Effect of enhanced iron chelation therapy on glucose metabolism in patients with β-thalassaemia major


Nicholas Angelopoulos, Division of Endocrinology and Metabolism, ‘Hippokrateion’ Hospital of Athens, Vassilisis Sophia's Avenue 108, 115 27 Athens, Greece. E-mail: drangelnick@hotmail.com.


Recently introduced chelation regimens that combine deferoxamine (DFO) and deferiprone have been shown to have greater efficacy in promoting iron excretion than either chelator alone and have been associated with rapid reduction of the iron load in the heart and liver, and with reversal of cardiac dysfunction. It is unclear whether this combined therapy could be associated with a reduction in iron load or decline in the severity of iron-induced endocrinopathies. Starting in January 2001, 42 patients with β-thalassaemia major, previously maintained on subcutaneous DFO only, were switched to combined treatment with DFO and deferiprone. The primary endpoint was to investigate the effects of this therapy on the glucose metabolism characteristics of this population. Combination therapy markedly decreased ferritin levels (638 ± 1345 vs. 2991 ± 2093 μg/l, P < 0·001). Glucose responses were improved at all times during an oral glucose tolerance test, particularly in patients in early stages of glucose intolerance. Glucose quantitative secretion also decreased significantly with combined therapy, while no significant change occurred in insulin levels in any group. Insulin secretion, according to the homeostasis assessment model, markedly increased in all groups, while overall reduction in insulin sensitivity did not reach statistical significance. This study showed that the combination of DFO and deferiprone was associated with an improvement in liver iron deposition and glucose intolerance.

Glucose metabolism disturbances, among other endocrinopathies, are a common feature of β-thalassaemia (Italian Working Group on Endocrine Complications in Non-endocrine Diseases, 1995; Cunningham et al, 2004; Angelopoulos et al, 2006a). The prevalence of diabetes among thalassaemia patients has been reported to range from 2·3% to 24% (Italian Working Group on Endocrine Complications in Non-endocrine Diseases, 1995; Arrigo et al, 1998; Gamberini et al, 1998; Chern et al, 2001).

Traditionally, insulin deficiency (Saudek et al, 1977) and long-standing insulin resistance (De Sanctis et al, 1988; Merkel et al, 1988; Pappas et al, 1996) that result from direct toxic damage by iron to pancreatic-beta cells are thought to be the main underlying mechanisms leading from mild glucose intolerance to overt diabetes. As these endocrine complications result from chronic iron overload, they are much more common in patients whose chelation therapy is insufficient.

On the other hand, although deferoxamine (DFO) is still widely used in chelation therapy, deferiprone has become a promising alternative for the treatment of iron overload. Another encouraging new approach to chelation therapy has been the combined administration of deferiprone and DFO (Link et al, 2001; Mourad et al, 2003). Several clinical studies have demonstrated that long-term combination therapy, in which deferiprone is given every day and DFO is given on 2–4 d/week, is effective in reducing serum ferritin and liver iron levels in patients who were inadequately chelated on DFO or deferiprone alone (Hoffbrand et al, 2003; Mourad et al, 2003). Considering that the efficacy in mobilising iron deposition seems to be tissue specific and differs from the results obtained with deferiprone or DFO alone (Saudek et al, 1977; Pappas et al, 1996), the overall effects of such a combination on beta-cell function remain unknown. Currently, there are no available data regarding the effects of deferiprone on glucose metabolism while poor compliance, iron overload and liver disease (cirrhosis or severe fibrosis) have emerged as the main risk factors associated with diabetes in patients treated with DFO (Gamberini et al, 2004). The present study aimed to determine the effects of combined therapy on glucose metabolism in β-thalassaemic patients without overt diabetes who had been inadequately chelated with DFO monotherapy. Possible correlations between the various insulin secretion and sensitivity indices during an oral glucose tolerance test (OGTT), with the treatment parameters were further assessed.

Materials and methods


Forty-two patients (22 males and 20 females) with transfusion-dependent β-thalassaemia, aged 8–42 years (25·16 ± 8·90, mean ± SD) took part in this study.

