Diagnostic strategies using DNA testing for hereditary haemochromatosis in at-risk populations: a systematic review and economic evaluation

Iron metabolism

Iron is vital for all living organisms as it is an essential component of a wide variety of metabolic reactions including transport of oxygen, DNA synthesis and electron transport. However, iron concentrations in the tissue need to be tightly regulated as excessive iron is toxic as a result of the formation of free radicals. The control of iron uptake and storage is therefore complex.

The majority of total body iron (60–70%) is present in haemoglobin in the erythrocyte pool. Another 10% is present in the form of myoglobins, cytochromes and iron-containing enzymes and, in a healthy individual, the remaining storage iron is sequestered by ferritin and haemosiderin in the liver, spleen and bone marrow. There is a constant turnover of iron for haemoglobin synthesis by erythroid precursor cells in the bone marrow; the majority of iron for this is recovered from the destruction of red blood cells. Iron is transported in the blood tightly bound to transferrin and, although this is less than 1% of the total body iron store, because of its high turnover it is the most significant body iron pool. 15–18

A constant balance between uptake, transport, utilisation and storage of iron is needed to maintain cellular iron homeostasis both at the level of the cell and at the level of the whole organism. Identification of the gene for HHC has led to increased understanding of the biological pathways underpinning this process. This tight regulation depends upon the constant movement of iron bound to transferrin in the plasma between the functional iron pool and the storage iron pool. Because of its low solubility iron is not excreted, although iron is lost through menstruation, other blood loss and shedding of epithelial cells from the gastrointestinal and urogenital systems and the skin. The primary level at which body iron content is controlled is by variation in the amount of iron absorbed from the diet at the level of the small intestine.

Iron is transported across the apical cell membrane of duodenal epithelial cells by the divalent metal transporter 1 (DMT1) protein. It is then either stored as ferritin or exported by ferroportin across the basolateral membrane where it enters the circulation bound by transferrin. Both DMT1 and ferroportin expression are dependent on cellular iron stores. In haemochromatosis duodenal DMT1 and ferroportin expression are raised, leading to excess iron absorption and gradual accumulation of iron. 18–20

The recently identified protein hepcidin is central to the pathophysiology of haemochromatosis. Hepcidin is a circulating hormone synthesised in the liver and levels of expression are regulated by iron stores and inflammation. Hepcidin is upregulated in the presence of iron and acts on ferroportin to inhibit iron transport, presumably resulting in decreased iron absorption. 21 In HHC hepcidin levels are inappropriately low.

Clinical features of haemochromatosis

Serum iron, total iron binding capacity and transferrin saturation

Measurement of serum iron alone is of little clinical use as there is considerable variation from hour to hour in normal individuals. More information is obtained by measuring serum iron concentration and total iron-binding capacity (serum iron plus unbound iron-binding capacity) as a surrogate for percentage saturation of transferrin. Transferrin may also be measured by immunological methods and TS calculated directly {TS = [serum iron/total iron-binding capacity] × 100 or [serum iron/(serum iron × unsaturated iron-binding capacity )] × 100}. TS is considered the ‘gold standard’ for assessment of iron overload; however, there remain issues relating to test standards and quality control, whichever technique is used, and problems with standardisation of these assays are recognised. 17,49 TS can also be decreased in inflammatory states, can be increased by alcohol consumption and can be artefactually changed by recent ingestion of iron or vitamins.

Serum ferritin (SF) is considered to correlate with the total amount of storage iron in normal individuals. Iron overload is correlated with a high SF; however, SF may also be high in other forms of liver disease, cancer, infection, inflammation and chronic disease. 17

Quantitative phlebotomy

This is used to measure iron stores. It provides a direct measurement of the amount of iron available for haemoglobin synthesis. Blood is removed weekly and after a number of venesections the patient is unable to maintain his or her normal haemoglobin level. At this point it is assumed that the available iron stores have been used and the amount of iron removed can be calculated, although there is no gold standard against which this can be evaluated. Individuals with normal iron stores become iron deficient after the removal of approximately 1.5–2 g of iron (i.e. four 500-ml units of blood). 17 Individuals with iron overload will require more venesections to deplete their storage iron. 50

Liver biopsy

Liver biopsy was the definitive test for a diagnosis of haemochromatosis and allows histochemical estimation of tissue iron, assessment of the extent of fibrosis or cirrhosis and chemical measurement of hepatic iron concentration. The degree of stainable liver iron is usually graded and the consensus is that grades 0–1 are normal and grades 2–4 represent increased parenchymal iron stores. Iron deposition in the liver may be increased in various forms of liver disease including alcoholic cirrhosis.

It has been accepted that a hepatic iron index (hepatic iron concentration divided by age) greater than 1.9 discriminates between hepatic iron overload caused by HHC and hepatic iron overload caused by other liver diseases. This is based on the concept that iron overload increases with age in haemochromatosis but is stable in other chronic liver diseases. A hepatic iron index greater than 1.9 was considered to be the gold standard test for haemochromatosis. 50,51 The identification of the genetic basis of haemochromatosis now means that the role of liver biopsy in the diagnostic pathway is being reduced. 52

Non-invasive imaging, such as magnetic resonance imaging or computerised tomography scans, are not widely recommended for diagnostic purposes.

UK epidemiology studies

Eight studies have been included in the assessment of the prevalence of the key genetic mutations within the general population and among those with haemochromatosis in the UK, Jersey and Ireland (Table 1 and see Chapter 3). 13,38,62–67 Six studies were prospective cohort studies,38,62,63,65–67 with two including a control group. 38,62 The other two studies were a retrospective cohort study13 and a cross-sectional case–control study. 64 The participants in the studies varied depending upon the rationale for the study, including people diagnosed with haemochromatosis,13,38,62 people attending a blood donation service,65,67 those on registers of newborn children63 or heart attack sufferers,64 and those attending for routine blood screening. 66 Control groups included healthy blood donors,38 people registered with general practitioners64 and people on bone marrow registers. 62

Biochemical testing and liver biopsy

Before the discovery of the common gene mutations, diagnosis of haemochromatosis was based on clinical suspicion including persistently raised TS with no other explanation followed by liver biopsy. Both the BSH69 and the US College of American Pathologists35 recommend the TS test as the initial diagnostic test for HHC. This test should be carried out on a fasting sample. Elevated fasting TS indicates iron accumulation and, if this has been demonstrated, the BSH guidelines recommend that the SF concentration is measured. SF concentrations are not usually abnormal in the early stages of iron accumulation but once liver iron concentrations are elevated they rise disproportionately with the degree of liver damage. Liver biopsy with assessment of iron deposition, although previously considered to be the gold standard for diagnosis, is no longer used as a diagnostic tool; however, it is still widely used to confirm cirrhosis and as the only way of determining fibrosis. It is worth noting that biochemical tests are unable to discriminate between different possible causes of iron overload, so although iron overload may suggest a diagnosis of HHC other possible diagnoses will need to be ruled out.

Genetic testing

DNA-based testing can identify people who are homozygous for mutations in the HFE gene before symptoms of iron overload become apparent. Although this test identifies people at risk of developing the symptoms of iron overload, because of the low penetrance of the disorder, only a proportion will go on to exhibit the HHC phenotype. This means that both the sensitivity for detecting phenotypic HHC and the PPV of genetic testing are low in the general population.

A wide range of DNA-based tests to identify the C282Y and H63D mutations have been described, most commonly polymerase chain reaction (PCR) followed by restriction enzyme digest, amplification refractory mutation system PCR, and PCR with single-strand conformation polymorphism (SSCP) analysis. 73

Family screening

Hereditary haemochromatosis is an autosomal recessive condition and, as such, children and siblings are at increased risk for inheriting susceptibility to the disease, with probabilities of at least 1 in 20 and at least 1 in 4 respectively. The purpose of testing family members is to detect those individuals at risk who would benefit from treatment, and to detect those at risk who do not currently require treatment but who will be monitored for a suitable period of time until the need for future treatment can be ascertained. Testing can also identify those not at risk who can be excluded from further investigations and reassured. Penetrance may be higher in families as they share other genetic and environmental factors, although the main risk factor appears to be possession of the at-risk genotype. 13

Psychosocial aspects of genetic testing

Concern about genetic testing and screening is evidenced by a number of reports that focus on issues of potential stigmatisation, discrimination, family implications and the possible psychological consequences. 74,75 However, the rationale for genetic exceptionalism may not be well established. 76 In addition, concerns relating to unfair discrimination will be influenced by the legal framework and welfare provision of the country of residence. For example, the situation regarding coverage and reimbursement for health care is very different in North America to that in the UK. In haemochromatosis the purpose of using the genetic test is to identify individuals who will benefit from treatment and to rule out disease in family members. Studies have investigated the psychosocial consequences of using DNA tests for population and targeted screening for haemochromatosis and found few adverse effects. 77–79

Clinical validity

• Intervention: DNA tests.

• Population: Caucasians with iron overload suggestive of haemochromatosis (north European populations).

• Comparator: control population (Caucasian).

• Outcomes: sensitivity and specificity (reported or calculable).

• Study type: controlled cohort or case–control.

Clinical utility

Psychosocial aspects of DNA testing

• Intervention: DNA tests.

• Population: diagnosed and at-risk individuals (people with suspected HHC and first-degree relatives).

• Outcomes: psychosocial aspects (treatment compliance, psychological outcomes, legal implications, quality of life, discrimination/stigmatisation).

• Study type: any quantitative or qualitative primary research.

Haemochromatosis cohorts

Eight of the 11 included studies reported clear selection criteria for the eligibility of haemochromatosis cases38,87–90,92,94,95 whereas the three remaining studies were unclear and did not report sufficient details. 62,91,93 The criteria used to define haemochromatosis cases varied between studies but most required the presence of two or more diagnostic criteria. Six studies38,87,88,90,92,95 have used a definition that is likely to correctly classify HHC whereas the remainder are unclear from the description given.

Seven studies87–89,91–94 included a requirement for both a high TS and a high SF concentration in their haemochromatosis patient definition. Five studies87–89,92,94 defined an elevated TS and this ranged from a cut-off of over 45% to a cut-off of over 62%. These five studies also defined an elevated SF and this ranged from a cut-off of over 200 μg/l (women only) to a cut-off of over 400 μg/l (men only). Two91,93 of the seven studies did not state what values of TS or SF would be considered high. One additional study95 used a high TS concentration (greater than 60%) without mention of SF levels. Three studies did not use TS and SF values. 38,62,90

In six studies62,87,88,90,92,93 all members of the haemochromatosis cohort had results from liver biopsy that showed liver iron deposition, and four studies reported that haemochromatosis cohort members had a hepatic iron index (HII) greater than 1.938,88 or greater than 2. 90,95 One study92 used a value of iron from wet weight of liver and reported that haemochromatosis cohort members had an HII of over 30. Phlebotomy treatment for either some or all of the haemochromatosis cohort was a feature of seven studies. 38,87–90,92,95 Only four studies38,87,88,90 reported that other causes of iron overload had been ruled out. One study62 provided few details about the diagnosis of the patients in the haemochromatosis cohort, reporting only that ‘haemochromatosis cases were both clinically assessed and pathologically diagnosed by liver biopsy’.

