Brian Schmitt, MD, MPH; Robert M. Golub, MD; Richard Green, MD
Grant Support: The paper was written under contract with the American College of Physicians.
Potential Financial Conflicts of Interest: None disclosed.
Requests for Single Reprints: Brian Schmitt, MD, MPH, Medicine and Neurology Service Line (111), Hines Veterans Affairs Hospital, Hines, IL 60141; e-mail, firstname.lastname@example.org.
Current Author Addresses: Dr. Schmitt: Medicine and Neurology Service Line (111), Hines Veterans Affairs Hospital, 5th Avenue and Roosevelt, Hines, IL 60141.
Dr. Golub: Northwestern University, 676 North St. Clair, Suite 200, Chicago, IL 60611.
Dr. Green: Northwestern University, 303 East Chicago Avenue, Chicago, IL 60611.
Schmitt B., Golub R., Green R.; Screening Primary Care Patients for Hereditary Hemochromatosis with Transferrin Saturation and Serum Ferritin Level: Systematic Review for the American College of Physicians. Ann Intern Med. 2005;143:522-536. doi: 10.7326/0003-4819-143-7-200510040-00011
Download citation file:
Published: Ann Intern Med. 2005;143(7):522-536.
Appendix: Search Methods and Article Selection
Therapeutic phlebotomy for hereditary hemochromatosis is relatively safe and presumably efficacious when offered before cirrhosis develops, so screening primary care patients is of substantial interest.
To conduct a systematic review of the evidence on 1) the prevalence of the disease in primary care, 2) the risk for morbid or fatal complications for untreated patients, 3) the diagnostic usefulness of transferrin saturation and serum ferritin level in identifying early disease, 4) the efficacy of early treatment, and 5) whether the benefits of screening outweigh the risks.
MEDLINE search from 1966 through April 2004, complemented by reference review of identified original studies and review articles published in English.
PubMed Clinical Queries filters search of prognosis, diagnosis, etiology, or treatment were used depending on the question. Two authors reviewed all titles and abstracts.
Two investigators independently reviewed extracted data.
The prevalence of hereditary hemochromatosis was 1 in 169 patients to 1 in 556 patients (nâ€‰= 3 studies). Uncontrolled, prospective studies of genetic homozygous patients did not consistently identify a link to overt hereditary hemochromatosis. A serum ferritin level less than 1000 Âµg/L was predictive of absence of cirrhosis. Six studies demonstrated reduced survival in patients with cirrhosis. Diagnostic studies varied with respect to case definition. No blinded, independent comparisons of screening tests with the gold standard (biopsy or results of quantitative phlebotomy) or randomized, controlled trials of phlebotomy were identified. Cost-effectiveness analysis was limited by lack of prospective data on the natural history of the disease.
Varied case definition and lack of prospective cohort studies or randomized trials.
The available evidence does not demonstrate that benefits outweigh the risks and costs of screening for hemochromatosis.
Hereditary hemochromatosis is a genetic disorder of iron metabolism resulting in excessive iron overload and is associated with clinically significant morbid and fatal complications related to tissue iron deposition (1). It is an autosomal recessive disorder linked to a mutation of the HFE gene on the short arm of chromosome 6. This gene is the result of a single base change in which tyrosine is substituted for cysteine at position 282 of the HFE protein (C282Y). Available evidence strongly supports an association of the C282Y mutation with hereditary hemochromatosis, although other mutations in HFE have been identified. The substitution of aspartic acid for histidine at position 63 (H63D) has been observed but has limited clinical effect (2). Several genetic mutations are also linked to juvenile hereditary hemochromatosis (3, 4).
The C282Y mutation is variably and unpredictably associated with phenotypic changes of iron overload that include elevated transferrin saturation and serum ferritin levels. An increasing transferrin saturation is the earliest detectable biochemical abnormality in hereditary hemochromatosis and is attributed to increased intestinal iron absorption (4). Enterocytes aberrantly continue to transfer iron from the gut into the bloodstream rather than store the iron as ferritin. The role of HFE in this pathophysiologic process is not fully clear. Some patients with the HFE gene never progress beyond this biochemical abnormality. However, progressive iron overload occurs in others. Marked elevation of serum ferritin level has been associated with histologic evidence of iron deposition (5). The morbid complications of hereditary hemochromatosis are the result of tissue deposition, and they develop late in its course. They include arthritis, diabetes mellitus, congestive heart failure, cirrhosis, and hepatocellular carcinoma (6). Liver iron deposition with cirrhosis is associated with reduced survival (7).
The estimated prevalence of hereditary hemochromatosis is 1 in 200 persons to 1 in 250 persons in the general population, making hereditary hemochromatosis one of the most common genetic disorders (8). However, the prevalence of hereditary hemochromatosis varies depending on the case definition of disease (9). With genetic testing of populations originating in northern Europe, approximately 0.5% is homozygous for the C282Y mutation (10). With phenotypic screening, 1% to 6% of the U.S. population have elevated transferrin saturation levels, and 11% to 22% of them have concomitant elevations of their serum ferritin levels (11).
Over the past decade, interest in promoting general population screening for hereditary hemochromatosis has increased (12, 13). Advances in genetic testing, changing definitions of the disease that include earlier stages of iron overload, increased appreciation of the prevalence and importance of the disease, and the presumed effectiveness of a simple intervention (phlebotomy) have prompted a debate on the benefits and risks of a screening intervention program in the United States. The use of genetic testing to screen family members of individuals identified with hereditary hemochromatosis seems to be cost-effective (14). The benefit in primary care is less clear. In 1997, the Centers for Disease Control and Prevention and the National Institutes of Health sponsored a meeting on iron overload, public health, and genetics (15). A result of the conference was a published review of the evidence for hemochromatosis screening (13). Evidence for screening was evaluated against the U.S. Preventive Services Task Force criteria (16). Little evidence was available to support the efficacy of genetic testing, and there were substantial ethical, legal, and social concerns. Likewise, evidence was insufficient to support the use of transferrin saturation, and few comparative data were available to establish the magnitude and clinical significance of risk in patients with various levels of iron overload. However, others argued for expanded screening and have targeted primary care physicians for educational intervention to improve awareness of the disease (12, 17). The Working Group on Research Priorities identified the evidence that is most needed to provide a scientific basis for population screening (18). They identified the need for research to characterize the natural history of the relationship between genotype and phenotype in hereditary hemochromatosis and other iron overload disorders. They also identified 3 additional priorities: 1) development of an optimal approach for screening for iron overload; 2) analyses of the cost-effectiveness of screening; and 3) assessment of the ethical, legal, and social implications of screening. Given this background, we must examine the current evidence and determine whether it now supports general population screening for hereditary hemochromatosis.
