Hereditary Hemochromatosis Variant Associations with Incident Nonliver Malignancies: 11-Year Follow-up in UK Biobank

Hereditary hemochromatosis is a common genetic disorder of iron overload, predominantly caused by the HFE p.C282Y mutation and to a lesser extent the HFE p.H63D mutation (1). The p.C282Y variant is carried by 10% to 15% of people with northern European ancestries, with approximately 1 in 150 being p.C282Y homozygous (1). Iron is an essential element that aids in cell proliferation and growth, but at elevated levels can be toxic. High iron levels have been linked to cancer due to its contribution to tumor initiation and growth (2, 3). The iron overload associated with hemochromatosis is well established in playing a causal role in the development of liver disease and liver cancer (4, 5). We previously showed in UK Biobank that male p.C282Y homozygotes have a 10-fold increase in the risk of hepatic malignancies compared with those without the mutations (4). Female p.C282Y homozygotes however have shown no significant increase in risk of hepatic malignancies (4) perhaps due to partial protection from severe iron overload through menstrual iron losses (6).

The reports of risk of other types of cancer in individuals with the hemochromatosis-risk mutations is variable, and studies have often not separately examined cancer risk in males and females. A population-based cohort study of almost 7,000 Swedish hemochromatosis patients followed for up to 48 years examined the risk of nonhepatic gastrointestinal cancers compared with the background population. Male and female patients with hereditary hemochromatosis had a significantly increased standardized incidence ratio (SIR) of esophageal squamous cell carcinoma [SIR = 3.2; 95% confidence interval (CI), 1.3–6.6] and colon adenocarcinoma (SIR = 1.4; 95% CI, 1.1–1.9; ref. 7). However, this study consisted of patients with diagnosed hereditary hemochromatosis, with information on HFE mutation status unavailable. The Melbourne Collaborative Cohort Study found that p.C282Y homozygotes had an increased risk of colorectal cancer (HR = 2.28; 95% CI, 1.22–4.25; P = 0.01) and female p.C282Y homozygotes had an increased risk of breast cancer (HR = 2.39; 95% CI, 1.24–4.61; P = 0.01), but male p.C282Y homozygotes were not at increased risk of prostate cancer (8). A study of Finnish patients with cancer showed that frequency of HFE mutations (p.C282Y and p.H63D) did not significantly differ between male patients with prostate cancer and the population-based controls (9). A large case–control study in the United States did not observe a significant association between hemochromatosis and risk of pancreatic cancer in men and women aged over 66 years (10).

Given the uncertainty about cancer risks (outside the liver) in community populations with hemochromatosis-related genotypes, we estimated the risk of several types of cancer by hereditary hemochromatosis genotype status (p.C282Y/p.H63D) compared with those without the mutations, stratified by sex in a large community genotyped sample of participants of European descent. Clarifying these risks is important for mutation carriers to know to monitor their own health.

Participants were from the UK Biobank cohort, a large prospective study of more than 500,000 community volunteers, who were assessed between 2006 and 2010 at baseline in 22 centers throughout England, Scotland, and Wales (11). Participants were aged 40 to 70 years at baseline and assessment included collecting information on demographics, lifestyle variables and disease history, including providing a blood sample for genotyping. UK Biobank gained ethical approval from the North West Multi-Centre Research Ethics Committee (Research Ethics Committee reference 11/NW/0382). At the baseline assessment, participants gave written informed consent for the collection of data, genotyping from blood samples, and linkage to electronic medical records for follow-up.

Data were available for 451,143 UK Biobank participants of European descent, with HFE p.C282Y (rs1800562) and HFE p.H63D (rs1799945) genotype information. European descent was defined as participants who self-reported as being “White European” and who were confirmed genetically as having European ancestry (12). UK Biobank used Affymetrix microarrays (800,000 markers directly genotyped); HFE p.H63D was directly genotyped on the microarray but HFE p.C282Y was imputed by the central UK Biobank team (13). 98.7% of participants were imputed with 100% confidence. 1.26% were recoded (i.e., estimated genotype dose between 0 and 0.25 set to 0, values between 0.75 and 1.25 set to 1, and finally between 1.75 and 2 to 2) and the remaining 0.04% of participants (n = 183) were excluded due to imprecise p.C282Y imputation.

