Qualitative and quantitative magnetic resonance imaging in haemoglobin H disease: Screening for iron overload

Clinical Radiology (1999) 54, 98-102

Qualitative and Quantitative Magnetic Resonance Imaging in Haemoglobin H Disease: Screening for Iron Overload G. C. OOI”,

F. E. CHENf,

K. N. CHANT, K. W. T. TSANGI, VIVIAN CHANT, H. NGAN”

Y. H. WONGT,

R. LIANGI,

Departments of *Diagnostic Radiology and I-Medicine, The University of Hong Kong, Queen Mary Hospital, Hong Kong Received: 18 May 1998

Revised: 21 August 1998

Accepted: 22 September 1998

OBJECTIVES: To evaluate the clinical utility of magnetic resonance imaging (MRI) in screening for iron overload in non-transfusion dependent Haemoglobin (Hb) H disease. PATIENTS AND METHODS: Thirty-six non-transfusion dependent HbH patients were evaluated with axial spin echo Tl and gradient echo T2 MRI of the abdomen and heart. The ratios of signal intensities (SIR) of the liver, spleen, pancreas and heart to paraspinous muscles were calculated. SIR ~1 was taken as indicative of iron overload. Qualitative grading (O-4 scale) of iron overload was also performed. The relationship between Tl and T2 SIR and serum ferritin, and that between qualitative grading and serum ferritin were examined using standard statistical methods. Comparisons were also made between qualitative grading and quantitative Tl and T2 SIR data in diagnosing iron overload. Six patients underwent liver biopsies. RESULTS: T2 SIR was more sensitive in detecting iron overload than Tl SIR. Thirty-three livers, 13 spleens, six pancreas and one heart were diagnosed as having iron overload with T2 SIR, including three patients with normal serum ferritin. A positive diagnosis by T2 SIR was more closely related to that of qualitative grading than Tl SIR. Serum ferritin was negatively correlated with hepatic SIR (Tl and T2), and with T2 SIR of the spleen and pancreas, even after adjustment for age. Liver haemosiderosis was confirmed in all six patients who underwent liver biopsies. Liver iron concentration of only one and a half times the normal was found in one patient with positive MR findings. CONCLUSION: MR is a non-invasive, effective method for early detection of iron overload particularly in the liver and spleen. Qualitative grading and quantitative T2 SIR data are equivalent in diagnosing iron overload. Routine screening of non-transfusion dependent HbH patients will identify high risk patients in whom early therapeutic intervention may prevent further complications and morbidity. Ooi, G. C. et al. (1999) Clinical Radiology 54, 98-102. Key words: haemoglobin H, iron, liver, magnetic resonance,screening.

Haemoglobin H (HbH) diseaseis a severeform of ol-thalassaemia commonly found in Southeast Asia. Unlike fl-thalassaemia major (Cooley’s anaemia) where early death is caused by complications of iron overload, HbH disease is generally regarded as a benign condition. Elevated serum ferritin levels have been reported in middle-aged Chinese HbH patients [I]. The clinical significance of this is unclear as serum ferritin is an inaccurate and indirect measure of body iron stores that is affected by infection, chronic inflammation and liver disease. The most accurate method of quantifying iron overload is by measuring liver iron concentration which requires a liver biopsy. This is an invasive procedure with inherent risks, making serial assessmentsdifficult to justify particularly in a largely asymptomatic HbH population. Magnetic resonance imaging (MRI) has been used for the Correspondence to: Dr Gaik Cheng Ooi, Department Radiology, The University of Hong Kong. Queen Mary 405, Block K, Pokfulam Road, Hong Kong. 0009-9260/99/020098+05

$12.00/O

of Diagnostic Hospital, Room

diagnosis [2-111 and follow-up of patients with hepatic iron overload [ 12,131. Earlier studies using spin echo (SE) sequences suffered from a lack of sensitivity for mild iron overload [2,7,14,15]. This deficiency has largely been rectified with high-field-strength systems which reduce background noise [11,16,17], and gradient echo (GRE) sequences which are more sensitive to the paramagnetic effects of iron [9,16,17]. Most studies on MRI of iron overload have been performed in primary haemochromatosis and chronic liver diseases [24,7,9-18]. Similar studies on thalassaemia have been limited, and largely confined to patients with fl-thalassaemia major [6,7,12,19]. This study was conducted to evaluate qualitative and quantitative MR surveillance of iron overload in non-transfusion dependent HbH disease. We used previously validated MR techniques to quantify iron overload by measuring signal intensity ratios of liver, spleen, pancreas and heart tissue using paraspinous muscle as an internal standard [4-6, 8,9,12,15,19]. 0 1999 The Royal