To promote screening within the primary care setting, one must demonstrate that the disease is common, the burden is substantial, the treatment is efficacious, the screening tests are accurate, the screening is effective, and the benefits of screening outweigh the risks (16). We evaluate the evidence to support screening within the primary care setting. Because of the controversy surrounding genetic testing as a screening tool (low penetrance of hereditary hemochromatosis and lack of complete identification of genetic mutations associated with hereditary hemochromatosis), our review focuses on the phenotypic measures that are most likely to be useful in primary care: transferrin saturation and serum ferritin level. To evaluate the evidence, we address each key question that is relevant to a screening intervention.
We conducted a systematic review for each question in MEDLINE for papers published from 1966 through April 2004 by using PubMed Clinical Queries filters for a sensitive search of prognosis, diagnosis, etiology, or treatment depending on the question. We included only English-language studies. Two reviewers independently reviewed all abstracts. A third reviewer resolved conflicts about inclusion of an article. We also manually searched references from included studies. The Appendix includes details for conducting the search for each subquestion. We assessed methodologic quality of studies for a specific question by using accepted epidemiologic criteria (19). We did not use any formal method of quality assessment or scoring.
This paper was written under contract with the American College of Physicians. The funding source had no role in the collection, analysis, or interpretation of the data or in the decision to submit the manuscript for publication.
We identified 3 prevalence studies in primary care settings (Table 1) and 12 studies in various general population settings (Table 2) (Appendix Figure 1). Because of variability among studies, we examined the prevalence data in 3 ways: use of the strict definition recommended in the Hemochromatosis and Iron Overload Screening (HEIRS) study (34); use of the prevalence reported by the investigator(s); and use of a ceiling estimate, which assumed that the many study participants with elevated transferrin saturation and serum ferritin levels who declined liver biopsy or therapeutic phlebotomy had the same probability of primary iron overload as those who were actually evaluated.
Using strict criteria, we found that the prevalence of hereditary hemochromatosis within a primary care setting was 0.18% to 0.59% (1 in 169 patients screened to 1 in 556 patients screened). The prevalence reported by the investigators was 0.2% to 1.8% (1 in 56 patients screened to 1 in 500 patients screened). We estimated the ceiling prevalence of hereditary hemochromatosis to be 0.37% to 0.79% (1 in 127 patients screened to 1 in 270 patients screened). Most patients with hereditary hemochromatosis did not have concomitant cirrhosis. Using the HEIRS study definition, we found that hereditary hemochromatosis prevalence was higher in white men (0.37% to 0.46%) than nonwhite men (0% to 0.17%) and was higher in men than women.
Using strict criteria, we estimated the prevalence of hereditary hemochromatosis from the 2 general population studies (23, 24) to be 0.16% to 0.28% (1 in 357 patients screened to 1 in 625 patients screened). The investigators reported higher rates—as high as 0.74% in Norwegian men. In screening employees or blood donors, the prevalence of hereditary hemochromatosis was higher in men than women and, when demographic data were available, was higher in white men than nonwhite men (Table 2).
Within the primary care setting (n = 23 055) and using strict criteria, we identified hereditary hemochromatosis in 59 patients (0.13%) (equivalent to 1 in 795 screened patients). Of these 59 patients, 55 patients (93%) were 40 years of age or older, 46 patients (78%) were men, and 38 patients (93%) in the 2 studies that provided racial characteristics were white. Fifty-two of 59 patients (88%) did not have cirrhosis. These data demonstrate that the prevalence of hereditary hemochromatosis varies across different subgroups determined by race, sex, and age. The highest prevalence of hereditary hemochromatosis detected by screening within primary care seems to be in white men, probably ranging between 1 of 185 screened patients and 1 of 219 screened patients (20, 21). Screening that targeted white men 40 years of age or older would further increase the prevalence of hereditary hemochromatosis.
The definition of iron overload varied across studies. Some studies evaluated the relationship between the asymptomatic elevation of transferrin saturation and serum ferritin levels and the development of disease, while other studies addressed the relationship between iron tissue deposition and the development of complications and death. Consequently, we separately aggregated the evidence to address these 2 related but different ways of characterizing iron overload.
We identified 11 studies (5, 35-44) for inclusion (Appendix Figure 2). None was a prospective cohort study comparing survival or complications in patients with and without hereditary hemochromatosis defined by an elevated serum ferritin level.
Three uncontrolled, prospective studies (35-37) examined the change of iron stores over time in individuals identified as being C282Y homozygotes (Table 3). One study (35) followed 12 Australian patients for 17 years. While the median transferrin saturation and ferritin level increased, some patients had constant levels and some had decreasing ferritin levels. In a Canadian study of 22 participants, Yamashita and Adams (36) observed a decrease in serum ferritin levels in 13 patients and serum ferritin levels remaining less than the upper limit of normal in 20 patients after a median of 4 years. In the Copenhagen City Heart Study (37), average transferrin saturation levels increased substantially after 25 years of follow-up, but no homozygote developed hereditary hemochromatosis. In summary, these studies did not consistently identify increasing transferrin saturation and serum ferritin levels over time and did not demonstrate a clear link to the overt clinical manifestation of hereditary hemochromatosis.
