- Email: [email protected]
Screening blood donors for hereditary hemochromatosis: decision analysis model comparing genotyping to phenotyping
THE AMERICAN JOURNAL OF GASTROENTEROLOGY © 1999 by Am. Coll. of Gastroenterology Published by Elsevier Science Inc.
Vol. 94, No. 6, 1999 ISSN 0002-9270/99/$20.00 PII S0002-9270(99)00176-8
Screening Blood Donors for Hereditary Hemochromatosis: Decision Analysis Model Comparing Genotyping to Phenotyping Paul C. Adams, M.D., and Leslie S. Valberg, M.D. Department of Medicine, London Health Sciences Center, and University Campus, University of Western Ontario, London, Ontario, Canada
OBJECTIVE: The identification of a gene for hereditary hemochromatosis in 69 –100% of typical hemochromatosis patients has resulted in a genotypic test to identify persons with the typical missense mutation. Population screening by genotyping has the potential to reduce screening costs because of a high specificity of the genetic test. METHODS: Decision analysis techniques are used to compare the outcome, utility, and incremental cost savings of a plan to screen voluntary blood donors and their siblings for hemochromatosis using a genotypic test (C282Y mutation) with phenotypic tests (transferrin saturation, serum ferritin). RESULTS: Genotypic screening is less expensive than phenotypic screening only if the cost of the initial genetic test is less than $20. The screening program saves money (dominant strategy) if the cost of the initial genetic test is less than $28. Incremental cost saving declines as the cost of the gene test increases. At a gene test cost of $173, it costs $109,358 to identify a homozygote with potential lifethreatening illness. Incremental cost saving also declines as the penetrance of the hemochromatosis gene in the population screened decreases. Phenotypic screening with confirmatory genetic testing results in a cost of $2,711 per homozygote with life-threatening complications. CONCLUSIONS: Population screening programs for hemochromatosis have the potential to save money. Optimal strategies for screening include initial testing for iron overload (phenotyping) with confirmatory genetic testing, or initial genetic testing if the test is less than $28. (Am J Gastroenterol 1999;94:1593–1600. © 1999 by Am. Coll. of Gastroenterology)
impotence, heart failure, cirrhosis, and hepatocellular carcinoma and result in long-term survival similar to the general population (6 – 8). The high prevalence, morbidity and mortality, and benefit of early diagnosis and treatment make hemochromatosis a prime target for screening (5, 9, 10). Previous screening studies have been based on the phenotypic expression of the disease (iron overload) (2– 4, 11–25) using sequential testing of transferrin saturation and/or ferritin. The problem is ideally suited to decision analysis because the clinical outcomes occur many years after the original diagnosis in a screening study. Decisionanalysis models suggest that population screening is a costeffective strategy that improves societal health and decreases third-party payments, but it has not been widely implemented. The hemochromatosis gene has been described on chromosome 6, 4.5 Mb telomeric to HLA-A. Recent studies have demonstrated that 69 –100% of typical hemochromatosis patients have a characteristic missense mutation (C282Y), which is readily determined with a genetic test (26 –28). The relationship of a second mutation (H63D) to hemochromatosis is not as clearly established. In this study of the cost-utility of screening for hemochromatosis, we compare the outcome and cost of a model to genotype blood donors and their siblings with a previously developed strategy for phenotypic screening (29, 30). For phenotypic testing, we used the transferrin saturation and serum ferritin (21, 31). Clinical data on the natural history of hemochromatosis and life-threatening events are predominantly from our database of 170 hemochromatosis patients (29). Therefore, a unique aspect to this study is that the clinical outcomes are based on actual clinical data obtained over a 30-yr period rather than clinical estimates.
INTRODUCTION Hereditary hemochromatosis is a common autosomal recessive disease in the Caucasian population with a phenotypic prevalence of approximately one in 300 (1–5). The diagnosis of hemochromatosis is often missed, and the disease is commonly discovered during the management of incidental illnesses or periodic health examinations (6). Early detection and treatment of the disease can prevent the development of
METHODS Target Population The target population is a hypothetical cohort of 10,000 voluntary blood donors and 50 siblings of the identified homozygotes. Sixty percent of donors are men. The advantages of using blood donors compared to the general population have been previously established (29). Screening of
1594
Adams and Valberg
AJG – Vol. 94, No. 6, 1999
Figure 1. Decision analysis tree: A comparison of the genotypic and phenotypic screening strategy. Siblings are tested in the same way as blood donors. The natural history of the unscreened hemochromatosis patients is similar to our previous study (29).
blood donors at their first blood donation is preferred to reduce the effects of blood donation on the screening parameters. Decision Tree STRUCTURE. Decision trees are constructed using decision-analysis software (Decision Analysis, TreeAge, DATA, Version 2.6, TreeAge Software Inc., Boston, MA) (Fig. 1). The decision tree for the natural history of unscreened donors is similar to our previous cost models (29, 30).
COST. The results are expressed in US$ (Appendix). Cost is estimated from a third-party payer perspective because the decision to implement the strategy will likely be made by private or government health insurance plans. It is based on life-threatening clinical outcomes: heart failure, cirrhosis, hepatocellular carcinoma, and diabetes. The presence of heart failure is based on clinical features, diabetes on the need for therapy for hyperglycemia, and both cirrhosis and hepatocellular carcinoma on histological examination of a liver biopsy specimen. The costs of nonlife-threatening dis-
AJG – June, 1999
orders such as arthritis, impotence, and infection are not considered. Future costs are discounted at a rate of 3%. Indirect costs such as earnings lost because of illness or therapy, or indirect benefits such as extra earnings gained due to prevention of hemochromatosis are not considered. The cost of tests and procedures includes overhead, but it does not include the cost of conducting the screening process. Cost of a nurse/administrator to manage the screening study is estimated at $45,000/annum, and overhead for administration is $11,000/annum. Personnel costs were for the duration of the screening study (1 yr). This amounts to a blood collection fee of $5.50 per specimen. The future outcomes, costs, and utilities of the various strategies are converted to net present value (NPV) using a financial function, XNPV, which returns the present value for a schedule of cash flows (Microsoft Excel, Version 7.0, Microsoft Corporation, Redmond, WA). The use of NPV allows a comparison of the outcome and cost of not screening blood donors with the results of screening them and incorporates a discount rate of 3%. Utility Utility estimates are based on assumptions dependent on symptoms of cirrhosis (0.8), diabetes (0.9), and heart failure (0.5) (30, 32, 33). Utility was estimated from quality of life studies in hemochromatosis at this center (32). The incremental cost saving of a screening program is compared to the incremental health improvement attributable to the program, where the health improvement is measured in qualityadjusted life years gained. Quality-adjusted life days or years are arrived at by adjusting the length of time affected by the health outcome by the utility value of the resulting level of health status (a measure of preference on a scale of 0 to 1). Clinical Hemochromatosis Database Clinical, biochemical, pathological, and long-term survival data were retrieved from a database of 170 hemochromatosis homozygotes as previously described. These patients were referred to this tertiary care center for diagnosis and treatment of hemochromatosis. They consist of symptomatic probands and asymptomatic family members. This forms the basis of the cost of not screening for hemochromatosis and paying for medical care associated with the complications of hemochromatosis. Estimates of survival and penetrance of disease are from actual patient data. Although the prevalence of life-threatening complications are drawn from a 30-yr experience, the cost of these complications is estimated from costs originally determined in 1995, and updated in 1998. The model assumes that 43% of men and 28% of women will develop life-threatening manifestations of hemochromatosis if unscreened (29). The probability of each lifethreatening complication is from our previous study (29). These rates of progression are based on patients presenting with clinical symptoms and evaluation of their family members and may differ in population-based screening.
Genetic Screening for Hemochromatosis
1595
Phenotypic Screening The phenotypic screening strategy is similar to our previous model. It begins with a transferrin saturation and, if abnormal, the patient returns for a subsequent fasting transferrin saturation to decrease the number of false-positive cases progressing to further investigation (3, 30). Threshold levels for transferrin saturation were 50% for women and 60% for men and for ferritin were 150 mg/L for women and 200 mg/L for men, as previously described (29). Some studies have advocated a screening threshold of 45%, which increases the cost slightly because it increases the number of investigations in unaffected people (10). Genotypic Screening In the genotype screening, donors begin with a genetic test to detect persons who are homozygous for the C282Y mutation (26 –28). Homozygous donors are tested with serum ferritin and those with an elevated value are assumed to have iron overload and begin therapeutic venesections. Homozygous donors with a normal serum ferritin are followed and retested for iron overload with serum ferritin at 5-yr intervals for 10 yr. Siblings of the proband case found through screening are tested for the C282Y mutation and treated in the same way as donors. Heterozygotes are considered not to be at risk for pathological iron overload and are not followed up in the model (34 –36). It is acknowledged that this strategy will miss patients who have been described to be heterozygotes for C282Y and heterozygous for the second mutation H63D. Unscreened Blood Donors and Siblings The outcome for unscreened donors is as previously described. The natural history of patients who present with symptoms of hemochromatosis may be different than asymptomatic patients discovered through screening (29), but we believe the estimates of symptom-specific survival to be the best available because they are based on real data accumulated over a 30-yr period at this medical center. The costs for unscreened donors are based on US$ costs (9). Missed Homozygotes In the phenotyping, 4.6 donor homozygotes are missed and in the genotyping screening model, 4.5 donor homozygotes are missed. The number missed decreases as the sensitivity of the genotyping test is increased. The individual cost of their illness is similar to unscreened homozygous blood donors.
