponding to the relevant disease was collected from a pharmacy by the patient. We used ATC codes in the prescribed drug register to define these outcomes as follows: For type 1 diabetes, we required a prescription for insulin (A10A, including subheadings) and age <40 years at the time of first prescription, and no code for other antidiabetic medications (A10B, including subheadings).
For T2D, we required no ICD diagnosis of type 1 diabetes at or prior to baseline, and an ATC code for other antidiabetics (A10B), or only a code for insulin (A10A) and age ≥40 years at the time of the first prescription. Hypothyroidism was defined as an ATC code for levothyroxine (H03AA01). For these outcomes, we excluded study persons with a baseline prior to 30  We evaluated prevalent and incident outcome diagnoses separately. Prevalent diagnoses were defined as having the diagnosis in question at or before the time of the HFE test, while incident outcomes were defined as receiving the diagnosis any time after the HFE test. This was done as the diagnosis in question might impact the reason for HFE testing, e.g. a person with cirrhosis is likely to be tested for HFE as part of the routine evaluation if other tests are negative. Persons with an outcome diagnosis present at or before baseline were excluded from analyses studying incident events for that particular outcome. We used the main cause of contact with the healthcare system to define outcomes, this was done to reduce the risk for differential misclassification.

| Sensitivity analyses
Analyses were performed in several subgroups. First, we tested the hypothesis that the risk for incident malignancies compared to the matched reference population would be higher in persons with a homozygous p.C282Y mutation than in persons with a compound heterozygote p.C282Y/p.H63D mutation.
Second, we investigated risks stratified on gender as men are known to have a higher risk of developing phenotypic HH.
Third, in the subpopulations with available data on ALT or ferritin, we investigated if these biomarkers were associated with increased risk of the studied outcomes. ALT and ferritin were analysed as continuous parameters. These analyses were not performed on reference individuals since data on ALT or ferritin were not available.
Fourth, to further reduce the risk that prevalent diagnoses at baseline were included in the analysis investigating incident outcomes, we started follow-up in 2006, thus having a period of up to 5 years to better exclude any prevalent diagnoses at baseline using the outpatient register.
Fifth, as prevalent diagnoses might be the reason for why HFE testing was performed, for instance in the workup of a patient with newly diagnosed cirrhosis, we performed an analysis excluding those with the outcome-defining diagnosis made within 90 days prior to baseline, only counting diagnoses made before that timeframe.
Finally, we investigated if the risk for incident outcomes differed depending on if HH was diagnosed or not. For this analysis, we assumed that a person with a HFE mutation that had led to phenotypical HH at the time of testing would receive a HH diagnosis at least 1 year after baseline. We thus stratified the cohort by presence of a HH diagnosis up to 1 year after baseline and examined the risk of incident outcomes compared to reference individuals after that period. This analysis excluded any persons that died or were diagnosed with the outcome under study during that year.

| Statistical analysis
Descriptive statistics are presented as medians and interquartile ranges for continuous data and as total numbers and percentages for categorical data. We used logistic regression to obtain odds ratios (OR) for prevalent disease at or before baseline, comparing persons with HFE mutations to reference individuals.
We used Cox proportional hazards regression to obtain hazard ratios (HR) as measures of relative risk in studying the association between the incident outcomes and the exposure variables.

| Ethical considerations
This study was approved by the regional ethics committee in Stockholm (reg no 2015/1731-31/4). Informed consent was waived by the committee due to the retrospective data collection process.

| Prevalent diagnoses at HFE testing
Diagnoses present at or before baseline in persons with HFE mutations and reference individuals with corresponding ORs are presented in Table 2

| Incident outcomes following a positive HFE test
Incident outcomes for the full cohort together with incidence rates per 1000 person-years of follow-up and adjusted HRs are presented in

| Incident outcomes in homozygous vs. compound heterozygous mutations
Incident outcomes together with incidence rates per 1000 personyears of follow-up and adjusted HRs are presented stratified by mutation type in Table S2a

| Impact of ALT and ferritin on incident outcomes
A higher ALT was associated with an increased risk for T2D (aHR  Table S4.

| Starting follow-up in 2006
The analysis with an extended wash-out period found similar risks as in the main analysis (Table S5).

