Deferasirox for the treatment of iron overload associated with regular blood transfusions (transfusional hemosiderosis) in patients suffering with chronic anaemia: a systematic review and economic evaluation

Article history

The research reported in this issue of the journal was commissioned and funded by the HTA Programme on behalf of NICE as project number 06/12/01. The protocol was agreed in May 2007. The assessment report began editorial review in December 2007 and was accepted for publication in June 2008. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the referees for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Professor Tom Walley is Director of the HTA Programme. His employer is reimbursed for time allocated to duties as Director. Dr Ade Olujohungbe has declared receiving honoraria from Novartis (manufacturer of deferoxamine and deferasirox) for consultancy, speaking and attending a symposium.

Copyright statement

© 2009 Queen’s Printer and Controller of HMSO. This monograph may be freely reproduced for the purposes of private research and study and may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NCCHTA, Alpha House, Enterprise Road, Southampton Science Park, Chilworth, Southampton SO16 7NS, UK.

2009 Queen’s Printer and Controller of HMSO

Description of health problem

For many patients with chronic anaemias, regular red blood cell transfusions are life saving. However, with each unit of transfused blood, 200–250 mg of iron is transferred to the patient. There are no natural means of removing excess iron from the body and so iron gradually accumulates (over 5–10 years) to toxic levels that affect major organs such as the heart and liver. 1 This condition, commonly known as iron overload or transfusional haemosiderosis, can cause organ damage and death. 2 Currently the only way to prevent this is by long-term chelation therapy.

Epidemiology

Evidence on the incidence and prevalence of iron overload in the UK is not currently available. Indirect estimates can be produced by calculating the size of the population undergoing frequent blood transfusions and hence at risk of iron overload. The population size will vary depending on the underlying anaemic condition. As discussed earlier, the most common conditions requiring frequent blood transfusions are beta-TM, SCD and MDS.

Beta-thalassaemia major

The most reliable and up-to-date estimates of the number of beta-TM patients in the UK are thought to be held in the UK Thalassaemia Register. The register contains data such as date of birth, ethnicity, UK region of origin, deaths and cause of death. The database was thought to be 97% complete but unfortunately became inactive at the end of 2003. For the purpose of this HTA report the authors were granted access to an anonymised copy of the register. The database has information on 850 patients diagnosed with beta-TM, of whom 696 were alive in 2003. In general, the majority of beta-TM patients in the UK are of Indian, Pakistani or Mediterranean origin (see Figure 1). There is wide geographic variance in the distribution of beta-TM in the UK, with the majority of patients residing in the south of England (see Figure 2). However, clinical experts indicate that most affected births now occur in the Midlands and the north.

FIGURE 1.

Distribution of beta-thalassaemia major in the UK by ethnicity.

FIGURE 2.

Geographic location of beta-thalassaemia major patients in the UK.

Of the 696 beta-TM patients alive in 2003, 72 had undergone bone marrow transplantation and thus were not considered to be undergoing chronic blood transfusions. The remaining 624 patients were assumed to be receiving chronic transfusions and hence to be at risk of suffering from iron overload. The Office for National Statistics estimated the UK population to be 59,533,800 in mid 2003. 24 Using these figures we estimate the prevalence of iron overload in beta-TM patients to be approximately 1 per 100,000 population in the UK. However, as shown by Figures 1 and 2, the prevalence of iron-overloaded beta-TM patients in the UK will vary significantly depending on the geographic location and the presence of certain ethnic groups.

The incidence of iron-overloaded beta-TM patients is a factor of both the number of affected individuals migrating to the UK and the number of affected births, which in turn is dependent on the uptake of screening programmes. There may be a lag between the date of birth and the diagnosis of beta-TM, and similarly between the diagnosis of beta-TM and the development of iron overload. However, for our purposes we will assume that the annual number of births reported to the UK Thalassaemia Register approximately equates to the incidence of beta-TM.

The UK Thalassaemia Register did not have any patients listed as being born in 2003. This is to be expected as patients are rarely diagnosed at birth. To calculate the incidence of iron-overloaded beta-TM patients in 2003, the number of patients born in each year between 1990 and 2003 (see Figure 3) was estimated. Taking the mean gives an incidence of 15 iron-overloaded beta-TM patients per year. Using 2003 UK population figures (59,533,800) this gives an incidence rate of 0.03 iron-overloaded beta-TM patients per 100,000 population in the UK.

FIGURE 3.

Estimation of the birth rate of beta-thalassaemia major patients.

Other beta-thalassaemias

Analysis of the UK Thalassaemia Register indicates that in 2003 there were 99 beta-TI patients and 63 haemoglobin E/beta-thalassaemia patients who had not had a bone marrow transplant. Only a small proportion of these are likely to be at risk of iron overload.

Sickle cell disease

Approximately 12,500 individuals are estimated to be living with SCD in the UK, and in the region of 318 infants are born with SCD annually in England (Allison Streetly, Sickle Cell and Thalassaemia Screening Programme, October 2007, personal communication). Approximately 5% of SCD patients receive chronic transfusions. 14 Applying this figure to the population of SCD patients in the UK (12,500) gives a prevalence of approximately 625 chronically transfused patients potentially suffering from iron overload. The incidence of iron-overloaded SCD patients in the UK can be calculated in a similar way (i.e. applying 5% to 318 infants born each year) and is estimated to be approximately 16 infants annually.

Using 2003 UK population figures (59,533,800) gives a prevalence rate of 1.04 and an incidence rate of 0.02 iron-overloaded SCD patients per 100,000 population in the UK. However, given recent evidence that transfusions can help prevent primary stroke in high-risk children25 it is likely that the prevalence and incidence rates may increase.

SCD is primarily found in black ethnicities (predominantly from sub-Saharan Africa). As such there is a very unequal geographic distribution of SCD in the UK, with the highest density being located in inner city areas with a high proportion of ethnic minority populations. 26

Myelodysplastic syndrome

Epidemiological data on MDS are sparse. There are no estimates of the prevalence of MDS. Several studies, both in the UK and elsewhere, have attempted to estimate the incidence of MDS and report rates ranging from 1 to 12.6 per 100,000 population. 27–38 The estimates were generally higher for the UK ranging from 3.6 (England and Wales only) to 12.6 per 100,000. 28,31,33,35 Using UK 2003 population estimates (59,533,800) this equates to an annual incidence of MDS in the UK of approximately 2143–7501. However, not all MDS patients require chronic transfusions and not all transfusion-dependent MDS patients are at risk of iron overload.

One study20 based on the WHO classification scheme ascertained that only RA/RARS patients receiving chronic blood transfusions are at risk of iron-induced morbidity and mortality, because of their prolonged survival. RA/RARs patients accounted for approximately 23% (110/467) of all MDS patients in the study. Approximately 10% of the total (48/467) were also transfusion dependent and therefore at risk of iron overload. Hence, the incidence of MDS patients requiring transfusions and at risk of iron overload can be estimated as approximately 0.36–1.26 per 100,000 population. Using 2003 UK population figures (59,533,800) this gives an incidence of approximately 214–750 iron-overloaded MDS patients per year in the UK.

As there was no estimate of the prevalence of iron overload in MDS patients, we attempted to calculate a rough estimate using the incidence rate and the median survival. Ideally mean survival would be used because survival distributions tend to be skewed, but when mean data are not available the median can provide a rough estimate. Malcovati et al. 20 calculated the median survival in RA/RARS as approximately 9 years (108 months). Given an incidence of 214–750 cases per year and a survival of 9 years this gives a prevalence of 1924–6750 patients in the UK (prevalence rate of 3.2–11.3 per 100,000 population).