RESULTS No-Screening Strategy Total future expenditure on life-threatening illness among 30 unscreened homozygotes in the blood donor population is $515,478 ($ NPV). Phenotypic Screening Strategy In phenotypic screening with transferrin saturation, we predict that 25.4 homozygous blood donors and 12.5 homozy-
1596
Adams and Valberg
AJG – Vol. 94, No. 6, 1999
Table 1. Benefit of Screening for Hemochromatosis Compared to No Screening
Initial Screening Test Phenotyping with transferrin saturation ($12) Genotyping ($20) Genotyping ($50) Genotyping ($100) Genotyping ($173) Genotyping ($200)
Incremental Cost Saving Per Person Screened (US$) 0.97 2.1 228 278 2151 2178
Cost Per Homozygote Identified (US$) 2333 2563 7,371 20,595 39,901 47,042
Cost Per Homozygote With Potential Life-Threatening Illness* (US$) 2775 21544 20,202 56,444 109,358 128,929
* Assumes that 43% of men and 28% of women progress to life-threatening illness. (2) means that there is cost savings; (1) means that costs are incurred.
gous siblings will be identified. We assume that two siblings are tested per homozygote discovered. Cost of the phenotypic screening strategy with transferrin saturation combined with the cost of 4.6 missed homozygotes is $502,843. This is a higher cost than our previous model because we used transferrin saturation rather than the less expensive UIBC (29) and because of the use of American costs. Compared to the no-screening strategy, the incremental cost saving of phenotyping is $0.97 for each person screened (Table 1). Genotyping Screening Strategy In genotypic screening, we predict that 25.5 homozygous blood donors and 12.5 homozygous siblings will be identified. Cost of the genotypic screening strategy with a genetic test costing of $173 combined with the cost of 4.6 missed homozygotes is $2,031,727 or $109,358 to detect one homozygote with potential life-threatening illness. Compared to the no-screening strategy, the incremental cost saving of genotyping is 2$151 for each person screened (Table 1). There are 2.75 quality life days saved per person screened for a cost-utility of $20,042 per quality-adjusted life years saved. Cost saving is only apparent in genotypic screening if the initial gene test costs less than $28 with a prevalence of hemochromatosis of 0.003 (Table 1, Fig. 2). The cost of identifying homozygous donors and siblings rises as the cost of the genetic test increases.
Figure 2. Effect of changes in the cost of the initial genotyping test on the incremental cost savings per person screened including subsequent investigation of siblings. A negative value indicates costs incurred. (Cost expressed as 1998 US$—net present value, prevalence of hemochromatosis 5 .003.)
Genotyping Versus Phenotyping The phenotypic screening strategy with transferrin saturation is dominant (saves money) (Table 2). When blood donors with a high transferrin saturation follow the genotyping strategy with the gene test cost set at $173, the total cost is somewhat higher than phenotyping alone (Table 2). More homozygotes are missed with the combined strategy (phenotype, then genotype), 6.5 compared to 4.5 for genotyping and 4.6 for phenotyping. When genotyping is used to confirm hemochromatosis in patients with a high transferrin saturation, C282Y negative iron loaded patients are missed. Familial iron overload in the absence of the C282Y mutation is extremely rare, and investigation of siblings will be of a lower yield in these cases and could be eliminated with the subsequent cost savings (37). The major difference in cost between phenotyping and genotyping is the difference in the cost of the initial test: $12 for transferrin saturation versus $173 for the gene test. Utility Analysis Both phenotypic and genotypic strategies led to a similar incremental increase in quality-adjusted life days saved for each person screened (utility) (Table 1). Screening also produced a greater utility benefit in men compared to women (Table 3). Sensitivity Analysis GENETIC TEST. If the sensitivity of the genetic test is increased to 0.95, only 1.5 homozygotes are missed, and the total cost of screening, $2,015,866, is slightly reduced. We have considered the genetic test to have no false-positives except in the rare case of sample mislabeling. We do not consider a person homozygous for the C282Y mutation without iron overload to be a false-positive but rather a nonexpressing homozygote. For the purpose of the analysis, we modelled the effect of including these nonexpressing homozygotes as false-positives and decreased the specificity from 1.0 to 0.97. Decreasing the specificity to 0.97 (sensitivity 0.85), markedly increases the cost, $3,533,815, and decreases the incremental cost saving per person screened. This is as a result of donors with false-positive tests proceeding to further testing and follow-up.
AJG – June, 1999
Genetic Screening for Hemochromatosis
1597
Table 2. Comparison of Population Screening Strategies Incremental Cost Saving Per Person Screened (US$)
Homozygotes Missed (Out of 10,000 Donors)
Screening Tests Phenotyping with transferrin saturation ($12), liver biopsy Phenotyping with transferrin saturation, then genotyping Genotyping ($173), then ferritin
Cost Per Homozygote Identified (US$)
Cost Per Homozygote With Potential Life-Threatening Illness* (US$)
4.6
0.97
2333
2775
6.5
23.74
989
2,711
39,901
109,358
2151
4.5
* 43% of men and 28% of women progress to life-threatening illness. (2) refers to cost savings.
PREVALENCE. Fewer homozygotes are found at low prevalence rates and thus the cost of identifying each homozygote is higher. This may be relevant in urban areas with significant non-Caucasian populations. Our baseline prevalence estimate of .003 is conservative and would allow for some ethnic diversity in the population tested. PENETRANCE. The effect of varying the rate of progression to life-threatening symptoms is shown in Table 3. When penetrance is lowered: 1) the total cost of not screening decreases; 2) the total cost of identifying homozygotes with potential to develop life-threatening illness rises; and 3) incremental cost saving declines. This effect is apparent in both men and women (Table 3).
DISCUSSION The description of a new genetic test for a missense mutation on chromosome 6 introduces the possibility of population screening for hemochromatosis using a true genetic test rather than a phenotypic test for iron overload using the transferrin saturation and serum ferritin. The value of screening with a new genetic test is closely related to several critical features of the test, as follows. Cost of the Initial Screening Test Based on our model and cost projections, if the cost of the new genetic test can be kept to less than $28, it will result
in cost savings in blood donor population screening studies (Fig. 2). However, significant costs rather than savings are apparent as the cost of the genetic test increases into the range that has been charged in several reference American laboratories (Table 1). We have restricted the genetic test in this analysis to the C282Y mutation. The inclusion of the screening for the second mutation (H63D) provides only a limited amount of additional information. The gene test charge of $173 currently includes the cost of H63D testing in the United States. The strategy outlined may miss compound heterozygotes (C282Y/H63D) or apparent heterozygotes or homozygotes that carry another undetected mutation associated with iron overload. The compound heterozygote has represented less than 1% of our ironloaded patients, and previous reports have suggested that these patients have mild-to-moderate iron overload. Sensitivity of the Genetic Test Available data on sensitivity comes from six independent studies. The initial report from Mercator Genetics demonstrated that 83% of hemochromatosis patients were homozygous for the C282Y mutation in 176 patients. These patients were collected from different American medical centers, and the precision of the phenotypic assignment and confirmatory family studies presumably varied considerably in the samples. Subsequent studies from France (38), United Kingdom (39), and Australia (28) gave a sensitivity of 90 –100%
Table 3. Effect of Penetrance of Hemochromatosis Gene on Benefits of Genotypic Screening Strategy (Gene Test $173) Progression to Life-Threatening Illness Men 0.43 Men 0.3 Men 0.2 Men 0.1 Women 0.28 Women 0.2 Women 0.15 Women 0.1
Incremental Cost Incurred Per Person Screened (US$)
Cost Per Homozygote Identified (US$)
Cost Per Homozygote With Potential Life-Threatening Illness (US$)
Quality Life Days Saved Per Person Screened
Cost-Utility (Incremental Cost/ QALY* Saved)
142 160 173 187 164 176 183 191
39,730 44,628 48,396 52,163 40,005 42,885 44,686 46,486
92,394 132,432 198,648 397,296 93,034 142,951 297,905 464,857
3.09 2.45 1.50 0.71 2.43 1.40 1.33 1.06
16,535 23,468 41,518 95,269 24,268 45,359 49,484 64,946
* QALY 5 quality-adjusted life years.
