Clinical Features of Genetic Hemochromatosis in Women Compared with Men

Methods

Patients

Patients in this study were referred to the Clinique des Maladies du Foie in Rennes, France (since 1970), or the London Health Sciences Centre, London, Canada (since 1972), for assessment and treatment of hemochromatosis. New patients, including many with mild iron overload who do not have hemochromatosis, are evaluated weekly at both centers. The annual referral load of patients suspected of having iron overload has increased at both centers. During our study, hemochromatosis was diagnosed in 485 French homozygotes (137 women and 348 men) and 213 Canadian homozygotes (57 women and 156 men). The diagnosis of homozygous hemochromatosis in probands was based on clinical history; physical examination; serum transferrin saturation; serum ferritin level; and results of liver biopsy with hepatic iron concentration and hepatic iron index [6, 7]. Hepatic iron concentration was measured either from fresh liver biopsy tissue or from tissue excised from paraffin blocks. Standard lymphocyte microcytotoxicity tests in Terasaki plates [8, 9] were used for HLA-A and HLA-B typing. In probands, HLA typing was not used for diagnosis.

For affected relatives in whom hemochromatosis was detected by family screening, the diagnosis was made on the basis of HLA typing that was identical to that of the proband in the same generation; diagnosis did not depend on the phenotypic expression of disease [10]. The presence of arthritis and pigmentation was determined on the basis of history and physical examination. We considered patients who required oral hypoglycemic agents or insulin to have diabetes. Cardiac disease was diagnosed on the basis of a history of congestive heart failure, physical examination, or arrhythmias requiring medical therapy. The presence of liver disease was determined according to abnormal findings on history, physical examination, blood chemistry, imaging studies of the liver, and biopsy of liver tissue. A presenting clinical event was considered to be incidental if it was not clearly related to hemochromatosis. Therefore, only patients presenting with fatigue, diabetes, liver disease, arthritis, cardiac disease, or pigmentation were considered to have had a clinical feature of hemochromatosis as the presenting symptom. Patients without these symptoms were considered to be asymptomatic. Alcohol consumption was assessed by patient history; consumption of more than 80 g of alcohol per day by men and more than 60 g per day by women was considered excessive.

So that we could compare the phenotypic expression of hemochromatosis, all women who met the diagnostic criteria were considered for the study and were matched for year of birth (within 1 year) with men who had hemochromatosis and were from the same country. Fourteen Canadian women were so old that we could not match them with a man who had the same year of birth. Medical reports were checked for information on age at menopause or hysterectomy. Additional data on reproductive function, including the number of pregnancies and the use of oral contraceptives or intrauterine devices, were obtained by a questionnaire administered to 77 French women.

All patients underwent weekly venesection, during which 300 to 500 mL of blood was drawn, until the serum ferritin level was approximately 50 µg/L. The amount of iron removed was calculated as the number of liters of blood removed to achieve iron depletion, multiplied by 0.5 g. Only patients treated at the two tertiary centers were analyzed for exchangeable body iron because detailed venesection records were not available for other patients.

Statistical Analysis

Data are expressed as the differences in means or proportions with 95% CIs. Although men and women were matched for year of birth for the purpose of assembling the study groups, nonpaired testing was used because no clinically meaningful relation existed between the pairs. In fact, the age at evaluation of iron measurements and the prevalence of symptoms almost always differed for a man and woman who were born in the same year. The objective of the matching was to obtain a group of men in whom hemochromatosis was diagnosed in an era similar to that in which the women received a diagnosis of hemochromatosis rather than to match each woman with male homozygote. Correlation coefficients were calculated for the relation between iron measurements and age at presentation.

Results

Patient Characteristics

The demographic and biochemical characteristics of all homozygotes with hemochromatosis are shown in Table 1. Two hundred sixty-six homozygotes (133 women and 133 men) were French, and 86 (43 women and 43 men) were Canadian. The median year of diagnosis was 1990 for French patients and 1988 for Canadian patients (range, 1961 to 1995). Severity of disease, as assessed by serum ferritin level, hepatic iron concentration, hepatic iron index, and percentage of patients who had cirrhosis, did not significantly differ between French and Canadian patients (data not shown). The mean age at presentation was similar in men and women, even when analysis was restricted to probands (50.9 years compared with 47.5 years; mean difference, −3.4 years [CI, −7.0 to 0.1 years]). Of 341 patients for whom HLA typing was done, 235 (69%) tested positive for HLA-A3.