In all studies it is not clear whether bias has been avoided in the sampling of participants in the haemochromatosis cohorts as the methods used have not been described. Most of the haemochromatosis patient cohorts comprised fewer than 100 individuals62,87,88,92,93,95 (range 18–92), three cohorts comprised 115–156 individuals,38,90,94 there was one patient cohort of 29689 and the largest contained 478 haemochromatosis patients. 91 Individuals in five of the haemochromatosis cohorts38,87,88,90,92 were reported to be unrelated to one another, and in one study93 although the majority of the cohort consisted of unrelated individuals it is not clear whether this is true for the entire cohort. One haemochromatosis cohort is reported to include some related individuals. 95 Only three studies88,91,95 specifically state that the patients are Caucasian, with a further study cohort94 being predominantly Caucasian (96.2%).

Only two studies92,94 report some additional characteristics for their haemochromatosis cohort, and three studies87–89 report on the age and sex of the cohort.

Control cohorts

Control subjects were drawn from different sources in the included studies and may not be representative of the population that the haemochromatosis patients were drawn from and may not be free from selection bias. In four studies87,90,91,94 the control cohort was drawn from the general population; only two of these studies91,94 clearly state that the control group subjects are Caucasian. Three studies recruited the control cohort from amongst hospital employees88,93 or hospital employees and students and their relatives. 89 Three studies38,62,92 recruited from particular groups of healthy persons, including blood donors38 and a bone marrow registry. 62 The remaining study recruited the control cohort from amongst several other patient groups. 95

Sampling of the control cohorts was described as random in four studies,87,90,91,93 with the sampling methods being unclear in five studies;38,88,89,94,95 the remaining studies involved volunteers. 62,92

All of the control cohorts comprised more than 100 individuals. One cohort comprised 200 individuals,95 one 250 individuals,89 and the two largest control cohorts contained 40462 and 41091 subjects. Only two studies38,92 reported that the individuals in the control cohort were unrelated to one another, and in one study89 the inclusion criteria allowed for related individuals to be included.

Only one study92 reported some additional characteristics for the control cohort, and one study89 reported age and sex of the cohort, noting that these parameters were significantly different from those of the haemochromatosis cohort.

Genotyping methodology

To analyse the HFE gene regions encompassing the C282Y and H63D mutation sites eight studies38,87,88,90–94 isolated genomic DNA and performed a PCR followed by restriction enzyme digest of the PCR products [restriction fragment length polymorphism (RFLP) analysis] to determine whether samples carried the two mutations of interest. One study95 used this method only for H63D mutation determination. The primers used for the PCR varied between studies. Six studies38,87,88,90–92 published the sequences of some or all of the primers used, two studies93,95 cited a reference and one study94 provided no information at all about the PCR primers used. To identify wild-type and mutant C282Y PCR products by RFLP the restriction enzymes Rsa138,90,93,94 or SnaB187,88,91,92 were used, and for H63D the restriction enzymes Bcl138,87,88,91,92,94,95 or Mbo190,93 were used. Only two of the studies38,93 using the PCR-RFLP method specifically mention that primers had been modified to include an internal control restriction site. One study90 had analysed the C282Y and H63D loci in DNA samples from haemochromatosis and control cohorts by fluorescent sequencing before conducting the PCR-RFLP analysis. Another study,88 in addition to the PCR-RFLP analysis, also performed a PCR-ELISA (enzyme-linked immunosorbent assay) to analyse the C282Y mutation, and a single-strand conformation polymorphism analysis for capillary electrophoresis (SSCP-CE) analysis for both C282Y and H63D mutations in the HFE gene.

Two studies performed a PCR and then immobilised the PCR products on membranes so that hybridisation with normal and mutant versions of C282Y62,95 and H63D62 could be carried out to identify whether samples carried the mutations of interest. One study62 published the sequences of the primers used for the PCR and the other study95 cited a reference. Both studies published the sequences of the probes used in the hybridisation. One study62 also stated that a number of the PCR products had been sequenced to confirm that the correct region of the HFE gene had been amplified, and samples of known genotype were also included as a control in the hybridisations.

In the 10 studies described above samples from the haemochromatosis cohort and the control cohort were analysed in the same way. In one study89 different methods were used for the two cohorts. DNA samples from the haemochromatosis cohort were sequenced to identify the mutations of interest; details of the primers used in the sequencing reaction are not provided. DNA samples from the control cohort were analysed by PCR-RFLP (references are cited for this method) and all substitutions detected by RFLP were confirmed. For C282Y the RFLP was repeated, and for H63D samples were sequenced to confirm the result.

Outcomes

All studies reported on the prevalence of the HFE mutations C282Y and H63D in the haemochromatosis and control cohorts. Two studies89,91 also reported on the HFE mutation S65C. Other commonly reported outcomes included allele frequencies in haemochromatosis87,91,94 and/or control cohorts,62,87,91,94,95 clinical data for haemochromatosis patients who were not C282Y homozygous,38,87,88,93 genotype–phenotype correlations for the haemochromatosis cohort89,94 or both cohorts,92 and haplotype analysis. 38,90

Withdrawals and dropouts were not applicable to these studies but one study reported missing SF and TS data for the haemochromatosis group. 89

Comparability of groups

In one study89 the haemochromatosis and control groups are not comparable in terms of age, sex and race/ethnicity and in all other studies it is not clear whether the groups are comparable.

Generalisability

The results of six studies38,87,88,90,92,95 may be generalisable in terms of them being from representative samples of unrelated Caucasian haemochromatosis patients. It is unclear whether the results of the other studies are generalisable.

Sensitivity

Clinical sensitivity ranges from 28.4% to 100% as determined from the included studies. The definitions of primary iron overload are variable, which may account for the wide range. When considering only the studies that are most likely to correctly define haemochromatosis,38,87,88,90,92,93,95 clinical sensitivity ranges from 72.2% to 100%. If the studies are further limited to those that have reported that other causes of iron overload have been ruled out and that patients are unrelated, the range is 72.2–92.4%. 38,87,88,90 The range is 91.3–92.4% when the small study in southern Germany, which may be on the north–south European divide, is excluded. 88

From two studies the estimated clinical sensitivity is low at 28.4%89 and 21.1%. 94 This may be explained in one study89 by the fact that HFE testing was carried out retrospectively in patients with clinical suspicion of iron overload based only on biochemical measurements, some of which were missing. The other study with low clinical sensitivity included patients referred to a hospital endocrinology and metabolism department with excessive alcohol intake and diabetes, in whom disturbed iron parameters may have been due to cirrhosis and insulin resistance, respectively, and not to genetic iron overload. 94

Clinical sensitivity was 100% in one small study of 18 patients, which included related subjects. 95

Specificity

Clinical specificity ranges from 98.8% to 100% from the included studies. In two38,89 of the five studies that give an estimate below 100%, the single individual homozygous for the C282Y mutation in the control group showed signs of evidence of iron overload. No details are given about the homozygous individuals in the control groups of the other three studies62,91,95 so it is not known if they showed signs of irregular iron loading or whether they would in the future. However, specificity may be less than 100% as some homozygotes in the control group may never exhibit symptoms.

Positive predictive value and negative predictive value

The PPV ranges from 84.3% to 100% and the NPV from 46.3% to 100%. As expected these values reflect those of the clinical specificity and clinical sensitivity, respectively, and show a range of values for the reasons given above. Considering the most appropriate studies the PPV ranges from 99% to 100% and the NPV from 91% to 94.4%.

Participating in testing

Only one study96 reported outcomes of participating in testing (Table 8). Most patients who could recall having the genetic test stated that they were satisfied with the information they had received before the test, had wanted to receive the test and had understood the rationale for being tested. People who could not recall having the genetic test (34%, n = 30) did not contribute to this outcome measure.

Emotional reactions

All three studies reported on the emotional reactions experienced by people in response to receiving the HFE genetic test (Table 9). However, the only common outcome measures among the studies were the STAI and the SF-36, which were both reported by two studies. 97,98

Understanding of test result

Two studies96,97 report outcomes related to participants’ understanding of their test result (Table 10). This information is often embedded within, and difficult to separate from, the reporting of outcomes on participants’ knowledge about haemochromatosis.

The overall recall of information by patients reported by Hicken and colleagues96 was not significantly different from that of control subjects (patients 65% ± 26% versus control subjects 59% ± 30%, p > 0.5). The percentage of participants providing the correct answer to each of the individual questions that made up this outcome are also reported but if any statistical comparisons were made between patients and control subjects these are not reported. Hicken and colleagues also asked the patient cohort some short-answer and true–false questions. The short-answer questions were correctly answered by 70% of patients (range 25–90%) and the true–false questions were answered correctly by 87% (range 42–99%). Only 25% of patients could correctly answer the short-answer question about the difference between HFE genotyping and the TS test. Unsurprisingly the patients who could not recall undergoing an HFE genotyping test were less likely to understand the difference between HFE genotyping and the TS test (p < 0.0001).

In total, 101 participants in the study reported by Meiser and colleagues97 answered seven true–false questions 2 weeks after their consultation. Two of the questions assessed their understanding of the significance of carrying one or two mutations, respectively, for haemochromatosis. Most participants responded correctly to these two questions (76.1% and 87.3%). In total, 93% of participants knew that regular removal of blood will avoid or reduce many of the symptoms of haemochromatosis. The question that fewest participants answered correctly assessed whether participants knew that the C282Y mutation is found in most people with haemochromatosis. Only 45.2% of participants knew that this was the case. Participants were asked these questions again at the 12-month follow-up. The results are presented in a figure but the study authors do not comment on whether these differed in any way to the results obtained at the 2-week follow-up. Genetic testing results were available for 95 of the 101 participants in the Meiser and colleagues study. Of the 95 participants with a genetic testing result, 80 (84.2%) had learnt of the outcome of their test before their first attendance at the clinic and entry into the study. Homozygotes and heterozygotes for C282Y and H63D, plus compound heterozygotes, were asked at the 2-week consultation whether they believed that they had one or two gene mutations, or if they could not remember. The results for 77 of these participants are reported: 69.3% were able to correctly state the number of mutations that they carried. There was no association between education level and having an accurate understanding of the number of gene changes associated with one’s particular genetic testing result (p = 0.29), and similarly there was no association between the presence of a family history of haemochromatosis and understanding of the number of gene changes associated with the genetic testing result (p = 0.53).

Benefits and problems of testing perceived after test carried out

The patients in the study by Hicken96 reported the positive benefits and negative detrimental outcomes that resulted from HFE genotyping on two short-answer questions (Table 11). The most commonly reported benefit of testing was improved health and prevention of future health problems (40% of patients). The other reported benefits were learning of the risk to self and family (19%), improved understanding of health (11%), and improved psychological well-being (12%). Eleven participants (19%) did not identify any benefit of testing. The majority of participants did not report any problems from genetic testing (88%). Decreased psychological well-being was reported by one woman, and one man reported that he had been denied health insurance because of his HFE genotype.