1598
Adams and Valberg
in patients, including pedigree studies (40). An Italian study had a much lower sensitivity at 69% (41). Many of their patients had confounding variables such as alcoholism and chronic viral hepatitis, but the prevalence of the C282Y mutation was also less in the Italian population. Patients at our center have been homozygous in 95% of cases (42). We chose a sensitivity value of 85% for this study to take the wide range of reported sensitivities into consideration. Other Mutations A potential shortcoming of initial screening for the C282Y mutation is that it will not detect iron-loaded blood donors who carry another mutation or another disease associated with iron overload (41). Non-HLA linked iron-overload diseases have been reported particularly in non-Caucasian populations (43– 47). We have not included any cost associated with patients missed with those disorders. The phenotypic strategy may also detect patients with iron deficiency. This may benefit the patient but may also increase the cost of this strategy (48). Specificity of Genotyping We assume that patients who are homozygous for the mutation have genetic hemochromatosis. Previous studies have suggested that many patients with hemochromatosis have incomplete penetrance without phenotypic expression. In our own series, we have observed five families with HLA identical siblings who are homozygous for the C282Y mutation without evidence of iron overload (42, 49). In these subjects, predisposition to iron overload rather than the identification of a disease has been determined by genotyping. Preliminary results from our population screening study have detected two iron-loaded siblings of nonexpressing homozygotes (31). Case definition is an essential part of this analysis. We have chosen to emphasize genotyping as our “gold standard” because we continue to believe that most of the described cases of iron overload that are negative for the C282Y mutation represent isolated cases (nonfamilial), which differ from genetic hemochromatosis. However, we have set the sensitivity of the genotyping at 0.85 to allow for some atypical iron-loaded cases that may be considered to be hemochromatosis. To define hemochromatosis solely on the basis of degree of iron overload would strongly favor the phenotypic screening model over the genotyping model. Penetrance The penetrance used in our baseline assumptions may be an overestimate of the penetrance of disease in population-based studies. If we had sampled a younger population with more women, there would have been fewer homozygotes detected by phenotypic screening. Furthermore, our penetrance baseline estimates introduce a referral bias because this is a tertiary care center of hemochromatosis. Reducing the penetrance increases the cost of identifying homozygotes with the potential to develop life-threatening illness and increases the cost-utility of genotyping (Table 3). The detection of iron overload by phenotypic screening identifies those who are iron loaded
AJG – Vol. 94, No. 6, 1999
and may be a more practical approach because it immediately identifies patients in need of venesection therapy, but it requires an initial inexpensive screening test. We tested a model using transferrin saturation followed by C282Y testing without liver biopsy. This approach has a cost profile similar to our phenotyping model, but an additional two homozygotes are missed in the screening process. The major cost of all population screening models is driven by the cost of the initial screening test. It is for this reason that several studies (31) are exploring the use of an inexpensive automated unbound iron-binding capacity as an alternative to transferrin saturation. We have not included utility adjustments for the knowledge of a potential future illness because the disease is treatable with appropriate follow-up monitoring. We have also not included the indirect costs associated with loss of insurance or employment as a result of genetic discrimination. By limiting our model to blood donors, we have eliminated many patients who could be false-positive by phenotypic screening (alcoholics, chronic viral hepatitis, iron-loading anemias). If the genetic test is proven to have a specificity close to 1.0, the screening of other patient groups such as those in ambulatory care clinics and hospitalized patients may become more attractive than phenotypic screening where many false-positives are detected and unnecessary invasive procedures are needed to establish the presence of iron overload. The genotyping model in this study eliminates the use of diagnostic liver biopsy when compared to the phenotypic model. This is because of the high specificity assigned to the genotype test. Quantitative phlebotomy has been used as a diagnostic criterion for hemochromatosis in the absence of liver biopsy. In the phenotypic model, the hepatic iron distribution, hepatic iron concentration, and hepatic iron index contribute significantly to the confidence of the phenotypic assignment. The cost of liver biopsy is not a significant factor because relatively few donors proceed to liver biopsy in the phenotypic screening model (Table 1). The presence of cirrhosis has been associated with a decreased survival in hemochromatosis (8, 33), and other studies have suggested that it is important to establish the presence of cirrhosis for prognosis and to consider screening for hepatocellular carcinoma (50). The role of liver biopsy in the diagnosis of hemochromatosis requires reassessment since the development of genetic testing. It is likely that it will move from a diagnostic test to a prognostic test in selected cases. A multivariate analysis of 362 C282Y homozygotes who underwent liver biopsy demonstrated that the risk of fibrosis was minimal in the patient without hepatomegaly, with a ferritin , 1000 mg/ml, and a normal AST (51). These parameters would apply to most young homozygotes found in a screening study, so liver biopsy would not be required except perhaps in a C282Y negative iron-loaded patient. We have previously emphasized that hemochromatosis is an ideal disease for population screening based on the World Health Organization criteria (29). The development of a genetic test that can be performed at reasonable cost (, $28) has the potential to replace phenotypic screening for hemo-
AJG – June, 1999
chromatosis. The utilization of a genetic screening test for a common treatable disease is one of the triumphs of genetic mapping and a goal of the Human Genome Project. Implementation of population screening is dependent on the willingness of private or government health agencies to pay the initial screening costs to achieve savings many years later.
ACKNOWLEDGMENTS The authors thank Ann Kertesz for her assistance in the development and maintenance of the hemochromatosis database and David Lloyd for providing the software for decision analysis. This paper is dedicated to the memory and contributions of Leslie Valberg who passed away during the preparation of the manuscript. Grant Support: Dr. Adams acknowledges the support of the Physician Services Incorporated Foundation of Ontario, Medical Research Council of Canada, and the Pomajba family.
Genetic Screening for Hemochromatosis
14. 15. 16. 17. 18.
19. 20.
Reprint requests and correspondence: Paul C. Adams, M.D., Department of Medicine, London Health Sciences Centre, 339 Windermere Rd., London, Ontario, Canada N6A 5A5. Received July 31, 1998; accepted Dec. 9, 1998.
21.
REFERENCES
23.
1. Simon M, Bourel M, Fauchet R, et al. HLA and “non-immunological” disease: Idiopathic haemochromatosis. Lancet 1976;1:973– 4. 2. Borwein ST, Ghent CN, Flanagan PR, et al. Genetic and phenotypic expression of hemochromatosis in Canadians. Clin Invest Med 1983;6:171–9. 3. Edwards CQ, Griffen LM, Goldgar D, et al. Prevalence of hemochromatosis among 11,065 presumably healthy blood donors. N Engl J Med 1988;318:1355– 62. 4. Hallberg L, Bjorn-Rasmussen E, Jungner I. Prevalence of hereditary haemochromatosis in two Swedish urban areas. J Intern Med 1989;225:249 –55. 5. Edwards CQ, Kushner JP. Screening for hemochromatosis. N Engl J Med 1993;328:1616 –20. 6. Niederau C, Fischer R, Sonnenberg A, et al. Survival and causes of death in cirrhotic and in noncirrhotic patients with primary hemochromatosis. N Engl J Med 1985;313:1256 – 62. 7. Adams PC, Kertesz AE, Valberg LS. Clinical presentation of hemochromatosis: A changing scene. Am J Med 1991;90:445–9. 8. Adams PC, Speechley M, Kertesz AE. Long-term survival analysis in hereditary hemochromatosis. Gastroenterology 1991;101:368 –72. 9. Phatak PD, Guzman G, Woll JE, et al. Cost-effectiveness of screening for hereditary hemochromatosis. Arch Intern Med 1994;154:769 –76. 10. Bassett M, Leggett BA, Halliday JW, et al. Analysis of the cost of population screening for haemochromatosis using biochemical and genetic markers. J Hepatology 1997;27:517–24. 11. Bassett ML, Doran TJ, Halliday JW, et al. Idiopathic hemochromatosis: Demonstration of homozygous-heterozygous mating by HLA typing of families. Hum Genet 1982;60: 352– 6. 12. Beaumont C, Simon M, Fauchet R, et al. Serum ferritin as a possible marker of the hemochromatosis allele. N Engl J Med 1979;301:169 –74. 13. Dadone MM, Kushner JP, Edwards CQ, et al. Hereditary
24.
22.
25.
26. 27. 28. 29.
30. 31. 32. 33. 34. 35.