College

of Radiologists

MRI SCREENING

PATIENTS

IN HbH DISEASE

AND METHODS

Thirty-six non-transfusion dependent patients [21 female, 15 male, mean age t SD (43.4 ? 14.5 years)] with HbH disease in our institution were studied. HbH disease was diagnosed on the basis of the presence of H band on Hb electrophoresis and DNA analysis of the a-globin genes by hybridization of enzymedigested genomic DNA to { gene probe [20]. Four patients had undergone a previous splenectomy. Serum ferritin was measured by chemical luminescence procedure (Chiron, ACS 180) within a week of the MR examination. The normal range of ferritin is 11%884pmol/l in men and 15-331pmol/l in women. Liver biopsies were performed in six patients with a 14-gauge tru-cut needle (11 cm; Banern Medical International, S.A. Santa Domingo, Dominican Republic) and liver iron concentrations (dry weight) determined in two. Histological grade of haemosiderosis was determined according to haemosiderotic granule size: Grade 0 = absent or barely discernable at 400 magnification; Grade I = barely discernible at 250 magnification or easily confirmed at 400 magnification; Grade II = easily confirmed at 100 magnification; Grade III = easily confirmed at 25 magnification; Grade IV = massesof granules seen at 10 x magnification or under naked eye. MRI was performed with a 1.5-T system (Signa; GE Medical Systems, Milwaukee, USA). Respiratory gated multislice axial abdominal scans were obtained with SE Tl-weighted (TRITE, 300-700/9 ms) and GRE T2-weighted (TR/TE, 500-80005 ms, excitation flip angle 20”) sequences. The number of signal averages was four and the acquisition matrix 256x 128. Cardiac gated axial images of the heart were also obtained with (a) SE Tl-weighted (TOTE, one r-r interval/8-9 ms) and (b) GRE T2-weighted (TIUTE, one r-r interval/l5 ms, excitation flip angle 20”) sequences, both with two signal averages, matrix 256x 128. Signal intensities (SI) were measured by placing operator-defined regions of interest (ROI, average area 9.5 mm2) in the liver, spleen, pancreas, myocardium and paraspinous muscles, taking care to avoid fat, and biliary and vascular structures. The average of three ROIs were obtained. The ratio of SI values (SIR) of liver, spleen, pancreas, and myocardium to paraspinous muscles were calculated. Iron overload was defined as SIR
@I Fig. 1 - (a) SE Tl-weighted and (b) GRE TZweighted MR images of the same patient showing marked iron overload in the liver (Qualitative grade 4). The spleen (Qualitative grade 3) and pancreas (Qualitative grade 3) are also involved.

correlation analysis. Multivariate analysis was used to examine the relationship between Tl and T2 SIR and serum ferritin, allowing for the effects of other possible confounding factors. Finally, the relationship between serum ferritin and qualitative grading of the MR images was evaluated using analysis of variance. As these data were log-normally distributed, they were therefore log-transformed before statistical analysis. The diagnosis of iron overload by qualitative evaluation was compared with both the Tl and T2 SIR diagnosis (SIR ~1) using Cohen’s kappa coefficients. These indices took account of chance agreement between observations on the same MR images [21].

Grades 2-4 were regarded as indicative of mild, moderate and severe iron overload respectively (Fig. 1). RESULTS

Data Analysis Tl and T2 SIR of the liver, spleen, pancreas and heart images were calculated and correlated with serum ferritin, haemoglobin concentration and age using conventional Pearson’s

Thirty-three livers, 13 spleens, six pancreas and one heart were identified as having iron overload by T2 SIR compared with 11 livers, four spleens, two pancreas and one heart by Tl SIR. For all organs except for the heart, T2 SIR was

100 Table

CLINICAL 1 - Comparison

of Tl and T2 signal

intensity

Tl SIR Geometric Mean (95%

Liver (n = 36) Spleen (n = 32)” Pancreas (n = 36) Heart (n = 36)

ratio

RADIOLOGY

(SIR)

T2 SIR Geometric Mean (95%

CI)

Tl SIR compared with T2 SIR Odds ratio (95% CI)

CI)

1.14 (0.36,

3.63)

0.16 (0.02,

1.25)

1.42 (0.81,

2.48)

0.71 (0.08,

6.55)

7.39 (5.70, 1.99 (1.45,

1.14 (0.22, 5.87) 1.77 (1.07, 2.92)