All patients were transfused with leuco-depleted, genotyped red blood cells every 2–4 weeks in order to maintain a pre-transfusion haemoglobin level above 9·5 g/dl. Prior to the initiation of this study, all patients received chelation monotherapy with subcutaneous DFO, 40 mg/kg; 8–12 h, subcutaneously, 5–6 d/week). Individual dosing and frequency of DFO infusions were determined by patients’ clinical and laboratory assessments, such as cardiac function and serum ferritin values. Between January 2001 and October 2003, all patients were gradually switched to an individually tailored combined regimen. Combination therapy [deferoxamine deferiprone combination (DFC)] consisted of both daily oral administration of deferiprone (Ferriprox, 75–90 mg/kg/d) and subcutaneous DFO (20–40 mg/kg; 8–12 h, 2–6 d/week). Deferiprone was given daily in three divided doses. Glucose metabolism characteristics were evaluated by OGTTs. The procedure was part of the long-term diagnostic follow up of endocrine complications, which is scheduled annually for all thalassaemic patients in our thalassaemia facility. Each patient was individually assessed before initiating combined therapy, and all patients were reassessed in June 2005. The period of combination therapy was individually calculated for each subject. Patients’ records were thoroughly reviewed to determine the overall transfusional iron overload and start of chelation therapy (age and time of first blood transfusion and start of iron chelation), as well as splenectomy and hepatitis infection status. Based on the previous annual OGTTs, none of the 42 patients was diagnosed as having diabetes before the current study began. Patients were classified by their glucose metabolism status according to the World Health Organization criteria (WHO Study Group, 1985) since these were shown to be superior to the American Diabetes Association criteria in patients with this haemoglobinopathy (Holl et al, 1998). Insulin sensitivity was assessed by calculated indices, using the most widely applied insulin sensitivity index – the homeostasis assessment model (HOMA) (Matthews et al, 1985): ISIHOMA = 22·5 × 18/G0 (mg/dl) × I0 (mU/l) (G0 is fasting glucose; I0 is fasting insulin). In order to assess beta-cell secretion in the fasting state, we used the insulin release index according to the HOMA model, which is based on fasting glucose and insulin alone (Matthews et al, 1985; Stumvoll et al, 2000): SCHOMA = I0 (pmol/l) × 3·33/G0 (mmol/l) × 3·5. The area under the curve (AUC) was also calculated for estimating integrated response during OGTT for both glucose and insulin. The study was approved by the local ethics committee and written informed consent was obtained from all participants. Patients were monitored weekly for absolute neutrophil count (ANC) and any adverse events, and serum alanine transaminase and creatinine levels were measured quarterly. Non-invasive hepatic iron quantification was assessed in each patient before and after combination therapy by magnetic resonance imaging (MRI), based on T2 contrast values.

Laboratory methods

Ferritin values were calculated as the mean of the last three monthly determinations before the start of combination therapy and before the final assessment. OGTTs were performed with 1·75 g glucose/kg body weight (maximum 75 g, Dextro-OGT; Roche Diagnostics, Mannheim, Germany) in the morning after a 12-h fasting period. Venous plasma was collected via a peripheral plastic catheter at 0, 30, 60, 90 and 120 min following ingestion of glucose, to determine blood glucose and insulin levels. Ferritin levels were measured in the same laboratory by two methods: microparticle enzyme immunoassay (Abbott AxSYM System; Abbott Diagnostics, Abbott Park, IL, USA) and latex-enhanced immunoturbidimetric immunoassay (RX Daytona; Randox, Antrim, UK) with the RANDOX analyser. The average of the measured values was then calculated. Glucose was measured by an enzymatic kit (Glucose GOD-PAP assay; Roche Diagnostics) and insulin by an electrochemiluminescence immunoassay (Insulin Kit; Roche Diagnostics). Both glucose and insulin were measured using an automated chemistry analyser (Modular Analytics SWA, Roche/Hitachi System; Roche Diagnostics). All imaging was conducted at 1·5 T (Signa CV/i, with 40 mT/m gradients; General Electric, Milwaukee, IL, USA), as previously described (Ooi et al, 2004; Wood et al, 2005). Imaging parameters for sequences were electrocardiogram gating, respiratory compensation, matrix of 160 × 256, bandwidth of 31·5 kHz, slice thickness of 8 mm, fat suppression and no flow compensation to ensure ‘black’ blood. Covering the upper two-thirds of the liver between 8 and 10 images were acquired.