1599
hemochromatosis. Analysis of laboratory expression of the disease by genotype in 18 pedigrees. Am J Clin Pathol 1982; 78:196 –207. Expert Scientific Working Group. Summary of a report on assessment of the iron nutritional status of the United States population. Am J Clin Nutr 1985;42:1318 –30. Tanner AR, Wright R. Hemochromatosis screening. Lancet 1988;2:286. Lindmark B, Eriksson S. Prevalence of hemochromatosis. N Engl J Med 1988;319:1548 –9 (letter). Leggett BA, Halliday JW, Brown NN, et al. Prevalence of haemochromatosis amongst asymptomatic Australians. Br J Haematol 1990;74:525–30. Wiggers P, Dalhoj J, Petersen PH, et al. Screening for hemochromatosis: Influence of analytical imprecision, diagnostic limit and prevalence on test validity. Scand J Clin Invest 1991;51:143– 8. Jonsson JJ, Johannesson GM, Sigfusson N, et al. Prevalence of iron deficiency and iron overload in the adult Icelandic population. J Clin Epidemiol 1991;44:1289 –97. Wiggers P, Dalhoj J, Kiaer H, et al. Screening for haemochromatosis: Prevalence among Danish blood donors. J Intern Med 1991;230:265–70. Baynes RD, Meyer TE, Bothwell TH, et al. A screening test for detecting iron overload in population studies. S Afr Med J 1988;74:167–9. Borwein S, Ghent CN, Valberg LS. Diagnostic efficacy of screening tests for hereditary hemochromatosis. Can Med Assoc J 1984;131:895–901. Bassett ML, Halliday JW, Bryant S, et al. Screening for hemochromatosis. Ann NY Acad Sci 1988;526:274 – 89. Bassett ML, Halliday JW, Ferris RA, et al. Diagnosis of hemochromatosis in young subjects: Predictive accuracy of biochemical screening tests. Gastroenterology 1984;87:628 – 33. Niederau C, Niederau CM, Lange S, et al. Screening for hemochromatosis and iron deficiency in employees and primary care patients in Western Germany. Ann Int Med 1998; 128:337– 45. Feder JN, Gnirke A, Thomas W, et al. A novel MHC class I-like gene is mutated in patients with hereditary hemochromatosis. Nat Genet 1996;13:399 – 408. Jazwinska EC, Cullen LM, Busfield F, et al. Haemochromatosis and HLA-H. Nat Genet 1996;14:249 –51. Jouanolle A-M, Gandon G, Jezequel P, et al. Haemochromatosis and HLA-H. Nat Genet 1996;14:251–2. Adams PC, Gregor JC, Kertesz AE, et al. Screening blood donors for hereditary hemochromatosis: Decision analysis model based on a thirty-year database. Gastroenterology 1995; 109:177– 88. Adams PC, Kertesz A, Valberg LS. Screening for hemochromatosis in children of homozygotes: Prevalence and costeffectiveness. Hepatology 1995;22:1720 –7. Adams PC, Kertesz AE, Barr R, et al. Population screening for hemochromatosis with the unbound iron binding capacity. Gastroenterology 1998;114:A1199 (abstract). Adams PC, Speechley M. The effect of arthritis on the quality of life in hereditary hemochromatosis. J Rheumatol 1996;23: 707–10. Niederau C, Fischer R, Purschel A, et al. Long-term survival in patients with hereditary hemochromatosis. Gastroenterology 1996;110:1107–19. Bulaj ZJ, Griffen L, Jorde LB, et al. Clinical and biochemical abnormalities in people heterozygous for hemochromatosis. N Eng J Med 1996;335:1799 – 805. Adams PC. Prevalence of abnormal iron studies in heterozy-
1600
36.
37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51.
Adams and Valberg
gotes for hereditary hemochromatosis—An analysis of 255 heterozygotes. Am J Hematol 1994;45:146 –9. Bassett ML, Halliday JW, Powell LW. HLA typing idiopathic hemochromatosis: Distinction between homozygotes and heterozygotes with biochemical expression. Hepatology 1981;1: 120 – 6. Pietrangelo A, Montosi G, Garuti C, et al. No HFE, No HLA, No 6p-linked adult hemochromatosis: A new genetic iron overload condition. Hepatology 1998;28:528A (abstract). Adams PC, Bradley C, Henderson AR. Evaluation of the hepatic iron index as a diagnostic criterion in hereditary hemochromatosis. J Lab Clin Med 1997;130:509 –14. The UK Haemochromatosis Consortium. A simple genetic test identifies 90% of UK patients with haemochromatosis. Gut 1997;41:841–5. Burke W, Thomson E, Khoury M, et al. Hereditary hemochromatosis: Gene discovery and its implications for population-based screening. JAMA 1998;280:172– 8. Carella M, D’Ambrosio L, Totaro A, et al. Mutation analysis of the HLA-H gene in Italian hemochromatosis patients. Am J Hum Genet 1997;60:828 –32. Adams PC, Chakrabarti S. Genotypic/phenotypic correlations in genetic hemochromatosis: Evolution of diagnostic criteria. Gastroenterology 1998;114:319 –23. Oliver M, Scully L, Guiraudon C, et al. Non-HLA linked hemochromatosis in a Chinese woman. Dig Dis Sci 1995;40:1589–91. Eason RJ, Adams PC, Aston CE, et al. Familial iron overload with possible autosomal dominant inheritance. Aust NZ J Med 1990;20:226 –30. Gordeuk V, Mukiibi J, Hasstedt SJ, et al. Iron overload in Africa. Interaction between a gene and dietary iron content. N Engl J Med 1992;326:95–100. Morita H, Ikeda S, Yamamoto K, et al. Hereditary ceruloplasmin deficiency with hemosiderosis: A clinicopathological study of a Japanese family. Ann Neurol 1995;37:646 –56. Wurapa RK, Gordeuk V, Brittenham G, et al. Primary iron overload in African Americans. Am J Med 1996;100:9 –18. Baer DM, Simons JL, Staples RL, et al. Hemochromatosis screening in asymptomatic ambulatory men 30 years of age and older. Am J Med 1995;98:464 – 8. Adams PC, Campion ML, Gwandon G, et al. Clinical and family studies in genetic hemochromatosis: Microsatellite and HFE studies in five atypical families. Hepatology 1997;26:986–90. Fargion S, Fracanzani AL, Piperno A, et al. Prognostic factors for hepatocellular carcinoma in genetic hemochromatosis. Hepatology 1994;20:1426 –31. Guyader D, Jacquelinet C, Moirand R, et al. Non-invasive prediction of fibrosis in C282Y homozygous hemochromatosis. Gastroenterology 1998;115:929 –36.