Compliance with treatment

Hicken and colleagues96 reported on patient compliance with treatment recommendations (Table 12). Compliance with therapy was defined as achieving an SF ≤ 20 ng/ml after undergoing serial phlebotomy for iron overload. Iron depletion was achieved in 99% of patients. However, after achieving iron depletion, adherence to maintenance therapy in the following years dropped. In the first year after achieving iron depletion 94% adhered to maintenance therapy but this dropped by 8% each year in subsequent years. Adherence to maintenance was not associated with demographic factors, barriers to compliance or knowledge of haemochromatosis (p > 0.05). The majority of patients (88%) reported few difficulties with obtaining annual serum measurements.

Meiser and colleagues97 reported that, at the 12-month follow-up, all participants for whom iron studies are recommended (homozygotes and compound heterozygotes) reported having ever had iron studies, and 96% reported having had iron studies in the past year. At the 12-month follow-up 62% of those participants who had increased SF at baseline reported ever having had a venesection, and 57% reported having undergone a venesection in the past year, which included any that may have been carried out at the actual clinic visit. However, it is important to note that 39 participants (38.6%) had been lost to follow-up by the 12-month time point.

Quantity and quality of research

The literature search identified two published economic evaluations that met the inclusion criteria. 99,100 Studies that were assessed and excluded are shown in Appendix 3 with their reasons for exclusion. El-Serag and colleagues99 estimated the cost-effectiveness of screening for HHC in family members using genotypic tests compared with phenotypic tests and no screening in the USA. Adams100 investigated the likely cost of genotyping spouses and whether this would reduce the number of investigations of children in Canada. The quality of these economic evaluations has been assessed using a standard checklist adapted from Drummond and Jefferson82 (Table 13) and the two studies are discussed in more detail below.

Both studies clearly defined the study question and explained the competing alternatives. They each used the correct comparator and the patient group of interest was clearly stated and appropriate. Furthermore the study type appeared reasonable: Adams100 used a cost-minimisation model and El-Serag and colleagues99 used a cost–utility model. Both studies were conducted in North America and so it is unclear how these studies relate to the UK NHS.

Adams100 did not consider long-term costs and consequences. El-Serag and colleagues100 used a lifetime horizon and estimated incremental cost-effectiveness with appropriate discounting rates. They also presented sensitivity analyses of the key parameters. It is unclear whether the studies valued the costs and consequences appropriately. Adams100 does not report the source of costs used in the study or how these costs were derived. El-Serag and colleagues100 seem to have included all of the costs relevant to HHC and screening; however, these have been taken from earlier studies and have not been adjusted for time. The data used in the model have not been discussed and in many cases the sources of the data have not been given. The conclusions in both of the studies appear credible from the results presented.

Results of the published economic evaluations

A summary of the results of the published economic evaluations is shown in Table 14.

Adams100 estimated the costs of genotyping spouses of homozygotes to reduce the number of investigations of children. Costs were estimated for genotyping all children of homozygotes compared with genotyping spouses and then genotyping children if the spouse was a heterozygote or homozygote. If the spouse was not a homozygote or heterozygote, investigations in the children would be unnecessary.

A total of 291 children of homozygotes were investigated using TS, SF and genetic tests. Costs incurred in the phenotypic strategy were CDN$58,200. In total, 13 of these children were found to be homozygotes in 116 families with the C282Y mutation. In the spousal strategy 116 spouses were genotyped with subsequent investigation of 22 children at a total cost of CDN$35,600. Therefore the genotyping of the spouses reduced the number of investigations in children from 291 to 22 with a cost saving of 39%.

El-Serag and colleagues99 developed a decision tree model to evaluate screening strategies for HHC in siblings and children of affected patients. They assumed that a proband had been confirmed to have HHC on the basis of standard phenotypic criteria. It was estimated that 5% of the children would be homozygous, assuming that the proband was homozygous and 25% of siblings were homozygous. The model estimated the life expectancy, qualify of life and costs for various screening strategies, and incremental cost-effectiveness ratios were calculated compared with no screening. For the no screening strategy, life expectancy was estimated based on previous studies, including that of Adams and colleagues. 11 The serum iron studies strategy entailed measuring TS and SF levels in relatives of the proband. Screening of children was assumed to start at 10 years of age and continue until 40 years of age or until an abnormal test result was found. Siblings were screened once with repeated testing in people with elevated values.

There were three strategies for genetic testing. In the first genetic testing strategy, the proband was tested first and, if found to be homozygous, the spouse was tested. Children were gene tested if the spouse was heterozygous. Children homozygous for C282Y underwent iron studies. If a child was found not to be homozygous no further screening was performed. If the proband was not homozygous the child would undergo iron studies.

The next strategy was similar to the first one except that the spouse was not gene tested. The proband was gene tested followed by gene testing of the children or siblings if the proband was homozygous. Relatives who were homozygous then underwent iron studies. As before, if the proband was not homozygous the relative would undergo iron studies.

In the final genetic testing strategy the relatives were gene tested before the proband. Those relatives who were homozygous underwent iron testing. If the relative was not homozygous, then the proband was gene tested and the relative underwent iron studies if the proband was homozygous.

Strategies using gene testing were less costly than serum iron studies. Compared with no screening, gene testing the proband followed by testing of a child was the least expensive and most cost-effective strategy for one child (incremental cost-effectiveness ratio US$508 per life-year saved). For screening two or more children, gene testing the spouse if the proband was homozygous was the most cost-effective strategy. Compared with the no screening strategies for siblings, all strategies cost less and yielded greater benefits.

El-Serag and colleagues99 performed sensitivity analyses for the costs of the tests, screening frequency and proportions of probands with HFE gene mutations. For all sensitivity analyses, HFE gene testing remained cost-effective.

Economic model structure

The addition of DNA testing to diagnostic algorithms in people suspected of having haemochromatosis

Before the discovery of the common gene mutations, diagnosis of HHC was based on clinical suspicion, including persistently raised TS and SF with no other diagnosis followed by liver biopsy. Since the identification of the gene it has been possible to use DNA testing to confirm the diagnosis in those in whom it is suspected. Liver biopsy then becomes a prognostic test in those suspected of having liver damage and can be avoided in those with raised iron levels and no biochemical evidence of liver damage.

Decision models were constructed to compare the costs and consequences of two diagnostic algorithms in people suspected of having HHC on the basis of persistently raised TS > 45% and SF > 300 μg/l. 72 The algorithms for people with suspected HHC are liver biopsy for all people and genetic testing for all people. The end point of both algorithms is detection of a case requiring treatment according to current clinical guidelines and the reported outcome will be cost per case detected. The goal of a diagnostic strategy is to improve patient management and ultimately patient outcomes. The advantage of the genetic testing strategy is that it avoids the use of liver biopsy.

The decision tree is shown in Figure 2. For the liver biopsy strategy all patients have liver biopsy and are either confirmed positive or negative phenotypic haemochromatosis. Those who are positive will be treated for haemochromatosis and those who are negative will not.

FIGURE 2.

Decision tree for DNA testing in people suspected of having haemochromatosis. HC, haemochromatosis; LB, liver biopsy; SF, serum ferritin; YD, compound heterozygous C282Y/H63D.

For the DNA testing strategy all people receive a DNA test, which will be either positive (YY, C282Y homozygous) or negative. Those who are YD compound heterozygous (C282Y/H63D) will be treated in the same way as those who have a negative DNA test. This assumption was made based on clinical guidelines and expert clinical opinion and was a pragmatic approach in view of the limited data on the long-term prognosis and treatment of YD compound heterozygous patients. All patients with raised SF (> 1000 μg/l) receive a liver biopsy to check for liver cirrhosis. All patients with a positive DNA test will be treated as they are assumed to have HHC. Patients with a negative DNA test and SF < 1000 μg/l are monitored and receive a repeat SF test. If their SF is stable or decreasing then they are not treated, whereas those with increasing SF have a liver biopsy to confirm haemochromatosis. Those with confirmed haemochromatosis will be treated.

The model can be run for different strategies for different TS and SF levels. For the purpose of the model a ‘typical’ patient is defined who is representative of all patients. For the baseline run it is assumed that a typical patient is a 45-year-old man because clinical HHC is more common in men and raised iron levels will typically appear during the middle of the fourth decade of life. The effect of patient age or gender is investigated in the model through the use of sensitivity analyses.

DNA testing for family members of people diagnosed with haemochromatosis

Separate decision models were constructed to compare the costs and consequences of two testing algorithms in family members of people diagnosed with HHC. The algorithms for testing family members are biochemical testing for all versus genetic testing for all. The end points of the algorithms are detection of a case requiring treatment according to current clinical guidelines, identification of family members at risk for HHC who need to be monitored and identification of family members who are not at risk for HHC. These outcomes will be incorporated into the model by considering unnecessary investigation of those with false-positive diagnoses and missed diagnoses.

The decision tree is shown in Figure 3. For the biochemical test strategy relatives have SF and TS tests. If they have raised iron levels (positive biochemical tests results, i.e. TS > 45% and SF > 300 μg/l) according to clinical guidelines,72 they are treated; if not they will be monitored to see if their iron levels increase. If the iron levels increase, they will be treated for HHC.

FIGURE 3.

Decision tree for the use of DNA testing in family members. SF, serum ferritin; TS, transferrin saturation.

For the DNA test strategy relatives have a DNA test. Those with a positive result (YY, homozygous C282Y) have biochemical tests and those with raised iron levels (positive biochemical tests results, i.e. TS > 45% and SF > 300 μg/l) are treated. 72 Those who do not have raised iron levels (negative TS and SF results) are monitored to see if their iron levels increase; if they do increase they will be treated for HHC, and if they do not they will not be treated. Those without a positive DNA test for C282Y will not need treatment or any further medical investigation for HHC.

For the purpose of the model a typical patient is defined who is representative of all patients. Siblings will be assumed to be over 45 years of age because most new probands will be over 45 years of age at diagnosis. Those who are monitored will be retested once according to clinical advice. Similarly, a typical child will be aged 25 years as there is a presumption against testing children for conditions that do not directly affect them during childhood. 102 At this age most will not have manifestations of iron overload but we assume that those who develop symptoms of iron overload will do so within 20 years and that the proportion of children with increased iron levels will rise linearly over this time period. The children will be tested every 5 years until iron overload is detected, i.e. a maximum of five times.

Data sources used in the models

This section describes the inputs to the models, provides justification for their use, details their respective sources and explains their role in the models. The data used in the models have been collected from systematic reviews and systematic searches discussed in more detail in Chapter 4 and Appendix 1. Data sources were chosen for the models on the basis of appropriateness to the UK and the quality of the data as assessed by the reviewers and in consultation with clinical experts.

The literature shows that a range of thresholds is used for the diagnostic tests, and there is a wide range of results in the accuracy of these tests. Table 15 shows the data used in the models and also the ranges of other data found in the systematic searches (Appendix 10). These ranges were used to inform appropriate sensitivity analysis ranges. The thresholds used here for modelling are TS > 45% and SF > 300 μg/l, as used in clinical guidelines. 72 The effect of using different thresholds has been shown as a sensitivity analysis below (see Table 22). The sensitivity and specificity for TS was also taken from the clinical guidelines. 72 The sensitivity and specificity of SF was taken from Moodie and colleagues61 as this UK study reported values for different thresholds of SF separately for men and women. Liver biopsy is used as the gold standard for confirming diagnosis from SF and TS tests and so the model assumed that liver biopsy was 100% accurate in diagnosing HHC.