Statistical analysis

All statistical analyses were carried out using the Statistical Package for Social Sciences (spss Release 13·0; SPSS Inc., Chicago, IL, USA). Unless indicated differently, data are presented as mean ± SD. The two-way analysis of variance procedure was used for statistical comparison of parametric data (trend analysis over time). Correlation study was performed using Spearman correlation coefficient. Proportions of patients between the treatment groups were compared by McNemar test. Homogeneity of variance was tested for each variable, and because the distributions of insulin, SCHOMA and ISIHOMA values were skewed, the values were log transformed before statistical evaluation. A P-value of <0·05 was considered significant. Compliance to chelation was determined by: Compliance Index = actual/proposed treatment, which was calculated on a monthly basis for each medication.



One patient stopped combination therapy after 3 months, because of a decreased ANC (0·6 × 109/l), and this normalised after therapy was discontinued. This patient was not included in the analysis of this study. During the year before the initiation of combined therapy, mean Compliance Index was 0·46. Twenty-two subjects were male. Mean age at first blood transfusion of the remaining patients was 21·29 ± 22·01 months, and mean age at start of iron chelation therapy was 78·63 ± 47·19 months. Sixty-eight per cent of the patients had been splenectomised and 12% had antibody against hepatitis C infection. The patients were initially stratified into three groups according to OGTT results: diabetes mellitus (diabetes, n = 6, 14·6%), impaired glucose tolerance (IGT, n = 15, 36%) or normal glucose tolerance (NGT, n = 20, 48·8%). To better investigate glucose metabolism characteristics in normoglycaemia, patients from the last group (NGT) were further divided into two subgroups according to their fasting glucose levels (American Diabetes Association, 2005): normal fasting glucose (NFG) if glucose levels were <5·6 mmol/l (n = 14, 34·2%) or impaired fasting glucose (IFG, n = 6, 14%) if fasting glucose was between 5·6 and 6·9 mmol/l.

The mean ± SD time on combination therapy was 44·5 ± 12·4 months (range 20–54 months). This regimen was associated with decreased ferritin levels and liver iron concentration, as assessed by MRI T2 (Table 1). Combination therapy resulted in a decreased number of patients with abnormal glucose tolerance [13/41 (32%) vs. 21/41 (51%) with DFO, P = 0·043]. The improvement was observed mainly in patients with IFG. All those with IFG at baseline had a NFG at the final glucose assessment. Thus, if patients with IFG were also considered to have ‘abnormal glucose metabolism’ among with patients in the diabetes and IGT groups, combined therapy clearly improved glucose disturbances (13 subjects with DFC vs. 27 with DFO, P = 0·004, Table I). Nevertheless, after combination therapy, the number of patients with diabetes was also reduced (to 4 from 6) and the Compliance Index was improved (0·82 during the last 6 months of the study). On the other hand, combination therapy was associated with an increased body mass index.

Table I.   Characteristics of glucose metabolism before and after combination therapy.
  1. Data are expressed as mean ± SD. Values in parentheses are expressed in percentage.

  2. BMI, body mass index; MRI, magnetic resonance imaging; IGT, impaired glucose tolerance; IFG, impaired fasting glucose; AUC, area under the curve; ISIHOMA, insulin sensitivity homeostasis model; SCHOMA, insulin secretion homeostasis model.

  3. *McNemar test.