Three studies examined the relationship between biochemical measures of iron overload and diabetes mellitus (Table 4). Jiang and colleagues (38) used blood samples obtained from 32 826 women in the Nurses' Health Study cohort between 1989 and 1990 who did not have diabetes mellitus, cardiovascular disease, or cancer. Multivariate relative risk increased across the quintiles of the initial ferritin levels (1.00, 1.09, 1.26, 1.30, and 2.68, respectively; P < 0.001 for trend). However, the results must be interpreted cautiously because cases and controls had several differences at baseline. Ellervik and colleagues (39) compared Danish patients with type 1 diabetes mellitus with controls who were selected from the Copenhagen City Heart Study. Nine patients with diabetes who were homozygous for C282Y had transferrin saturation greater than 50%, and 6 of these patients had cirrhosis by liver biopsy. Ellervik and colleagues did not directly compare transferrin saturation or serum ferritin levels between those with and without diabetes mellitus to allow a direct determination of the association, and they did not present comparison data for cases and controls. Mainous and colleagues (40) used the National Health and Nutrition Examination Survey I (NHANES I) database. The adjusted odds ratio of diabetes (controlling for age, sex, race, cholesterol level, obesity, and hypertension) was 0.89 (95% CI, 0.59 to 1.34) for transferrin saturation greater than 45%, 0.95 (CI, 0.53 to 1.70) for transferrin saturation greater than 50%, and 1.03 (CI, 0.44 to 2.43) for transferrin saturation greater than 55%. However, Mainous and colleagues did not actually identify patients with sustained elevations in transferrin saturation and serum ferritin levels and did not measure the association with diabetes mellitus.
Three studies examined the relationship between biochemical measures of iron excess and cirrhosis (Table 5). All were cross-sectional studies. Two studies (41, 42) developed a prediction rule for patients homozygous for the C282Y mutation to diagnose or exclude cirrhosis and to more clearly define those needing liver biopsy. Both studies had separate derivation and validation populations. The earlier study (41) aimed to predict the absence of cirrhosis. It was derived in a French population and validated in a Canadian population. Only 1 of 105 patients (0.9%) with a ferritin level of 1000 µg/L or less had cirrhosis. In combination with a normal aspartate aminotransferase (AST) level and no hepatomegaly, 0 of 94 patients had cirrhosis. Findings in the validation population were similar. The more recent follow-up report (42) found that ferritin levels greater than 1000 µg/L, platelet counts less than 200 × 109 cells/L, and elevated AST levels led to a correct diagnosis of cirrhosis in 77% of the Canadian participants and in 90% of the French participants who were tested. Morrison and colleagues (5) found that patients with ferritin levels less than 1000 µg/L were unlikely to have cirrhosis on liver biopsy (1 of 93 patients). These 3 studies strongly suggest that patients who are at high risk for hereditary hemochromatosis (homozygous C282Y mutation) with serum ferritin levels of 1000 µg/L or less are unlikely to have cirrhosis.
Mahon and colleagues (43) examined the relationship between C282Y mutations and idiopathic dilated cardiomyopathy (Table 6). This case–control study demonstrated that 31 of 207 cases (15%) carried the C282Y mutation compared with 24 of 200 controls (12%) (adjusted odds ratio, 1.2 [CI, 0.7 to 2.2]). With respect to the H63D mutation, the odds ratio was 1.6 (CI, 1.1 to 2.5). Serum iron and transferrin saturation did not correlate with disease severity or survival. The clinical significance of the link of the H63D mutation with cardiomyopathy is unclear.
Mainous and colleagues (44) examined the association between transferrin saturation and all-cause mortality by using the NHANES I database (Table 6). After controlling for comorbid disease, smoking, and cholesterol level, all-cause mortality statistically significantly increased for transferrin saturation greater than 55% compared with those with lower transferrin saturation (hazard ratio, 1.60 [CI, 1.17 to 2.21]). Although no one who died had hereditary hemochromatosis listed as a cause of death, they were more likely to have cirrhosis or diabetes mellitus.
We identified 13 original studies (7, 45-56)(Appendix Figure 3). Seven studies explored the relationship of hereditary hemochromatosis and survival, and the remainder evaluated the association of this disease with other complications (Table 7).
Wojcik and colleagues (45), studying patients from a tertiary care facility, found that actuarial survival at 5, 10, and 20 years was 95%, 93%, and 66%, respectively, with a mean follow-up of 7.3 years. At the time of diagnosis, 36% of men and 19% of women had life-threatening diseases. Cirrhosis and diabetes mellitus were the only major factors that affected survival. Milman and colleagues (46) found that after a median of 8.5 years, survival in patients with cirrhosis and diabetes was statistically significantly lower than survival in those without either complication (who had a survival similar to that in the general population on the basis of a nonconcurrent, historical control). The main causes of death were liver failure secondary to cirrhosis and cirrhosis with liver cancer. Although not clearly stated, Niederau and colleagues' studies (7, 56) probably shared some of the same patients. Their more recent publication (7) also found that survival in noncirrhotic and nondiabetic patients with hereditary hemochromatosis was similar to survival in those in the healthy population (nonconcurrent, historical control). The adjusted relative risk for death for cirrhosis and diabetes mellitus was 4.3 and 2.4, respectively. Fargion and colleagues (47) found that over a median of 44 months, 44 of 146 patients with cirrhosis died (20 with hepatocellular carcinoma and 10 with liver failure). No deaths occurred in patients without cirrhosis. Adams and colleagues (48) found that patients with hereditary hemochromatosis and cirrhosis were 5.5 times more likely to die than those without cirrhosis. Diabetes mellitus did not increase the risk. Yang and colleagues (49) conducted a retrospective analysis of the Multiple-Cause Mortality Files compiled by the National Center for Health Statistics. Patients who died of hereditary hemochromatosis were more likely to have liver neoplasms, liver disease, and cardiomyopathy. Conversely, patients who died of liver neoplasms, liver disease, and cardiomyopathy were more likely to have hereditary hemochromatosis than those without these conditions. These studies consistently demonstrated a link between hereditary hemochromatosis-associated cirrhosis and reduced survival and strongly suggest a link between hereditary hemochromatosis–related diabetes mellitus and reduced survival.