APPENDIX: VARIABLE DEFINITIONS Age at Death ● heart failure (female) 5 60 ● heart failure (male) 5 55 ● cirrhosis (female) 5 72 ● cirrhosis (male) 5 67 ● diabetes (female) 5 73 ● diabetes (male) 5 68 ● hepatocellular carcinoma (female) 5 69 ● hepatocellular carcinoma (male) 5 64 ● life-threatening event homozygote (female) 5 71 ● life-threatening event homozygote (male) 5 66 ● normal (female) 5 78
AJG – Vol. 94, No. 6, 1999
●
normal (male) 5 72
Age at First Symptoms ● female homozygote 5 59 ● male homozygote 5 54 Medical Care Costs in US$ ● annual ambulatory cirrhosis care 5 1,500 ● annual ambulatory diabetes care 5 1,500 ● probability of compliance with follow-up 5 0.80 ● physicians consultation 5 200 ● physicians hospital visit 5 50 ● physicians office visit 5 50 ● daily inpatient care 5 500 ● hepatocellular carcinoma, 1998 5 50,000 ● serum ferritin 5 20 ● genotyping 5 173 (this charge currently includes C282Y and H63D testing in the USA) ● liver biopsy 5 600 ● probability of a complication 5 0.002 ● probability of dying from complication 5 0.05 ● transferrin saturation 5 12 ● unbound iron binding capacity 5 1.40 Donors ● age at initial screening 5 37 ● male probability 5 0.6 ● number 5 10,000 ● homozygous probability 5 0.003 ● probability of life-threatening manifestations (female) 5 0.28 ● probability of life-threatening manifestations (male) 5 0.43 Siblings ● probability HLA identical 5 0.25 ● female probability 5 0.5 ● homozygote probability 5 0.25 ● number 5 50 Probability of Developing Complications ● cirrhosis 5 0.303 ● cirrhosis and diabetes 5 0.273 ● diabetes 5 0.182 ● heart failure 5 0.045 ● die from heart failure 5 0.1 ● heart failure and cirrhosis 5 0.045 ● heart failure, cirrhosis, and diabetes 5 0.107 ● heart failure and diabetes 5 0.045 ● hepatocellular carcinoma 5 0.25 Venesection Therapy ● venesection cost 5 $45 ● number of initial venesections (donor) 5 53 ● maintenance number of venesections per year 5 3 ● number of initial venesections (siblings) 5 6 (females) and 30 (males)
Vol. 94, No. 6, 1999 ISSN 0002-9270/99/$20.00 PII S0002-9270(99)00176-8
Screening Blood Donors for Hereditary Hemochromatosis: Decision Analysis Model Comparing Genotyping to Phenotyping Paul C. Adams, M.D., and Leslie S. Valberg, M.D. Department of Medicine, London Health Sciences Center, and University Campus, University of Western Ontario, London, Ontario, Canada
OBJECTIVE: The identification of a gene for hereditary hemochromatosis in 69 –100% of typical hemochromatosis patients has resulted in a genotypic test to identify persons with the typical missense mutation. Population screening by genotyping has the potential to reduce screening costs because of a high specificity of the genetic test. METHODS: Decision analysis techniques are used to compare the outcome, utility, and incremental cost savings of a plan to screen voluntary blood donors and their siblings for hemochromatosis using a genotypic test (C282Y mutation) with phenotypic tests (transferrin saturation, serum ferritin). RESULTS: Genotypic screening is less expensive than phenotypic screening only if the cost of the initial genetic test is less than $20. The screening program saves money (dominant strategy) if the cost of the initial genetic test is less than $28. Incremental cost saving declines as the cost of the gene test increases. At a gene test cost of $173, it costs $109,358 to identify a homozygote with potential lifethreatening illness. Incremental cost saving also declines as the penetrance of the hemochromatosis gene in the population screened decreases. Phenotypic screening with confirmatory genetic testing results in a cost of $2,711 per homozygote with life-threatening complications. CONCLUSIONS: Population screening programs for hemochromatosis have the potential to save money. Optimal strategies for screening include initial testing for iron overload (phenotyping) with confirmatory genetic testing, or initial genetic testing if the test is less than $28. (Am J Gastroenterol 1999;94:1593–1600. © 1999 by Am. Coll. of Gastroenterology)
impotence, heart failure, cirrhosis, and hepatocellular carcinoma and result in long-term survival similar to the general population (6 – 8). The high prevalence, morbidity and mortality, and benefit of early diagnosis and treatment make hemochromatosis a prime target for screening (5, 9, 10). Previous screening studies have been based on the phenotypic expression of the disease (iron overload) (2– 4, 11–25) using sequential testing of transferrin saturation and/or ferritin. The problem is ideally suited to decision analysis because the clinical outcomes occur many years after the original diagnosis in a screening study. Decisionanalysis models suggest that population screening is a costeffective strategy that improves societal health and decreases third-party payments, but it has not been widely implemented. The hemochromatosis gene has been described on chromosome 6, 4.5 Mb telomeric to HLA-A. Recent studies have demonstrated that 69 –100% of typical hemochromatosis patients have a characteristic missense mutation (C282Y), which is readily determined with a genetic test (26 –28). The relationship of a second mutation (H63D) to hemochromatosis is not as clearly established. In this study of the cost-utility of screening for hemochromatosis, we compare the outcome and cost of a model to genotype blood donors and their siblings with a previously developed strategy for phenotypic screening (29, 30). For phenotypic testing, we used the transferrin saturation and serum ferritin (21, 31). Clinical data on the natural history of hemochromatosis and life-threatening events are predominantly from our database of 170 hemochromatosis patients (29). Therefore, a unique aspect to this study is that the clinical outcomes are based on actual clinical data obtained over a 30-yr period rather than clinical estimates.
INTRODUCTION Hereditary hemochromatosis is a common autosomal recessive disease in the Caucasian population with a phenotypic prevalence of approximately one in 300 (1–5). The diagnosis of hemochromatosis is often missed, and the disease is commonly discovered during the management of incidental illnesses or periodic health examinations (6). Early detection and treatment of the disease can prevent the development of
METHODS Target Population The target population is a hypothetical cohort of 10,000 voluntary blood donors and 50 siblings of the identified homozygotes. Sixty percent of donors are men. The advantages of using blood donors compared to the general population have been previously established (29). Screening of
1594
Adams and Valberg
AJG – Vol. 94, No. 6, 1999
Figure 1. Decision analysis tree: A comparison of the genotypic and phenotypic screening strategy. Siblings are tested in the same way as blood donors. The natural history of the unscreened hemochromatosis patients is similar to our previous study (29).
blood donors at their first blood donation is preferred to reduce the effects of blood donation on the screening parameters. Decision Tree STRUCTURE. Decision trees are constructed using decision-analysis software (Decision Analysis, TreeAge, DATA, Version 2.6, TreeAge Software Inc., Boston, MA) (Fig. 1). The decision tree for the natural history of unscreened donors is similar to our previous cost models (29, 30).
COST. The results are expressed in US$ (Appendix). Cost is estimated from a third-party payer perspective because the decision to implement the strategy will likely be made by private or government health insurance plans. It is based on life-threatening clinical outcomes: heart failure, cirrhosis, hepatocellular carcinoma, and diabetes. The presence of heart failure is based on clinical features, diabetes on the need for therapy for hyperglycemia, and both cirrhosis and hepatocellular carcinoma on histological examination of a liver biopsy specimen. The costs of nonlife-threatening dis-
AJG – June, 1999
orders such as arthritis, impotence, and infection are not considered. Future costs are discounted at a rate of 3%. Indirect costs such as earnings lost because of illness or therapy, or indirect benefits such as extra earnings gained due to prevention of hemochromatosis are not considered. The cost of tests and procedures includes overhead, but it does not include the cost of conducting the screening process. Cost of a nurse/administrator to manage the screening study is estimated at $45,000/annum, and overhead for administration is $11,000/annum. Personnel costs were for the duration of the screening study (1 yr). This amounts to a blood collection fee of $5.50 per specimen. The future outcomes, costs, and utilities of the various strategies are converted to net present value (NPV) using a financial function, XNPV, which returns the present value for a schedule of cash flows (Microsoft Excel, Version 7.0, Microsoft Corporation, Redmond, WA). The use of NPV allows a comparison of the outcome and cost of not screening blood donors with the results of screening them and incorporates a discount rate of 3%. Utility Utility estimates are based on assumptions dependent on symptoms of cirrhosis (0.8), diabetes (0.9), and heart failure (0.5) (30, 32, 33). Utility was estimated from quality of life studies in hemochromatosis at this center (32). The incremental cost saving of a screening program is compared to the incremental health improvement attributable to the program, where the health improvement is measured in qualityadjusted life years gained. Quality-adjusted life days or years are arrived at by adjusting the length of time affected by the health outcome by the utility value of the resulting level of health status (a measure of preference on a scale of 0 to 1). Clinical Hemochromatosis Database Clinical, biochemical, pathological, and long-term survival data were retrieved from a database of 170 hemochromatosis homozygotes as previously described. These patients were referred to this tertiary care center for diagnosis and treatment of hemochromatosis. They consist of symptomatic probands and asymptomatic family members. This forms the basis of the cost of not screening for hemochromatosis and paying for medical care associated with the complications of hemochromatosis. Estimates of survival and penetrance of disease are from actual patient data. Although the prevalence of life-threatening complications are drawn from a 30-yr experience, the cost of these complications is estimated from costs originally determined in 1995, and updated in 1998. The model assumes that 43% of men and 28% of women will develop life-threatening manifestations of hemochromatosis if unscreened (29). The probability of each lifethreatening complication is from our previous study (29). These rates of progression are based on patients presenting with clinical symptoms and evaluation of their family members and may differ in population-based screening.
Genetic Screening for Hemochromatosis
1595
Phenotypic Screening The phenotypic screening strategy is similar to our previous model. It begins with a transferrin saturation and, if abnormal, the patient returns for a subsequent fasting transferrin saturation to decrease the number of false-positive cases progressing to further investigation (3, 30). Threshold levels for transferrin saturation were 50% for women and 60% for men and for ferritin were 150 mg/L for women and 200 mg/L for men, as previously described (29). Some studies have advocated a screening threshold of 45%, which increases the cost slightly because it increases the number of investigations in unaffected people (10). Genotypic Screening In the genotype screening, donors begin with a genetic test to detect persons who are homozygous for the C282Y mutation (26 –28). Homozygous donors are tested with serum ferritin and those with an elevated value are assumed to have iron overload and begin therapeutic venesections. Homozygous donors with a normal serum ferritin are followed and retested for iron overload with serum ferritin at 5-yr intervals for 10 yr. Siblings of the proband case found through screening are tested for the C282Y mutation and treated in the same way as donors. Heterozygotes are considered not to be at risk for pathological iron overload and are not followed up in the model (34 –36). It is acknowledged that this strategy will miss patients who have been described to be heterozygotes for C282Y and heterozygous for the second mutation H63D. Unscreened Blood Donors and Siblings The outcome for unscreened donors is as previously described. The natural history of patients who present with symptoms of hemochromatosis may be different than asymptomatic patients discovered through screening (29), but we believe the estimates of symptom-specific survival to be the best available because they are based on real data accumulated over a 30-yr period at this medical center. The costs for unscreened donors are based on US$ costs (9). Missed Homozygotes In the phenotyping, 4.6 donor homozygotes are missed and in the genotyping screening model, 4.5 donor homozygotes are missed. The number missed decreases as the sensitivity of the genotyping test is increased. The individual cost of their illness is similar to unscreened homozygous blood donors.
