Reversibility of hepatic fibrosis in treated genetic hemochromatosis: A study of 36 cases


  • Potential conflict of interest: Nothing to report.


The current study was undertaken to assess whether fibrosis could regress under venesection therapy in patients with C282Y homozygous genetic hemochromatosis. The 36 patients studied were recruited from a subfile of our database consisting of 125 C282Y homozygotes with either severe fibrosis or cirrhosis (F3 or F4 fibrosis stage, respectively, according to the METAVIR grading system). The second liver biopsy was performed for management of liver cancer, extrahepatic surgery, or assessment of liver fibrosis. All paired biopsies were reviewed by two pathologists without knowledge of clinical data. Among the 13 patients who had F3 fibrosis on their initial liver biopsy, 3 had F0, 6 had F1, and 2 had F2 on their second liver biopsy. Among the 23 patients with cirrhosis on their initial liver biopsy, 1 had F0, 4 had F1, 3 had F2, and 2 had F3 on their second liver biopsy. When defining regression of fibrosis as a decrease of at least 2 METAVIR units, fibrosis regressed in 9 of 13 (69%) F3 and in 8 of 23 (35%) F4. When the ratio of gammaglobulins (g/L) to (platelets [n/mm3] × prothrombin activity [%]) was greater than 7.5, fibrosis never regressed. In conclusion, these data extend the concept of regression of fibrosis to patients with treated genetic hemochromatosis and suggest that some simple biochemical tests would be predictive of further regression of fibrosis as a result of venesection therapy. If confirmed on larger series, this could modify the ultrasound screening policy of hepatocellular carcinoma in genetic hemochromatosis. (HEPATOLOGY 2006;44:472–477.)

Recent evidence indicates that fibrosis could be reversible even when at an advanced stage. In experimentally induced fibrosis, secondary biliary cirrhosis was shown to be reversible with no functional consequences.1 In humans, spontaneous resolution of liver fibrosis may occur after successful treatment of the underlying disease. This was described in a small series of patients with chronic hepatitis B2–5 and D,6 hemochromatosis,7–9 primary or secondary biliary cirrhosis,10–12 nonalcoholic steatohepatitis,13 Indian childhood cirrhosis,14 and autoimmune hepatitis.15–17 Chronic hepatitis C virus infection has been the most extensively studied, and therapy with interferon alone or in association with ribavirin has been shown to induce regression of fibrosis irrespective of its virological effects.18–23

Genetic hemochromatosis (GH) benefits from an efficient treatment based on regular phlebotomies. However, data on regression of fibrosis in this disease are scarce, and no study of well-defined patients using a reproducible grading method has been published. The aim of the current study was to assess whether fibrosis could regress under venesection therapy in patients with C282Y homozygous GH.


GH, genetic hemochromatosis; HCC, hepatocellular carcinoma.

Patients and Methods

Inclusion Criteria.

Patients of our database who fulfilled the following criteria were included: (1) homozygosity for the C282Y mutation; (2) availability of pretherapeutic liver biopsy of at least 1 cm in length and six portal tracts; (3) either severe fibrosis or cirrhosis, defined as F3 or F4 fibrosis stage, respectively, according to the METAVIR grading system on their initial liver biopsy; (4) minimal lag period of 2 years between the end of phlebotomy therapy and the second liver biopsy; and (5) age between 18 and 75 years at the time of the study.

Eighty-nine patients fulfilled these criteria and were selected to enter the study. Of these, 19 had already undergone a second liver biopsy for management of liver cancer (n = 5) or follow-up of fibrosis (n = 14). From these 19 patients, 1 was excluded due to the small size of the second liver biopsy so that 18 were finally included into the study. From the remaining 70 subjects, 6 had died and 64 were contacted by mail. Eighteen accepted undergoing a second liver biopsy and gave their written informed consent for that.

A total of 36 patients were included. All patients were initially treated by weekly phlebotomies to achieve low body iron stores (serum ferritin < 50 ng/mL) and then placed on maintenance therapy on the basis of 3 to 12 phlebotomies per year.

Data Recorded.