Adams (50), using receiver-operating characteristic curve analysis to identify the threshold of hepatic iron concentration for predicting the presence of cirrhosis, derived an optimal threshold of 283 mmol/kg dry weight, with a sensitivity of 85% and a specificity of 84%. Fargion and colleagues (51) evaluated the prognostic factors for hepatocellular carcinoma in hereditary hemochromatosis. Hepatocellular carcinoma developed in 29% of patients with cirrhosis compared with 0% in those without cirrhosis. The 2 studies by Deugnier and colleagues (52, 53) identified a specific preneoplastic change (iron-free foci) in patients with hereditary hemochromatosis that was associated with the development of hepatocellular carcinoma. Ammann and colleagues (54) found that out of 36 patients with hereditary hemochromatosis, 5 cases of hepatoma and 6 cases of extrahepatic carcinoma developed over 8 years.
The case–control study by Salonen and colleagues (55) found that men with high iron stores were 2.4 times more likely to have diabetes than men with lower iron stores.
We identified 3 studies (20-22) conducted in the primary care setting (Appendix Figure 4). No study independently and blindly compared the screening tests for iron overload (transferrin saturation and serum ferritin level) with the gold standard (liver biopsy or mobilizable iron by phlebotomy) in all screened patients.
Phatak and colleagues (21) screened for sustained transferrin saturation of 45% or greater (n = 311). Of 50 patients with transferrin saturation greater than 55% and serum ferritin levels greater than 200 µg/L, 35 patients were offered liver biopsy and 21 patients underwent the procedure. Eighteen patients (86%) were found to have hereditary hemochromatosis. Of 29 patients who declined or were not offered liver biopsy, 17 patients were labeled as having clinically proven hereditary hemochromatosis. Of this group, only 2 patients had mobilizable iron stores of 2 g or more. Using the strict HEIRS study criteria and these diagnostic cutoff levels, we identified hereditary hemochromatosis in 20 patients (40%). Using the authors' criteria, 36 patients (72%) had hereditary hemochromatosis. For those with transferrin saturation between 45% and 55% and serum ferritin levels greater than 200 µg/L (n = 78), biopsy was recommended in 13 patients and hereditary hemochromatosis was identified in 7 patients. Of those who did not have a recommendation for biopsy (n = 65), 2 patients had mobilizable iron stores of 2 g or more. Using strict criteria and these cutoff levels, 9 of 78 screened patients (11.5%) had hereditary hemochromatosis. Using the authors' criteria, we identified hereditary hemochromatosis in 11 patients (14%).
Niederau and colleagues (22) studied 3027 healthy outpatients in West Germany. They used thresholds of transferrin saturation greater than 50% in men and greater than 60% in women and serum ferritin levels of 250 µg/L in men and 350 µg/L in women. Using these cutoff levels, 235 patients (7.8%) had elevated serum ferritin levels, 139 patients (4.6%) had elevated transferrin saturation, and 44 patients (1.5%) had both. Of this latter group who were retested, 31 patients (1% of total) had persistently elevated transferrin saturation and serum ferritin levels. Twenty-three patients continued with further evaluation. We identified hereditary hemochromatosis in 18 patients (78%) by using these diagnostic cutoff levels for transferrin saturation and serum ferritin levels.
Baer and colleagues (20) studied 3977 consecutive men 30 years of age or older who belonged to the Oakland Kaiser Permanente health maintenance organization. Those with transferrin saturation of 62% or greater underwent repeated testing in the fasting state. Patients with persistently elevated transferrin saturation (≥62%) and serum ferritin levels of 500 µg/L or greater were offered a liver biopsy. Forty patients (1%) had transferrin saturation that exceeded the cutoff value. Thirty-six patients were available for follow-up. Of these, 14 patients with persistent elevations were referred for liver biopsy and 12 patients underwent the procedure. One of the 12 patients had a serum ferritin level less than 500 µg/L. These 12 patients had hereditary hemochromatosis defined by strict criteria. All patients with transferrin saturation of 62% or greater and serum ferritin levels of 500 µg/L or greater had hereditary hemochromatosis.
These 3 studies demonstrate substantial variation in the testing sequence; the decision thresholds for transferrin saturation, serum ferritin level, and their combined results; and the application of the gold standard. Patients with levels less than the threshold values at any location in the testing sequence were not assessed by the gold standard or followed for the development of disease. Not all patients with levels greater than the threshold values underwent the gold standard assessment. Consequently, sensitivity and specificity cannot be determined. The proportion of patients with levels greater than the threshold values who underwent the gold standard assessment does provide a limited estimate of the positive predictive value.
Other researchers have reviewed the proportion of screened patients in a general population with initial and repeatedly positive transferrin saturation test results (57). Initial positive test results were found in 1.1% to 6.2% of patients screened. A relationship between the proportion positive and the diagnostic cutoff level was not apparent. The highest proportion was in blood donors with transferrin saturation cutoff values of 50% or greater (29), and the lowest proportion was in those with cutoff values of 55% or greater (28). This differs from the 3 primary care studies in which the more stringent criteria were associated with a higher proportion of patients identified with hereditary hemochromatosis.
We did not identify any randomized trial. We identified only 1 study for inclusion because it compared patients who were adequately phlebotomized with those were not adequately phlebotomized (46). We identified an additional study (7) by manual search because of its before–after phlebotomy comparison of liver histology (Appendix Figure 5).