Other rare anaemias

The prevalence and incidence of iron overload in other anaemic conditions is difficult to estimate because of the rarity of the conditions and/or the spectrum of transfusional requirements; however, numbers are likely to be extremely small. For example, estimates of DBA indicate that there are in the region of 125 patients in the UK,22 not all of whom will be suffering from iron overload.

Current service provision

Deferasirox

Deferasirox (Exjade^®; Novartis) is an orally active iron-chelating agent that binds iron in a 2:1 ratio and is primarily excreted in faeces. It is given once daily as an oral suspension (usually in water or fruit juice) at a dose of 10–30 mg/kg. 47

Deferasirox may be of particular value in treating patients with iron overload who cannot tolerate DFO and who are not suitable for, or who are intolerant of, deferiprone. The ease of administration of deferasirox (oral) compared with DFO (infusional) might improve patient adherence to therapy7 and, if effective, may also improve quality and quantity of life.

Previous reviews of effectiveness

Seven published systematic reviews were identified by our search strategy. All of the reviews attempted to address the role of iron chelation therapy for iron overload, of which four also included a meta-analysis (see Appendix 3).

The review by Addis et al. 55 was limited to deferiprone only, with no consideration of comparators. This review was carried out when the use of deferiprone was still relatively rare and consequently the number of patients included in the studies is small and limited to cohort studies. Based on findings from fewer than 100 patients it reported that half of all patients given a dose of deferiprone of 75 mg/kg or more achieved negative iron balance and three-quarters of patients reduced their levels of serum ferritin, on average by almost one-quarter. This review concluded that deferiprone has clinical efficacy in achieving negative iron balance and reducing body iron burden in highly iron-overloaded patients.

The Addis et al. 55 review was subsequently included in the far broader review undertaken by the Malaysian Health Technology Assessment Unit,56 which, in addition to chelation therapy, considered other aspects of thalassaemia management such as screening, transplantation and bone marrow treatment. In terms of chelation therapy it presented evidence from studies showing beneficial impacts of DFO in terms of a wide range of factors including endocrine function and growth, cardiac disease, liver disease, survival, quality of life and cost effectiveness; and of deferiprone in terms of safety, increasing urinary iron excretion, decreasing serum ferritin levels and reducing liver iron. However, in all instances, the number of studies cited to support the evidence was small (and many of the studies that were listed as included in the review in the appendix were not referred to in the text, including the review by Addis et al. 55). Nevertheless, it was concluded that there was sufficient evidence to conclude that both DFO and deferiprone are effective in preventing or improving serious complications of the disease.

The review by Caro et al. 57 was the first to include studies that directly compared one iron chelator with another. Most of the studies included were case series and clinical trials, with only one randomised controlled trial (RCT). The findings from this review suggested that DFO was more effective than deferiprone in reducing LIC. It should however be noted that, in general, baseline LIC values were greater in patients receiving DFO, which could arguably bias in favour of DFO in terms of the chances of being able to achieve a greater reduction in LIC. Thus, to account for these differences, an analysis of covariance (ANCOVA) was conducted controlling for LIC at baseline, but this did not affect the results. Other potential sources of bias were also noted in the review. First, deferiprone patients had often failed DFO in the past (including for non-adherence) and so may also have been more prone to fail on deferiprone (for similar reasons). Second, deferiprone doses were generally low compared with DFO doses. Finally, LIC was only one of a number of outcomes included and often only in a small subset of study patients (generally a subset of those who continued treatment for a prolonged period of time and for whom long-term information on changes in iron load was available). Therefore, as the authors concede, the ‘methodological caveats and the heterogeneity of study characteristics’ raise questions about the appropriateness of pooling the data.

A larger and more up-to-date review was carried out in 2004 by Franchini and Veneri,58 which primarily focused on deferiprone and combination therapy but also considered subcutaneous bolus DFO injections and two initial phase I RCTs of deferasirox. 59,60 The meta-analysis of non-comparative studies indicated that deferiprone was effective in reducing levels of serum ferritin (overall mean reduction of around 25%), which in some studies was maintained for 3–4 years. A number of adverse events were commonly reported (GI symptoms, arthropathy, neutropenia, agranulocytosis and hepatoxicity) although only in a few cases (8.7%) did these necessitate permanent discontinuation of the drug. It was therefore concluded that deferiprone was a safe and effective oral chelator but that further studies were required to evaluate the impact on cardiac and liver disease. The authors also recommended long-term follow-up studies of bolus DFO injections because of safety concerns. With regards to deferasirox the authors concluded that the results of the phase I trials were promising in terms of safety and efficacy but that more studies were required.

The 2005 Cochrane review by Roberts et al. 61 included a comparison of different iron chelators. This was the only identified review that exclusively included RCTs. However, it was found that very few trials measured the same outcomes, which limited the ability of the review to conduct meta-analysis. Based on the outcomes that were available, the study findings did not suggest that any one chelator was better than the other and so it was concluded by the authors that there was no evidence to change current practice.

The 2006 review by VanOrden and Hagemann7 focused on deferasirox. Despite stating that this review was confined to evidence in phase III trials, the review includes evidence from phase I and phase II trials as well as pharmacokinetic studies in both humans and non-humans. Three-quarters of the patients in the efficacy analysis are from a single phase III RCT. 62 No attempt was made to pool the data from the trials and so the findings are presented narratively. The authors conclude that the results presented in the review suggest that deferasirox is as safe and effective as DFO. However, most of the patients included in this review had thalassaemia (and all of the patients in the single phase III trial had this disease) and so further studies are needed to assess the use of deferasirox in patients with other diseases such as SCD and MDS.

Finally, Abetz et al. 63 considered the impact of iron overload and its treatment on patients’ quality of life. This was concerned entirely with DFO, although it is noted that all of the included studies focused on the impacts of disease on quality of life rather than the impacts of iron chelation in particular. Nevertheless, it was reported that the degree of discomfort associated with DFO treatment was a strong predictor of a negative perception of quality of life. The authors of this review concluded that an oral iron chelator that is at least as efficacious and well tolerated as DFO is needed to improve quality of life, increase adherence and ultimately reduce morbidity and mortality due to iron overload.

During the conduct of this HTA review another review64 was published in July 2007 comparing the effects of deferiprone versus DFO and combination therapy (deferiprone and DFO) versus DFO or deferiprone. As in the 2005 Cochrane review61 this second review for the Cochrane Collaboration only included RCTs. Because of the different outcomes used as well as difficulties in assessing baseline characteristics of the included trials the reviewers only pooled data for mean changes in serum ferritin for deferiprone versus DFO. As before, no evidence was found to suggest that any one chelator was better than the other and thus the same conclusion was reached that there was no evidence to suggest change to current clinical practice, i.e. deferiprone is indicated for treating iron overload in people with thalassaemia when DFO is contraindicated or inadequate.

Identification of evidence: clinical effectiveness

Identification of evidence: cost-effectiveness

Identification of evidence: longer-term adverse event data

Selection of included trials

A total of 884 non-duplicate records was identified by our search strategy (see Chapter 3 and Appendix 5) and subsequently screened for inclusion in the review. Of these, 213 were identified to which the inclusion criteria were applied. These included 14 trials (reported in 31 publications) making comparisons between deferasirox, DFO, deferiprone and combination therapy (deferiprone and DFO) (see Table 5). Data for all of these trials were published in peer-reviewed journals (although two were only presented as abstracts70,71) with additional information derived from contacting authors. In the case of deferasirox versus DFO, additional data were retrieved from the US FDA clinical review. 72

Quality assessment of included trials

The methodological quality of the included trials is presented in Table 6 using the criteria based on the Centre for Reviews and Dissemination Report No. 4,66 which include key aspects of RCT design and quality. It should be noted that Ha et al. 93 reported on two trials (of well-chelated and poorly-chelated patients) in one paper and so for the purposes of quality assessment there were only 13 trials (although the information was still derived from 31 publications).