1.52 (1.22,

1.90)

0.96 (0.89,

1.04)

1.73 (1.03, 2.92) 1.84 (0.84, 4.01)

P value

9.58) 2.75)

0.0001

0.0001 0.0004 0.29

95% CI: 95% confidence intervals. *Four patients had had splenectomy

Table

2 - Comparison

of diagnosis

of iron

Diagnosis qualitative n (%I Liver (n = 36) Spleen (n = 32) Pancreas (n = 36) Heart (n = 36) t Qualitative f Qualitative

grading grading

compared compared

overload

by grading

by Tl

and T2 signal

Diagnosis Tl SIR n (%I

33 (91.7)

11 (30.5)

14 (43.7) 4 (11.1) 1 (2.8)

4 (12.5) 2 (5.5) 2 (5.6)

intensity

ratios

(SIR)

and qualitative

evaluation

by

Diagnosis T2 SIR n (%)

kappa coefficient? 0.08 0.31 0.64 0.65

by kappa coefficient*

33 (91.7) 13 (40.6) 6 (16.7)

1.00 0.94 0.77

1 (2.8)

1 .oo

with Tl SIR. with T2 SIR.

significantly lower than that of Tl SIR (Table I). Qualitative MR grading identified 33, 14, six and one case(s) of iron overload (Grade 2 or above) in the liver, spleen, pancreas and heart, respectively. T2 SIR diagnosis of iron overload was more closely related to that of qualitative grading than Tl SIR (Table 2), irrespective of the organs involved. There was a close relationship between MR qualitative grading of iron overload and serum ferritin level (Fig. 2), with increasing levels of serum ferritin corresponding to increasing grades of iron overload. There was also a significant inverse relationship between T2 SIR and qualitative grading (data not shown). Serum ferritin [mean 2305 2 2115 pmol/l (range 220-8900 pmol/l)] was negatively correlated with Tl and T2 SIR of the liver and spleen (Fig. 3), T2 SIR of the pancreas (Y= -0.53, P= O.OOOS), and Tl SIR of the heart (Y= -0.37, P = 0.03). Age was also negatively correlated with both Tl and T2 SIR of the liver (v= -0.44, P= 0.007; Y= -0.38, P=O.O2) and spleen (Y= -0.48, P = 0.006; r = -0.63, P = O.OOOl),and with Tl SIR of the heart (Y= -0.35, P= 0.04). There was also a positive correlation between serum ferritin and age (u= 0.59, P= 0.0001). Using multiple regression analysis, after adjustment for effects of age, serum ferritin remained negatively correlated with Tl and T2 SIR of liver, and with T2 SIR of spleen and pancreas (Table 3). Haemoglobin concentration [mean 8.9 g/d1 (range 4.4-12.8 g/dl)] was not a significant factor after allowing for the effect of serum ferritin and age in the analysis. Of the six patients with pancreatic iron overload, two suffered from diabetes mellitus, the rest were normoglycaemic. All six patients also had hepatic and splenic iron overload, with myocardial overload being an additional feature in one patient.

Three out of six patients with normal serum ferritin levels had MR evidence of liver iron overload on both qualitative (two with Grade 2, and one with Grade 3 iron overload) and T2 quantitative data. Liver haemosiderosis was confirmed in another six patients who underwent liver biopsy. Histological grade II, III and IV haemosiderosis were each present in two patients. Cirrhosis and fibrosis were concomitant features in two patients, while chronic hepatitis B infection was present in one. Liver iron concentration per dry liver weight calculated in two

0

4096

.

k

I

256 I

o-1

I 2

I

I

3

4

Qualitative grading for iron status in liver Fig. 2 - Relationship between MR qualitative evaluation of iron status of liver and sernm ferritin. Data shown are the geometric mean and 95% confidence intervals of serum ferritin for the various grades of iron overload.

MRI SCREENING

2.0000

-

1.0000

- (b)o

0

p: 0.5000 ;i E: 0.2500

-

2 0.1250

-

3

0.0625

-

0.0313

-

0

0.25

2 128

HbH DISEASE

.

o\S”

c

IN

\

r= -0.76 p < 0.0001

00

101

0

0

-

r = -0.73

p < 0.0001

00

0.0156~ 256

512

1024

2048

4096

8192

16384

128

256

512

Ferritin (pmoIil)

128

256

512

1024

2048

4096

8192

16384

128

256

512

Ferritin (pmol/l) (b‘, liver T2 SIR and serum

patients were 2.3 mg/g (41 pmollg) and 5.76 mg/g (103 PmoVg), respectively, the upper limit of normal being 1.5 mg/g (27 kmol/g).

