Background
Hereditary hemochromatosis (HH) is a common autosomal recessive disorder in populations of northern European heritage [
1,
2]. It is characterised by increased iron absorption caused by a defect in the HFE gene. Several mutations have been identified: C282Y, H63D and S56C [
3‐
5]. C282Y homozygosity accounts for 80 to 90 % of people diagnosed with iron-overload, with the other mutations uncommonly associated with iron overload [
6,
7]. It has been hypothesised that HH is most prevalent in northern European populations due to a mutation occurring in Central Europe, hence the description ‘Celtic mutation” [
8]. Prevalence of C282Y homozygosity has been reported to be between 1 in 150 to 200 persons of Northern European ancestry. Amongst populations of different heritage, prevalence is much lower: 1 in 300 Hispanics; 1 in 1,000 Native Americans; 1 in 1,000, 000 Asians [
9‐
13]. Whilst prevalence of other genotypes is more common (1 in 50 C282Y/H63D compound heterozygotes), the burden of disease associated with these mutations is low [
4,
14].
In a proportion of C282Y homozygotes, elevated hepcidin production increases the absorption of dietary iron, which is stored in the parenchymal tissues of the heart, liver and pancreas. If left untreated, iron overload can be a cause of morbidity and mortality, including multiple arthropathies, type 2 diabetes, liver disease and heart disease [
15‐
17]. HH and iron overload is commonly diagnosed by conducting iron studies (transferrin saturation and serum ferritin) with confirmatory genotyping. Treatment consists of regular therapeutic venesection.
Rates of clinical penetrance (i.e. expression of disease) reported in the literature vary, in part due to different definitions. Some authors have defined penetrance as irreversible organ damage, such as cirrhosis or hepatocellular carcinoma, whilst other have included a spectrum of health states, from elevated iron stores and serum iron through to irreversible organ damage. Recent studies have reported rates of cirrhosis of the liver amongst C282Y homozygotes to be between 2 and 6 % [
10,
18,
19]. When penetrance is defined as elevated iron stores and serum iron through to irreversible organ damage, rates of 28.4 % for male and 1.2 % for female C282Y homozygotes have been recently reported [
10].
Whilst diagnosis and prevention of iron overload in genetically susceptible patients is relatively straightforward, the non-specific nature of early symptomatology, in that this can be experienced by people with clinically normal iron levels, contributes to some patients being diagnosed only after irreversible organ damage has occurred [
20‐
23]. Effective treatment is readily available, therefore early diagnosis and timely treatment leads to substantial improvements in patient outcomes. Population screening strategies have been proposed as an approach to increase early identification of people with HH, thereby reducing the potential burden of disease associated with iron overload [
24‐
28].
Whilst HH is a condition that fulfils several of the criteria set out by the World Health Organisation for population screening programs [
29], a lack of robust health economic data has been cited as a hurdle to implementing such a program [
24,
25,
30,
31]. Considerable limitations have been identified in the economic evaluations of HH screening programs that have been published to date [
32,
33].
Cost effectiveness analyses and cost utility analyses give rise to a ratio of the difference in costs and effectiveness between two or more health interventions. The cost of an intervention is measured in monetary units and effectiveness may be measured unidimensionally for cost effectiveness analyses (e.g. life years gained) or by means of a multidimensional instrument (such as the EQ-5D, SF-6, AQOL-4D) for cost utility analyses. Importantly, multi-attribute utility instruments allow for calculation of an individual’s utility (HSUV): a measure of the strength of preference for a particular health state. Utilities are measured on a scale of zero to one, with one representing full health, and zero, death. Some instruments such as the AQOL-4D and the EQ-5D allow for negative values, as certain states may be considered worse than death [
36,
37]. When a utility is combined with life years gained (LYG), the outcome reflects both morbidity and mortality: quality adjusted life years (QALYs). A cost per QALY can then be reported, the preferred unit of measurement of many decision makers, such as the UK’s National Institute for Health and Care Excellence (NICE) [
34] and the Australian Pharmaceutical Benefits Advisory Committee (PBAC) [
35].