Overall, the methodological quality of the included trials was poor. The published papers all stated that patients were randomly allocated to treatment groups; however, only four77,81,91,93 described the method of randomisation used and only two of these81,91 noted whether or how allocation was concealed. One other83 gave details of allocation concealment but did not adequately document the randomisation process. Blinding of administrators or participants was acknowledged to be difficult or unethical given the administration route of the main comparator DFO, but the blinding of assessors was generally addressed inadequately, with only two trials91,94 providing information in this respect. Intention to treat (ITT) analyses were carried out in four trials;81,83,91,96 one trial62 was a non-inferiority trial in which ITT analysis may increase the risk of falsely concluding non-inferiority and thus per protocol analysis (as presented) may be preferable. 100 Baseline characteristics including age and gender, along with outcome variables such as serum ferritin, LIC and other potentially significant factors (number of transfusions or patients who had splenectomies), were provided in eight trials. 46,62,77,81,91,94,96,97 Comparability between groups was achieved in six trials46,62,81,91,93,97 and partially achieved in four. 71,77,94,96 All trials specified the number of patients originally randomised and provided full or partial details of eligibility criteria. All trials reported outcomes for 80% or more of the patients originally randomised; one95 failed to adequately account for withdrawals.

Trial characteristics

The included trials involved a total study population of 1480, ranging in size from 1393 to 586. 62 Two trials81,91 reported on study populations of up to 200 patients but the majority were populated by less than 100 patients. All but three83,95,96 were designed as multicentred trials.

Most trials were designed as parallel and open-label studies. Of these, two were three-arm trials. 70,95 There was one double-blind, placebo-controlled, parallel trial46 and one randomised crossover design. 83

The duration of each trial varied between 5 days83 and 2 years71 with the majority46,62,77,81,91,94,96,97,101 continuing for approximately 12 months. Three trials were halted prematurely, the two Ha et al. RCTs93 because of the unexpected death of a patient in one of these trials and the third trial71 because of withdrawal of support from the pharmaceutical company funding the trial.

Outcome measures varied across trials and were surrogate measures of iron overload: serum ferritin; LIC determined by biopsy, SQUID or liver T2*; heart iron content assessed by myocardial T2*.

Serum ferritin was the most commonly utilised measure, set as the primary outcome in six trials91,93,95–97 and the secondary outcome in seven others. 46,62,70,71,77,81,94

One trial used LIC determined by biopsy to measure the primary outcome;70 another trial62 employed LIC determined by biopsy and SQUID as the primary outcome. Of the two trials in which LIC by biopsy was a secondary outcome, one trial set out to measure all patients93 and one a subset of patients. 91 Two trials employed LIC by SQUID to measure the primary outcome71,97 and three others used it as a secondary outcome. 77,81,94

Success in terms of change in LIC was an outcome in two trials. 62,77 In Cappellini et al. 62 success was defined in patients with a baseline LIC of < 10 mg Fe/g dw as an end-of-study LIC value of 1–7 mg Fe/g dw and in patients with a baseline LIC of ≥ 10 mg Fe/g dw as a decrease in LIC of ≥ 3 mg Fe/g dw. In Piga et al. 77 success was defined as a fall in baseline LIC of > 10%.

Myocardial T2* was the primary outcome in two trials. 46,94 Other outcomes included liver T2*,46 a range of safety measures,60,77,81,94,96,97 urinary or faecal iron excretion83,91,95,96,101 and adherence. 46,70,91,93,94,96,97

Trials differed in respect of the lower age limits of participants. One trial did not specifically state ages but described patients as ‘children’. 95

At least half (7/14) of the trials received pharmaceutical support. 46,62,71,77,81,94,97

Trial characteristics are presented in Table 7.

Participant characteristics

The majority of trials included patients with beta-TM or thalassaemia (Table 8). Two patients with beta-TI were included in one trial77 and there were two patients with DBA in another. 83 One trial included only patients with SCD. 81 The youngest patient was aged 2 years62 and the oldest was aged 54 years. 81 Trials were evenly balanced in terms of male and female participants but there were differences across trials in terms of baseline LIC and serum ferritin.

Data analysis

Results have been grouped by treatment(s) and comparators as follows: deferasirox versus DFO; deferiprone versus DFO; combination therapy (deferiprone and DFO) versus DFO and/or versus deferiprone. The following sections provide an overview of the data available from the trials, the comparability across trials and, when possible and appropriate, the results of any meta-analysis conducted. When meta-analyses were not carried out a narrative summary of the study results is provided. Study outcome data are presented in Table 9.

Deferiprone versus DFO

Five trials compared deferiprone with DFO,71,83,91,93,94 one of which was a crossover trial. 83 In addition, two three-arm trials70,95 compared both deferiprone and DFO with combination therapy (deferiprone and DFO) and therefore are included in this section as well as in the section on combination therapy versus DFO or deferiprone.

Population

All trials included patients with thalassaemia with the majority explicitly stating that patients had beta-TM. 70,71,83,91,94 Olivieri and Brittenham71 intended to include patients with SCD according to an early report by Basran et al. ;87 however, subsequent reports refer to patients with beta-TM,86,88,89 thalassaemia90 or make no explicit reference to any disease. 71

Five trials included both children and adults, whereas Gomber et al. 95 recruited only children and Pennell et al. 94 included only patients aged 18 years and over.

Comparison of baseline LIC is problematic because of differences in how this was measured. Three trials70,91,93 measured LIC by biopsy, Olivieri and Brittenham71 measured it by biopsy or SQUID, and Pennell et al. 94 measured it by SQUID; the remaining two trials83,95 did not measure LIC at all. In the three trials that measured LIC by biopsy it was notable that the baseline LIC was higher in Aydinok et al. 70 (at least 16 mg Fe/g dw) than in Maggio et al. 91 (around 3.5 mg Fe/g dw or less) or Ha et al. 93 (7 mg Fe/g dw or less). In Ha et al. 93 the baseline LIC was not actually presented but to be included in this trial it was stated that patients had to be well-chelated, which was defined as having a baseline LIC of 7 mg Fe/g dw or less.

Baseline mean serum ferritin concentrations were measured and reported in all but one study (Ha et al. 93) and were varied. In Aydinok et al. 70 baseline levels were higher in both groups (deferiprone, 3.84 mg/l; DFO, 3.34 mg/l) and in Gomber et al. 95 they differed between the two groups (deferiprone, 2.67 mg/l; DFO, 5.08 mg/l). In Maggio et al. 91 inclusion criteria stated that patients must have a baseline serum ferritin of between 1.50 mg/l and 3.00 mg/l, although 11 patients in the deferiprone group and seven patients in the DFO group had baseline levels of 3.00 mg/l or above.

Interventions/comparators

Deferiprone target doses were the same (75 mg/kg/day) in five of the seven trials but in Pennell et al. 94 a higher dose (75 mg/kg/day rising to a target dose of 100 mg/kg/day) was given and in Olivieri et al. 83 a lower dose was given (50 mg/kg/day). DFO target doses were relatively similar in all of the trials (40 mg/kg/day,95 50 mg/kg/day71,83,91,94 or a range between 40 and 50 mg/kg/day70 or 30 and 60 mg/kg/day93).