Presenting Clinical Features

The clinical event that led the patient to seek medical attention was incidental to the diagnosis of hemochromatosis in 36% of female probands and 45% of male probands. Among patients who presented with a symptom of hemochromatosis, women presented more frequently with fatigue and men presented more frequently with symptoms of liver disease. Presenting symptoms did not statistically differ for French and Canadian patients.

Clinical Findings

The prevalence of clinical findings was similar for men and women. The prevalence of cirrhosis and diabetes was higher in men, whereas the prevalence of fatigue and pigmentation was higher in women (Table 2).

Clinical Features Found on Testing

We compared the clinical and biochemical features of the 30 men and 65 women who were found to be HLA identical to an affected sibling by family screening. Serum ferritin levels and transferrin saturation were normal in 11 (6.2%) of these women (range, 28 to 59 years of age) and none of the men. Normal serum ferritin levels and abnormal transferrin saturation were more common in women (4.0%) than in men (1.1%). The process of assembling the study cohort resulted in the inclusion of more women than men in whom hemochromatosis was discovered through family screening, but a difference in the proportion of women without biochemical phenotypic expression was apparent when only cases discovered through family screening were considered (11 of 65 women compared with 0 of 30 men). Serum ferritin levels were lower in these women (433 µg/L) than in men (1181 µg/L) (mean difference, 748 µg/L [CI, 467 to 1029 µg/L]). In women and men, hepatic iron concentration (243 µmol/g compared with 262 µmol/g; mean difference, 19 µmol/g [CI, −70.3 to 108 µmol/g]) and amount of iron removed by venesection (3.9 g compared with 5.7 g; mean difference, 1.8 g [CI, −0.88 to 4.5 g]) did not differ significantly.

Cirrhosis and Related Factors

Cirrhosis occurred in only 21 women (13.8%) compared with 43 men (26%). Most patients with cirrhosis were probands; only five cases of cirrhosis were discovered in patients in whom hemochromatosis was diagnosed by family screening. Although excessive alcohol consumption was more common in male patients (36 of 176) than in female (8 of 176) patients, it was not more common in men with cirrhosis than in women with cirrhosis (Table 3). The correlation coefficient for hepatic iron concentration and amount of iron removed was 0.69 in men and 0.55 in women (P < 0.001).

Effects of Age at Menopause, Hysterectomy, and Pregnancy

In 81 women, menses had ceased by the time that hemochromatosis was diagnosed; the other 95 women were premenopausal. Median age at menopause or hysterectomy was 50 years (range, 25 years to 57 years). Hepatic iron concentration and hepatic iron index were higher in women in whom menstruation stopped before 50 years of age than in those in whom menstruation stopped after 50 years of age (Table 4). The correlation coefficient between age at menopause or hysterectomy and hepatic iron concentration was −0.36(P < 0.01); the coefficient between hepatic iron index and age at menopause or hysterectomy was −0.48(P < 0.01). Additional data on reproductive function obtained in 77 French women showed that the mean number of pregnancies was 2.7 ± 2.3. No significant correlation was found between the number of pregnancies and hepatic iron concentration or amount of iron removed by venesection or between the duration of use of oral contraceptives or an intrauterine device and serum ferritin levels, hepatic iron concentration, or hepatic iron index.

Discussion

We describe a large series of women with genetic hemochromatosis who were seen at two tertiary referral centers for hemochromatosis. Our study provides detailed clinical, biochemical, and pathologic evidence that women with genetic hemochromatosis can have significant clinical expression of the disease and symptoms (such as marked iron overload and cirrhosis) similar to those found in men.

An unexpected observation was that less iron was removed by venesection in women than men despite similar hepatic iron concentrations. This finding is not readily explained by differences in liver size and may indicate that men have greater extrahepatic deposition of iron. In addition, serum ferritin levels were higher in men than in women, which may also suggest that men have increased extrahepatic iron stores. The accuracy of venesection records is difficult to ascertain; therefore, we restricted our study to patients who underwent therapy at the two tertiary hospitals. We considered the amount of iron removed by venesection to be the most representative measure of the total body iron stores of the patient. If this is true, then our data are consistent with a lower iron burden in women. However, it is important to emphasize that some women presented with a degree of iron overload that was similar to that seen in the most severely affected men.

It was previously assumed that menstruation reduces body iron stores in women, thereby minimizing tissue damage. Therefore, premature menopause and hysterectomy were considered to be risk factors for enhanced iron accumulation in women with hemochromatosis. Women whose menses ceased before 50 years of age (the median age of menopause in this study) had a higher hepatic iron concentration and hepatic iron index than women in whom menopause occurred later (Table 4). Women probably accumulate iron more rapidly after menstruation ceases. Although we could not determine the rate of iron accumulation in individual patients, our results show that women can have severe hemochromatosis before menopause.