To date, just four cost utility studies of HH screening programs have been published [
33]. The studies did not report the sources of the utilities used, and the estimates employed for conditions such as healthy state, heart disease and cirrhosis of the liver were markedly higher than reported for comparable populations [
33]. Such use of elevated utility values is likely to result in underestimates of the potential gains associated with screening programs, which in turn may impact on policy decisions regarding provision of HH screening programs.
The purpose of this study was to assess the utilities for a sample of people with HH with different stages of disease severity using a multi-attribute utility instrument.
Discussion
This is the first study that reports HSUV measured directly from a cohort with HH. This is of importance, as a lack of robust health economic data has been cited as a barrier to implementing population screening programs for HH [
25,
30,
31,
42]. The utility values calculated in this study provide robust estimates that can be used in future economic models of screening interventions. Whilst the sampling strategy may have introduced bias, this has been mitigated by reporting utility values for categories of HH rather than across the study population in general. These values can then be used in combination with penetrance rates in economic models for HH interventions.
Symptomatic stages of HH (categories three and four [
25]) were associated with lower utility than asymptomatic stages. The values for all four categories are useful, as they incorporate all aspects of HH and related conditions and can be used to populate health economic models. Previous CUA models have only incorporated specific comorbidities which are associated with significant morbidity and mortality: cirrhosis, diabetes and heart failure, with no consideration of common comorbidities such as arthritis, or symptoms such as fatigue. This may be related to the relatively high prevalence of both fatigue and arthritis amongst other populations, and the difficulties surrounding the aetiologies of both, however there is evidence suggesting that the prevalence of both is higher amongst some groups of HH patients. The prevalence of fatigue amongst general practice patients has been estimated to be between 1.4 and 7.0 % of encounters [
43‐
46]. Work by Allen and colleagues has reported a much higher rate of 22 % for C282Y homozygotes with elevated serum ferritin levels (greater than 1,000 μg/l) [
10]. Similarly, arthritis, specifically osteoarthritis, is prevalent in Australia, with 9 % reporting this condition [
47]. Allen and colleagues reported use of arthritis medication as a proxy measure for arthritis, noting that 20 % of C282Y homozygotes with serum ferritin greater than 1,000 μg/l reported use of these medications. In combination, these data guided the decision to include arthritis and fatigue in the current study.
To date, just four cost utility analyses have been published on HH screening programs, none of which cited the sources of the utility values employed [
7,
48‐
50]. Values were assigned for cirrhosis, diabetes and heart failure, and in some cases, combinations of these. In a Norwegian study [
48], a basal utility value of 1.00 was assumed for all HH conditions except cirrhosis, which was assigned a utility of 0.95, values that are substantially higher than those reported here. Two Canadian studies, by the same research group, used utilities of 0.8 for cirrhosis, 0.9 for diabetes, 0.5 for heart failure, 0.72 for cirrhosis and diabetes, 0.78 for cirrhosis and heart failure, 0.87 for diabetes and heart failure and 0.70 for a combination of cirrhosis, diabetes and heart failure [
7,
49]. A fourth study did not provide the utility values used in the modelling [
50]. Some concerns arise in respect of these estimates. First, in comparing these utility values to US population normative data, a disparity appears: the mean utility derived from the SF-6D ranged from 0.79 to 0.81 for persons aged 35 to 74, and similarly, using the EQ-5D, mean utility ranged between 0.87 and 0.89 [
51]. The fact that the utility estimates that were used in cost utility analyses for participants with health conditions such as cirrhosis and diabetes are similar to or higher than those reported for the general US population indicates these estimates may be incorrect. The likely overestimates of HSUV for HH-related conditions are likely to lead to underestimates of potential utility gains associated with screening interventions.