Table 16 shows the key inputs to the SHTAC decision tree models and the ranges of these parameter values found in the data searches. The decision tree for people suspected of having haemochromatosis required an estimate of the prevalence of HHC in those who are referred with symptoms of HHC. The data for prevalence for this population was scarce with only one relevant study found. 61 The prevalence of HHC in a population with suspected iron overload was estimated from this study of 427 patients referred for investigation of liver disease from an ethnically mixed population in south London. 61 The prevalence was estimated by excluding those of Afro-Caribbean, African, Asian or Mediterranean origin, including only those with northern European or Celtic origins. The prevalence of HHC in relatives is estimated using simple genetic theory.

Although liver biopsy was associated with a small risk of death and other complications, for the base case we assumed that there were no deaths or major events from liver biopsy as in other modelling studies. 104

Vantyghem and colleagues94 described the makeup of 156 subjects recruited in the Endocrinology and Metabolism Department of Lille University who were referred because of general symptoms of iron overload and abnormal iron levels (SF > 300 μg/l or TS > 45%). Amongst other data they reported the numbers who were homozygous for the C282Y mutation with high SF (> 1000 μg/l).

Penetrance is the proportion of people homozygous for C282Y who go on to develop manifestations of the disease. The penetrance depends on the definition of disease (biochemical variables or fibrosis, etc.) and hence the value for the penetrance varies widely.

Results

Each of the decision tree models shown in Figures 2 and 3 were run with the parameters discussed in the sections above. The models estimated the number of patients detected with HHC and the numbers treated and monitored as appropriate. Cost data were used to estimate the total resource costs for each strategy and the strategies were compared according to cost per case of HHC detected.

Diagnostic tests in people suspected of having haemochromatosis

The results for the diagnostic test decision tree are for an average man of 45 years of age (as it is assumed that HHC will have become manifest by this age). The flow diagram (Figure 4) illustrates the diagnostic process and shows the numbers of people that are treated. The results in this section are presented per 100 people who have a positive TS and SF result (shown in dotted lines in the flow diagram). Based on the chosen accuracy of the diagnostic tests, there will be 100 positive test results for TS and SF for every 2880 people tested for suspected iron overload. Out of these 2880 people, 109 actually have HHC. Of those 100 people with positive test results, 75.1 are true positives (i.e. the PPV of the combined test is 75.1%). A total of 2780 people have a negative test result. Of these, 34.3 actually have the disease and are missed by the biochemical tests.

FIGURE 4.

Flow diagram of diagnostic tests (based on TS sensitivity = 94%, TS specificity = 94%, SF sensitivity = 73%, SF specificity = 85%, prevalence = 0.038, see Tables 15 and 16). SF, serum ferritin; TS, transferrin saturation.

Table 18 shows the results for the diagnostic tests decision tree, comparing liver biopsy with DNA testing for 100 people with signs and symptoms of iron overload and a positive test result for TS and SF. Each strategy detects a similar number of cases (75.1) and misses 1.2 per 100 tested with the initial TS and SF test as shown in Figure 4. Of those with a positive test result for TS and SF, 75.1% actually have HHC. In the liver biopsy strategy everyone receives a liver biopsy, whereas in the DNA strategy only those with a negative DNA test for HHC or high SF receive a liver biopsy. Some of those with a negative DNA test are monitored to see if their SF increases. Thus, the DNA strategy has fewer liver biopsies performed but more patients will be monitored than in the liver biopsy strategy. The extra costs for liver biopsy are more than the extra costs of monitoring patients and DNA tests and so the DNA strategy will be cost saving.

TABLE 18 - Base-case results for the diagnostic pathways decision tree, per 100 people with a positive test

	Liver biopsy, n	Cases detected, n	Monitor, n	Total cost, £	Cost saving/person tested, £	Cost saved/case detected, £
Liver biopsy	100.0	75.1	0	83,068
DNA strategy	41.1	75.1	22.9	73,823	92.45	123

The results vary according to the prevalence of the disease and the accuracy of the biochemical tests. This is illustrated by varying the PPV of the tests. PPV is the proportion of those with a positive biochemical test who have the disease, and it is dependent upon the prevalence of disease and the sensitivity and specificity of the tests:110

PPV = ( sensitivity ) ( prevalence ) / [ ( sensitivity ) ( prevalence ) + ( 1 – specificity ) ( 1 – prevalence ) ]

If the PPV increases, the number of cases detected increases and the number of cases monitored in the DNA strategy decreases, because there are fewer people with a negative DNA test. If the PPV decreases, the converse happens. The number of liver biopsies performed in the DNA strategy stays fairly constant irrespective of the PPV even though they may be in different arms of the decision tree shown in Figure 2. The total cost saving for both scenarios is similar at around £90 per person referred. Thus, the cost saving per case detected is higher when fewer cases are detected.

This is illustrated in Table 19 with a hypothetical PPV of 80%, for example through an increased prevalence of 0.05, in which case the cost saved per case detected is £115. With a hypothetical PPV of 44%, for example through a prevalence of 0.01, the cost saved per case detected is £216 (Table 20). The results are not greatly influenced by changes in costs related to the other parameters.

TABLE 19 - Base-case results for the diagnostic pathways decision tree with a positive predictive value of 0.8

	Liver biopsy, n	Cases detected, n	Monitor, n	Total cost, £	Cost saving/person tested, £	Cost saved/case detected, £
Liver biopsy	100.0	80.1	0	85,687
DNA test	41.8	80.1	19.4	76,467	92.20	115

TABLE 20 - Base-case results for the diagnostic pathways decision tree with a positive predictive value of 0.44

	Liver biopsy, n	Cases detected, n	Monitor, n	Total cost, £	Cost saving/person tested, £	Cost saved/case detected, £
Liver biopsy	100.0	43.5	0	66,465
DNA test	37.1	43.5	45.2	57,060	94.05	216

Sensitivity analyses

The parameters in the diagnostic pathways decision tree were varied in a series of sensitivity analyses and the results are shown in Table 21 and Figure 5. When possible the parameters were varied according to the ranges from the confidence intervals of these parameters, otherwise a suitable range was chosen after discussion with experts. The sensitivity analyses show that the conclusions from the decision tree are robust across all reasonable parameter ranges, that is, the DNA strategy is cost saving compared with the baseline strategy using liver biopsy. The results were most sensitive to the specificity of the TS test, the cost of the liver biopsy test and the proportion of people with a positive DNA test for the C282Y mutation and raised SF. Results were also estimated for cost per person tested. The cost saved per person tested varied between £47 and £138 for changes in the cost of the liver biopsy but it changed little for changes in TS specificity.

TABLE 21 - Sensitivity analyses for the diagnostic pathways decision tree

SF, serum ferritin; TS, transferrin saturation.

Results in the ‘low’ and ‘high’ columns use inputs from corresponding columns to give the range; the base-case results are shown in Table 18.

Variable	Inputs			Cost saved/case detected, £
	Base case	Lowa	Higha	Lowa	Higha	Range
Proportion raised SF > 1000 μg/l, DNA +ve	0.39	0.22	0.56	190	57	133
TS specificity, %	94	75	99	224	97	127
Costs liver biopsy test, £	388.05	310.44	465.66	62	184	122
Prevalence in suspected iron overload	0.038	0.016	0.06	169	111	58
Costs DNA test, £	100	80	120	150	97	53
SF specificity, %	85	75	99	144	93	51
Prevalence of genetic mutation	91.3	90	100	119	150	31
SF sensitivity, %	73	50	99	151	122	29
Proportion raised SF > 1000 μg/l, DNA –ve	0.24	0.16	0.32	132	115	17
TS sensitivity, %	94	75	99	131	122	9

FIGURE 5.

Tornado plot for sensitivity analyses for the diagnostic pathways decision tree. SF, serum ferritin; TS, transferrin sensitivity.

Changing the threshold for the biochemical tests

The threshold values for positive TS and SF tests in the decision tree were based upon those suggested in guidelines. However, other thresholds have been suggested in the literature. Therefore a sensitivity analysis was performed to investigate the effect of changing the threshold values for the biochemical tests. Adams and Chakrabarti111 reported the accuracy of the TS test for different thresholds of 40%, 46% and 55%. Table 22 shows the effect of using these different thresholds for TS. With a higher threshold there are fewer positive test results for TS, fewer people have SF tests and thus more people with the disease are missed. In fact, more of those with positive test results have the condition (PPV). For example, with a TS threshold of 55%, the PPV is 100% and the cost saved per case detected is £91. On the other hand, with a threshold of 40%, the PPV is 70% and the cost saved per case detected is £135.

TABLE 22 - Model results using different thresholds for transferrin saturationa

PPV, positive predictive value; SF, serum ferritin; TS, transferrin saturation.

For 1000 patients tested with TS and SF.

Threshold	Sensitivity of TS, %	Specificity of TS, %	Positive TS and SF, n	Negative TS and SF, n	True positive, n (PPV)	False negative	Cost saved/case detected, £
40%	92	92	37	963	26 (70%)	12	135
46%	89	95	32	968	25 (78%)	13	119
55%	77	100	21	979	21 (100%)	17	91

The decision tree model was run for 45-year-old women. In this group a threshold for SF of 200 μg/l has been suggested in the literature. 72 Using this threshold with other parameters unchanged there were 60.1 cases detected and there was a cost saving per case detected for the DNA strategy of £155.

Family testing

Siblings

Table 23 shows the results from the family testing decision tree model (Figure 3) for male siblings of 45 years of age. The biochemical strategy tests all siblings with biochemical tests whereas the alternative strategy uses DNA tests. Each strategy detects 19% and misses 6% of cases of HHC. More people without HHC are treated and monitored in the biochemical testing strategy than with the DNA test strategy because those with negative biochemical tests are monitored. In the DNA strategy, although fewer people are monitored, the cost is higher than the cost of the biochemical strategy because the cost of monitoring is much cheaper than the cost of the DNA test. If the cost of the DNA test were to fall from £100 to £60, the DNA strategy would be the cheaper one. If those who were monitored were tested twice (instead of only once), the DNA strategy becomes cost saving (£79 per case detected).

TABLE 23 - Base-case results for the family testing decision tree for male siblings, per 100 people tested

	Cases detected, n	Cases treated, n	Monitor, n	DNA test, n	Total cost, £	Additional cost/person tested, £	Additional cost/case detected, £
Biochemical	19.0	19.7	86.2	0.0	23,628
DNA testing	19.0	19.1	11.9	100.0	27,423	37.95	200

An alternative strategy was run for the biochemical tests. In this case, if relatives have raised iron levels, i.e. TS > 45% and SF > 300 μg/l, they will be treated as before. If the SF is < 200 μg/l, they will be monitored for a number of months to see if their iron levels increase and, if the iron levels increase, they will be treated for HHC, otherwise they will be discharged. If the SF is between 200 and 300 μg/l they will receive a DNA test. Those with a positive DNA test will be treated and those with a negative DNA test will be monitored. This scenario is very similar to the baseline biochemical test strategy because there are few people with a SF between 200 and 300 μg/l; it cost an extra £3 per case detected compared with the baseline biochemical test strategy.