BMI (kg/m2)21·33 ± 2·8222·38 ± 2·51<0·001
Ferritin (μg/l)2991 ± 2093638 ± 1345<0·001
Liver MRI T2 (ms)22·39 ± 5·1234·35 ± 6·9<0·001
Patients with abnormal glucose tolerance21 (51·2)13 (31·8)0·049*
Patients with abnormal glucose metabolism27 (57·2)13 (31·8)<0·001*
Patients with diabetes6 (14·6)4 (9·8)0·68*
Patients with IGT15 (36·6)9 (22)0·21*
Patients with IFG6 (14·6)0 (0)0·03*
ISIHOMA1·127 ± 1·760·896 ± 0·760·4
SCHOMA102·96 ± 127·8183·41 ± 134·480·004
AUC glucose (mg/dl/min)18 218 ± 415516 366 ± 51970·019
AUC insulin (μU/ml/min)5284 ± 54545276 ± 38950·345

A descriptive analysis of patients’ characteristics according to their glucose metabolism before and after combination therapy is illustrated in Table II.

Table II.   Patients’ characteristics stratified according to oral glucose tolerance test results in two groups (normal versus abnormal) before and after combined therapy.
 DeferoxamineCombined therapy
Abnormal (n = 27)Normal (n = 14)P-valueAbnormal (n = 13)†Normal (n = 28)‡P-value
  1. Data are expressed as mean ± SD.

  2. Abnormal, all patients with IGT, IFG and diabetes; normal, patients with normal fasting glucose and normal response in glucose tolerance test; BMI, body mass index; MRI, magnetic resonance imaging; IGT, impaired glucose tolerance; IFG, impaired fasting glucose; AUC, area under the curve; ISIHOMA, insulin sensitivity homeostasis model; SCHOMA, insulin secretion homeostasis model.

  3. *Fisher's exact test.

  4. †One patient initially stratified as ‘normal’ developed IGT and one developed diabetes.

  5. ‡Includes four patients with prior IFG, 10 patients with IGT and two patients with diabetes at the initial evaluation.

  6. §Mann–Whitney nonparametric test.

  7. ¶Concerns log values.

Age (years)26·80 ± 8·8822·01 ± 8·340·1033·77 ± 6·6926·61 ± 8·870·01
BMI (kg/m2)21·55 ± 3·0520·89 ± 2·360·4823·40 ± 1·8921·91 ± 2·640·07
Splenectomy (%)19 (70)9 (64)0·73*10 (77)18 (64)0·49*
Hepatitis C infection (%)2 (7)3 (21)0·31*2 (15)3 (10)0·64*
Mean age at first blood transfusion (months)19 ± 1926 ± 290·5521·50 ± 20·8221·25 ± 23·080·98
Mean age at start of iron chelation therapy (months)74 ± 5387 ± 330·6186·40 ± 57·1775·85 ± 45·240·68
Mean duration of transfusions without chelation therapy (months)63·67 ± 49·1261·00 ± 41·430·9161·20 ± 48·0763·38 ± 46·450·93
Duration of combined therapy (months)39·38 ± 16·5246·93 ± 8·650·33§
Ferritin (μg/l)3373 ± 21912254 ± 17280·13¶1395 ± 2097287 ± 5700·70§
Liver MRI T2 (ms)22·1 ± 5·2222·9 ± 5·140·6331·04 ± 9·1235·77 ± 5·30·17§
SCHOMA82·67 ± 66·61154·1 ± 196·50·09135·03 ± 79·1195·28 ± 607·680·81
ISIHOMA1·19 ± 2·140·93 ± 0·380·661·13 ± 1·230·78 ± 0·350·88§
AUC glucose (mg/dl/min)19893 ± 402814988 ± 1902<0·001§21533 ± 585613968 ± 2441<0·001¶
AUC insulin (μU/ml/min)4996 ± 41355832 ± 76040·706668 ± 47565416 ± 27530·414