It has also been reported [11] that the clinical onset of hemochromatosis is delayed in women. In our study, however, age at presentation did not differ between women and men, even when analysis was restricted to probands. A comparison of all male and female homozygotes without matching may lead to overrepresentation of older women because of the increased longevity of women in general. The lack of a statistically significant correlation between hepatic iron concentration and age at presentation (r = 0.1) is probably related to the heterogeneous expression of disease; this feature has been described previously, even within families [12-14].

The overall phenotypic expression of the disease has also been thought to be substantially less marked in women [15-18]. The clinical presentation of the disease differed between women and men: Liver disease and diabetes were more common in men, and fatigue and pigmentation were more common in women (Table 2). It is difficult to interpret the relation between clinical manifestations and the iron burden in men and women because the relation between iron burden and extrahepatic symptoms has not been clearly established [14]. Furthermore, the prevalence of such nonspecific symptoms as fatigue, arthritis, and diabetes was similar in women and men and thus led to a similar prevalence in symptomatic patients. We did not determine the prevalence of symptoms in the general population.

Without a control group, it is difficult to conclude that hemochromatosis is expressed equally in men and women. A study of Canadian patients with hemochromatosis showed that arthritis and diabetes were three times more common in patients with hemochromatosis than in controls with chronic liver diseases [19]. However, women can not only develop a degree of hepatic iron overload similar to that of men but can also develop cirrhosis. Cirrhosis was associated with an increased hepatic iron concentration and an increased serum ferritin level, as has been shown elsewhere [14, 20, 21]. Among the 95 patients in whom hemochromatosis was diagnosed because the patients were found to be HLA identical to a sibling with iron overload, 11 of the 65 women (who were phenotypically unaffected) and none of the 30 men had normal serum ferritin levels and transferrin saturation at the time of diagnosis. We observed no other obvious confounding factor, such as a source of pathologic blood loss or history of frequent blood donation.

Our findings suggest a slight underexpression of hemochromatosis in women. This concurs with the predominance of men in studies of population screening for hemochromatosis [1, 3]. The matching in our study creates an illusion of a similar prevalence of clinical hemochromatosis in men and women; the ratio of male probands to female probands registered at our two study centers is, in fact, approximately 3 to 1. This suggests that the disease is underdiagnosed in the general population of women; the detection of fewer female probands despite a similar prevalence of nonspecific symptoms in the population of affected women suggests that the threshold for pursuing further diagnostic investigations may be higher in women.

The major limitation of our retrospective study is the possibility of sampling bias. A selection of patients from tertiary referral centers for hemochromatosis may not be representative of patients in the general population. Furthermore, although matching by year of birth allowed for the assessment of clinical symptoms at the same time point, patients were evaluated over a 30-year period during which knowledge about the clinical expression of hemochromatosis was evolving. Interpretation of our findings is further confounded by the possibility that mild, asymptomatic diabetes was underdiagnosed and by the difficulty of correcting for the background rates of diabetes, cardiac disease, and other findings that were caused by a condition other than hemochromatosis. Similarly, the limitation of data on iron removal to patients who underwent venesection only at the tertiary referral centers may have introduced selection bias. Although data were carefully collected over this period, the degree of investigation and documentation varied among patients. Despite these limitations in study design, age at diagnosis did not differ in men and women. We considered the lack of phenotypic expression in HLA-identical women to be related to incomplete penetrance, but we have not excluded the somewhat unlikely possibility of genetic recombination [22, 23].

We conclude that homozygous hemochromatosis is slightly underexpressed in women, although severe disease can be present and the clinical features are different than those seen in men. We show that women with genetic hemochromatosis can and do develop progressive iron overload and clinical symptoms despite menstruation and pregnancy. Clinicians should have a substantial index of suspicion for hemochromatosis in women with unexplained clinical features that are compatible with the disease (such as fatigue, arthralgia, and pigmentation) and should consider testing these women for hemochromatosis by measuring transferrin saturation and serum ferritin levels. This approach allows for earlier diagnosis and avoids progressive tissue damage. Our data also indicate that women should not be excluded from population screening programs [24].

Dr. Adams: Department of Medicine, University Hospital, University of Western Ontario, Box 5339 A, London, Ontario N6A 5A5, Canada.