RESULTS No-Screening Strategy Total future expenditure on life-threatening illness among 30 unscreened homozygotes in the blood donor population is $515,478 ($ NPV). Phenotypic Screening Strategy In phenotypic screening with transferrin saturation, we predict that 25.4 homozygous blood donors and 12.5 homozy-
1596
Adams and Valberg
AJG – Vol. 94, No. 6, 1999
Table 1. Benefit of Screening for Hemochromatosis Compared to No Screening
Initial Screening Test Phenotyping with transferrin saturation ($12) Genotyping ($20) Genotyping ($50) Genotyping ($100) Genotyping ($173) Genotyping ($200)
Incremental Cost Saving Per Person Screened (US$) 0.97 2.1 228 278 2151 2178
Cost Per Homozygote Identified (US$) 2333 2563 7,371 20,595 39,901 47,042
Cost Per Homozygote With Potential Life-Threatening Illness* (US$) 2775 21544 20,202 56,444 109,358 128,929
* Assumes that 43% of men and 28% of women progress to life-threatening illness. (2) means that there is cost savings; (1) means that costs are incurred.
gous siblings will be identified. We assume that two siblings are tested per homozygote discovered. Cost of the phenotypic screening strategy with transferrin saturation combined with the cost of 4.6 missed homozygotes is $502,843. This is a higher cost than our previous model because we used transferrin saturation rather than the less expensive UIBC (29) and because of the use of American costs. Compared to the no-screening strategy, the incremental cost saving of phenotyping is $0.97 for each person screened (Table 1). Genotyping Screening Strategy In genotypic screening, we predict that 25.5 homozygous blood donors and 12.5 homozygous siblings will be identified. Cost of the genotypic screening strategy with a genetic test costing of $173 combined with the cost of 4.6 missed homozygotes is $2,031,727 or $109,358 to detect one homozygote with potential life-threatening illness. Compared to the no-screening strategy, the incremental cost saving of genotyping is 2$151 for each person screened (Table 1). There are 2.75 quality life days saved per person screened for a cost-utility of $20,042 per quality-adjusted life years saved. Cost saving is only apparent in genotypic screening if the initial gene test costs less than $28 with a prevalence of hemochromatosis of 0.003 (Table 1, Fig. 2). The cost of identifying homozygous donors and siblings rises as the cost of the genetic test increases.
Figure 2. Effect of changes in the cost of the initial genotyping test on the incremental cost savings per person screened including subsequent investigation of siblings. A negative value indicates costs incurred. (Cost expressed as 1998 US$—net present value, prevalence of hemochromatosis 5 .003.)
Genotyping Versus Phenotyping The phenotypic screening strategy with transferrin saturation is dominant (saves money) (Table 2). When blood donors with a high transferrin saturation follow the genotyping strategy with the gene test cost set at $173, the total cost is somewhat higher than phenotyping alone (Table 2). More homozygotes are missed with the combined strategy (phenotype, then genotype), 6.5 compared to 4.5 for genotyping and 4.6 for phenotyping. When genotyping is used to confirm hemochromatosis in patients with a high transferrin saturation, C282Y negative iron loaded patients are missed. Familial iron overload in the absence of the C282Y mutation is extremely rare, and investigation of siblings will be of a lower yield in these cases and could be eliminated with the subsequent cost savings (37). The major difference in cost between phenotyping and genotyping is the difference in the cost of the initial test: $12 for transferrin saturation versus $173 for the gene test. Utility Analysis Both phenotypic and genotypic strategies led to a similar incremental increase in quality-adjusted life days saved for each person screened (utility) (Table 1). Screening also produced a greater utility benefit in men compared to women (Table 3). Sensitivity Analysis GENETIC TEST. If the sensitivity of the genetic test is increased to 0.95, only 1.5 homozygotes are missed, and the total cost of screening, $2,015,866, is slightly reduced. We have considered the genetic test to have no false-positives except in the rare case of sample mislabeling. We do not consider a person homozygous for the C282Y mutation without iron overload to be a false-positive but rather a nonexpressing homozygote. For the purpose of the analysis, we modelled the effect of including these nonexpressing homozygotes as false-positives and decreased the specificity from 1.0 to 0.97. Decreasing the specificity to 0.97 (sensitivity 0.85), markedly increases the cost, $3,533,815, and decreases the incremental cost saving per person screened. This is as a result of donors with false-positive tests proceeding to further testing and follow-up.
AJG – June, 1999
Genetic Screening for Hemochromatosis
1597
Table 2. Comparison of Population Screening Strategies Incremental Cost Saving Per Person Screened (US$)
Homozygotes Missed (Out of 10,000 Donors)
Screening Tests Phenotyping with transferrin saturation ($12), liver biopsy Phenotyping with transferrin saturation, then genotyping Genotyping ($173), then ferritin
Cost Per Homozygote Identified (US$)
Cost Per Homozygote With Potential Life-Threatening Illness* (US$)
4.6
0.97
2333
2775
6.5
23.74
989
2,711
39,901
109,358
2151
4.5
* 43% of men and 28% of women progress to life-threatening illness. (2) refers to cost savings.
PREVALENCE. Fewer homozygotes are found at low prevalence rates and thus the cost of identifying each homozygote is higher. This may be relevant in urban areas with significant non-Caucasian populations. Our baseline prevalence estimate of .003 is conservative and would allow for some ethnic diversity in the population tested. PENETRANCE. The effect of varying the rate of progression to life-threatening symptoms is shown in Table 3. When penetrance is lowered: 1) the total cost of not screening decreases; 2) the total cost of identifying homozygotes with potential to develop life-threatening illness rises; and 3) incremental cost saving declines. This effect is apparent in both men and women (Table 3).
DISCUSSION The description of a new genetic test for a missense mutation on chromosome 6 introduces the possibility of population screening for hemochromatosis using a true genetic test rather than a phenotypic test for iron overload using the transferrin saturation and serum ferritin. The value of screening with a new genetic test is closely related to several critical features of the test, as follows. Cost of the Initial Screening Test Based on our model and cost projections, if the cost of the new genetic test can be kept to less than $28, it will result
in cost savings in blood donor population screening studies (Fig. 2). However, significant costs rather than savings are apparent as the cost of the genetic test increases into the range that has been charged in several reference American laboratories (Table 1). We have restricted the genetic test in this analysis to the C282Y mutation. The inclusion of the screening for the second mutation (H63D) provides only a limited amount of additional information. The gene test charge of $173 currently includes the cost of H63D testing in the United States. The strategy outlined may miss compound heterozygotes (C282Y/H63D) or apparent heterozygotes or homozygotes that carry another undetected mutation associated with iron overload. The compound heterozygote has represented less than 1% of our ironloaded patients, and previous reports have suggested that these patients have mild-to-moderate iron overload. Sensitivity of the Genetic Test Available data on sensitivity comes from six independent studies. The initial report from Mercator Genetics demonstrated that 83% of hemochromatosis patients were homozygous for the C282Y mutation in 176 patients. These patients were collected from different American medical centers, and the precision of the phenotypic assignment and confirmatory family studies presumably varied considerably in the samples. Subsequent studies from France (38), United Kingdom (39), and Australia (28) gave a sensitivity of 90 –100%
Table 3. Effect of Penetrance of Hemochromatosis Gene on Benefits of Genotypic Screening Strategy (Gene Test $173) Progression to Life-Threatening Illness Men 0.43 Men 0.3 Men 0.2 Men 0.1 Women 0.28 Women 0.2 Women 0.15 Women 0.1
Incremental Cost Incurred Per Person Screened (US$)
Cost Per Homozygote Identified (US$)
Cost Per Homozygote With Potential Life-Threatening Illness (US$)
Quality Life Days Saved Per Person Screened
Cost-Utility (Incremental Cost/ QALY* Saved)
142 160 173 187 164 176 183 191
39,730 44,628 48,396 52,163 40,005 42,885 44,686 46,486
92,394 132,432 198,648 397,296 93,034 142,951 297,905 464,857
3.09 2.45 1.50 0.71 2.43 1.40 1.33 1.06
16,535 23,468 41,518 95,269 24,268 45,359 49,484 64,946
* QALY 5 quality-adjusted life years.