Milman and colleagues (46) retrospectively identified a cohort of 158 patients with hereditary hemochromatosis by querying departments of medicine and pediatrics in Denmark and by reviewing a Danish death registry. Patients were followed for a median period of 8.5 years. Survival of patients who were adequately phlebotomized (n = 66) was greater than survival of those were not adequately phlebotomized (n = 62). The estimated Kaplan–Meier survival was 93% versus 48% at 5 years and 78% versus 32% at 10 years. Furthermore, adequately treated patients with cirrhosis or diabetes had better survival than those who were not adequately treated. However, we could not determine the clinical comparability of patients with cirrhosis who were or were not adequately treated.
Niederau and colleagues (7) studied a cohort of 251 German patients with hereditary hemochromatosis for a mean follow-up of 14.1 years. Patients underwent an initial liver biopsy and were then phlebotomized until serum ferritin level was normalized. At this point, they underwent a second biopsy. A before–after phlebotomy comparison was performed in 185 patients. However, no methodologic safeguards minimized ascertainment bias. Forty-two patients had improved liver histology after phlebotomy, and 2 patients had deterioration in histology. Improvement or deterioration was based on changes in fibrosis stage. Except for 1 patient, all improvements were by 1 stage. One patient had a 2-stage improvement. The 2 patients who deteriorated had a 1-stage decrement. The remaining 141 patients did not change. However, given the possibility of sampling error from one liver biopsy to the next, any change in histologic status must be interpreted cautiously.
No studies met a standard of evidence (blinded, randomized, controlled trial) that clearly establishes the efficacy of therapeutic phlebotomy. However, they do support the existing model of disease and suggest a benefit. Given the current opinion and the lack of clinically significant side effects, a randomized, controlled trial will probably not be performed.
Without the ability to establish the efficacy of therapeutic phlebotomy in a research setting, we could not establish the effectiveness of a screening approach directly in the primary care setting. The electronic literature search for a decision-analytic model of a cost-effectiveness analysis identified 4 relevant postings. Two models (58, 59) were previously reviewed in Annals of Internal Medicine(13). Cogswell and colleagues (13) identified the major determinants of screening cost-effectiveness: prevalence and disease burden; sensitivity and specificity of the screening tests; adherence to screening, diagnosis, and therapy; and costs of screening, diagnosis, and therapy. The lack of data on natural history led decision analysts to use data from hospital registries of patients affected with hemochromatosis, increasing the possibility of overestimation of morbidity and mortality.
Two studies (60, 61) have been published since Cogswell and colleagues' comprehensive review, and we evaluated them for new insights. In their decision-analytic model, Asberg and colleagues (60) based the prevalence of hereditary hemochromatosis and the risk for liver cirrhosis on cross-sectional data obtained from a population of 30 509 men (24). If their model accurately reflects reality, phenotypic screening of a cohort of 10 000 young (30 years of age), predominantly white men from the general population would generate a gain of 8 quality-adjusted life-years compared with waiting for symptomatic disease to occur. They estimated the costs to be $250 per quality-adjusted life-year saved.
The model used a screening approach that began with an initial nonfasting transferrin saturation (cutoff level, >55%), followed by a fasting transferrin saturation if the first result was positive. If this second test result was greater than 55%, they measured the serum ferritin level. If the serum ferritin level was elevated (>200 µmol/L), they referred the patient for clinical examination. They did not base the diagnosis of hereditary hemochromatosis on invasive liver biopsy results or the amount of mobilizable iron removed. Therefore, they did not assign risk to the clinical examination. If they could not identify a secondary cause of iron overload, they labeled the patients as having hereditary hemochromatosis. Consequently, they actually incorporated the screening tests into the gold standard of diagnosis, which explain the high estimated sensitivity and specificity of the screening tests used in their model. In their theoretical cohort of 10 000 men, Asberg and colleagues estimate that 53 people with hereditary hemochromatosis would be identified. However, if we take the available data from primary care (21) and use a strict case definition (biopsy or mobilizable iron), we estimate that only 18 patients with hereditary hemochromatosis would be identified. These marked differences illustrate the effect of differing case definitions on the models.
Furthermore, Asberg and colleagues (60) used Markov models to estimate outcomes for men with hereditary hemochromatosis with or without treatment (phlebotomy) and men without hereditary hemochromatosis with or without phlebotomy (4 health states). The only complication of hereditary hemochromatosis used to model survival was the presence or absence of cirrhosis. They estimated the incidence of cirrhosis from cross-sectional data of men with cirrhosis at the time of initial diagnosis across various age groups in 2 studies (21, 24). Asberg and colleagues assumed that men with hereditary hemochromatosis but without cirrhosis at diagnosis and treated with phlebotomy would not develop cirrhosis and estimated their quality of life to be 1.0. However, the only available data on the before–after phlebotomy effect on liver histology suggest that 77% of patients had no change during the before–after time period (from initial diagnosis to normalization of serum ferritin level) and 1% progressed (7). Although presumed to be true, on the basis of this information, whether phlebotomy will prevent the development of cirrhosis in all treated patients over longer periods of time is not certain. Furthermore, Asberg and colleagues did not consider the effect of being labeled with hereditary hemochromatosis (psychologically or economically).
The study by Adams and Valberg (61) suggested that screening blood donors for hereditary hemochromatosis may save money. They used a decision tree for the natural history of unscreened donors that was similar to their previous cost models (62, 63). The target population for this decision analysis was a hypothetical cohort of 10 000 voluntary blood donors and 50 siblings of identified homozygotes. They compared genotypic screening with phenotypic screening and modeled the effect of diabetes mellitus and heart failure on outcomes, as well as cirrhosis. As in previous analyses, they used a database of 170 homozygous patients with hereditary hemochromatosis who were referred to a tertiary care facility. An optimal screening strategy used phenotypic testing followed by genotyping—unless the cost of genetic testing was less than $28.