Second, disease specific HSUV used in these cost utility analyses are also higher than suggested in published literature. A meta-analysis of utility values for liver diseases using a range of approaches to measure utility reported a mean of 0.75 for compensated cirrhosis (range 0.65–0.90) and 0.67 for decompensated cirrhosis (range 0.57–0.81) [
52]. Whilst our study did not differentiate cirrhosis in this manner, amongst the small number of participants reporting this condition (
n = 5), the mean utility (0.61) was slightly lower than reported for decompensated cirrhosis but within the range reported. In contrast, the published cost utility analyses used values of 0.95 [
48] and 0.8 [
7,
49], higher than the mean values reported for both compensated and decompensated cirrhosis [
52]. Similarly, a meta-analysis of utility values for diabetes reported a mean of 0.76 (range 0.53–0.88) [
53]. In our study, a mean of 0.52 was reported (
n = 4), slightly lower than the lower range reported in this meta-analysis. In the three HH cost utility analyses, one used a utility value for diabetes of 1.00 [
48], and two used a value of 0.9 [
7,
49], both notably higher than published estimates.
Mean utility for heart failure varies depending on the severity of the condition. From a large, multi-site trial that used the EQ-5D, mean utility for different levels of severity based on the New York Heart Association (NYHA) classifications were: class I: 0.815, class II: 0.720, class III: 0.590, class IV: 0.508 [
54]. Our study reported a mean of 0.58, however data were available for only three participants, and all were in different NYHA classes. The two Canadian CUA models used a utility value of 0.5 [
7,
49], which is similar to the NYHA class 4. In contrast, the Norwegian study assumed a utility of 1.00, which is not in keeping with estimates in the current literature [
48].
To date, no economic analysis has incorporated HSUV related to arthritis. This is surprising as arthritis related to iron overload is commonly reported amongst patients diagnosed with HH [
10,
55‐
57]. Whilst HRQoL is not synonymous with HSUV, it can serve as an indicator. A study examining the effects of a range of HH-related comorbidities using the SF-36 found that, compared to cirrhosis and diabetes, arthritis was the single strongest factor that impacted on HRQoL [
58]. Whilst the paper was published in 1996, no subsequent studies have incorporated utility values for arthritis. Hawthorne and colleagues, using the AQOL-4D, reported the Australian normative utility value for arthritis as 0.69 (SD 0.26). Our study reported a lower mean value of 0.52 (SD 0.25,
n = 35). In the current study, both self-reported diagnosis of arthritis related to HH and symptoms suggestive of arthritis were associated with lower mean utility than the sample mean (0.52, 0.48, 0.66 respectively).
Limitations of this study include cross-sectional design and use of convenience sampling. Convenience sampling, which was used as a result of available resourcing, may limit the generalizability of these results. Further, the majority of the respondents were female, despite higher penetrance amongst males. To minimise sampling bias, we have focused on utility values for categories of disease and symptomatology for males and females separately. Whilst an overall sample mean HSUV is likely to be affected by under- or over-reporting from participants with more health problems, the mean values for each category are not affected. This allows for these values, in combination with penetrance estimates from robust epidemiological studies, to be used in HH health economic models.
A further limitation of this study was the reliance on participants’ self-report regarding experience of HH related comorbidities and symptoms. Whilst participants were asked if the comorbidities were related to HH, even with clinical verification, it is difficult to be certain of the aetiology of these. Whilst it can be argued that there may be some over-reporting of symptoms and comorbidities believed to be caused by HH, to minimise this possibility, cases in which participants were unsure of the aetiology have been excluded. Symptoms and comorbidities were only included when participants stated that they were related to HH. Lastly, the small number of participants reporting HH-related comorbidities was also a limitation. Whilst utility values were calculated wherever possible, the small number of respondents means that these data should be interpreted with caution and that no meaningful comparisons can be made between these comorbidities.
Competing interests
Barbara de Graaff, Amanda Neil, Kristy Sanderson, Kwang Chien Yee, and Andrew J. Palmer have no financial or non-financial competing conflicts of interest to declare that are directly relevant to the content of this manuscript.
Authors’ contributions
BdeG: planning of study, development of survey, ethics submission, recruitment, data analysis, preparation of manuscript. AN: contributed to study design, assisted with preparation of manuscript, statistics advice. KS: contributed to study design, assisted with preparation of manuscript. KCY: contributed to study design, assisted with preparation of manuscript, provided medical opinion; AP: planning of study, contributed to study design and development of survey, assisted with preparation of manuscript. All authors read and approved the final manuscript.