Exome sequence data was available for 49,772 of the 451,143 participants, from the Illumina NovaSeq 6000 platform (75 base pair paired-end sequencing with 20 times coverage of targeted panel; refs. 14, 15). HFE p.C282Y genotype calls were extracted from the data (Functional Equivalence pipeline; refs. 16) and compared with imputed genotype calls, as described above. The genotypes from imputation and sequencing were highly correlated (R² = 0.998); only 0.4% (1/231) of the imputed p.C282Y homozygotes was incorrectly classified (the sequenced genotype was p.C282Y heterozygote).

Prevalent cancer diagnoses were from baseline self-reported questionnaires, plus hospital inpatient data (National Health Service Hospital Episode Statistics) coded with International Classification of Diseases 10^th revision (ICD-10) codes from 1996 to baseline assessment. Incident cancer diagnoses were from baseline assessment (2006 to 2010) to the end of follow-up from hospital inpatient data (to March 2021) and national cancer registries (to July 2019). Information on cancer deaths were obtained from death registrations available until March 2021. We examined any cancer (other than nonmelanoma skin cancer and liver cancer) and 16 common cancer types: bladder, blood (with subanalyses of leukemia and lymphoma), bone, brain, breast, colorectal, kidney, lung, melanoma, esophageal, ovarian, pancreatic, prostate, and stomach (see Supplementary Table S1 for ICD-10 coding used). We did not examine liver cancer as we have previously published data on this outcome (4). Primary care follow-up data were not included for cancer diagnoses as the completeness of case ascertainment in cancer registry data in England is high at around 98% to 99% (17).

Cox proportional hazards regression models were used to test genotype associations with the risk of incident cancer. Each incident cancer outcome excluded participants with any respective baseline prevalent cancer diagnosis. All models were stratified by sex, and adjusted for age, assessment center, genotyping array and 10 genetic principal components generated in European-descent participants, accounting for population genetics substructure. We report significant associations based on simple P values without correction for multiple testing (P < 0.05). In addition, Bonferroni-corrected P values of 0.003 (dividing the 0.05 significance level by 17 cancer types) were considered statistically significant. We estimated life table probabilities of incident cancer outcomes from age 40 to 75 years in 5-year bands by sex and HFE genotypes, applying observed incidence rates in each age group to a notional cohort, estimating cumulative incidence proportions from age 40 to 75 years. All analyses were performed in Stata version 15.1

Data are available on application to the UK Biobank (https://www.ukbiobank.ac.uk/enable-your-research/register).

We included 451,143 participants of European descent in analyses. The mean age of participants was 56.8 years (SD 8.0; range, 40 to 70 years) and 54.3% were female. Two thousand eight hundred and ninety (0.6%) of the cohort were p.C282Y homozygotes. At baseline, 12.1% of male and 3.4% of female p.C282Y homozygotes were diagnosed with hereditary hemochromatosis. Participants were followed-up from baseline for incident cancer diagnoses for a mean period of 11.6 years (maximum 15.0 years). Nonmelanoma skin cancer was not included in our analyses.

In males, 15.9% of p.C282Y homozygotes had an incident diagnosis of any cancer (other than nonmelanoma and liver cancer) compared with 13.3% in those without mutations. In males, prostate cancer was the most common cancer in both p.C282Y homozygotes (6.8%) and in those with no mutations (5.4%; Table 1A). In females, 9.8% of p.C282Y homozygotes had an incident diagnosis of any cancer compared with 10.2% in those without mutations. In females, breast cancer was the most common cancer in both p.C282Y homozygotes (3.6%) and in those with no mutations (3.9%; Table 1B).