Unfortunately, these 2 new analyses are still subject to the limitations that we have noted in the primary studies. The lack of data on natural history has forced analysts to rely on cross-sectional studies or databases at large tertiary care referral sites. More recent data have raised additional questions about the usefulness of genetic testing because many people with HFE mutations do not progress to overt disease, and we have no prospective data on the incidence of cirrhosis or diabetes mellitus in patients with elevations of transferrin saturation and serum ferritin levels but without disease at the time of diagnosis.
On the basis of our review of the literature, hereditary hemochromatosis is a common genetic disease within the primary care setting, especially in white men older than 40 years of age. Given the relatively high estimated prevalence of hereditary hemochromatosis (1 in 127 patients to 1 in 270 patients), we must understand the magnitude of the burden that will likely occur. Ideally, the answer would come from a large prospective cohort study identifying patients with and without hereditary hemochromatosis at an early and uniform point in time (sustained elevations of transferrin saturation and serum ferritin levels without clinically significant liver iron deposition) and followed at regular intervals with blinded assessment to detect the important outcomes of cirrhosis, hepatocellular carcinoma, diabetes mellitus, congestive heart failure, arthritis, and death. Unfortunately, no prospective cohort studies early in the course of the disease are available to provide this information. The available data demonstrate low penetrance of HFE mutations and suggest that the magnitude of burden may be less than commonly believed. Specifically, 3 small longitudinal studies (34-37) of patients homozygous for the C282Y mutation did not demonstrate predictable progression to overt clinical hereditary hemochromatosis over long periods of follow-up.
The association between phenotypic measures of iron overload (transferrin saturation and ferritin levels) and the development of iron-related complications is inconsistent. No prospective cohort studies of patients without diabetes mellitus at baseline with and without elevated biochemical measures of iron are available to determine differences in the incidence of new disease. The available retrospective studies did not ensure balanced comparison groups, making selection bias likely (38-40). Consequently, the interpretation of an association between serum iron measures and diabetes mellitus was difficult at best. The prospective studies of patients with hereditary hemochromatosis based on liver iron deposition demonstrated a high prevalence of co-existing diabetes mellitus, and diabetes mellitus was independently associated with an increased risk for death in many of these cohorts. Given its high prevalence in patients with overt hereditary hemochromatosis, diabetes mellitus is probably associated, but the magnitude of risk is not clear. Associations with arthritis and congestive heart failure were not well supported.
The strongest link between iron overload and disease is the association of serum ferritin levels with cirrhosis. Two separate research groups consistently demonstrated the usefulness of the test for identifying a low-risk group for the presence of cirrhosis (serum ferritin level < 1000 µg/L). In addition, a prediction rule for the presence of cirrhosis has been developed and validated. Serum ferritin level is useful for predicting the prevalence of cirrhosis at a single point in time. A patient with a serum ferritin level less than 1000 µg/L without hepatomegaly and a normal AST level is unlikely to have cirrhosis. On the other hand, patients with serum ferritin levels greater than 1000 µg/L, platelet counts less than 200 × 109 cells/L, and elevated AST levels have a high probability of cirrhosis. The available literature suggest that the presence of cirrhosis in patients with hereditary hemochromatosis at the time of diagnosis portends a poor prognosis (a substantial reduction in survival and an increased risk for hepatocellular carcinoma). Only 5% of those with hereditary hemochromatosis identified by screening in a primary care setting had cirrhosis. However, the available data did not permit determination of the incidence of new cases of cirrhosis in patients with hepatic iron deposition at the time of original diagnosis.
The identification of the HFE gene and development of a genetic test to detect the presence of C282Y mutation has made the case definition of hereditary hemochromatosis vary (9). This has resulted in several different approaches for studying the diagnostic efficacy of available tests. Some investigators used genotyping as the gold standard and determined the sensitivity and specificity of phenotypic tests in predicting the presence or absence of homozygous C282Y genotypes. Without convincing data to demonstrate that patients with HFE mutations will progress to disease or early death, this definition may not have clinical utility. Other investigators have used persistently elevated transferrin saturation and serum ferritin levels, without biopsy or quantitative phlebotomy, to diagnose hereditary hemochromatosis. This results in a diagnostic incorporation bias (19). Consequently, we selected a gold standard for hereditary hemochromatosis for our review that required independent demonstration of iron overload (liver deposition or the amount of iron removed by phlebotomy). The use of this case definition limits the ability to determine the diagnostic efficacy of phenotypic tests in the primary care setting because patients with transferrin saturation or a second serum ferritin level interpreted as normal do not undergo liver biopsy, and no prospective follow-up data are available for those with negative test results to identify false-negative results. Further, only those with sustained elevations of transferrin saturation and serum ferritin levels are offered the gold standard test. Given these difficulties, determining the likelihood of disease in those with positive test results defined as sustained elevations of phenotypic measures (positive predictive value) was the best existing measure available. The diagnostic cutoff levels for transferrin saturation and serum ferritin have varied across studies as well. The higher cutoff levels (transferrin saturation ≥ 62% and serum ferritin levels ≥ 500 µg/L) identified a subgroup in which all patients had hereditary hemochromatosis. The least stringent criteria (transferrin saturation ≥ 45% and serum ferritin levels > 200 µg/L) identified a group in which only 11.5% had hereditary hemochromatosis. Given the predictive usefulness of serum ferritin level (<1000 µg/L) for excluding the presence of cirrhosis and the high prevalence of hereditary hemochromatosis in patients screened with higher cutoff levels, using these cutoff levels in a screening program is not unreasonable. However, some patients with hereditary hemochromatosis probably would be missed, and no data on the use of repeated tests or an appropriate interval for repeated testing are available. More important, no data are available on the incidence of cirrhosis in patients with levels greater than these cutoff values and without cirrhosis at the time of hereditary hemochromatosis diagnosis.