The parameters in the family testing decision tree for male siblings were varied in sensitivity analyses and the results are shown in Table 24. The sensitivity analyses show that the conclusions from the decision tree are robust across all reasonable parameter ranges, i.e. the DNA strategy is more expensive than the baseline biochemical strategy. The most sensitive parameters were the cost of the DNA test and the specificity of the TS test. Results were also estimated for cost per person tested, and show that the additional cost per person tested varied between £18 and £58 for changes in the price of the DNA test.

TABLE 24 - Sensitivity analyses for the family testing decision tree for male siblings

SF, serum ferritin; TS, transferrin saturation.

Results in the ‘low’ and ‘high’ columns use inputs from corresponding columns to give range; base-case results are shown in Table 23.

Variable	Base case	Inputs		Additional cost per case detected, £
		Lowa	Higha	Lowa	Higha	Range
Cost of DNA test, £	100	80	120	94	305	211
Cost of monitoring, £	71.15	56.92	85.38	256	144	112
Penetrance, %	76	60	93	253	163	90
TS specificity, %	94	75	99	149	211	62
SF specificity, %	85	75	99	189	215	26

A sensitivity analysis was run for women of 45 years of age with a threshold for SF of 200 μg/l and a penetrance of 0.32%. 60 In this case there were eight cases detected and an additional cost per case detected for the DNA strategy of £436 compared with the biochemical strategy.

Offspring

Table 25 shows the results from the family testing decision tree model for offspring of 25 years of age (Figure 3). The biochemical strategy tests all offspring with biochemical tests whereas the alternative strategy uses DNA tests. The DNA test strategy is cheaper than the baseline biochemical testing strategy and there are a similar number of HHC cases detected using both strategies. In the biochemical test strategy there are many more people monitored than in the DNA strategy. In this case, people are monitored five times, once every 5 years, and thus the monitoring cost is higher than the DNA test cost and so there is a cost saving for using DNA tests. These results assumed that 20% of offspring with HHC showed manifestations of the disease at the time of testing. The results are not very sensitive to this assumption.

TABLE 25 - Base-case results for the family testing decision tree for offspring, per 100 people tested

	Cases detected, n	Cases treated, n	Monitor, n	DNA test, n	Total cost, £	Cost saved/person tested, £	Cost saved/case detected, £
Biochemical	3.5	4.4	98.6	0.0	46,753
DNA testing	3.5	3.6	4.4	100.0	18,638	281.15	7982

The parameters in the family testing decision tree for offspring were varied in sensitivity analyses and the results are shown in Table 26 and Figure 6. The sensitivity analyses show that the conclusions from the decision tree are valid across all reasonable parameter ranges, i.e. the DNA strategy is cost saving compared with the baseline strategy. The most sensitive parameters were the number of times that the offspring are monitored and the penetrance of HHC. Results were also estimated for cost per person tested and show that the cost saved per person tested varied between £139 and £423 for changes in the frequency of monitoring.

TABLE 26 - Sensitivity analyses for the family testing decision tree for offspring

Results in the ‘low’ and ‘high’ columns use inputs from corresponding columns to give range; base-case results are shown in Table 25.

Variable	Base case	Inputs		Cost saved per case detected, £
		Lowa	Higha	Lowa	Higha	Range
Number of monitorings	5	3	7	4144	11,786	7642
Penetrance, %	76	60	93	10,111	6523	3588
TS specificity, %	94	75	99	9867	7430	2437
Cost of DNA test, £	100	80	120	8550	7415	1135
SF specificity, %	85	75	99	8406	7365	1041
Cost of monitoring, £	71.15	56.92	85.38	8366	7599	767
Initial proportion with iron overload	0.2	0	0.5	8141	7756	385

FIGURE 6.

Tornado plot of sensitivity analyses for family testing decision tree for offspring.

A sensitivity analysis was run for women of age 45 years with a threshold for SF of 200 μg/l and a penetrance of 0.32%. 60 In this case, there were 1.5 cases detected and a cost saved per case detected for the DNA strategy of £18,958.

Systematic review of economic evaluations

• Two studies were identified that met the inclusion criteria for the review.

• One study estimated the cost-effectiveness of screening for HHC in family members using genetic tests compared with a phenotypic test and no screening using a cost–utility model.

• The other study investigated the likely cost of genotyping spouses using a cost-minimisation model and whether this would reduce the number of investigations of children.

• Both studies were of reasonable quality when assessed against standard criteria, but as both studies were conducted in North America their generalisability to the UK is not clear.

• Gene testing was found to be a cost-effective method of screening relatives of patients with HHC in one study.

• In the other study genotyping the spouse of a homozygote was found to be the most cost-efficient strategy in family testing because it leads to more selective investigation of children for the HFE gene.

De novo model (SHTAC model)

Clinical validity

Eleven studies were identified that could be used to estimate the clinical validity of genotyping for the C282Y mutation for the diagnosis of HHC. The quality of the studies using the criteria developed for this review was variable and the studies used a range of definitions for the clinical phenotype. The clinical sensitivity of C282Y homozygosity for HHC ranged from 28.4% to 100% in the 11 studies. When using only the studies most likely to have correctly defined haemochromatosis, which reported ruling out other causes of iron overload and related patients and which included the most northerly populations, sensitivity ranged from 91.3% to 92.4%. Clinical specificity ranged from 98.8% to 100%.

Clinical utility

No clinical effectiveness studies meeting the inclusion criteria for the review were identified. Two cost-effectiveness studies were identified. One study estimated the cost-effectiveness of screening for HHC in family members using genetic tests compared with phenotypic tests and no screening using a cost–utility model. The other study investigated the likely cost of genotyping spouses using a cost-minimisation model. Both studies were of reasonable quality when assessed against standard criteria, but as they were conducted in North America their generalisability to the UK is not clear. Gene testing was found to be a cost-effective method of screening relatives of patients with HHC. Genotyping the spouse of a homozygote was the most cost-efficient strategy in family testing because it leads to more selective investigation of children for the HFE gene.

Psychosocial aspects of DNA testing

Three cohort studies met the inclusion criteria for the review. Each study assessed and reported on the psychosocial outcomes of genetic testing for HFE in a different way. All of the studies had methodological limitations and the generalisability of these studies is difficult to determine. Generally the results suggest that genetic testing in the case of haemochromatosis is well accepted, is accompanied by few negative psychosocial outcomes and may lead to reduced anxiety. Control subjects in the one study that had a control group anticipated greater anxiety, depression, anger and difficulty in affording the genetic test than was reported by patients. In one study, clinically affected participants had significantly lower health-related quality of life as measured by the SF-36 PCS before genetic testing than unaffected participants, but this was no longer significantly different at 12 months post consultation. Another study reported significant improvements in the vitality subscale of the SF-36 and the PCS after participants were informed of their genetic test result. For generalised anxiety scores or intrusive thoughts, one study reported no statistically significant differences between clinically affected and clinically unaffected participants before and after genetic testing; another study reported that anxiety fell significantly in C282Y homozygotes and heterozygotes once they received their genetic testing results.

Economic evaluation

The de novo economic model found that for people suspected of having haemochromatosis the DNA strategy is cost saving compared with the baseline strategy (cost saved per case detected £123). This is because the cost savings that result from the reduced number of liver biopsies being performed are greater than the increased costs of monitoring. For family testing the DNA strategy is not cost saving in the case of testing siblings as the DNA test costs are higher than the reduced monitoring costs (additional cost per case detected £200). If the cost of the DNA test were to fall from £100 to £60, the DNA strategy would be the cheaper one. For family testing in the case of offspring of people with haemochromatosis, the DNA test strategy is cheaper than the baseline biochemical testing strategy (cost saved per case detected £7982). Sensitivity analyses show that the conclusions in each case are robust across all reasonable parameter values.

General

• As mentioned in Chapter 3 the method used to evaluate DNA testing for HHC was largely based on the ACCE model developed by the Office of Genomics and Disease Prevention (Center for Disease Control, Atlanta, USA) and consists of systematic reviews of clinical validity, clinical utility and psychosocial aspects, with the development of a de novo economic model. The literature on the use of DNA testing for haemochromatosis is extensive; however, studies mostly consider gene frequencies in different populations, phenotypic and genotypic associations/correlations and population screening. The type of study employed is not always obvious from the title and abstract of publications. Finding studies that could be used for the different elements of the review was problematic, particularly for assessing the clinical validity of a genetic test. As such, a pragmatic approach was taken by developing different inclusion criteria for the different systematic reviews to ensure that relevant information was identified whilst retaining the focus of the review and allowing manageable synthesis of evidence.

Clinical validity

• The traditional diagnostic test assessment study that estimates the clinical sensitivity and clinical specificity of a test requires the new test to be compared with a reference standard (gold standard). However, this does not apply in the case of genetic tests for which the gold standard entails gene sequencing to detect mutations. 113 Potential alternative gold standards in the case of haemochromatosis are no longer used for diagnostic purposes, such as liver biopsy, which is used mostly for prognosis, and haplotyping, which has been superseded since the discovery of the HFE gene. As such, traditional diagnostic test assessment studies are inappropriate and are not available or applicable in this case.

• The ideal way to assess the clinical validity of DNA testing for haemochromatosis would be to follow a large group of individuals through to expression of phenotypic disease and perform genetic testing. No such population-based cohort studies are available and such studies could be considered unethical and are therefore unlikely. As such, clinical validity has been assessed here by considering studies that identify a group of individuals who have the primary iron overload phenotype and then determining the proportion who are C282Y homozygous and by comparing them with a control group. The control subjects are individuals who do not have the phenotype of interest according to the disease case definition. 114 It is only through the use of such control subjects that an assessment of specificity can be made. The use of such studies has limitations though in that individuals in the control group who test positive may yet go on to develop the phenotypic disease. Although this risk is likely to be small because of the prevalence and penetrance of HHC, it must be acknowledged that calculation of the specificity of the genetic mutation (C282Y) for HHC from such studies may not be accurate. Specificity of 100% for the homozygous C282Y mutation in patients with phenotypic expression has been reported. 115

• Studies that find the percentage of individuals who have the primary iron overload phenotype and who have a positive test result for C282Y homozygosity to give the clinical sensitivity of the mutation are in effect gene frequency studies. As such, there is some overlap between the epidemiology studies and those included in the clinical validity section.

• An associated difficulty in the case of HHC is specifying the exact diagnostic criteria or reference standard used to define the condition. The disease may sometimes be defined phenotypically by the presence of certain signs and symptoms and at other times genotypically by reference to the mutations that give rise to the disease. It has been suggested that the definition of a genetic disease should require both the clinical manifestations and the presence of the mutation; however, it has also been suggested that the presence of either the clinical features or the mutation suffices for a definitive diagnosis. In assessing genetic testing, as in the case of haemochromatosis, the definition of the disease should be by reference to signs and symptoms or clinical features but not by reference to genotype. 114 The definition of cases may vary from signs and symptoms to confirmed iron overload with clinical manifestations and this can affect the characteristics of the genetic test and the reported clinical validity (see below).