The following clinical and biochemical data were recorded at the time of the first and the second liver biopsies: age, sex, smoking habits, alcohol consumption with excessive drinking defined as consumption of more than 21 doses of alcohol per week for men and 14 for women, features of the metabolic syndrome defined according to ATPIII criteria (body mass index expressed as kg/m2, blood pressure, and serum cholesterol, triglycerides, and glucose), liver function tests (serum albumin, gamma-globulins, aspartate amino transferase, alanine aminotransferase, alkaline phosphatase, gamma-glutamyltransferase, and prothrombin time), markers of iron stores (serum iron, ferritin, and transferrin saturation), platelets and red blood cell counts, hepatitis B core antibody, and hepatitis C virus antibody. Data on liver sonography (signs of portal hypertension, liver surface, size of the spleen) were also included when available. The number of grams of iron necessary to obtain initial iron depletion was calculated according to the ratio “blood volume removed/2,” and the patients were classified into two groups according to whether they had both serum ferritin levels lower than 50 and transferrin saturation lower than 50% during maintenance therapy.

Histological Data.

All initial liver samples were obtained by needle biopsy. Second liver biopsies were performed either percutaneously (n = 31) or during surgery (n = 5). Specimens were routinely processed after fixation in 10% formaldehyde and embedding in paraffin. Three-micron-thick slides were stained using hematoxylin-eosin-saffron, Sirius red, or reticulin for evaluation of connective tissue and Perls' for iron assessment. Histological examinations were performed by two pathologists (B.T. and Y.D.) who have been trained in the same institute and are referral pathologists in France for liver diseases. Both were blind to clinical and biological data and to biopsy rank. All liver biopsy specimens were randomly sorted for separate reading by each pathologist. The following items were assessed: (a) portal fibrosis according to the METAVIR grading system24; (b) iron amount according to the semiquantitative grading system of Deugnier et al.25; and (c) steatosis according to a four-grade scale. In case of vascular disturbances as defined by Desmet and Roskams,26 the diagnosis of cirrhosis was made even in the absence of bridging fibrosis on the basis of parenchymal nodularity, loss of zonal gradient, thickening of hepatocellular plates, and disturbance of vascular architecture including abnormal location and number of vessels, especially centrolobular veins, and vascular thromboses. Then, results of both readings were compared and, in case of disagreement, a joint assessment was made without any additional bio-clinical knowledge. Liver iron concentration was measured using the Barry and Sherlock's method.27 Regression of fibrosis was defined as a decrease of at least 2 METAVIR units.

Statistical Assessment.

Results are expressed as mean ± SD. Mann-Whitney and chi-squared (Fisher) tests were used for the estimation of statistical significance. Calculations and figures were performed with Statview 5 software. A predictive index for regression of fibrosis was defined. Its accuracy was assessed by receiving operating curve analysis performed using MedCalc software. A result was considered significant at P < .05.


The main characteristics of the 36 patients at the time of the first and second biopsies are given in Table 1. Most subjects were middle-aged men with severe iron overload as indicated by the mean liver iron concentration of 388 ± 165 μmol/g (N < 36), serum ferritin levels greater than 1,000 ng/mL in 93.3%, and a mean quantity of removed iron of 17 ± 7 g. The duration between the two biopsies ranged from 3 to 28 years with a mean of 9.7 ± 5.2. All but four patients had been regularly venesected for maintenance therapy, and long-term low body iron stores had been achieved in 15 (42.8 %). The percentages of excessive drinkers and of patients with at least one feature of the metabolic syndrome did not differ between the first and the second liver biopsies, respectively: 57.6% versus 55.0%, and 69.0% versus 67.6%.

Table 1. Main Characteristics of the 36 Patients, 30 Men and 6 Women, at the Time of the First and Second Biopsies (Mean ± Standard Deviation)
 First Liver BiopsySecond Liver Biopsy
  1. BMI, body mass index; AST, aspartate aminotransferase; ALT, alanine aminotransferase.