In Olivieri et al. 83 patients acted as their own controls. Thus, they were admitted to hospital and given either deferiprone or DFO on days two, three and four. On day six patients were discharged and readmitted 3–4 weeks later following their next blood transfusion and given the alternative drug in the same manner. In all of the other trials deferiprone was taken orally three times a day, 7 days a week and DFO was infused, often overnight, between five and seven times a week.

Outcomes

Five trials measured mean changes in LIC70,71,91,93,94 but different measures of LIC were used across the trials. Changes were also assessed over different time periods, varying from 12 months70,94 to a median of 18 months93 to a mean of around 30 months or more. 71,91 These variations made it impossible to pool these data.

Findings were not consistent across the trials. Olivieri and Brittenham71 and Ha et al. 93 reported increases in LIC for both the deferiprone and DFO groups over a period of 18 months or more, with smaller increments in the DFO group, whereas Maggio et al. 91 and Pennell et al. 94 reported decreases in LIC that were reasonably similar for both groups. Aydinok et al. 70 also reported decreases in both groups but the decrease was larger in the DFO group.

Six trials measured mean changes in serum ferritin70,86,91,94–96 although again over different time periods varying from 6 months95 to 12 months,70,91,94 a median of 18 months93 and a mean of 22 months. 86 Again, findings were not consistent across the trials. At 12 months or more most trials reported a decrease amongst patients in both groups whereas, at 6 months, both Gomber et al. 95 and Pennell et al. 94 reported decreases in the DFO group as opposed to increases in the deferiprone group.

Data could be pooled for two trials at 6 months (Figure 4)94,95 and three trials at 12 months (Figure 5). 70,91,94 The pooled estimate was not significant but does appear to favour DFO at 6 months [random effects, weighted mean difference (WMD) 1.18, 95% confidence interval (CI) –0.42 to 2.78]; there was no significant difference in serum ferritin between the deferiprone and DFO groups at 12 months (random effects, WMD –0.10, 95% CI –0.57 to 0.38). However, the trials showed statistical heterogeneity at both time points (6 months: χ² = 6.27, df = 1, p = 0.01, I² = 84.0%; 12 months: χ² = 8.09, df = 2, p = 0.02, I² = 75.3%).

FIGURE 4.

Pooled mean changes in serum ferritin (mg/l) in trials comparing deferiprone with DFO at 6 months.

FIGURE 5.

Pooled mean changes in serum ferritin (mg/l) in trials comparing deferiprone with DFO at 12 months.

Only Pennell et al. 94 measured myocardial iron using T2*. This study reported both deferiprone and DFO to be efficacious in removing myocardial iron and the authors also reported that the difference between drugs was significant in favour of deferiprone at both 6 months (ratio of geometric mean, 1.09; p = 0.040) and 12 months (ratio of geometric mean, 1.12; p = 0.023).

All of the trials were concerned with measuring the control of iron overload except for the Olivieri et al. crossover trial. 83 Thus, this trial did not report on any relevant outcomes, although it did measure serum ferritin concentrations. However, given the short-term nature of this trial (5 days), any reported outcomes of this measure would have been of limited clinical value.

Summary: deferiprone versus DFO

Comparing patients receiving deferiprone with those receiving DFO is problematic because:

although five trials included patient populations consisting of a mixture of children and adults,70,71,83,91,93 one study focused only on children95 and another only on adults94

not all trials measured LIC and, in those that did, different methods and time points were used.

Based on mixed populations of children and adults the findings suggest that there was no significant difference between deferiprone and DFO in terms of changes in serum ferritin at 6 months94,95 or 12 months70,91,94 although statistical heterogeneity was evident.

Myocardial iron was assessed by T2* in one study94 (of adults) and reported deferiprone to be significantly superior to DFO, suggesting a superior outcome in terms of removing iron from the heart.

Combination therapy (deferiprone and DFO) versus DFO or deferiprone

Six trials evaluated combination therapy versus DFO,46,70,93,95–97 of which two also considered combination therapy versus deferiprone. 70,95

Population

All trials included patients with thalassaemia with most explicitly stating that patients had beta-TM. 46,70,96,97 The majority included a mix of children and adults although Gomber et al. 95 included only children and Tanner et al. 46 included only patients aged 18 years and over. Thus, mean ages at baseline ranged from around 13 years in Aydinok et al. 70 to nearly 29 years in Tanner et al. 46 Average ages in the other trials (except Gomber et al. 95 in which the average age of patients was not stated) were between 16 and 20 years depending on treatment group.

Comparison of baseline LIC remains difficult because of differences in measurement. Two trials measured LIC by biopsy,70,93 Galanello et al. 97 measured LIC by SQUID and Tanner et al. 46 used liver T2*. The remaining two trials95,96 did not measure LIC.

Of the two trials using biopsy, baseline LIC was 16 mg Fe/g dw or higher in Aydinok et al. ,70 whereas Ha et al. 93 simply reported that patients had to be poorly chelated, which was defined as having a baseline LIC of greater than 7 mg Fe/g dw.

Baseline serum ferritin varied across the trials, reflecting varied inclusion/exclusion criteria. Thus, the baseline serum ferritin ranged between 2.67 mg/l (deferiprone) and 5.08mg/l (DFO) in Gomber et al. 95 and between 4.15mg/l (combination therapy) and 5.51mg/l (DFO) in Mourad et al. 96 and was around 2.00 mg/l (in both groups) in Galanello et al. 97 Baseline levels were close to 2.00 mg/l (in both groups) in Tanner et al. 46 and 3.50 mg/l in Aydinok et al. 70 Ha et al. 93 did not report baseline levels.

Interventions/comparators

In all trials deferiprone target doses were the same (75 mg/kg/day) for combination therapy46,70,93,95–97 and, when applicable, for deferiprone monotherapy. 70,95 DFO doses were also comparable across the trials (40–50 mg/kg/day), either as monotherapy or in combination with deferiprone. Tanner et al. 46 was the only study in which a placebo pill was given with DFO as a comparator to combination therapy.

Outcomes

All of the trials that measured LIC46,70,93,97 reported similar changes in LIC between the groups irrespective of how LIC was measured. In Aydinok et al. 70 the mean fall in LIC was greater in the combination therapy group than in the deferiprone monotherapy group. Change in LIC data could not be pooled because of the different methods and time points of measurement.

All six trials46,70,93,95–97 measured mean change in serum ferritin. Over 6 months Gomber et al. 95 and Mourad et al. 96 reported findings supporting DFO and combination therapy, respectively, but overall the pooled estimate from the two trials suggested no significant difference (random effects, WMD 0.52, 95% CI –1.33 to 2.37; Figure 6). Over 12 months three trials reported combination therapy to be marginally superior to DFO in terms of the mean reduction in serum ferritin concentrations,46,70,97 whereas another trial reported DFO to be superior. 96 Data could only be pooled for three of these trials70,96,97 (Figure 7) because change in standard deviation data were not available and could not easily be calculated for Tanner et al. 46 The meta-analysis found a significantly larger decrease in mean serum ferritin in the combination therapy group than in the DFO group (fixed effects, WMD –0.71, 95% CI –1.01 to –0.41).

FIGURE 6.

Pooled mean changes in serum ferritin (mg/l) in trials comparing combination therapy with DFO at 6 months.

FIGURE 7.

Pooled mean changes in serum ferritin (mg/l) in trials comparing combination therapy with DFO at 12 months.