In male p.C282Y homozygotes, there was a significant increase in the risk of any cancer, other than nonmelanoma and liver cancer (HR = 1.22; 95% CI, 1.06–1.40; P = 0.01; n = 194). Male p.C282Y homozygotes were also at a significantly increased risk of prostate cancer (HR = 1.32; 95% CI, 1.07–1.63; P = 0.01; n = 87) compared with those with no mutations (Fig. 1). Excluding prostate cancer from the “any cancer” variable, there was no longer a significantly increased risk in male homozygotes (HR = 1.15; 95% CI, 0.95–1.38; P = 0.15). Information on whether participants had undergone prostate cancer screening at baseline was available for 94.7% of the males in this study (195,345/206,330). Within those with available data, male p.C282Y homozygotes were more likely to have been screened for prostate cancer (33.1%, 406/1,227) compared with those with no mutations (31.2%, 36,266/116,311; OR = 1.18; 95% CI, 1.03–1.34; P = 0.01). However, further sensitivity analysis limited to only UK Biobank study participants who had already been screened at baseline showed that the significantly increased risk of prostate cancer in p.C282Y homozygotes remained (HR = 1.39; 95% CI, 1.03–1.89; P = 0.03). There is not a significantly increased risk in prostate cancer deaths in male p.C282Y homozygotes, but power was limited due to low case numbers. (HR = 0.73; 95% CI, 0.27–1.94; P = 0.53; n = 4). Male p.C282Y heterozygotes also had a borderline significant increase in risk of prostate cancer (HR = 1.06; 95% CI, 1.00–1.13; P = 0.04; n = 1,381). However, following Bonferroni correction for multiple testing, none of the above findings remained significant. For example, the original P for the association between male p.C282Y homozygous status (P = 0.01) became nonsignificant after Bonferroni adjustment (P = 0.17).

There was no excess of any type of cancer in male p.H63D homozygotes or p.C282Y/p.H63D compound heterozygotes. There was also no increased risk of any cancers in female p.C282Y homozygotes, female p.H63D homozygotes female p.C282Y/p.H63D compound heterozygotes. Unexpectedly, male p.C282Y heterozygotes had a significantly decreased risk of blood cancer (and also the subtypes leukemia and lymphoma). Female p.C282Y/p.H63D compound heterozygotes had a significantly decreased risk of blood cancer and pancreatic cancer. Female p.H63D homozygotes had a significantly lower risk of colorectal cancer (Tables 2A and B).

In life table estimates based on observed incidence rates in 5-year age-band from 40 to 75 years, 14.4% (95% CI, 11.7–17.7) of male p.C282Y homozygotes were projected to develop prostate cancer by age 75, compared with 10.7% (95% CI, 10.4–10.9) in those with no mutations: this is an excess proportion of 3.8% (95% CI, 1.3–6.8).

Our study represents the largest study of community genotyped hemochromatosis participants reported to date, with over 10-fold more p.C282Y homozygotes than previously analyzed (18). In this UK Biobank study, male HFE p.C282Y homozygotes had previously reported markedly increased liver cancer rates (4) but we also see a modest 22% increase in “any cancer” (other than nonmelanoma and liver cancer) and a 32% increase in the risk of prostate cancer compared with participants without the mutations. There was no longer a significantly increased risk of “any cancer” in male p.C282Y homozygotes after excluding prostate cancer so the excess of any cancers seems to be due to prostate cancer. In female HFE p.C282Y homozygotes, and in male and female p.C282Y/p.H63D compound heterozygotes and p.H63D homozygotes, there were no significant increases in risk of any types of cancer.

The increased risk of prostate cancer in male p.C282Y homozygotes observed in our study is a novel result which has not been seen previously in other studies. The association was not statistically significant after Bonferroni correction for studying 17 cancer types, so it is possible this was a chance finding and replication is needed in other large community genotyped cohorts. However, Bonferroni correction is known to be conservative and risks missing a real association (19). Prostate cancer was the most common cancer in UK Biobank males and is also the most common cancer in the UK for males accounting for more than one quarter of male cases (27%; ref. 20). We found an excess 3.8% of male p.C282Y homozygotes projected to get prostate cancer by age 75 compared with those without mutations (14.4% compared with 10.7%). In contrast, a study of Finnish cancer patients showed that male p.C282Y homozygotes were not at an increased likelihood of prostate cancer compared with those without the mutations (OR = 1.66; 95% CI, 0.45–6.16; P = 0.45; ref. 9). However, this study comprised of a smaller number of 843 p.C282Y homozygotes, of which only 9 had prostate cancer (9) compared with 87 of 1,294 male p.C282Y homozygotes with prostate cancer in our study. Similarly, the Melbourne Collaborative Cohort Study found no increased risk of prostate cancer in male p.C282Y homozygotes (HR = 0.96; 95% CI, 0.43–2.15; P = 0.92; ref. 8). This study also had a much smaller number of 193 p.C282Y homozygotes (of which only 6 had prostate cancer) but were a very similar age with 99.3% also being aged 40 to 69 years (8). We found that male p.C282Y homozygotes were more likely to be screened for prostate cancer at baseline. However, this did not explain the increased risk of prostate cancer in p.C282Y homozygotes as the significant association remained in sensitivity analysis of only those who had been screened.