1598
Adams and Valberg
in patients, including pedigree studies (40). An Italian study had a much lower sensitivity at 69% (41). Many of their patients had confounding variables such as alcoholism and chronic viral hepatitis, but the prevalence of the C282Y mutation was also less in the Italian population. Patients at our center have been homozygous in 95% of cases (42). We chose a sensitivity value of 85% for this study to take the wide range of reported sensitivities into consideration. Other Mutations A potential shortcoming of initial screening for the C282Y mutation is that it will not detect iron-loaded blood donors who carry another mutation or another disease associated with iron overload (41). Non-HLA linked iron-overload diseases have been reported particularly in non-Caucasian populations (43– 47). We have not included any cost associated with patients missed with those disorders. The phenotypic strategy may also detect patients with iron deficiency. This may benefit the patient but may also increase the cost of this strategy (48). Specificity of Genotyping We assume that patients who are homozygous for the mutation have genetic hemochromatosis. Previous studies have suggested that many patients with hemochromatosis have incomplete penetrance without phenotypic expression. In our own series, we have observed five families with HLA identical siblings who are homozygous for the C282Y mutation without evidence of iron overload (42, 49). In these subjects, predisposition to iron overload rather than the identification of a disease has been determined by genotyping. Preliminary results from our population screening study have detected two iron-loaded siblings of nonexpressing homozygotes (31). Case definition is an essential part of this analysis. We have chosen to emphasize genotyping as our “gold standard” because we continue to believe that most of the described cases of iron overload that are negative for the C282Y mutation represent isolated cases (nonfamilial), which differ from genetic hemochromatosis. However, we have set the sensitivity of the genotyping at 0.85 to allow for some atypical iron-loaded cases that may be considered to be hemochromatosis. To define hemochromatosis solely on the basis of degree of iron overload would strongly favor the phenotypic screening model over the genotyping model. Penetrance The penetrance used in our baseline assumptions may be an overestimate of the penetrance of disease in population-based studies. If we had sampled a younger population with more women, there would have been fewer homozygotes detected by phenotypic screening. Furthermore, our penetrance baseline estimates introduce a referral bias because this is a tertiary care center of hemochromatosis. Reducing the penetrance increases the cost of identifying homozygotes with the potential to develop life-threatening illness and increases the cost-utility of genotyping (Table 3). The detection of iron overload by phenotypic screening identifies those who are iron loaded
AJG – Vol. 94, No. 6, 1999
and may be a more practical approach because it immediately identifies patients in need of venesection therapy, but it requires an initial inexpensive screening test. We tested a model using transferrin saturation followed by C282Y testing without liver biopsy. This approach has a cost profile similar to our phenotyping model, but an additional two homozygotes are missed in the screening process. The major cost of all population screening models is driven by the cost of the initial screening test. It is for this reason that several studies (31) are exploring the use of an inexpensive automated unbound iron-binding capacity as an alternative to transferrin saturation. We have not included utility adjustments for the knowledge of a potential future illness because the disease is treatable with appropriate follow-up monitoring. We have also not included the indirect costs associated with loss of insurance or employment as a result of genetic discrimination. By limiting our model to blood donors, we have eliminated many patients who could be false-positive by phenotypic screening (alcoholics, chronic viral hepatitis, iron-loading anemias). If the genetic test is proven to have a specificity close to 1.0, the screening of other patient groups such as those in ambulatory care clinics and hospitalized patients may become more attractive than phenotypic screening where many false-positives are detected and unnecessary invasive procedures are needed to establish the presence of iron overload. The genotyping model in this study eliminates the use of diagnostic liver biopsy when compared to the phenotypic model. This is because of the high specificity assigned to the genotype test. Quantitative phlebotomy has been used as a diagnostic criterion for hemochromatosis in the absence of liver biopsy. In the phenotypic model, the hepatic iron distribution, hepatic iron concentration, and hepatic iron index contribute significantly to the confidence of the phenotypic assignment. The cost of liver biopsy is not a significant factor because relatively few donors proceed to liver biopsy in the phenotypic screening model (Table 1). The presence of cirrhosis has been associated with a decreased survival in hemochromatosis (8, 33), and other studies have suggested that it is important to establish the presence of cirrhosis for prognosis and to consider screening for hepatocellular carcinoma (50). The role of liver biopsy in the diagnosis of hemochromatosis requires reassessment since the development of genetic testing. It is likely that it will move from a diagnostic test to a prognostic test in selected cases. A multivariate analysis of 362 C282Y homozygotes who underwent liver biopsy demonstrated that the risk of fibrosis was minimal in the patient without hepatomegaly, with a ferritin , 1000 mg/ml, and a normal AST (51). These parameters would apply to most young homozygotes found in a screening study, so liver biopsy would not be required except perhaps in a C282Y negative iron-loaded patient. We have previously emphasized that hemochromatosis is an ideal disease for population screening based on the World Health Organization criteria (29). The development of a genetic test that can be performed at reasonable cost (, $28) has the potential to replace phenotypic screening for hemo-
AJG – June, 1999
chromatosis. The utilization of a genetic screening test for a common treatable disease is one of the triumphs of genetic mapping and a goal of the Human Genome Project. Implementation of population screening is dependent on the willingness of private or government health agencies to pay the initial screening costs to achieve savings many years later.
ACKNOWLEDGMENTS The authors thank Ann Kertesz for her assistance in the development and maintenance of the hemochromatosis database and David Lloyd for providing the software for decision analysis. This paper is dedicated to the memory and contributions of Leslie Valberg who passed away during the preparation of the manuscript. Grant Support: Dr. Adams acknowledges the support of the Physician Services Incorporated Foundation of Ontario, Medical Research Council of Canada, and the Pomajba family.
Genetic Screening for Hemochromatosis
14. 15. 16. 17. 18.
19. 20.
Reprint requests and correspondence: Paul C. Adams, M.D., Department of Medicine, London Health Sciences Centre, 339 Windermere Rd., London, Ontario, Canada N6A 5A5. Received July 31, 1998; accepted Dec. 9, 1998.
21.
REFERENCES
23.
1. Simon M, Bourel M, Fauchet R, et al. HLA and “non-immunological” disease: Idiopathic haemochromatosis. Lancet 1976;1:973– 4. 2. Borwein ST, Ghent CN, Flanagan PR, et al. Genetic and phenotypic expression of hemochromatosis in Canadians. Clin Invest Med 1983;6:171–9. 3. Edwards CQ, Griffen LM, Goldgar D, et al. Prevalence of hemochromatosis among 11,065 presumably healthy blood donors. N Engl J Med 1988;318:1355– 62. 4. Hallberg L, Bjorn-Rasmussen E, Jungner I. Prevalence of hereditary haemochromatosis in two Swedish urban areas. J Intern Med 1989;225:249 –55. 5. Edwards CQ, Kushner JP. Screening for hemochromatosis. N Engl J Med 1993;328:1616 –20. 6. Niederau C, Fischer R, Sonnenberg A, et al. Survival and causes of death in cirrhotic and in noncirrhotic patients with primary hemochromatosis. N Engl J Med 1985;313:1256 – 62. 7. Adams PC, Kertesz AE, Valberg LS. Clinical presentation of hemochromatosis: A changing scene. Am J Med 1991;90:445–9. 8. Adams PC, Speechley M, Kertesz AE. Long-term survival analysis in hereditary hemochromatosis. Gastroenterology 1991;101:368 –72. 9. Phatak PD, Guzman G, Woll JE, et al. Cost-effectiveness of screening for hereditary hemochromatosis. Arch Intern Med 1994;154:769 –76. 10. Bassett M, Leggett BA, Halliday JW, et al. Analysis of the cost of population screening for haemochromatosis using biochemical and genetic markers. J Hepatology 1997;27:517–24. 11. Bassett ML, Doran TJ, Halliday JW, et al. Idiopathic hemochromatosis: Demonstration of homozygous-heterozygous mating by HLA typing of families. Hum Genet 1982;60: 352– 6. 12. Beaumont C, Simon M, Fauchet R, et al. Serum ferritin as a possible marker of the hemochromatosis allele. N Engl J Med 1979;301:169 –74. 13. Dadone MM, Kushner JP, Edwards CQ, et al. Hereditary
24.
22.
25.
26. 27. 28. 29.
30. 31. 32. 33. 34. 35.