Two trials compared combination therapy with deferiprone monotherapy. 70,95 Both reported combination therapy to be superior in terms of change in serum ferritin; over 6 months in Gomber et al. 95 and 12 months in Aydinok et al. 70

Only one study46 measured myocardial iron using T2*. Tanner et al. 46 reported significant improvements in myocardial T2* over 6 and 12 months in both the combination therapy group and the DFO group with the combination therapy group performing significantly better than the DFO group (increase of 10%, 95% CI 2–19%; p = 0.02).

Summary: combination therapy versus DFO or deferiprone

Comparing patients and measuring changes in LIC in those receiving combination therapy with those receiving DFO monotherapy or deferiprone monotherapy is problematic because:

• although four trials included patient populations consisting of a mixture of children and adults,70,93,96,97 one study focused only on children95 and another only on adults46

• only Aydinok et al. 70 and Gomber et al. 95 made direct comparisons between combination therapy and DFO but in populations of children and adults and children only respectively

• in trials that measured LIC, different methods and time points were reported

• only two trials compared combination therapy with deferiprone and at different follow-up periods.

Data that could be pooled for change in serum ferritin at 6 months95,96 and 12 months70,96,97 suggested there were no significant differences between combination therapy and DFO at 6 months but that combination therapy was significantly superior at 12 months.

Myocardial iron by T2* was assessed in one study (of adults) and reported combination therapy to be significantly superior to DFO. 46

Adverse events from RCTs

Inconsistent reporting of adverse events (AEs) in the included trials made it difficult to compare these events across the trials (Table 10).

Adverse events from longer-term follow-up studies of deferasirox

Longer-term follow-up data of patients receiving deferasirox, found by the additional literature search, were limited to three studies with a median period of up to two and a half years of follow-up, all of which were published as conference abstracts (Table 11). 105–107 All of these patients were previously participants in clinical trials, including RCTs described in the review above. Most patients had beta-TM, although one-fifth were suffering from SCD. In a cohort of just over 1000 patients (including over 400 paediatric patients) SAEs were rare in both adults and children. 106,107 In total there were 15 deaths, of which only one (child) was felt to be possibly drug related by an investigator but not by the Program Safety Board. GI disorders and skin rashes were the most common drug-related AEs. Discontinuation of deferasirox because of AEs was relatively uncommon. No notable effects on liver or renal function were noted. In a cohort of 480 patients,105 progressive creatinine rises were identified, but these were reported as being generally reversible with dose modification/interruption.

More recently we became aware that a further three abstracts were presented to the 49th American Society of Haematology Annual Meeting in December 2007. 108–110 Two of these108,109 contained data on the same cohort of just over 1000 patients after a further 12 months; no notable differences in adverse events were reported. The other study110 is an extension of the Vichinsky et al. 81 RCT of patients with SCD. 81 The most common AEs were GI and skin rash; there were no significant changes in markers of liver or renal function; no cases of progressive increases in serum creatinine were reported.

However, post-marketing AE data identified cases of fatal, acute, irreversible renal failure and cytopenias (including agranulocytosis and thrombocytopenia). 111 In September 2007 the FDA112 published more detailed information on these AEs; the most common involved the GI (including hepatic), renal and haematological systems (Table 12). Of 115 suspected AEs, 108 reported a serious outcome including death. Some of these records were duplicates, for example the number of deaths was reported to be 19, of which 17 were unduplicated. There were 24 unduplicated reports of hepatic AEs including increased ALT, increased bilirubin, jaundice, ascites, subclinical and clinical hepatitis, liver failure, hepatic encephalopathy and cholecystitis. Reports of renal AEs in 16 patients included renal failure (of which two AEs were fatal), acute renal failure, glomerulonephritis, interstitial nephritis and renal tubulopathy. There were 15 unduplicated reports of haematological AEs included agranulocytosis, neutropenia and thrombocytopenia. It was reported that ‘some’ of these patients died. In many of these cases the extent to which AEs may be caused by deferasirox is not known. Three patients with hepatic failure had a significant history of hepatic disease and/or use of concomitant medication with known hepatic AEs, four patients with renal AEs had a history of renal disease and ‘most’ of the patients with haematological AEs had pre-existing haematological disorders that are frequently associated with bone marrow failure. As a result, product labelling has been updated. In addition, the FDA requested health-care professionals to report any SAEs in association with deferasirox therapy (e.g. renal failure and cytopenias).

Other considerations

At the time of the literature search none of the RCTs included in this review presented quality of life (QoL) outcomes. One trial explicitly stated that it attempted to measure QoL but then failed to present any findings. 87 One of the authors involved in this trial was contacted for further information regarding the various published abstracts but failed to respond to our emails.

Eight trials reported on adherence with deferiprone versus DFO or combination therapy versus DFO and/or versus deferiprone. 46,70,71,91,93,94,96,97 It should, however, be noted that RCTs are not ideal for measuring adherence and tend to overestimate this in clinical practice.

In Olivieri and Brittenham71 adherence in the deferiprone group was measured with computerised bottles and was reported to be significantly better than adherence in the DFO group measured using ambulatory pumps [deferiprone, mean (SD) = 94.9% (1.1%); DFO, mean (SD) = 71.6% (22.5%); p < 0.005).

Adherence with the trial treatment was assessed in Maggio et al. 91 by counting the pills in each returned bag of deferiprone, by assessing the total dose of DFO consumed each week and by interviewing the patients’ relatives. A total of 55 patients in each trial group (deferiprone = 77.5%; DFO = 75.3%) took the prescribed dose of the trial treatment during the trial period and four patients (5.6%) in the deferiprone group and seven (9.6%) in the DFO group took a reduced dose because of low adherence.

In Pennell et al. ,94 Tanner et al. 46 and Galanello et al. ,97 deferiprone adherence was measured using the Medication Event Monitoring System device (Aardex, Zug, Switzerland) and DFO adherence was measured using Crono pumps (supplemented by the use of diary cards and weekly physical examination of infusion sites in Galanello et al. ). Adherence in Pennell et al. 94 was similar between groups (p = 0.81) being 94% (SD 5.3%) in the deferiprone group and 93% (SD 9.7%) in the DFO group. Tanner et al. 46 also reported similar rates of DFO adherence in the combination therapy and DFO groups (91.4% compared with 92.6%, p = 0.7). This trial also compared adherence with deferiprone tablets in the combination therapy group with placebo tablets in the DFO group and reported adherence with placebo to be significantly better than adherence with deferiprone (89.8% compared with 82.4%, p = 0.04); no reason is given for this result. In Galanello et al. 97 only adherence with DFO is reported and this was reported to be similar in both the combination therapy (96.1%) and DFO monotherapy groups (95.7%).

Mourad et al. 96 defined adherence as either ‘excellent’ (taking over 90% of the recommended doses) or ‘good’ (75–90% of recommended doses) and reported adherence to be better with combination therapy [excellent = 10/11 (90.1%)] than with DFO [excellent = 11/14 (78.6%)].

In Ha et al. 93 adherence was determined by diary entry and the use of tablets provided between visits and was reported to be ‘excellent’ in the combination therapy group with 75% of patients being compliant in taking both deferiprone and DFO. Adherence with DFO alone was considered to be ‘good’ for 60% of patients and ‘poor’ for 40% of patients; the criteria as to how these terms were defined are not reported.

In Aydinok et al. 70 adherence was assessed by adherence to treatment at weeks 12, 26, 38 and 54; adherence was reported to be better with both combination therapy [30/30 (100%)] and with deferiprone monotherapy [26/26 (100%)] than with DFO [22/25 (88.0%)].

Because of the small sample sizes in the above trials all results should be interpreted with extreme caution.