Age (y)47.1 ± 8.656.7 ± 8.6
BMI (kg/m2)24.5 ± 3.225.0 ± 3.1
METAVIR fibrosis3.6 ± 0.52.4 ± 1.5
LIC (μmol/g)388.4 ± 164.836.1 ± 85.5
Serum iron (μmol/L)41.6 ± 29.421.1 ± 10.9
Transferrin saturation (%)83.3 ± 15.842.6 ± 26.4
Serum ferritin (μg/L)3448 ± 1691179 ± 637
AST (UI/L)66.3 ± 50.527.6 ± 13.6
ALT (UI/L)88.1 ± 56.726.9 ± 12.5
Prothrombin activity (%)83.8 ± 12.987.6 ± 8.9
Platelet count (/mm3)190,000 ± 69,600216,000 ± 102,000
Serum albumin (g/L)43.5 ± 6.644.2 ± 4.3
G-Globulinemia (g/L)10.3 ± 5.29.5 ± 2.8
Fasting serum glucose (mmol/L)6.1 ± 1.66.0 ± 2.0
Uricemia (μmol/L)327.8 ± 63.4300 ± 64.8
Iron removed (g)17.0 ± 7.1

With respect to fibrosis grading, a total agreement between both pathologists was found in 30 of 36 cases (83%) in the series of the first biopsies and in 15 of 36 cases (42%) in the series of the second liver biopsies (difference of 1 METAVIR unit in 19 of 21 and of more than 1 METAVIR unit in 2 of 21). From the six discordant results in the first series, four were finally graded as F4 and 2 as F3 and from the 21 discordant results in the second series, 10 were downgraded to the lowest score and 11 were upgraded to the highest score. Finally, 13 patients (36.1 %) had F3 fibrosis on their initial liver biopsy, and 23 (63.9 %) had cirrhosis. The outcome of fibrosis is presented in Table 2. Among the 13 patients who had F3 fibrosis on their initial liver biopsy, fibrosis decreased in 11 (84.6 %) to stage 0 (n = 3), stage 1 (n = 6), and stage 2 (n = 2). Among the 23 patients with cirrhosis on their initial biopsy, fibrosis decreased in 10 (43.5 %) to stage 0 (n = 1), stage 1 (n = 4), stage 2 (n = 3), and stage 3 (n = 2). As a whole, 17 (47 %) patients presented with regression of fibrosis of at least 2 METAVIR units. Among the five patients with liver cancer, three had hepatocellular carcinoma (HCC), and two had cholangiocarcinoma. All but one with cholangiocarcinoma presented with no regression of fibrosis.

Table 2. Outcome of Fibrosis After Phlebotomy Therapy in Patients With Initial Fibrosis Grade 3 or Cirrhosis
Fibrosis AssessmentSecond Liver BiopsyTotal
F 0F 1F 2F 3F 4
First liver biopsy     
 F 33620213
 F 414321323

The results of the ultrasound (US) examination performed at the time of the second liver biopsy were recorded for patients with regression of fibrosis. Data were available for 16 of 17: liver and spleen size, texture, and shape and portal vein diameter were normal in nine cases, liver size was increased in six cases, of which two presented with a heterogeneous pattern and one had liver cancer (cholangiocarcinoma).

No clinical data were associated with regression of fibrosis (Table 3), except for compliance with maintenance therapy, which was more often observed in compliant (5/15; 33%) than in noncompliant (12/20; 60%) patients without reaching statistical significance. The only initial biochemical data associated with regression of fibrosis of at least two METAVIR units were serum gamma-globulins, prothrombin activity, and platelet count as shown in Table 3. When the ratio of gamma-globulins (g/L) to (platelets [n/mm3] × prothrombin activity [%]) was greater than 7.5, fibrosis never regressed (Fig. 1). The best compromise between sensitivity (100%-95% CI: 73-100) and specificity (90%-95% CI: 56-98) was found for an index threshold of 5.5 at receiver operating characteristic analysis (Fig. 2).