1599
hemochromatosis. Analysis of laboratory expression of the disease by genotype in 18 pedigrees. Am J Clin Pathol 1982; 78:196 –207. Expert Scientific Working Group. Summary of a report on assessment of the iron nutritional status of the United States population. Am J Clin Nutr 1985;42:1318 –30. Tanner AR, Wright R. Hemochromatosis screening. Lancet 1988;2:286. Lindmark B, Eriksson S. Prevalence of hemochromatosis. N Engl J Med 1988;319:1548 –9 (letter). Leggett BA, Halliday JW, Brown NN, et al. Prevalence of haemochromatosis amongst asymptomatic Australians. Br J Haematol 1990;74:525–30. Wiggers P, Dalhoj J, Petersen PH, et al. Screening for hemochromatosis: Influence of analytical imprecision, diagnostic limit and prevalence on test validity. Scand J Clin Invest 1991;51:143– 8. Jonsson JJ, Johannesson GM, Sigfusson N, et al. Prevalence of iron deficiency and iron overload in the adult Icelandic population. J Clin Epidemiol 1991;44:1289 –97. Wiggers P, Dalhoj J, Kiaer H, et al. Screening for haemochromatosis: Prevalence among Danish blood donors. J Intern Med 1991;230:265–70. Baynes RD, Meyer TE, Bothwell TH, et al. A screening test for detecting iron overload in population studies. S Afr Med J 1988;74:167–9. Borwein S, Ghent CN, Valberg LS. Diagnostic efficacy of screening tests for hereditary hemochromatosis. Can Med Assoc J 1984;131:895–901. Bassett ML, Halliday JW, Bryant S, et al. Screening for hemochromatosis. Ann NY Acad Sci 1988;526:274 – 89. Bassett ML, Halliday JW, Ferris RA, et al. Diagnosis of hemochromatosis in young subjects: Predictive accuracy of biochemical screening tests. Gastroenterology 1984;87:628 – 33. Niederau C, Niederau CM, Lange S, et al. Screening for hemochromatosis and iron deficiency in employees and primary care patients in Western Germany. Ann Int Med 1998; 128:337– 45. Feder JN, Gnirke A, Thomas W, et al. A novel MHC class I-like gene is mutated in patients with hereditary hemochromatosis. Nat Genet 1996;13:399 – 408. Jazwinska EC, Cullen LM, Busfield F, et al. Haemochromatosis and HLA-H. Nat Genet 1996;14:249 –51. Jouanolle A-M, Gandon G, Jezequel P, et al. Haemochromatosis and HLA-H. Nat Genet 1996;14:251–2. Adams PC, Gregor JC, Kertesz AE, et al. Screening blood donors for hereditary hemochromatosis: Decision analysis model based on a thirty-year database. Gastroenterology 1995; 109:177– 88. Adams PC, Kertesz A, Valberg LS. Screening for hemochromatosis in children of homozygotes: Prevalence and costeffectiveness. Hepatology 1995;22:1720 –7. Adams PC, Kertesz AE, Barr R, et al. Population screening for hemochromatosis with the unbound iron binding capacity. Gastroenterology 1998;114:A1199 (abstract). Adams PC, Speechley M. The effect of arthritis on the quality of life in hereditary hemochromatosis. J Rheumatol 1996;23: 707–10. Niederau C, Fischer R, Purschel A, et al. Long-term survival in patients with hereditary hemochromatosis. Gastroenterology 1996;110:1107–19. Bulaj ZJ, Griffen L, Jorde LB, et al. Clinical and biochemical abnormalities in people heterozygous for hemochromatosis. N Eng J Med 1996;335:1799 – 805. Adams PC. Prevalence of abnormal iron studies in heterozy-
1600
36.
37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51.
Adams and Valberg
gotes for hereditary hemochromatosis—An analysis of 255 heterozygotes. Am J Hematol 1994;45:146 –9. Bassett ML, Halliday JW, Powell LW. HLA typing idiopathic hemochromatosis: Distinction between homozygotes and heterozygotes with biochemical expression. Hepatology 1981;1: 120 – 6. Pietrangelo A, Montosi G, Garuti C, et al. No HFE, No HLA, No 6p-linked adult hemochromatosis: A new genetic iron overload condition. Hepatology 1998;28:528A (abstract). Adams PC, Bradley C, Henderson AR. Evaluation of the hepatic iron index as a diagnostic criterion in hereditary hemochromatosis. J Lab Clin Med 1997;130:509 –14. The UK Haemochromatosis Consortium. A simple genetic test identifies 90% of UK patients with haemochromatosis. Gut 1997;41:841–5. Burke W, Thomson E, Khoury M, et al. Hereditary hemochromatosis: Gene discovery and its implications for population-based screening. JAMA 1998;280:172– 8. Carella M, D’Ambrosio L, Totaro A, et al. Mutation analysis of the HLA-H gene in Italian hemochromatosis patients. Am J Hum Genet 1997;60:828 –32. Adams PC, Chakrabarti S. Genotypic/phenotypic correlations in genetic hemochromatosis: Evolution of diagnostic criteria. Gastroenterology 1998;114:319 –23. Oliver M, Scully L, Guiraudon C, et al. Non-HLA linked hemochromatosis in a Chinese woman. Dig Dis Sci 1995;40:1589–91. Eason RJ, Adams PC, Aston CE, et al. Familial iron overload with possible autosomal dominant inheritance. Aust NZ J Med 1990;20:226 –30. Gordeuk V, Mukiibi J, Hasstedt SJ, et al. Iron overload in Africa. Interaction between a gene and dietary iron content. N Engl J Med 1992;326:95–100. Morita H, Ikeda S, Yamamoto K, et al. Hereditary ceruloplasmin deficiency with hemosiderosis: A clinicopathological study of a Japanese family. Ann Neurol 1995;37:646 –56. Wurapa RK, Gordeuk V, Brittenham G, et al. Primary iron overload in African Americans. Am J Med 1996;100:9 –18. Baer DM, Simons JL, Staples RL, et al. Hemochromatosis screening in asymptomatic ambulatory men 30 years of age and older. Am J Med 1995;98:464 – 8. Adams PC, Campion ML, Gwandon G, et al. Clinical and family studies in genetic hemochromatosis: Microsatellite and HFE studies in five atypical families. Hepatology 1997;26:986–90. Fargion S, Fracanzani AL, Piperno A, et al. Prognostic factors for hepatocellular carcinoma in genetic hemochromatosis. Hepatology 1994;20:1426 –31. Guyader D, Jacquelinet C, Moirand R, et al. Non-invasive prediction of fibrosis in C282Y homozygous hemochromatosis. Gastroenterology 1998;115:929 –36.
APPENDIX: VARIABLE DEFINITIONS Age at Death ● heart failure (female) 5 60 ● heart failure (male) 5 55 ● cirrhosis (female) 5 72 ● cirrhosis (male) 5 67 ● diabetes (female) 5 73 ● diabetes (male) 5 68 ● hepatocellular carcinoma (female) 5 69 ● hepatocellular carcinoma (male) 5 64 ● life-threatening event homozygote (female) 5 71 ● life-threatening event homozygote (male) 5 66 ● normal (female) 5 78
AJG – Vol. 94, No. 6, 1999
●
normal (male) 5 72
Age at First Symptoms ● female homozygote 5 59 ● male homozygote 5 54 Medical Care Costs in US$ ● annual ambulatory cirrhosis care 5 1,500 ● annual ambulatory diabetes care 5 1,500 ● probability of compliance with follow-up 5 0.80 ● physicians consultation 5 200 ● physicians hospital visit 5 50 ● physicians office visit 5 50 ● daily inpatient care 5 500 ● hepatocellular carcinoma, 1998 5 50,000 ● serum ferritin 5 20 ● genotyping 5 173 (this charge currently includes C282Y and H63D testing in the USA) ● liver biopsy 5 600 ● probability of a complication 5 0.002 ● probability of dying from complication 5 0.05 ● transferrin saturation 5 12 ● unbound iron binding capacity 5 1.40 Donors ● age at initial screening 5 37 ● male probability 5 0.6 ● number 5 10,000 ● homozygous probability 5 0.003 ● probability of life-threatening manifestations (female) 5 0.28 ● probability of life-threatening manifestations (male) 5 0.43 Siblings ● probability HLA identical 5 0.25 ● female probability 5 0.5 ● homozygote probability 5 0.25 ● number 5 50 Probability of Developing Complications ● cirrhosis 5 0.303 ● cirrhosis and diabetes 5 0.273 ● diabetes 5 0.182 ● heart failure 5 0.045 ● die from heart failure 5 0.1 ● heart failure and cirrhosis 5 0.045 ● heart failure, cirrhosis, and diabetes 5 0.107 ● heart failure and diabetes 5 0.045 ● hepatocellular carcinoma 5 0.25 Venesection Therapy ● venesection cost 5 $45 ● number of initial venesections (donor) 5 53 ● maintenance number of venesections per year 5 3 ● number of initial venesections (siblings) 5 6 (females) and 30 (males)