Following the completion of the literature searches a subsequent paper was published by Cappellini et al. 113 relating, albeit indirectly, to QoL and adherence. This paper presents findings on patients’ experiences of treatment (reported satisfaction, convenience, preferences and willingness to continue trial treatment) from the Cappellini et al. 62 trial. In this trial, which excluded patients with previously poor adherence of DFO, at baseline, one-third of patients in both the deferasirox and DFO groups reported that they were dissatisfied with DFO treatment [94/289 (32.5%) and 93/282 (33.0%) respectively] whereas, at the end of the trial, 38.3% (108/282) of patients in the DFO group and 5.9% (17/289) of patients in the deferasirox group were dissatisfied with their respective trial treatment. 113 Similarly, two-thirds of patients in both groups reported that DFO treatment was inconvenient at baseline [198/289 (68.5%) and 193/282 (68.4%) respectively] but, at the end of the trial, 72.7% (205/282) of patients in the DFO group were inconvenienced by DFO compared with 1.0% (3/289) of patients in the deferasirox group. Amongst patients who had experience of using both deferasirox and DFO, at the end of the trial 0.7% (2/289) of patients stated they preferred DFO to deferasirox compared with 96.9% (280/289) of patients who reported that they preferred deferasirox to DFO. Finally, 85.8% (248/289) of patients receiving deferasirox reported that they would be willing to continue trial treatment compared with 13.8% (39/282) of patients receiving DFO (p < 0.001), whereas 3.5% (10/289) and 64.5% (182/282) of patients indicated that they would be unwilling to continue their trial treatment with deferasirox and DFO respectively.

Clinical discussion

The aim of this clinical review was to evaluate the effectiveness of a relatively new oral chelator, deferasirox, in the treatment of iron overload due to transfusional haemosiderosis in patients suffering with chronic anaemia. To achieve this comparisons with other iron chelators were necessary, i.e. DFO, deferiprone and combination therapy (deferiprone and DFO).

The range of outcome markers described in the review may be confusing and the different techniques used to measure LIC made meaningful comparisons problematic. Moreover, there was a diversity of inclusion and exclusion criteria, there were different follow-up periods and both the quality of reporting and the methodological quality of the trials were generally poor. Limited availability of information in two trials presented only as conference abstracts made it difficult to assess the methodological quality and extract data. 114,115 Most trials included in the review were small in size with only three62,81,91 including more than 100 participants; around half of the patients in the review were in the three trials that compared deferasirox with DFO. 62,77,81 As a result it was possible to undertake meta-analyses with data from only a small subset of the papers included in the review and, in most cases, there was evidence of statistical heterogeneity. It is therefore difficult to interpret the results of the review with any certainty.

The largest trial, which was designed to test for non-inferiority of deferasirox compared with DFO (utilising trial-specific measures of ‘success’ in terms of changes in LIC), included 586 patients and found that, at licensed doses of 20 mg/kg/day or above, deferasirox was not inferior to DFO. 62 However, some patients in the trial (those with baseline LIC < 7 mg Fe/g dw) appeared to be underdosed, particularly in comparison with patients receiving DFO. Thus, the main claim to non-inferiority is based on post hoc subgroup analysis, which raises concern although it is supported by prespecified secondary subgroup analysis which found that, in terms of mean changes in LIC, deferasirox was not inferior to DFO in patients with a baseline LIC ≥ 7 mg Fe/g dw (i.e. those patients who received deferasirox at a dose of 20 mg/kg/day or above). 72 In the smaller trials of deferiprone versus DFO, deferiprone was less efficacious in reducing LIC levels than DFO; however, a combination of deferiprone and DFO generally produced comparable results between the two chelators. Only two trials compared combination therapy with deferiprone.

Based on individual trial data, including the large Cappellini et al. trial,62 deferasirox and DFO appear to be generally similar in reducing serum ferritin concentrations. Based on both individual trial data and pooled data there are no significant differences between deferiprone monotherapy and DFO, and combination therapy is superior to DFO. Meta-analyses found no significant differences between deferiprone and DFO or combination therapy and DFO at 6 months. However, there was statistical heterogeneity in the deferiprone trials at both follow-up periods and in the combination therapy trials at 12 months; reasons for the heterogeneity are unknown but this could be accounted for by differences in the study populations. In addition, it should be noted that the number of patients included in each of the pooled analyses was still relatively small. Only two trials compared combination therapy with deferiprone and this was over different follow-up periods.

Only two trials46,94 included in this review measured changes in cardiac T2* (an indirect measure of myocardial iron overload), neither of which considered deferasirox. These show a statistically significant difference in cardiac T2* levels between the treatment arms favouring deferiprone and combination therapy over DFO.

The outcomes measured in these trials are relatively short term and only measure surrogate end points of the real outcomes of clinical importance (e.g. the effects of iron chelation on morbidity including end-organ damage or other toxicity such as cardiac dysfunction and liver fibrosis). Long-term retrospective studies on morbidity and mortality have reported cardiac events and mortality risk to be lower in thalassaemia patients with good adherence to iron chelation therapy and in those treated with deferiprone as opposed to DFO. 116–118 Similar studies have yet to be conducted, or at least published, in patients receiving deferasirox.

The RCTs suggest that generally deferasirox is safe, but post-marketing follow-up data involving patients receiving deferasirox have identified AEs of considerable concern (e.g. fatal, acute, irreversible renal failure and cytopenias). 111,112 It is currently unclear if these AEs are drug related. Thus, further longer-term observational studies are needed. In the meantime the updating of the product labelling for deferasirox to reflect the current information regarding the cases of acute renal failure and cytopenias has been recommended by the FDA.

Only the Cappellini et al. trial62 reported data on QoL; this appeared in a paper identified following the completion of the review. 113 This study reported that patients prefer deferasirox to DFO in terms of reported satisfaction, convenience, preferences and willingness to continue study treatment. QoL as measured by patient perceptions of their treatment is clearly important with regard to adherence.

Adherence with DFO was measured only in the RCTs that considered this as a comparator to deferiprone and/or combination therapy. Although DFO adherence was not as poor in these trials as would be expected from clinical practice,119 it should be noted that different methods were used for measuring adherence across the trials and that RCTs by their very nature are not the most adequately designed studies for measuring adherence in the real world. Adherence is an important issue because, although trials may suggest that all chelators are reasonably similar in terms of efficacy, in practice lack of adherence to treatment protocols will clearly limit the likelihood of the treatment being effective. 120 Adherence is more likely to be high in children for whom this is the responsibility of the parent. However, during adolescence, when chelation is becoming the patient’s (rather then the parent’s) responsibility, anecdotal evidence suggests that there can often be disruption of DFO treatment leading to long-term avoidance (UK Thalassaemia Society, 2007, personal communication).

With the exception of one trial of SCD patients,81 all of the RCTs included patients with thalassaemia (nearly always beta-TM), with no subgroup analyses by underlying disease. Patients with SCD may start blood transfusions later in life and with less frequency than those with beta-TM. Patients with MDS are typically older than those with thalassaemia or SCD, being in their 50s and 60s; in the current review the average age of each trial population was much lower than this. Furthermore, it has been shown in MDS patients that serum ferritin levels in excess of 1.0 mg/l are related to reduced survival20 as opposed to 2.5 mg/l in thalassaemia patients. 5

Recent data from non-RCTs are only partially illuminating. An open-label study121 reported that deferasirox reduced mean serum ferritin by around 0.8 mg/l at 12 months, with levels still on average 2.6 mg/l; around one-third of patients developed thrombocytopenia but new cytopenias were considered by the authors to be consistent with haematological progression of MDS. A separate prospective trial122 of patients with MDS (n = 47), DBA (n = 30) and other rare anaemias (n = 22), as well as thalassaemia (n = 85), reported mean reductions in LIC in all disease groups from deferasirox (dose depended on baseline LIC as in the large Cappellini et al. RCT,62 with most patients receiving 20–30 mg/kg); mean changes in LIC correlated to changes in serum ferritin, were dose respondent and greatest in the MDS and smallest in the DBA groups. However LIC was measured by a combination of biopsy or SQUID, with around half of all MDS and DBA patients being assessed by SQUID compared with fewer than one-quarter of patients with thalassaemia or other rare anaemias. There were no differences in the most common AEs (GIs, skin rash and non-progressive creatinine increases) across the disease groups although all deaths [5/184 (2.7%)] were reported in MDS and DBA patients; these were not considered to be drug related. There were 9/184 (4.9%) cases of neutropenia, all felt by the investigators to be related to the disease and not the drug; these were more prevalent in non-thalassaemia patients.