This first systematic MR evaluation on patients with nontransfusion dependent HhH diseasehas shown a high prevalence of iron overload in the liver (89%), spleen (41%) and pancreas (17%). Negligible myocardial iron overload (3%) was found. In three cases (8%) with normal serum ferritin levels, iron overload was evident on both qualitative and quantitative MR evaluations. We found T2 quantitative data to be more sensitive in detecting iron overload compared with Tl quantitative data. Both qualitative evaluation by our visual scoring system and T2 SIR data were equivalent in diagnosing iron overload. T2 SIR of the liver, spleen and pancreas were negatively correlated

Organ

variable

between

Tl

and T2 signal

intensity

Regression

model

ln(T1 ln(T2 ln(T2 ln(T2

= = = =

Liver Tl SIR Liver T2 SIR Spleen T2 SIR Pancreas T2 SIR Fenitin

is in pmol/l

and age in years;

SIR) SIR) SIR) SIR)

In indicates

4096

8192

16 384

1024

2048

4096

8192

16 384

ratio

(SIR)

3.59-0,5X*ln(ferritin) 4.04-0,85*ln(ferritin) 7.53-0,74*ln(ferritin) 2.28-0.58*ln(ferritin)

natural

logarithm.

ferritin

(c) splenic

Tl and serum ferritin,

and (d) splenic

T2 and

with both serum ferritin and age indicating that T2 SIR could be influenced by both factors. Using multivariate analysis, the relationship between T2 SIR and serum ferritin remained highly significant even after allowing for the effect of age (Table 3). Correlations between liver iron concentrations and SIR of liver using paraspinous muscle or subcutaneous fat are well documented [4,6,8-lo]. Studies in which a poor correlation existed were due to poor-signal-to-noise ratio with MR systems operating at much lower field strengths (0.35-0.5 Tesla) and using SE sequences with longer TE times [7,8,14,15]. Our MR protocol was tailored to maximize signal-to-noise ratio and to avoid the limitations of previous studies by using short TE SE and GRE T2 sequences. However, inaccurate quantification of iron overload has been reported with GRE! T2 sequences particularly in patients with severe disease [2,7,19]. This is due to exponential signal reduction as liver iron concentration increases [2,7,19]. A maximum threshold limit of 20mg/g of

DISCUSSION

3 - Relationship

2048

Fkriitin (pmolil)

Fig. 3 - Correlations between (a) liver Tl SIR and serum ferritin, serum ferritin are depicted in these graphs.

Table

1024

Ferritin (pmol/l)

and serum

+ + +

ferritin

0.20*ln(age) O.O9*ln(age) 0.66*ln(age) 0.58*ln(age)

after

adjusting

for the effects

of age

I?

P

0.66 0.52 0.62 0.34

0.0001 0.0001 0.0001 0.002

102

CLINICAL

dry liver weight (357.1 pmol/g) has been suggested for accurate MR quantification of transfusional iron overload in /3-thalassaemia major patients 1191. Iron overload above this level together with liver fibrosis may cause imprecise MR iron quantification [ 191. These limitations are, however, not relevant to this study in two respects. Firstly, the range of iron overload in nontransfusional HbH patients would lie well below that of transfusion dependant /3-thalassaemia major. Secondly, this study aims to evaluate MRI as a screening tool for early iron overload detection and not for precise transfusional iron overload monitoring when the degree of iron overload is severe. One would also argue that direct measurement of iron content in the liver would have been a better guide to iron overload than serum ferritin. However, as most of our patients were largely asymptomatic, ethical constraints and patient reluctance for invasive procedures made recruitment for liver biopsies difficult. GRE T2 sensitivity for mild to moderate iron overload in this study is illustrated by the three cases who had normal serum ferritin levels but whose MR scans were positive for iron overload. In another case of moderate hepatic iron overload (Grade 3), the serum ferritin was only mildly elevated, and the liver iron concentration only one and half times the normal (41 pmol/g dry weight). This sensitivity to mild iron overload compares favourably with other studies in which the lower threshold of detection lies between three to five times the normal upper limit of liver iron concentration [7,9,14,15,18]. MR surveillance of iron overload in our study has revealed a combination of parenchymal (primarily liver and pancreatic) and reticuloendothelial (primarily splenic) iron distribution in non-transfusion dependent HbH disease [18]. This study has also, for the first time in this group of patients, documented pancreatic iron overload that is associated with hepatic and splenic iron overload. This may signify severe disease, and an increased risk of developing diabetes mellitus and other complications in a condition hitherto thought to be ‘benign’. In the light of our findings, qualitative and quantitative MR screening should be performed in non-transfusion dependent HbH patients in whom we have demonstrated a significant prevalence of hepatic and splenic iron overload. Although the prevalence of pancreatic iron overload is low, its presence may indicate an increased risk of complications. Early identification of high risk patients may help to delay further complications and reduce morbidity by prompt therapeutic intervention. Acknowledgements. We would like to thank the radiographers of the MRI unit at Queen Mary Hospital for their support and help, and the patients without whose cooperation, this study could not have been possible.