• Clinical sensitivity from the included studies ranges from 28% to 100%. It can be seen from these results that the reliability of estimating clinical sensitivity of C282Y homozygosity for HHC is particularly susceptible to the definition of the clinical phenotype used in the studies. When the definition is more rigorous and strict criteria for phenotypic expression are used, the clinical sensitivity increases.

• There are also difficulties associated with assessing a prognostic or predictive test when the genotype is more highly prevalent than the phenotype. In classic rare single gene disorders the genotype is an accurate predictor of the phenotype because both genotype and phenotype are sufficiently rare that if they occur together the predictive value is high. This is not the case for haemochromatosis and will not be the case for the other common complex genetic diseases.

• Quality issues of diagnostic test studies are different from those in effectiveness studies and the quality assessment of clinical validity studies in the context of DNA are even further removed from the usual issues. Therefore, as the clinical validity studies did not conform to typical diagnostic accuracy studies or observational studies, a modified Spitzer assessment tool, incorporating elements from QUADAS, was used, which concentrated on the issues of particular relevance to the studies in question. That is, aspects considered for quality assessment included whether there was adequate description of the haemochromatosis group of patients; the likelihood that the definition of disease was an accurate reflection of the disease and that the patients were a representative sample; whether the control population was adequately described and appropriate; whether the groups were comparable in terms of factors such as age, sex and ancestry; whether the DNA tests were adequately described and appropriate; whether outcomes were fully reported; whether there was mention of missing data; and the generalisablity of results.

Clinical utility

• The best evidence for the clinical utility of a diagnostic test is an RCT with patients randomly assigned to alternative diagnostic strategies with clinical or cost-effectiveness reported. 85 In the absence of RCTs it was intended to assess the clinical effectiveness of DNA testing by using the highest level of evidence available that considered suspected cases of HHC (or relatives of cases) and comparative testing strategies and reported patient-based outcomes. No such studies with appropriate designs and outcomes were found. Some studies that purported to be clinical utility studies were problematic in that they reported gene frequency without any clinical or cost-effectiveness measure. Another example of difficulties associated with the literature was that some studies which initially appeared to be comparing diagnostic strategies in fact compared different groups of patients using biochemical and DNA testing algorithms rather than suspected HHC patients tested by different diagnostic algorithms with or without DNA tests to assess the utility of DNA testing.

• Two cost-effectiveness studies were identified. Although these were conducted in North America and used different approaches from that used in the current project, they support the case that incorporating DNA testing in diagnostic algorithms is likely to be a cost-saving strategy.

Psychosocial aspects

• The aims of the psychosocial section of the review were very specific: to compare the psychosocial benefits and harms of adding DNA testing to diagnostic algorithms. However, there were few studies that could be included in the review. An ideal study might have randomised people with suspected HHC and first-degree relatives into a trial in which participants in one arm received standard biochemical tests to diagnose HHC and those in the second arm received DNA testing in addition to the standard biochemical tests. With data collection on psychosocial outcomes carried out both before and after the intervention and disclosure of test results, such a study could provide an indication of whether the prospect of DNA testing for HHC is any more stressful than that of biochemical testing. Additionally, such a study might also indicate whether receiving DNA test results has a greater positive or negative psychosocial impact than receiving biochemical test results. Longer-term follow-up would be essential to capture data for outcomes such as treatment compliance, discrimination and stigmatisation.

• The three studies included in this report took place in the USA, Canada and Australia and therefore it is important to consider the nature of the health-care systems in these countries as this might impact on the transferability of the results to the UK setting. In the USA patients must either have health insurance or pay for their medical care directly themselves. Therefore, in the USA people might be expected to have more worries about financial and insurance issues in relation to a diagnosis of haemochromatosis than those in Canada and Australia, which have publicly funded health insurance plans, or those in the UK, where health care is provided by the NHS.

• The three included cohort studies employed different methods and reported results in different ways, which made it difficult to synthesise the evidence and draw any firm conclusions from the review. Nevertheless a common theme emerges. After people were informed of their test results anxiety levels fell or remained at pretest result levels and this was mirrored in the results for general health-related quality of life, which either improved in some aspects or stayed constant with respect to pretest result values. This suggests that when the genetic test result confirms a diagnosis of HHC this does not have a negative impact in terms of anxiety or health-related quality of life, and indeed there is some evidence to suggest that once the test result is received anxiety levels may fall and health-related quality of life can improve.

• A second common theme that emerges from two of the included studies is that most patients are able to correctly recall the information that they have been given about haemochromatosis. However, there were areas in which recall and/or understanding were poor, for example understanding the difference between HFE genotyping and the TS test result, and knowing that the C282Y mutation is found in most people with haemochromatosis. Of particular potential concern is the finding in one study that 34% of participants did not recall having received a genetic test result, and in the second study that between 16% and 21% of homozygotes, heterozygotes and compound heterozygotes were unable to remember their mutation status. This highlights the importance of comprehensive pretest and post-test counselling to ensure that patients fully understand the test results that they receive.

• The limited evidence base in this area does not address whether there might be different psychosocial effects of testing, depending on whether an individual is referred for genetic testing because they have signs and symptoms suggestive of HHC or whether they are asymptomatic and have been referred for testing following diagnosis of HHC in a family member.

• Population screening studies were not included in this review and a comprehensive search for such studies was not undertaken. However, those population screening studies that were excluded from this review (listed in Appendix 3) report findings that are similar to those reported here. For instance, when phenotypic and genotypic screening strategies have been compared, both have been acceptable79,116 and little psychological disturbance has been apparent in the short term with either strategy. 78,117 Negative psychosocial consequences have been reported to be rare. 118

• Although the evidence shows that there are not likely to be detrimental psychological effects of genetic testing it has been suggested that patients would benefit from routine assessment of psychological distress and that referral to a mental health professional should be available for those whose levels suggest a need for clinical intervention. 97

Economic evaluation

Genetic exceptionalism and phenotype versus genotype

• It is sometimes argued that information derived from a genetic source has special properties over that derived from a phenotypic source. If this were the case it could be argued that the identification of genetic susceptibility in haemochromatosis – the ‘at-risk genotype’ – has different properties to the identification of the iron overload state by phenotypic means. However, the diagnosis of haemochromatosis is made by combining the genetic information with the phenotypic information. It is suggested that for clinical purposes the case definition should be homozygosity for the common C282Y mutation with a raised SF. The person with homozygosity and normal iron studies is at risk of developing HFE-related iron overload but does not have haemochromatosis. The person who is a compound heterozygote with raised SF with no other explanation could be regarded as having HFE-related haemochromatosis. There will be a small group of patients who have iron overload with no other explanation who are negative for the common mutations in the HFE gene. These may be classified as atypical haemochromatosis and in a routine clinical setting further genetic analysis is impractical. Using this combination of genotypic and phenotypic information is probably the best scenario for clinical decision-making but does of course make it difficult to conduct classic studies comparing genotype and phenotype. 121

• The special properties of information derived from DNA-based analysis are said to relate to issues such as concern for kin, potential for discrimination and stigmatisation, the potential for long-term storage and ease of access to samples for future analysis, and predictive value. However, all of these characteristics can be applied to medical and health information derived from non-genetic sources. More careful analysis suggests that it is not the method by which the information is derived that gives it special status but the context in which that information is used. 76 In the case of haemochromatosis, as outlined above, the clinical definition relies on both genotypic and phenotypic information. The context of the information is what defines the property of that information, not how it is derived. The results of the genetic analysis can be used to diagnose those who might benefit from treatment, to predict those at risk of developing iron overload and to rule it out in those not at risk.

Published

Bryant J, Cooper K, Picot J, Clegg A, Roderick P, Rosenberg W, et al. A systematic review of the clinical validity and clinical utility of DNA testing for hereditary haemochromatosis type 1 in at-risk populations. J Med Genet 2008;45:513–18.

Cooper K, Bryant J, Picot J, Clegg A, Roderick P, Rosenberg W, et al. A decision analysis model for diagnostic strategies using DNA testing for hereditary haemochromatosis in at-risk populations. QJM 2008;101:631–41.

Accepted for publication

Picot J, Bryant J, Cooper K, Clegg A, Roderick P, Rosenberg W, et al. Psychosocial aspects of DNA testing for hereditary hemochromatosis in at-risk individuals: a systematic review. Genet Test Mol Biomark 2009;13:7–14.

Methods for reviewing effectiveness

The a priori methods used for the review are outlined in the following sections. The sources of information used are outlined in Appendix 2.

Clinical validity and utility

Adams PC, Chakrabarti S. Genotypic/phenotypic correlations in genetic hemochromatosis: evolution of diagnostic criteria. Gastroenterology 1998;114:319–23. (Location Canada.)

Adams PC. Implications of genotyping of spouses to limit investigation of children in genetic hemochromatosis. Clin Genet 1998;53:176–8. (Location Canada: included under cost-effectiveness studies.)

Adams PC, Kertesz AE, McLaren CE, Barr R, Bamford A, Chakrabarti S. Population screening for hemochromatosis: a comparison of unbound iron-binding capacity, transferrin saturation, and C282Y genotyping in 5,211 voluntary blood donors. Hepatology 2000;31:1160–4. (Location Canada.)

Asberg A, Hveem K, Thorstensen K, Ellekjter E, Kannelonning K, Fjosne U, 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. (No control group and no strategy comparison.)

Bacon BR, Olynyk JK, Brunt EM, Britton RS, Wolff RK. HFE genotype in patients with hemochromatosis and other liver diseases. Ann Intern Med 1999;130:953–62. (Location USA.)

Bartolo C, McAndrew PE, Sosolik RC, Cawley KA, Balcerzak SP, Brandt JT, et al. Differential diagnosis of hereditary hemochromatosis from other liver disorders by genetic analysis: gene mutation analysis of patients previously diagnosed with hemochromatosis by liver biopsy. Arch Pathol Lab Med 1998;122:633–7. (Location USA.)

Benn HP, Nielsen P, Fischer R, Schwarz D, Engelhardt R, Darda C, et al. Screening for hereditary hemochromatosis in prospective blood donors. Beitr Infusionsther Transfusionsmed 1994;32:314–6. (No control group and no strategy comparison.)

Beutler E, Felitti V, Gelbart T, Ho N. The effect of HFE genotypes on measurements of iron overload in patients attending a health appraisal clinic. Ann Intern Med 2000;133:329–37. (Location USA.)

Cadet E, Capron D, Gallet M, Omanga-Leke ML, Boutignon H, Julier C, et al. Reverse cascade screening of newborns for hereditary haemochromatosis: a model for other late onset diseases? J Med Genet 2005;42:390–5. (No control group and no strategy comparison.)

Cadet E, Capron D, Perez AS, Crepin SN, Arlot S, Ducroix JP, et al. A targeted approach significantly increases the identification rate of patients with undiagnosed haemochromatosis. J Intern Med 2003;253:217–24. (No strategy comparison.)

Cavanaugh JA, Wilson SR, Bassett ML. Genetic testing for HFE hemochromatosis in Australia: the value of testing relatives of simple heterozygotes. J Gastroenterol Hepatol 2002;17:800–3. (Location Australia and no control group.)