OGTT results

Fasting glucose as well as glucose response at all times during OGTT significantly decreased with combination therapy (Fig 1). Different effects were observed in various groups, also shown in Fig 1. In diabetic patients, glucose levels were similar during the fasting stage, but glucose response significantly improved at all other OGTT assessment times. In the IGT group, glucose levels were moderately, but significantly, decreased at 0, 90 and 120 min, P < 0·05. Glucose levels in patients with IFG were markedly declined at 0, 30 and 60 min, P < 0·05. There was no change in insulin levels after combination therapy for any of the groups (data not shown). Total glucose quantitative secretion significantly decreased in the total population (909 ± 289 mmol/l/min with DFC vs. 1012 ± 231 mmol/l/min with DFO, P < 0·05). This was due to a glucose decrease in IFG patients (727 ± 164 mmol/l/min with combination therapy versus 888 ± 160 mmol/l/min with DFO, P < 0·05), IGT patients (973 ± 263 mg/dl/min with combination therapy versus 1095 ± 176 mmol/l/min with DFO, P < 0·05) and diabetic patients (1067 ± 108 mmol/l/min with combination therapy versus 1345 ± 60 mmol/l/min with DFO, P = 0·11).

Figure 1.

 Time course of venous plasma glucose, after oral glucose challenge in pooled data. Levels shown before (black circle) and after the combination therapy (grey triangles). Values are given as mean ± SD; *statistical significance <0·05; **statistical significance <0·001.

A slight increase in the area under the insulin response curve was observed in all groups of patients subsequent to the combined therapy. It increased from 5832 ± 7604 to 6257 ± 4303 mU/l/min in NFG patients, from 4606 ± 2108 to 4834 ± 1980 mU/l/min in IGT patients, and from 3216 ± 2036 to 5748 ± 3627 mU/l/min in diabetic patients.

Insulin secretion and sensitivity indices

Insulin secretion according to the HOMA model (SCHOMA), increased from 102·96 ± 127·82 to 183·41 ± 134·48 (P = 0·004) in the overall group of patients (Fig 2A). This increase was mainly due to the increases in insulin secretion in the IGT patients (from 83·26 ± 58·25 to 159·41 ± 95·86, P = 0·039). No important change was noted in the other groups (from 75·51 ± 46·751 to 85·87 ± 157·98, P = 0·13 in IFG patients; from 100·47 ± 65·23 to 282·28 ± 198·34, P = 0·10 in the diabetes patients; and from 135·50 ± 203·01 to 163·99 ± 121·64, P = 0·54, in the NFG group).

Figure 2.

 Changes of insulin secretion (SCHOMA, A) and insulin sensitivity (ISIHOMA, B) homeostasis models with therapy in total population and in various groups separately. Deferoxamine (open columns) and combined therapy (filled columns); *statistical significance <0·05.

Overall insulin sensitivity (ISIHOMA) declined with therapy in the total population (Fig 2B). Intragroup analysis revealed an increase in insulin sensitivity in the IFG (1·42 ± 1·65 vs. 1·30 ± 0·97 with DFO, P = 0·60) and diabetes groups (0·75 ± 0·60 vs. 0·62 ± 0·33 with DFO, P = 0·68). Conversely, we observed a decrease in the NFG (0·93 ± 0·55 vs. 1·50 ± 2·89 with DFO, P = 0·42) and IGT groups (0·69 ± 0·25 vs. 0·88 ± 0·43 with DFO, P = 0·12). Nevertheless, none of these differences reached statistical significance. Interestingly, ferritin levels did not show any correlation with indices of glucose metabolism, even after excluding the two outliers (with ferritin levels of 6135 and 4803 μg/l).


Deferoxamine has been an important chelator (Brittenham et al, 1994), but its large, positively charged and relatively lipophobic molecule that undergoes conformational change upon binding iron, makes it unsuitable for chelation of intracellular iron, except when an active excretion pathway is present (Piga et al, 2003). Deferiprone, on the other hand, can penetrate the cell membrane and chelate toxic intracellular iron species (Shalev et al, 1995), leading to a more potent removal of myocardial iron (Shalev et al, 1995; Piga et al, 2003; Pennell et al, 2006; Borgna-Pignatti et al, 2006).