Table 3. Clinical and Biochemical Comparative Data, at the First Biopsy, According to Regression of Fibrosis or Not (Mean ± Standard Deviation)
 Regression of FibrosisNo Regression of FibrosisP
Age (y)46.0 ± 7.748.0 ± 9.3.32
BMI (kg/m2)24.4 ± 4.224.7 ± 1.8.51
ALAT (UI/L)81.1 ± 52.395.1 ± 61.6.42
ASAT (UI/:)61.2 ± 58.071.4 ± 42.9.23
Serum cholesterol (mmol/L)5.4 ± 1.35.6 ± 1.0.65
Serum triglycerides (mmol/L)1.3 ± 0.51.0 ± 0.0.40
Fasting serum glucose (mmol/L)5.7 ± 0.96.7 ± 2.2.33
G-Globulinemia (g/L)7.7 ± 2.113.1 ± 6.1.01
Serum ferritin (μg/L)3231 ± 13513695 ± 2038.39
Transferrin saturation (%)86.4 ± 6.478.7 ± 23.5.28
Prothrombin activity (%)91.3 ± 11.675.7 ± 8.7.002
Platelet count (/mm3)236,200 ± 57,300136,500 ± 36,100<.0001
Iron removed (g)17.1 ± 6.516.8 ± 8.0.92
LIC (μmol/g)440.9 ± 146.8339.1 ± 169.9.11
Size of liver biopsy (cm)2.5 ± 1.11.8 ± 1.1.8
Total iron score (0-60)35.2 ± 12.632.3 ± 14.6.59
Figure 1.

Regression of fibrosis according to the predictive index defined by the ratio gammaglobulins [g/L] to (platelets [n/mm3] × prothrombin activity [%]) at the first biopsy.

Figure 2.

Receiving operating curve of the regression predictive index in 22 patients. Area under the curve was 0.96 (95% confidence interval: 0.78-0.99; P < .0001).


The current study demonstrates that hepatic fibrosis may regress along venesection therapy in approximately half the patients with GH initially complicated with either severe fibrosis or cirrhosis. Niederau et al,9 in a follow-up study of a large series of hemochromatotics, have already reported regression of fibrosis in treated subjects; however, the authors did not give any methodological and pathological details, and the study was performed at a time when the diagnosis of hemochromatosis did not rely on genetic testing. The same remarks can be made about the small number of single case reports on regression of fibrosis in hemochromatosis.7, 8 Our study is the based on strict diagnostic criteria of both hemochromatosis and fibrosis; only C282Y homozygotes were included, and fibrosis assessment was carefully performed using the well-validated METAVIR scoring system. Because of these strict diagnostic and methodological criteria, a large number of cases (n = 15) could not be included, which reduced the sample size of patients but increased the reliability of our results.

However, several types of bias in the current study need to be discussed:

First, a recruitment bias may have been present because 46 patients either did not respond to our mail or declined liver biopsy in their response letter. This high number of non-included patients likely constitutes a bias with respect to the calculation of the exact prevalence of regression of fibrosis in hemochromatotic patients but cannot affect the main message of the paper that hepatic fibrosis regresses in an unexpected number of patients. Indeed, even when admitting that all the 46 patients who did not enter the study would have remained with cirrhosis—which is very unlikely—regression of fibrosis would have been found in 17 of 82 cases (20%). Then, regression of fibrosis cannot be considered as anecdotal in treated hemochromatosis patients.

Second, a possible sample bias may have occurred because of biopsy size. However, we have to stress that biopsy size was 2.15 ± 1.14 cm (range, 1-4 cm) in the initial series and 2.72 ± 1.37 cm (range, 1-5.9 cm) in the second series. More precisely, among the 17 second liver biopsies presenting with regression of fibrosis in our study, 3 and 12 were longer than 15 and 25 mm, respectively, and 2 were 10 to 15 mm long. In hepatitis C patients, that sampling error was 35% and 25% in 15-mm and 25-mm liver biopsy specimens with respect to fibrosis grading,28 and more recently, Schiano et al.29 have shown that liver biopsy specimens with a length of at least 10 mm usually reflect the stage of fibrosis in patients with chronic hepatitis C. Then, we can assume the full representativity of our samples. Moreover, to limit sampling error, we defined regression of fibrosis by the decrease of at least 2 METAVIR units.

Third, a reading bias may have occurred. The strength of agreement between both pathologists was almost perfect for the initial biopsies (83%). It was moderate (42%) for the second biopsies, but in 19 of 21 cases discordance was 1 METAVIR unit only. Moreover, the lowest score was finally kept in 10 of 21 and the highest in 11 of 21. Therefore, that joint assessment would have resulted in favoring the hypothesis of regression of fibrosis is unlikely.