Much remains unknown about the natural history of hereditary hemochromatosis. Although white men older than 40 years of age screened by phenotypic tests have the highest prevalence of hereditary hemochromatosis (established by liver biopsy or mobilized iron) in a primary care setting, how many without cirrhosis or other organ involvement at the time of diagnosis will progress to overt disease remains unclear. Further, no clearly defined patient or laboratory characteristics are available to risk-stratify patients into groups that are more or less likely to develop overt disease. The recent attempts to model benefits and risks of screening were not based on natural history studies and did not consider the effect of disease labeling (including insurance and social and psychological well-being).
McDonnell and Parrish (64) have reviewed and integrated the available data into a model of natural history to estimate progression to overt disease. They estimate that, in screening a theoretical cohort of 1 million people from the general population with either phenotypic or genotypic tests, only 2 per 1 million persons screened by HFE screening and 3 per 1 million persons screened by transferrin saturation screening would be identified with cirrhosis. Given the potential social, psychological, and financial harms from early hereditary hemochromatosis diagnosis of an asymptomatic patient, the benefit of early diagnosis relative to the risk remains unclear. Our knowledge deficit about natural history remains an important limitation to recommendations about screening patients in primary care or the general population. The HEIRS study (34) is currently being conducted and may help to shed light on some of these issues. This primary care-based study of 100 000 adults over 5 years will provide valuable information related to the prevalence of hereditary hemochromatosis, value of phenotypic and genotypic screening tests, and benefits and risks for primary care-based screening. The results of the screening stage have been recently reported (65). The prevalence of the C282Y mutation was more common (0.44%) in non-Hispanic white persons than in other groups. For those C282Y homozygous individuals, the serum ferritin levels were elevated in 88% of men and 57% of women. However, the actual prevalence of hereditary hemochromatosis based on liver biopsy or therapeutic phlebotomy was not presented.
Hereditary hemochromatosis seems to be relatively common in certain ethnic, sex, and age groups. In patients with overt clinical manifestations, hereditary hemochromatosis can shorten survival and substantially reduce the quality of life. Our current lack of information on natural history severely limits our understanding about the burden related to this disease from a societal perspective. Phenotypic tests can identify subgroups of patients with differing probabilities of hereditary hemochromatosis by liver biopsy and can be used to predict the presence or absence of cirrhosis. No data based on randomized trials indicated the true efficacy of phlebotomy. The limited quality of the evidence base precludes phenotypic testing for a judgment that hereditary hemochromatosis in the primary care setting fulfills the necessary criteria for a screening test (16).
We searched the medical literature to determine the prevalence of asymptomatic hereditary hemochromatosis in primary care settings or in the general population defined by increased hepatic iron concentration on liver biopsy or by evidence of substantial iron overload on the basis of the response to therapeutic phlebotomy. We used the current and ongoing HEIRS study (34) definitions for primary iron overload (hepatic iron concentration > 30 mmol/kg dry weight or by removal of >2000 mg of storage iron by phlebotomy to achieve a serum ferritin level < 200 µg/L). We considered only individuals with evidence of phenotypic primary iron overload to have hereditary hemochromatosis. We conducted a MEDLINE search using the terms hereditary hemochromatosis, hemochromatosis, prevalence, primary care, general population, and screening. We restricted the review to original studies with at least 1000 patients. For a study to be included, the patients identified as having primary iron overload by screening (transferrin saturation and serum ferritin level) must have had either a liver biopsy or a measured response to therapeutic phlebotomy. Because many study participants with elevated transferrin saturation and serum ferritin levels declined liver biopsy or therapeutic phlebotomy, we calculated “ceiling estimates” of prevalence, assuming the same probability of primary iron overload as in those who were actually evaluated. Consequently, we presented prevalence data in 3 ways: use of the strict definition recommended in the HEIRS study, use of the reported prevalence by the investigator, and use of a ceiling estimate.
An electronic search of PubMed identified 345 unique postings through 3 March 2004. After reviewing titles and abstracts and manually reviewing references, we considered 15 studies (20-33, 66) for inclusion (Appendix Figure 1). Four studies (20-22, 66) were conducted in an outpatient clinical care setting. However, we excluded 1 of these studies (66) because it probably included inpatients and outpatient specialty clinics. The other 3 studies were conducted in a primary care setting. Two studies (23, 24) screened patients in the general population, 5 studies (22, 25-28) screened employees, and 5 studies (29-33) screened blood donors.
In 1998, the need for information on the natural history of hereditary hemochromatosis at various stages of iron overload was identified (18). By using the PubMed Clinical Queries filters for prognosis and the terms hereditary hemochromatosis, hemochromatosis, transferrin saturation, serum ferritin, cirrhosis, hepatocellular carcinoma, diabetes mellitus, congestive heart failure, arthritis, and natural history, we examined the literature for new evidence on the transition from states of asymptomatic iron overload (by elevated transferrin saturation and serum ferritin level) to tissue iron deposition (by liver biopsy), clinical disease, and death. We looked for prospective studies that allowed us to determine the incidence of new complications or comparative studies (retrospective or prospective cohort and case–control) that allowed us to determine the association of iron overload with the risk for complications (by relative risk or odds ratio). For a study to be included, patients had to be identified as having primary iron overload with repeated elevations of both transferrin saturation and serum ferritin levels. We determined the outcome of diabetes mellitus by using the Diabetes Expert Committee criteria (67). The outcomes of cirrhosis and hepatocellular carcinoma required tissue confirmation. The outcome of congestive heart failure required symptoms, signs, and objective assessments of reduced left ventricular systolic function (left ventricular ejection fraction 0.45). Arthritis required clinical assessment of signs and symptoms and at least evidence of synovial swelling.
Regarding the relationship between transferrin saturation–serum ferritin level and hereditary hemochromatosis–related disease, an electronic search through 20 April 2004 identified 170 unique postings. After reviewing the titles and abstracts, we identified 10 studies (5, 36-43) for inclusion (Appendix Figure 2). We identified 1 additional study by manually searching references (37). Of these 11 studies, none was a prospective cohort study comparing survival, complications, or both in patients with and without hereditary hemochromatosis defined by an elevated serum ferritin level.