Because of these differences in chelation properties of DFO and deferiprone, combination therapy with these chelators has been an attractive alternative for the treatment of transfusional iron overload (Mourad et al, 2003). Indeed, the combined use of both chelators is associated with additive or synergistic iron excretion, a rapid reduction of the iron load in the heart and liver and reversal of cardiac dysfunction (Alymara et al, 2004; Origa et al, 2005; Kattamis et al, 2006). Whether the overall improvement in glucose metabolism that was achieved in our patients can be solely attributed to the intensification of chelation therapy and better compliance or if there is a tissue-specific effectiveness of deferiprone in iron removal from pancreas is debatable.

Although iron deposition on pancreatic parenchyma seems to play a major role in the onset of diabetes, no link has been shown between the development of the disease and ferritin levels in long-term chelated patients (De Sanctis et al, 1986; Telfer et al, 2000). This indicates that additional factors, such as genetic susceptibility and insulin resistance, may also be important in its underlying aetiology. It is noteworthy that insulin resistance may originate from iron deposition in both the liver (where iron deposits may interfere with insulin's ability to suppress hepatic glucose production) and muscle tissue (where iron deposits may decrease glucose uptake because of muscle damage) (Merkel et al, 1988).

Based on the above, a reduction in liver iron concentration could be associated with increased insulin sensitivity. Interestingly, such an improvement in the current cohort of patients was scant and only observed in the IFG and diabetes groups, supporting the concept that a defect in insulin secretion would be more crucial in the progress of glucose intolerance than would a decrease in insulin sensitivity. In support of the above speculation, we did not detect significant differences in insulin sensitivity between normal and abnormal glucose metabolism, either before or after combination therapy.

Chern et al (2001) reported that chronic hepatitis C infection was an independent risk factor for abnormal glucose tolerance. In contrast, we found no relationship between hepatitis C infection and glucose intolerance, probably because none of our patients had evident hepatic dysfunction or signs of cirrhosis. Although liver biopsies were not available in the present study, none of our patients showed increased liver enzymes during combination therapy, in accordance with previous reports (Wonke et al, 1998; Mourad et al, 2003; Origa et al, 2005).

To our knowledge, this is the first report documenting the beneficial effect of long-term combined therapy on glucose metabolism in thalassaemic patients. Not only was NGT achieved in a significant proportion of our patients, but also the cumulative glucose response significantly improved with this regimen. A profound improvement in glucose disturbances (reflected by the AUC calculation) was found in IGT and IFG patients, although the improvement in diabetic patients was less impressive. Our interpretation of this finding is that beta-cell dysfunction in diabetic patients is more severe, and thus probably irreversible by intensive chelation, in the majority of this population. Both insulin secretion and sensitivity was noticeably increased in this group, but did not sufficiently affect hyperglycaemia.

Ferritin levels did not accurately predict development of abnormal glucose metabolism, and this was also the case prior to initiation of combined therapy. It was recently demonstrated that the degree of iron overload, at least reflected by ferritin levels, was not associated with the development of other endocrine complications (Cario et al, 2003; Angelopoulos et al, 2005, 2006a,b,c,d,e). This indicates that long-term iron balance, rather than current iron status, is related to the development of glucose intolerance. In view of that observation, insulin secretion decreased, whereas the integrated glucose concentration during the OGTT was increased, with age. Early onset of chelation therapy also seems to have played an important role in ameliorating the detrimental effect of iron accumulation. We speculate that this excessive iron overload initially led to an immediate impairment of beta-cell function. However, the impairment was temporary in some patients, and their secretory capacity improved, suggesting that iron-induced toxicity is mainly time dependent.

In conclusion, we found that combination therapy with deferiprone and DFO results in an overall improvement in glucose metabolism disorders in β-thalassaemia. However, prospective trials are needed to compare the safety and efficacy of combined therapy to that of DFO or deferiprone alone in resolving other iron-induced endocrinopathies. Nevertheless, physicians caring for patients with thalassaemia major should be particularly alert to glucose metabolism, given that subclinical glucose intolerance may be present for a long period before overt diabetes develops.