Datz C, Lalloz MRA, Vogel W, Graziadei I, Hackl F, Vautier G, et al. Predominance of the HLA-H Cys282Tyr mutation in Austrian patients with genetic haemochromatosis. J Hepatol 1997;27:773–9. (Location Austria.)

Emery J, Rose P, Harcourt J, Livesey K, Merryweather-Clarke A, Pointon JJ, et al. Pilot study of early diagnosis of hereditary haemochromatosis through systematic case finding in primary care. Community Genet 2002;5:262–5. (No control group and unclear case definition.)

Gleeson F, Ryan E, Barrett S, Crowe J. Clinical expression of haemochromatosis in Irish C282Y homozygotes identified through family screening. Eur J Gastroenterol Hepatol 2004;16:859–63. (No control group, no strategy comparison)

Guttridge MG, Carter K, Worwood M, Darke C. Population screening for hemochromatosis by PCR using sequence-specific primers. Genet Test 2000;4:111–14. (No control group and no strategy comparison.)

Hannuksela J, Niemela O, Leppilampi M, Parkkila AK, Koistinen P, Nieminen P, et al. Clinical utility and outcome of HFE-genotyping in the search for hereditary hemochromatosis. Clin Chim Acta 2003;331:61–7. (No control group and no strategy comparison.)

Jackson HA, Bowen DJ, Worwood M. Rapid genetic screening for haemochromatosis using heteroduplex technology. Br J Haematol 1997;98:856–9. (No control group and no strategy comparison.)

Jezequel P, Bargain M, Lellouche F, Geffroy F, Dorval I. Allele frequencies of hereditary hemochromatosis gene mutations in a local population of west Brittany. Hum Genet 1998;102:332–3. (No defined HHC group and no strategy comparison.)

Jorquera F, Dominguez A, az-Golpe V, Espinel J, Munoz F, Herrera A, et al. C282Y and H63D mutations of the haemochromatosis gene in patients with iron overload. Rev Esp Enferm Dig 2001;93:298–302. (Location Spain; excludes southern Europeans.)

Lombard M, Craven C, Spencer S, Crowe J. Predictive accuracy of biochemical tests for idiopathic hemochromatosis in an Irish population. Ir J Med Sci 1986;155:248–9. (No control group and no strategy comparison.)

Ludwig J, Batts KP, Moyer TP, Baldus WP, Fairbanks VF. Liver biopsy diagnosis of homozygous hemochromatosis: a diagnostic algorithm. Mayo Clin Proc 1993;68:263–7. (Liver biopsy study.)

McCune CA, Ravine D, Carter K, Jackson HA, Hutton D, Hedderich J, et al. Iron loading and morbidity among relatives of HFE C282Y homozygotes identified either by population genetic testing or presenting as patients. Gut 2006;55:554–62. (No control group and no strategy comparison.)

McGrath JS, Deugnier Y, Moirand R, Jouanolle AM, Chakrabarti S, Adams PC. A nomogram to predict C282Y hemochromatosis. J Lab Clin Med 2002;140:6–8. (Location Canada.)

Milman N, Koefoed P, Pedersen P, Nielsen FC, Eiberg H. Frequency of the HFE C282Y and H63D mutations in Danish patients with clinical haemochromatosis initially diagnosed by phenotypic methods. Eur J Haematol 2003;71:403–7. (No control group and no strategy comparison.)

Moodie SJ, Ang L, Stenner JM, Finlayson C, Khotari A, Levin GE, et al. Testing for haemochromatosis in a liver clinic population: relationship between ethnic origin, HFE gene mutations, liver histology and serum iron markers. Eur J Gastroenterol Hepatol 2002;14:223–9. (No control group and no strategy comparison.)

Mundy L, Merlin T, Bywood P, Parrella A. MRI for the assessment of liver iron stores: management of patients with suspected or diagnosed haemochromatosis. Horizon Scanning Prioritising Summary – Volume 4. Adelaide: Adelaide Health Technology Assessment (AHTA) on behalf of National Horizon Scanning Unit (HealthPACT and MSAC); 2004. (MRI for liver iron.)

Murtagh LJ, Whiley M, Wilson S, Tran H, Bassett ML. Unsaturated iron binding capacity and transferrin saturation are equally reliable in detection of HFE hemochromatosis. Am J Gastroenterol 2002;97:2093–9. (Location Australia and no control group and no strategy comparison.)

Njajou OT, Houwing-Duistermaat JJ, Osborne RH, Vaessen N, Vergeer J, Heeringa J, et al. A population-based study of the effect of the HFE C282Y and H63D mutations on iron metabolism. Eur J Hum Genet 2003;11:225–31. (No control group and no strategy comparison.)

O’Hara R, Cavanagh N, Cassidy M, Cullina M. The role of transferrin saturation as a screening test for hereditary haemochromatosis in an Irish population seeking medical care. Ann Clin Biochem 2003;40:169–74. (No control group and no strategy comparison.)

Palomaki GE, Haddow JE, Bradley LA, Richards CS, Stenzel TT, Grody WW. Estimated analytic validity of HFE C282Y mutation testing in population screening: the potential value of confirmatory testing. Genet Med 2003;5:440–3. (Location USA.)

Parks SB, Popovich BW, Press RD. Real-time polymerase chain reaction with fluorescent hybridization probes for the detection of prevalent mutations causing common thrombophilic and iron overload phenotypes. Am J Clin Pathol 2001;115:439–47. (Location USA.)

Patch C, Rosenberg W, Roderick P. Comparison of two screening strategies for haemochromatosis: a pilot study investigating uptake, feasibility and cost. J Med Genet 2001;38:S31. (No control group and population screening setting.)

Patch C, Roderick P, Rosenberg W. Prevalence and burden of disease in hemochromatosis: estimates derived from routine data. Nurs Health Sci 2006;8:128–9. (No control group and no strategy comparison.)

Poullis A, Moodie SJ, Ang L, Finlayson CJ, Levin GE, Maxwell JD. Routine transferrin saturation measurement in liver clinic patients increases detection of hereditary haemochromatosis. Ann Clin Biochem 2003;40:521–7. (No control group and no strategy comparison.)

Press RD, Flora K, Gross C, Rabkin JM, Corless CL. Hepatic iron overload: direct HFE (HLA-H) mutation analysis vs quantitative iron assays for the diagnosis of hereditary hemochromatosis [see comment]. Am J Clin Pathol 1998;109:577–84. (Location USA.)

Rossi E, Henderson S, Chin CY, Olynyk J, Beilby JP, Reed WD, et al. Genotyping as a diagnostic aid in genetic haemochromatosis. J Gastroenterol Hepatol 1999;14:427–30. (Location Australia and no control group and no strategy comparison.)

Ryan E, Byrnes V, Coughlan B, Flanagan AM, Barrett S, O’Keane JC, et al. Underdiagnosis of hereditary haemochromatosis: lack of presentation or penetration? Gut 2002;51:108–12. (No control group and no strategy comparison.)

Stave GM, Mignogna JJ, Powell GS, Hunt CM. Evaluation of a workplace hemochromatosis screening program. Am J Prev Med 1999;16:303–6. (Location USA.)

Watkins S, Thorburn D, Joshi NG, Neilson M, Joyce T, Spooner R, et al. Biochemical and clinical penetrance of individuals diagnosed with haemochromatosis by predictive genetic testing. Gut 2004;53:A86–7. (No control group and no strategy comparison.)

Whiting PW, Fletcher LM, Dixon JK, Gochee P, Powell LW, Crawford DH. Concordance of iron indices in homozygote and heterozygote sibling pairs in hemochromatosis families: implications for family screening. J Hepatol 2002;37:309–14. (Australia and no control group and no strategy comparison.)

Wrede CE, Hutzler S, Bollheimer LC, Buettner R, Hellerbrand C, Schoelmerich J, et al. Correlation between iron status and genetic hemochromatosis (codon C282Y) in a large German population. Isr Med Assoc J 2004;6:30–3. (No control group and no strategy comparison.)

Psychosocial aspects

Anderson RT, Press N, Tucker DC, Snively BM, Wenzel L, Ellis SD, et al. Patient acceptability of genotypic testing for hemochromatosis in primary care. Genet Med 2005;7:557–63. (Subjects had not been diagnosed with haemochromatosis; population screening context.)

Bassett M, Dunn C, Battese K, Peek M. Acceptance of neonatal genetic screening for hereditary hemochromatosis by informed parents. Genet Test 2001;5:317–20. (Subjects had not been diagnosed with haemochromatosis; prenatal screening context.)

Gason AA, Aitken MA, Metcalfe SA, Allen KJ, Delatycki MB. Genetic susceptibility screening in schools: attitudes of the school community towards hereditary haemochromatosis. Clin Genet 2005;67:166–74. (Subjects had not been diagnosed with haemochromatosis; school pupils attitudes to screening.)

Hall MA, McEwen JE, Barton JC, Walker AP, Howe EG, Reiss JA, et al. Concerns in a primary care population about genetic discrimination by insurers. Genet Med 2005;7:311–6. (Subjects had not been diagnosed with haemochromatosis; population screening context.)

Hicken BL, Calhoun DC, Tucker DC. Genetic testing for hemochromatosis: attitudes and acceptability among young and older adults. Genet Test 2003;7:235–9. (Subjects had not been diagnosed with haemochromatosis.)

Hicken BL, Foshee A, Tucker DC. Perceptions and attitudes about HFE genotyping among college-age adults. J Genet Couns 2005;14:465–72. (Subjects had not been diagnosed with haemochromatosis.)

Paglierani LM, Kalkwarf HJ, Rosenthal SL, Huether CA, Wenstrup RJ. The impact of test outcome certainty on interest in genetic testing among college women. J Genet Couns 2003;12:131–50. (Subjects had not been diagnosed with haemochromatosis.)

Patch C, Roderick P, Rosenberg W. Comparison of genotypic and phenotypic strategies for population screening in hemochromatosis: assessment of anxiety, depression, and perception of health. Genet Med 2005;7:550–6. (Subjects had not been diagnosed with haemochromatosis; population screening context.)

Patch C, Roderick P, Rosenberg W. Factors affecting the uptake of screening: a randomised controlled non-inferiority trial comparing a genotypic and a phenotypic strategy for screening for haemochromatosis [see comment]. J Hepatol 2005;43:149–55. (Subjects had not been diagnosed with haemochromatosis; population screening context.)

Patch C, Roderick P, Rosenberg W. Feasibility and acceptability of two screening strategies for haemochromatosis, report of phase one of a randomised controlled trial. Eur J Hum Genet 2002;10:309. (Subjects had not been diagnosed with haemochromatosis; population screening context.)

Patch C, Roderick P, Rosenberg WM. Psychological effects of genetic and biochemical population screening for hemochromatosis. Hepatology 2004;40:576A. (Subjects had not been diagnosed with haemochromatosis; population screening context.)

Power TE, Adams PC. Psychosocial impact of genetic testing for hemochromatosis in population screening and referred patients. Hepatology 2000;32:413A. (Abstract; full paper available and is included.)

Salkovskis PM, Dennis R, Wroe AL. An experimental study of influences on the perceived likelihood of seeking genetic testing: ‘nondirectiveness’ may be misleading. J Psychosom Res 1999;47:439–47. (Subjects had not been diagnosed with haemochromatosis; population screening context.)