Beyond the previously reported increase in liver cancer (4) and the increase in prostate cancer seen here, we found no significant increase in risk of any other type of cancer in male or female p.C282Y homozygotes. We found no association between HFE-genotypes and the risk of pancreatic cancer, despite the previous suggestion that higher serum iron is associated with pancreatic cancer (21). In contrast to our results, a case control study in a Chinese Han population found that with each additional copy of p.H63D allele there was a 21% increase in risk of pancreatic cancer (OR  =  1.21; 95% CI, 1.05–1.39; P =  7.72 × 10⁻³; ref. 22). However, in line with our results a large case–control study in the United States also did not find a significant association between hemochromatosis and risk of pancreatic cancer, in older men and women (10). We also found no association between p.C282Y homozygous status and colorectal cancer, although the Melbourne Collaborative Cohort Study found over a two-fold increase in risk. However, the Melbourne study included just 193 p.C282Y homozygotes compared with 2,890 in this study (8). In male p.C282Y heterozygotes we found a significant but modest decrease in the risk of blood cancers which has not been seen before, and therefore warrants further investigation. In women, we found no association between HFE-genotypes and the risk of breast cancer in our study despite a previous study in Sweden showing a significant positive association between serum iron and breast cancer risk in postmenopausal women (23), and an Australian cohort study which showed a 2.4-fold increase in the risk of breast cancer in female p.C282Y homozygotes (8).

This was a large study of community genotyped participants, with more than 10 times the number of p.C282Y homozygotes compared with the previous largest study (18). The study had objective classification of cancer diagnoses from hospital inpatient data, and also cancer registries which have high levels of case ascertainment completeness (98%–99%; ref. 17). There was a long prospective follow-up period of over a mean of 11 years, in 40- to 70-year–olds. However, the study has some limitations. Firstly, at baseline UK Biobank participants were healthier than the general population (24); however risk estimates are from incident cancer during follow-up so this should be less susceptible to any baseline bias. Secondly, we included participants of European ancestry so results may not be applicable to other populations. Third, data on ferritin concentrations and transferrin saturation were not available in UK Biobank so we do not have information to comment on iron loading specifically. Fourth, although our findings were adjusted for chronologic age, we did not evaluate heterogeneity between individuals in terms of extent or rate of biological aging which may interact with alterations in iron-mediated cellular metabolism associated with specific hereditary hemochromatosis genotypes in determining cancer risk. Finally, information on stage of diagnosis of prostate cancer is not currently available within UK Biobank but it would be interesting to explore this in future work.

In male p.C282Y homozygotes, beyond the known increase in risk of liver cancer, there may be a modest increase in the risk of prostate cancer. No excess risks were found for other studied cancers in male or female p.C282Y homozygotes, or in any other p.C282Y/p.H63D genotype groups of either sex. Further work is needed to replicate these findings in other large community genotyped cohorts.

No disclosures were reported.

J.L. Atkins: Conceptualization, data curation, formal analysis, writing–original draft, project administration, writing–review and editing. L.C. Pilling: Data curation, writing–review and editing. S.V. Torti: Writing–review and editing. F.M. Torti: Writing–review and editing. G.A. Kuchel: Writing–review and editing. D. Melzer: Conceptualization, supervision, writing–review and editing.

This research was funded by an award to D. Melzer by the UK Medical Research Council (MR/S009892/1). J.L. Atkins is supported by an NIHR Advanced Fellowship (NIHR301844). D. Melzer and L.C. Pilling are supported by the University of Exeter (Exeter, UK) Medical School. S.V. Torti and F.M. Torti acknowledge support from the NCI (R01 CA188025, to S.V. Torti and R01 CA233636, to F.M. Torti). G.A. Kuchel acknowledges support from the National Institute on Aging (P30 AG067988; R33 AG061456) as well as the Travelers Chair in Geriatrics and Gerontology. The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. This research was conducted using the UK Biobank resource, under application 14631. We thank the UK Biobank participants and coordinators for the dataset.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.