Regarding the relationship between primary iron tissue deposition and hereditary hemochromatosis–related disease, an electronic search through 20 April 2004 identified 259 unique postings. After reviewing the titles and abstracts, we identified 13 original studies (7, 45-56)(Appendix Figure 3). Seven studies explored the relationship of hereditary hemochromatosis and survival, and the remainder evaluated the association of hereditary hemochromatosis with other complications.
We examined the medical literature to determine the diagnostic efficacy of common tests used to assess iron overload (elevated transferrin saturation on 2 occasions, elevated serum ferritin level, or both) in predicting the presence or absence of hereditary hemochromatosis as previously defined. We used the PubMed Clinical Queries filters for diagnosis and the terms hereditary hemochromatosis, hemochromatosis, transferrin saturation, and serum ferritin. We included original studies that were conducted in a primary care setting that addressed the diagnostic usefulness of transferrin saturation or serum ferritin level in predicting hereditary hemochromatosis defined by liver biopsy or response to therapeutic phlebotomy.
The electronic search identified 375 unique postings. Of these, we identified 7 for possible inclusion (21-23, 25, 27, 28, 68). We identified 2 additional studies by reviewing references (20, 29). Only 3 studies (20-22) were conducted in the primary care setting (Appendix Figure 4). No studies independently and blindly compared the screening tests for iron overload (transferrin saturation and serum ferritin) with the gold standard (liver biopsy or mobilizable iron by phlebotomy) in all screened patients.
On the basis of large uncontrolled case series, most hepatologists have accepted the efficacy of phlebotomy for improving survival in patients without cirrhosis (7). This acceptance has made it difficult to conduct prospective, controlled comparisons (cohort studies or randomized, controlled trials) for ethical reasons. However, we assessed the rigor of the available literature for answering this question. We searched for controlled comparisons of phlebotomy in asymptomatic patients with iron overload or in patients with liver iron deposition but without cirrhosis. We used the PubMed Clinical Queries filters for treatment and the terms hereditary hemochromatosis, hemochromatosis, survival, and diabetes mellitus, congestive heart failure, arthritis, cirrhosis, and hepatocellular carcinoma. We included original studies conducted in a primary care setting or in a representative sample of the general population that addressed the usefulness of therapeutic phlebotomy in patients with hereditary hemochromatosis. We looked for differences in diabetes mellitus, congestive heart failure, arthritis, cirrhosis, hepatocellular carcinoma, and survival in patients receiving or not receiving therapeutic phlebotomy.
An electronic search of MEDLINE identified 409 unique postings. We identified only 1 study for inclusion because it compared those adequately phlebotomized with those inadequately phlebotomized (46). We identified 1 additional study by manual search (7) because of its before–after phlebotomy comparison of liver histology (Appendix Figure 5).
A review for evidence of the effectiveness of hereditary hemochromatosis screening in the primary care setting was contingent on identification of studies to establish the efficacy of screening within a research setting. To assess the balance of benefits and risks identified in subquestion 4, we evaluated all integrative secondary data analyses (decision analysis or cost-effectiveness analysis) that modeled the benefits and harms of a screening program in primary care. To be included, the model had to compare some form of phenotypic hereditary hemochromatosis screening with no screening.
The electronic literature search for a decision analytic model of a cost-effectiveness analysis identified 4 unique postings. We included 1 (60) in our review. The 3 excluded postings were not decision-analytic comparisons. Manual search of references identified 8 additional references for possible inclusion (14, 20, 58, 59, 61-63, 66). We excluded the studies by Balan and colleagues (66) and Baer and colleagues (20) because they did not model costs or effects across 2 or more screening options. They simply provided the costs involved in screening patients in a clinic setting. We excluded 2 additional studies because they evaluated the cost-effectiveness of screening family members of heterozygotes (14, 63). No studies specifically evaluated the cost-effectiveness of screening within the primary care setting. The remaining studies addressed general population screening, and we included them for review. Two (58, 59) were previously reviewed in Annals of Internal Medicine(13). Cogswell and colleagues (13) identified the major determinants of screening cost-effectiveness: prevalence and disease burden; sensitivity and specificity of the screening tests; adherence to screening, diagnosis, and therapy; and costs of screening, diagnosis, and therapy. The lack of data on natural history led decision analysts to use data from hospital registries of patients affected with hemochromatosis, increasing the possibility of overestimation of morbidity and mortality.
We evaluated 2 studies (61, 62), which have been published since Cogswell and colleagues' comprehensive review (13), for new insights.
The In the Clinic® slide sets are owned and copyrighted by the American College of Physicians (ACP). All text, graphics, trademarks, and other intellectual property incorporated into the slide sets remain the sole and exclusive property of the ACP. The slide sets may be used only by the person who downloads or purchases them and only for the purpose of presenting them during not-for-profit educational activities. Users may incorporate the entire slide set or selected individual slides into their own teaching presentations but may not alter the content of the slides in any way or remove the ACP copyright notice. Users may make print copies for use as hand-outs for the audience the user is personally addressing but may not otherwise reproduce or distribute the slides by any means or media, including but not limited to sending them as e-mail attachments, posting them on Internet or Intranet sites, publishing them in meeting proceedings, or making them available for sale or distribution in any unauthorized form, without the express written permission of the ACP. Unauthorized use of the In the Clinic slide sets will constitute copyright infringement.
May 18, 2006
Targeted testing for hemochromatosis
The review and guidelines for hemochromatosis screening published in the Annals identified several gaps in our knowledge of the disease. However, one particular scenario was not addressed: the patient with incidentally noted hepatic dysfunction and abnormal iron indices. This type of patient is often encountered in the course of routine clinical practice, but has not been the focus of any study. Would these supposed high-risk patients warrant genetic testing, as suggested in the previous AASLD guidelines from 2001?
Results provided by:
Copyright © 2016 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use
This PDF is available to Subscribers Only