Shaheen NJ, Lawrence LB, Bacon BR, Barton JC, Barton NH, Galanko J, et al. Insurance, employment, and psychosocial consequences of a diagnosis of hereditary hemochromatosis in subjects without end organ damage. Am J Gastroenterol 2003;98:1175–80. (Not about impact of genetic testing; impact of HC diagnosis.)

Stuhrmann M, Hoy L, Nippert I, Schmidtke J. Genotype-based screening for hereditary hemochromatosis: II. Attitudes toward genetic testing and psychosocial impact – a report from a German pilot study. Genet Test 2005;9:242–54. (Population screening context.)

Tucker DC, Acton RT, Press N, Ruggiero A, Reiss JA, Walker AP, et al. Predictors of belief that genetic test information about hemochromatosis should be shared with family members. Genet Test 2006;10:50–9. (Subjects had not been diagnosed with haemochromatosis; population screening context.)

Epidemiology

Andrikovics H, Kalmar L, Bors A, Fandl B, Petri I, Kalasz L, et al. Genotype screening for hereditary hemochromatosis among voluntary blood donors in Hungary. Blood Cells Mol Dis 2001;27:334–41. (Location Hungary.)

Barton JC, Cheatwood SM, Key TJ, Acton RT. Hemochromatosis detection in a health screening program at an Alabama forest products mill. J Occup Environ Med 2002;44:745–51. (Location USA.)

Barton JC, Acton RT, Dawkins FW, Adams PC, Lovato L, Leiendecker-Foster C, et al. Initial screening transferrin saturation values, serum ferritin concentrations, and HFE genotypes in whites and blacks in the Hemochromatosis and Iron Overload Screening Study. Genet Test 2005;9:231–41. (Location USA.)

Barton JC, Acton RT, Lovato L, Speechley MR, McLaren CE, Harris EL, et al. Initial screening transferrin saturation values, serum ferritin concentrations, and HFE genotypes in Native Americans and whites in the Hemochromatosis and Iron Overload Screening Study. Clin Genet 2006;69:48–57. (Location USA.)

Bell H, Thordal C, Raknerud N, Hansen T, Bosnes V, Halvorsen R, et al. Prevalence of hemochromatosis among first-time and repeat blood donors in Norway. J Hepatol 1997;26:272–9. (Location Norway.)

Bradley LA, Johnson DD, Palomaki GE, Haddow JE, Robertson NH, Ferrie RM. Hereditary haemochromatosis mutation frequencies in the general population. J Med Screen 1998;5:34–6. (Location USA.)

Burt MJ, George PM, Upton JD, Collett JA, Frampton CM, Chapman TM, et al. The significance of haemochromatosis gene mutations in the general population: implications for screening. Gut 1998;43:830–6. (Location New Zealand.)

Distante S, Berg JP, Lande K, Haug E, Bell H. High prevalence of the hemochromatosis-associated Cys282Tyr HFE gene mutation in a healthy Norwegian population in the city of Oslo, and its phenotypic expression. Scand J Gastroenterol 1999;34:529–34. (Location Norway.)

Hallberg L, Bjorn-Rasmussen E, Jungner I. Prevalence of hereditary haemochromatosis in two Swedish urban areas. J Intern Med 1989;225:249–55. (Location Sweden.)

Holmstrom P, Marmur J, Eggertsen G, Gafvels M, Stal P. Mild iron overload in patients carrying the HFE S65C gene mutation: a retrospective study in patients with suspected iron overload and healthy controls. Gut 2002;51:723–30. (Location Sweden.)

Milman N, Pedersen P, Steig T, Melsen GV. Frequencies of the hereditary hemochromatosis allele in different populations. Comparison of previous phenotypic methods and novel genotypic methods. Int J Hematol 2003;77:48–54. (Review.)

Moirand R, Jouanolle AM, Brissot P, Le Gall JY, David V, Deugnier Y. Phenotypic expression of HFE mutations: a French study of 1110 unrelated iron-overloaded patients and relatives. Gastroenterology 1999;116:372–7. (Location France.)

Nielsen P, Carpinteiro S, Fischer R, Cabeda JM, Porto G, Gabbe EE. Prevalence of the C282Y and H63D mutations in the HFE gene in patients with hereditary haemochromatosis and in control subjects from Northern Germany. Br J Haematol 1998;103:842–5. (Location Germany.)

Patch C, Roderick P, Rosenberg W. Prevalence and burden of disease in hemochromatosis: estimates derived from routine data. Nurs Health Sci 2006;8;128–9. (Estimates derived from routine data.)

Willis G, Wimperis JZ, Smith K, Fellows IW, Jennings BA. HFE mutations in the elderly. Blood Cells Mol Dis 2003;31:240–6. (Restricted to elderly population.)

Merryweather-Clarke AT, Pointon JJ, Shearman JD, Robson KJ. Global prevalence of putative haemochromatosis mutations. J Med Genet 1997;34:275–8. (Review.)

Liver biopsy

Bicknell SG, Richenberg J, Cooperberg PL, Tiwari P, Halperin L. Early discharge after core liver biopsy: is it safe and cost-effective? Can Assoc Radiol J 2002;53:205–9. (Not an RCT.)

Chuah SY. Liver biopsy – past, present and future. Singapore Med J 1996;37:86–90. (Not an RCT.)

Farrell RJ, Smiddy PF, Pilkington RM, Tobin AA, Mooney EE, Temperley IJ, et al. Guided versus blind liver biopsy for chronic hepatitis C: clinical benefits and costs. J Hepatol 1999;30:580–7. (Participants had hepatitis C.)

Froehlich F, Lamy O, Fried M, Gonvers JJ. Practice and complications of liver biopsy. Results of a nationwide survey in Switzerland. Dig Dis Sci 1993;38:1480–4. (Not an RCT.)

Gilmore IT, Burroughs A, Murray-Lyon IM, Williams R, Jenkins D, Hopkins A. Indications, methods, and outcomes of percutaneous liver biopsy in England and Wales: an audit by the British Society of Gastroenterology and the Royal College of Physicians of London. Gut 1995;36:437–41. (Not an RCT.)

Janes CH, Lindor KD. Outcome of patients hospitalized for complications after outpatient liver biopsy. Ann Intern Med 1993;118:96–8. (Not an RCT.)

Kavanagh G, McNulty J, Fielding JF. Complications of liver biopsy: the incidence of pneumothorax and role of post biopsy chest x-ray. Ir J Med Sci 1991;160:387–8. (Not an RCT.)

Pasha T, Gabriel S, Therneau T, Dickson ER, Lindor KD. Cost-effectiveness of ultrasound-guided liver biopsy. Hepatology 1998;27:1220–6. (Not an RCT.)

Piccinino F, Sagnelli E, Pasquale G, Giusti G. Complications following percutaneous liver biopsy. A multicentre retrospective study on 68,276 biopsies. J Hepatol 1986;2:165–73. (Not an RCT.)

Pokorny CS, Waterland M. Short-stay, out-of-hospital, radiologically guided liver biopsy [see comment]. Med J Aust 2002;176:67–9. (Not an RCT.)

Pongchairerks P. Ultrasound-guided liver biopsy: accuracy, safety and sonographic findings. J Med Assoc Thailand 1993;76:597–600. (Not an RCT.)

Smith BC, Desmond PV. Outpatient liver biopsy using ultrasound guidance and the Biopty gun is safe and cost-effective. Aust N Z J Med 1995;25:209–11. (Not an RCT.)

Stone MA, Mayberry JF. An audit of ultrasound guided liver biopsies: a need for evidence-based practice. Hepatogastroenterology 1996;43:432–4. (Not an RCT.)

Stotland BR, Lichtenstein GR. Liver biopsy complications and routine ultrasound. Am J Gastroenterol 1996;91:1295–6. (Editorial.)

Terjung B, Lemnitzer I, Dumoulin FL, Effenberger W, Brackmann HH, Sauerbruch T, et al. Bleeding complications after percutaneous liver biopsy. An analysis of risk factors. Digestion 2003;67:138–45. (Not an RCT.)

Tobkes AI, Nord HJ. Liver biopsy: review of methodology and complications. Dig Dis 1995;13:267–74. (Not an RCT.)

Van Thiel DH, Gavaler JS, Wright H, Tzakis A. Liver biopsy. Its safety and complications as seen at a liver transplant center. Transplantation 1993;55:1087–90. (Not an RCT.)

Vivas S, Palacio MA, Rodriguez M, Lomo J, Cadenas F, Giganto F, et al. Ambulatory liver biopsy: complications and evolution in 264 cases. Rev Esp Enferm Dig 1998;90:175–82. (Not an RCT.)

Wawrzynowicz-Syczewska M, Kruszewski T, Boron-Kaczmarska A. Complications of percutaneous liver biopsy. Rom J Gastroenterol 2002;11:105–7. (Not guided liver biopsies.)

Younossi ZM, Teran JC, Ganiats TG, Carey WD. Ultrasound-guided liver biopsy for parenchymal liver disease: an economic analysis. Dig Dis Sci 1998;43:46–50. (Not an RCT.)

Cost-effectiveness studies

Adams PC, Kertesz AE, Valberg LS. Screening for hemochromatosis in children of homozygotes: prevalence and cost-effectiveness. Hepatology 1995;22:1720–7. (Not DNA testing.)

Killeen AA, Miller PJ. Universal screening for hereditary hemochromatosis in the United States: a cost–benefit analysis. Clin Chem 2002;48:A175. (Screening study.)

Moodie SJ, Ang L, Stenner JM, Finlayson C, Khotari A, Levin GE, et al. Testing for haemochromatosis in a liver clinic population: relationship between ethnic origin, HFE gene mutations, liver histology and serum iron markers. Eur J Gastroenterol Hepatol 2002;14:223–9. (Not economic evaluation.)

Pardo A. Cost-effectiveness of genetic diagnosis for hereditary hemochromatosis screening. Gastroenterology 2000;118:A1405. (Screening study.)

Phatak PD, Guzman G, Woll JE, Robeson A, Phelps CE. Cost-effectiveness of screening for hereditary hemochromatosis. Arch Intern Med 1994;154:769–76.(Screening – transferrin saturation; serum ferritin.)

Talwalkar J, Torgerson H, Malinchoc M, Brandhagen D. Health-related quality of life assessment in hereditary hemochromatosis. Hepatology 2003;38:663A. (Not economic evaluation.)

UK Women’s Cohort Study phase 2

NRR data provider: Leeds Teaching Hospitals NHS Trust.

Region: Northern/Yorkshire Regional Office.

Project number: N0436165676.

Principal research question: To determine the relationship between iron intake, iron status and the risk of iron overload in subjects who are heterozygous (and homozygous) for two genetic mutations (C282Y and H63D) associated with haemochromatosis compared with subjects without these mutations.

Lead centre name: University of Leeds.

Start date: 1 May 2000.

End date: 30 December 2007 [ongoing at time of writing].

Project status: Ongoing.

Funding organisation name: Food Standards Agency.

Funding amount: £250,000.

Funding organisation name: NHS R&D Support Funding.

Funding reference number: 2007/08.

Quality assessment of clinical validity studiesa