All but three RCTs included a mix of children and adults in their patient populations, with no subgroup analyses by age group. There are pharmacokinetics data which demonstrate that children appear to metabolise deferasirox more rapidly than adults. 77,123 This in turn may have implications with regard to both safety and efficacy although the long-term data to date have shown no apparent differences between children and adults with regard to AEs.

A final factor to be considered which may decrease the value of the studies included is publication bias and the fact that most studies in this area were conducted with pharmaceutical company involvement; such studies in the past in other therapeutic areas have been shown to contain a bias towards the drugs of the sponsor. 124,125

In summary, there is some evidence in the clinical review to support the use of all three iron chelators in people with iron overload but, for reasons discussed above, these must be interpreted with caution. There is a need to strengthen the evidence base with further research of clinical outcomes, particularly cardiac T2* in patients receiving deferasirox.

Introduction

This chapter explores the published literature on the costs and benefits of iron chelation therapy for the treatment of iron overload in chronically transfused patients. The aim of this review was to identify published cost-effectiveness studies of deferasirox versus DFO, deferiprone or placebo; or deferiprone versus DFO (alone or in combination with deferiprone). Because of limitations in the availability of published information (many abstract-only studies) this review is more a descriptive presentation than a critical appraisal.

Identification of studies

Details of the search strategy and the methods for selecting evidence are presented in Chapter 3. A total of 884 records was identified by the search strategy; five were subsequently singled out as pertaining to the economics of chelation therapy and obtained in full text format to facilitate the application of inclusion/exclusion criteria. Of these, four records were selected for inclusion in the review. A further five abstracts were identified from searching conference proceedings [American Hematology Association (AHA) and European Haemotology Association (EHA)] and one full text article was identified from hand-searching activities, equating to a total of 10 articles. 126–135 From this, eight distinct cost-effectiveness studies were identified: one full paper131 and seven studies in abstract-only form. 126,127,130,132–135

Quality assessment

Quality assessment of the abstract-only studies would be meaningless because of word limit constraints. Hence, the decision was taken to apply detailed quality assessment criteria to the single full text article only. 131 In general the quality of this study was high (Table 13).

Study characteristics

All eight studies undertook a cost–utility analysis, presenting results as cost per quality-adjusted life year (QALY), and all compared deferasirox with DFO (Table 14). Four studies considered only beta-TM patients,130–133 one study considered SCD patients,126 one study included only MDS patients127 and two studies considered beta-TM, SCD and MDS patients as a group. 134,135 Only two studies had a UK perspective, three studies had a US perspective and the remaining studies were Canadian, Brazilian and European. The four studies in beta-TM patients adopted a long-term time frame (lifetime/50 years);130–133 the remaining studies appeared to be of 1 year in duration. All of the studies had industry author affiliations, and there was a large degree of overlap, in terms of both data sources and authors, between many of the studies.

Economic models

In the studies two distinct modelling approaches were adopted: long-term modelling (lifetime) and short-term (1 year) modelling. See Table 15 for full details of each of the models.

Costing

The majority of studies expressed costs in US dollars; the remaining studies utilised UK pounds, Canadian dollars and Euros. Only half of the studies provided a price year, which ranged from 2004 to 2006. In the long-term studies, discounting of costs was undertaken using rates appropriate to the country of origin. The price of chelators was presented in only three studies,131,132,134 all of which were in different currencies making comparison difficult, although the price of deferasirox was consistently greater than that of DFO. None of the studies presented resource use separately from costs, making it impossible to validate the estimated costs (see Table 16).

Health outcomes

The incorporation of health outcomes was highly dependent upon the time period chosen for the analysis (see Table 17).

Results and sensitivity analyses

The results and sensitivity analyses (SA) of the published economic evaluations are presented in Tables 18 and 19.

Summary

The results of this literature review appear to demonstrate the cost-effectiveness of deferasirox compared with DFO for the treatment of iron overload in a number of different patient populations and study locations. However, it must be noted that, because of the large proportion of information that was only available in abstract form, the validity of these studies in terms of data sources, methods and assumptions could not be verified, hence conclusions based on their results must be viewed with caution.

That being said, it was still possible to establish two distinct trends in modelling approaches: long-term and short-term modelling, and identify some of the shortcomings of each of the approaches.

The short-term modelling studies (1 year) in SCD patients, MDS patients and beta-TM, SCD and MDS patients as a composite rely on QoL differences solely. Given the chronic nature of iron overload, a 1-year time frame seems short, especially for SCD and beta-TM patients who can survive for many decades if treated appropriately. However, the authors of the short-term studies defend their choice of time frame by acknowledging the lack of long-term data to inform modelling (particularly in SCD and MDS).

The long-term modelling studies (lifetime) in beta-TM patients rely on a number of assumptions concerning adherence and survival to extrapolate to the long term. Although it is justifiable to attempt to determine the long-term effects of chelation therapy, heavy reliance on inference and assumptions is dubious. Without suitable data to validate these assumptions it is difficult to ascertain the reliability of the cost-effectiveness results.

This literature review highlights the difficulties of matching up the needs of a long-term model that will capture all of the important factors and issues associated with a chronic condition such as iron overload (this is especially complex given the different patient populations) with the constraints of limited data, which is no doubt due to the rarity of iron overload.

Health outcomes

Our systematic review of RCT clinical data was unable to discern a statistically and clinically important difference in terms of reductions in LIC and serum ferritin between deferasirox and DFO. Little could be gleaned on the comparative effectiveness of deferasirox and deferiprone because of the lack of data for this comparison; however, it must be acknowledged that the analysis was severely limited by a high degree of heterogeneity between trials and reported outcomes. Nevertheless, although absence of evidence is not evidence of absence,142 it seems plausible for the purposes of our economic analysis to assume that the three chelators are equivalent in terms of LIC and serum ferritin in the short term (1 year).

It is impossible to use this short-term RCT data to make inferences on long-term health outcomes such as myocardial iron loading, cardiac disease (which is especially important for beta-TM patients) and survival. As deferasirox is a relatively new compound, long-term data from non-RCT sources are not yet available to assess the safety and survival profiles in any patient population. There are some limited survival data from beta-TM patients treated with DFO but how reliable these data are as a proxy for the survival of any patient population treated with deferasirox is questionable. Considering that the adverse event and adherence profiles are known to be different for the two agents, and that the effects of deferasirox on myocardial iron loading and cardiac death in the long term are unknown, making assumptions regarding the long-term benefits of deferasirox seems at best highly speculative and at worst potentially misleading.

Given that it is not possible to determine the long-term health outcomes for patients treated with deferasirox the analysis reduces to a short-term (1 year) assessment. As there is no discernable difference between the three agents in terms of LIC or serum ferritin, the health benefits appear to be restricted to differences in quality of life.