Fibrosis course was accurately predicted using three simple variables, allowing for the construction of a “regression predictive index” calculated as the ratio of gamma-globulin (g/L) to [platelet count (n/mm3) × prothrombin time (%)]. This index appeared to be, in fact, an index of liver disease severity, which suggests that, above a certain threshold of severity, fibrosis is unlikely to regress. This is in accordance with data published in patients with hepatitis C,21, 22 in whom regression of fibrosis was mainly observed in “young” and, likely, less severe cirrhosis. The use of noninvasive tests—such as elastometry (Fibroscan™) or serum hyaluronic acid assay—allowing for a continuous assessment of fibrosis should help to identify, more appropriately than histological examination, the type of fibrosis susceptible to regress.

The mean follow-up duration between the two liver biopsies was 9.7 ± 5.3 years. During this long period, various factors may have acted as modulators of fibrosis course. Compliance to treatment was not found to interact significantly with fibrosis course; however, fibrosis regressed twice more in compliant than in noncompliant subjects. In addition, even when classified as noncompliant, most patients of our series had been treated and then maintained at a lower iron burden than initially. Surprisingly, initial liver iron concentration was higher in patients whose fibrosis regressed along venesection therapy than in those whose fibrosis did not regress, even if the difference did not reach statistical significance. In GH, iron deposition is less abundant within fibrous septa than in parenchyma.25 This could explain a lower initial liver iron concentration in the most severe cases of cirrhosis that did not regress. A role of hepatitis C and B virus can be ruled out in this series, as all patients were hepatitis C virus and hepatitis B core antibody negative at the time of both biopsies. Similarly, body mass index was not found to play a role in fibrosis course, contrary to what has been reported in chronic hepatitis30–32 and alcoholic patients.33 With respect to the role of alcohol on fibrosis course, the current study did not bring a clear answer, because precise assessment of alcohol consumption during maintenance therapy was lacking in the 16 oldest patients files. However, among the 15 patients who either maintained (n = 13) or developed (n = 2) cirrhosis, 5 were regular excessive drinkers, 9 presented with at least one feature of metabolic syndrome (of whom 4 were also excessive drinkers), and 1 was infected with human immunodeficiency virus. This could suggest, despite the absence of statistical significance in the current series, that alcohol consumption and development of metabolic abnormalities may impair regression of fibrosis in treated GH patients, as recently pointed out by Powell et al.34

The current data as well as those from Powell et al.35 show that fibrosis in the absence of cirrhosis may regress along venesection therapy in a large number of hemochromatotic patients. However, the question of the complete reversal of cirrhosis, which is clinically and basically relevant, remains debated. As suggested by Desmet,26, 36 cirrhosis-associated vascular disturbances are unlikely to regress in parallel to fibrosis, and there is a risk to spoil the diagnosis of remaining macronodular cirrhosis, which is always difficult to exclude on needle biopsy specimens, especially in its incomplete septal variant.37 However, none of the eight patients whose cirrhosis regressed had abnormal vascular architecture on their second liver biopsy, of which the mean size was 32 ± 16 mm. This means that, in samples considered as representative, we failed to find any lesion suggestive of macronodular cirrhosis according to the strict criteria we used. In addition, ultrasound examination was either strictly normal or nearly normal in six of eight.

The question remains whether the risk of liver cancer is still present after regression of fibrosis. It is an important issue with respect to ultrasound screening policy in hemochromatosis. In the current series, no patient with HCC at the time of the second biopsy had regression of fibrosis; all patients still had cirrhosis. The only case of liver cancer arising in a liver with regressing fibrosis (F4 to F2) was a cholangiocarcinoma, a tumor for which systematic screening has not proved to be effective. However, 10 cases of HCC have been reported in treated patients with no cirrhosis.38–42 Moreover, Blumberg et al.43 described a well-documented case of HCC complicating hemochromatosis after “cirrhosis reversibility.” Based on these data, giving up systematic ultrasound screening in the subset of patients with initial cirrhosis that further regressed would be unethical, even if regression of fibrosis is likely associated with a decreasing risk of liver cancer.

In conclusion, the current study extends the concept of regression of severe fibrosis to patients with treated GH. In fact, among our 36 patients with GH, 47% presented regression of fibrosis of at least 2 METAVIR units, after phlebotomy therapy. These data also suggest that some simple biochemical tests could be predictive of further improvement in fibrosis stage along venesection therapy. If confirmed on larger series, this could modify the ultrasound screening policy of HCC in GH.