RADIOLOGY

REFERENCES 1 Tso SC, Loh TT, Todd D. Iron overload in patients with haemoglobin H disease. Stand .I Huematol 1984;32:391-394. 2 Kaltwasser JP, Gottschalk R, Schalk KP et al. Non-invasive quantitation of liver iron-overload by magnetic resonance imaging. Br J Huematol 1990;74:360-363. 3 Gomori JM, Horev G, Tarnary H et al. Hepatic iron overload: quantitative MR imaging. Radiology 1991;179:367-369. 4 Jensen PD, Jensen FT, Christensen T et al. Non-invasive assessment of tissue iron overload in the liver by magnetic resonance imaging. Br J Haematol 1994;87:171-184. 5 Villari N, Caramella D, Lippi A et al. Assessment of liver iron overload in thalassemic patients by MR imaging. Acta Radiologica 1992;33: 347-350. 6 Bonetti MG, Castriota-Scanderberg A, Criconia GM et al. Hepatic iron overload in thalassemic patients: proposal and validation of an MRI method of assessment. Pediat Radio1 1996;26:650-656. 7 Guyader D, Gandon Y, Robert JY et al. Magnetic resonance imaging and assessment of liver iron content in genetic hemochromatosis. J Hepatol 1992;15:304-308. 8 Hernandez RJ, Samaik SA, Lande I et al. MR evaluation of liver iron overload. J Col?zp Assist Tomogr 1988;12:91-94. 9 Gandon Y, Guyader D, Heautot JF et al. Hemochromatosis: diagnosis and quantification of liver iron with gradient-echo MR imaging. Radiology 1994;193:533-538. 10 Lawrence SP, Caminer SJ, Yavorski RT et al. Correlation of liver density by magnetic resonance imaging and hepatic iron levels. A noninvasive means to exclude homogeneous hemochromatosis. J Clin Gastroenterol 1996;23:113-117. 11 Rocchi E, Cassanelli M, Borghi A et al. Magnetic resonance imaging and different levels of iron overload in chronic liver disease. Hepatology 1993;17:997-1002. 12 Jensen PD, Jensen FT, Christensen T et al. Evaluation of transfusional iron overload before and during iron chelation by magnetic resonance imaging of the liver and determination of serum ferritin in adult non-thalassaemic patients. Br / Haematol 1995;89:880-889. 13 Andersen PB, Birgeg&d G, Nyman R et al. Magnetic resonance imaging in idiopathic hemochromatosis. Eur J Haematol 1991;47: 174-178. 14 Chezmar JL, Nelson RC, Malko JA et al. Hepatic iron overload: diagnosis and quantification by noninvasive imaging. Gastrointest Radial 1990;15:27-31. 15 Bonkovsky HL, Slaker DP, Bills EB et al. Usefulness and limitations of laboratory and hepatic imaging studies in iron-storage disease. Gastroenterology 1990;99:1079-1091. 16 Stark DD. Hepatic iron overload: paramagnetic pathology. Radiology 1991;179:333-335. 17 Siegelman ES, Mitchell DG, Outwater E et al. Idiopathic hemochromatosis: MR imaging findings in cirrhotic and precirrhotic patients. Radiology 1993;188:637-641. 18 Siegelman ES, Mitchell DG, Rubin R et al. Parenchymal versus reticuloendothelial iron overload in the liver: distinction with MR imaging. Radiology 1991;179:361-366. 19 Angelucci E, Giovagnoni A, Valeri G et al. Limitations of magnetic resonance imaging in measurement of hepatic iron. Blood 1997;90: 4736-4742. 20 Chan V, Chan TK, Cheng MY et al. Organization of the l--o1 gene in the Chinese. Br J Haematol 1986;64:97-105. 21 Armitage P, Berry G. Statistical method in medical research, 3rd ed. Oxford: Blackwell Scientific Publications, 1994;443-447.