| Excluding prevalent diagnoses up to 90 days prior to baseline
The ORs for these sensitivity analyses were generally reduced com-

| D ISCUSS I ON
In this large cohort study, we could confirm the findings from a recent population-based study, linking the homozygous p.C2828Y or compound heterozygous p.C2828Y/p.H63D genotypes to increased risk of cirrhosis, HCC, T2D and osteoarthritis, 11 but did not detect any increased risk for colorectal or breast cancer. In absolute terms, however, liver-related outcomes were rare, affecting only around 1% of the cohort. Men with HFE mutations had an increased mortality compared to reference individuals, while no increased risk was found in women. Our finding of no significant increased risk for colorectal or breast cancer in HFE mutations carriers is in contrast to previous studies. 12,[16][17][18][19]23,24 Likewise, the risk of developing Parkinson's disease was not altered in our main analysis, in contrast to other findings describing a decreased risk. 27,28 This discrepancy could be due to the size of this study together with systematic ascertainment of outcomes from national registers.
Our risk estimates were modified by mutation type and gender for some outcomes, which might be due to lower power of each analysis but could also have biological causes. For instance, the elevated risk of overall mortality was not found in women, which is biologically plausible since women develop symptomatic HH at a later age compared to men. These results can be used to inform care of patients with HFE mutations regarding the future risk of the studied outcomes.
In an analysis of participants in the UK biobank, of 2890 persons found to be homozygous for p.C282Y, 22% of men and 10% of women developed HH after 7 years of follow-up. 11 This is lower compared to our findings, where 53% of men and 46% of women not being diagnosed with HH at baseline were diagnosed with incident HH after a similar follow-up of 7 years. This high penetrance of HH could possibly be due to examining a slightly different population, since patients in our study were likely tested as part of a liver work-up or a suspicion of HH, while participants in the UK biobank, by contrast, could be more healthy than the general population since they volunteered to participate. 37 The risk of overall mortality was slightly increased in our study as compared to non-significant after adjustments in the UK biobank study, and we found that the risk for excess mortality at around 7 years of follow-up was restricted to men. With respect to HH, diabetes, osteoarthritis, cirrhosis and HCC, our findings of an increased risk are comparable to those of the UK biobank study, that did not examine the risk of colorectal and breast cancer.
Our study is one of the largest cohorts on the long-term out- A main limitation of this study is that we do not know the precise reason for HFE testing. This could be part of a work-up for suspected liver disease or HH, or family screening of HH patients. Indeed, the data suggests that for cirrhosis, HCC and osteoarthritis, symptomatic disease was likely to be the reason for the HFE testing as in many cases the HFE testing was performed within 90 days of the diagnosis in question. The high penetrance of a HH diagnosis in this cohort could possibly also be attributed to the level of care given to these patients, although we cannot know if the HH diagnosis is correct in all cases. Further, we lack detailed data on the magnitude of iron overload, in which cases where treatment with phlebotomy was initiated, and if this affected the outcomes.
Taken together, our results should be generalizable to HFE carriers diagnosed at a secondary or tertiary level of care and thereby relevant for clinicians working in such hospitals. Multiple comparisons across subgroups increases the risk of type 1 and type 2 errors, which is why our results from the subgroup analyses should be interpreted carefully, particularly for rare outcomes. Other limitations include the lack of serum ALT and ferritin data in around 70% of the cohort, as well as in the reference population. Finally, it is possible that there is residual confounding in parameters such as alcohol consumption between subgroups (perhaps in particular between men and women) that could affect the estimates.

| CON CLUS IONS
In a large cohort of individuals with homozygous or compound heterozygous HFE mutations diagnosed at secondary or tertiary level of care, approximately 50% were diagnosed with hereditary HH after in mean 7.9 years of follow-up. An increased risk of cirrhosis and liver cancer was found, but these outcomes were rare, affecting only 1% of the cohort. The risk of overall mortality was increased also in HFE carriers without a diagnosis of hereditary HH. We found increased risks for development of diabetes type 2 and osteoarthritis, but the risk of colorectal and breast cancer development was not increased.
The findings can be used to inform clinicians about the natural history of patients with HFE mutations and generate hypotheses for future research.

ACK N OWLED G EM ENTS
We acknowledge the work performed by staff at each participating laboratory in providing patient data.

CO N FLI C T O F I NTE R E S T
HH reports consulting fees from Novo Nordisk and Gilead. Research grants from Gilead, Astra Zeneca, Intercept. Advisory Board, Bristol