All but one of the published economic evaluations appear to have used the same Australian TTO study137 to estimate the quality of life gain of oral deferasirox compared with infusional DFO (utility scores of 0.85 and 0.61 respectively). As discussed in Chapter 5, Health outcomes, this study may have a number of problems but in general the methodology was sound and hence the results appear credible. A recent UK study (unpublished), which was used in the recent UK economic evaluation (Karnon et al. 135), verified these results (0.84 deferasirox; 0.66 DFO). Personal communication with Novartis, the manufacturer of deferasirox and DFO (Karen Jewitt, July 2007), confirmed that this study employed the same methodology as the Australian study ‘but the vignettes describing the different modes of administration were reviewed by UK clinicians and patients and amended to make sure that they were more relevant to the UK setting. Health states were then elicited using the TTO technique in a cross-section of the UK general population.’ Hence, for our analysis we chose to use the UK figures to estimate the utility of deferasirox (0.84) and DFO (0.66).

No data were identified regarding the utility of deferiprone therapy. Assumptions regarding the utility conferred by deferiprone are required; in view of the high degree of uncertainty, the spectrum of utility values (ranging from 0.66 to 0.84) needs to be explored.

Numerous adverse events are associated with chelation therapy. Of the published economic evaluations, the majority did not include the costs and consequences of adverse events. This is no doubt because of the added complexity of costing and valuing a large number of adverse events, together with the fact that they do not appear to add significantly to the costs or incur large disutilities as demonstrated by those economic evaluations that did include adverse events. However, it is worth noting that none of the economic evaluations considered deferiprone, which has been linked with neutropenia and agranulocytosis, which can incur substantial health costs and disutilities.

For the purposes of our analysis we have not included the costs and health outcomes associated with adverse events. This is primarily because of the inconsistent reporting of adverse events in the trials (see Chapter 4, Combination therapy versus DFO or deferiprone), which makes it almost impossible to estimate the rates of adverse events. Furthermore, it would be difficult to assign disutilities to these adverse events. The end result of including such arbitrary adverse event data would be at best meaningless and at worst misleading. Our decision not to include speculative adverse event data in the model should not greatly alter the results for the comparison of DFO and deferasirox as they do not appear to have major side effects. However, it may affect the results when considering the comparison of deferasirox with deferiprone (because of its link with agranulocytosis), potentially in favour of deferiprone.

Patient populations

Our previous description of the various anaemic conditions that may be at risk of iron overload (see Chapter 2) clearly demonstrates that the different anaemic conditions represent distinct patient populations with regards to aetiology, morbidity and mortality. The strongest evidence of the benefits of chelation therapy comes from patients suffering from beta-TM, followed by SCD patients. There is almost no evidence from MDS patients and very little with regards to other rare anaemias.

Considering that only beta-TM, and to a lesser degree SCD, patients have sufficient evidence of the harms of iron overload and the benefits of chelation therapy, our economic analysis will only consider these two patient populations. Beta-TM and SCD represent distinct populations and may not have the same pattern of organ damage and long-term health benefits (see Chapter 2). However, our short-term model only includes the QoL benefits of oral versus infusional therapy (rather than long-term morbidity and mortality), which should not be dependent on the patient’s underlying disease. Therefore, for the purposes of this economic evaluation we will group SCD and beta-TM patients together.

Costs

There are numerous costs associated with iron overload but for the purposes of this review we have only considered those costs borne by the NHS. Hence, only the costs of chelation therapy, monitoring and administration are discussed. As mentioned earlier, adverse events were not included in our analysis.

Overview of our economic model

We developed a simple 1-year analysis that estimates the costs and benefits of chelation therapy for SCD and beta-TM patients, split into males and females and stratified by age, ranging from 2 to 18 years plus in yearly intervals. The model makes comparisons between deferasirox and DFO and between deferasirox and deferiprone.

The only benefits that could be attributed to the different agents were the utility benefits associated with an oral therapy over infusional chelation therapy. As the analysis was of 1 year in duration only, no discounting was undertaken. The three agents are assumed to be equally effective with regards to removing iron from the body and the analysis does not consider issues of adherence or adverse events.

Only the costs outlined in the previous sections were included in the model. This represents an NHS perspective and, once again, considering the short time frame, it was not appropriate to undertake discounting. All costs are based on 2007 prices apart from DFO administration costs, which are based on 2004/5 prices (Karen Jewitt, Novartis, August 2007, personal communication).

Because of uncertainties regarding the utility associated with deferiprone and the usage of balloon infusers to administer DFO, a range of sensitivity analyses or ‘scenarios’ are presented rather than a single base case. These scenarios are outlined below, split into two comparisons: deferasirox versus DFO and deferasirox versus deferiprone.

This analysis is highly speculative and, given the dearth of data, must be interpreted with caution. As the results of our model are effectively a range of sensitivity analyses, no separate sensitivity analyses were undertaken.

Results

The results of our economic model are presented below, split into the various scenarios. Note that, because of space constraints and the fact that there was virtually no difference in terms of cost-effectiveness between male and female patients, only the results for male patients are shown. Please also be aware that all results are incremental, which is in line with NICE guidance on performing cost-effectiveness analysis. This means that interventions are compared with the most appropriate ‘current treatment’ rather than no therapy.

Economic discussion

We developed a simple short-term (1 year) model to assess the cost-effectiveness of deferasirox versus DFO and deferasirox versus deferiprone. Because of data constraints a range of cost-effectiveness scenarios are presented rather than a single base case. These scenarios are split into the comparisons of deferasirox versus DFO and deferasirox versus deferiprone.

The model suggests that in the short term deferasirox may be a cost-effective strategy compared with DFO administered via the balloon infuser; however, this is dependent on the benefit conferred by the balloon infuser. If it assumed to offer the same utility as the traditional pump then deferasirox is cost-effective for all ages and both SCD and beta-TM. If the balloon infuser offers more utility than the standard pump then deferasirox may not be cost-effective for adults suffering from SCD.

If DFO is administered via the traditional pump, which is cheaper than the balloon infuser, deferasirox may not be cost-effective once patients reach adolescence. This is simply attributed to the fact that as the patients mature they require more of the drug (as it is dosed according to weight), which increases the costs of deferasirox to a point at which the costs exceed the benefits.

When deferasirox is compared with deferiprone it is a less clear-cut picture and depends upon the utility benefit attributed to deferiprone in relation to deferasirox. However, given the large price differential between deferasirox and deferiprone it is unlikely that deferasirox will be generally cost-effective for the majority of patients (short term). In all scenarios deferasirox appears to be cost-effective only in the youngest patients (as the lower doses required incur less extra cost); for older children and adults in all scenarios deferiprone appears to be economically more attractive.

Taken as a whole the results could be interpreted as indicating that in the short term deferiprone is more cost-effective than deferasirox and deferasirox is more cost-effective than DFO. However, there are a number of issues that must be considered.

We have not attempted to assess the costs and consequences of adverse events in our model. Of the eight published economic analyses, only one study included adverse events. The adverse events associated with DFO and deferasirox appear to incur minimal costs (less than £25 for deferasirox and £6 for DFO including monitoring) and have very little impact on utility. However, no studies have attempted to estimate the costs and consequences of adverse events associated with deferiprone. Given that deferiprone has been linked with neutropenia and agranulocytosis, the costs and disutilities associated with deferiprone complications could be expected to be greater than those associated with DFO and deferasirox. This would impact upon the cost-effectiveness of deferiprone and may mean that it is less economically attractive when compared with deferasirox. However, there have been recent warnings that deferasirox may also be associated with neutropenia and agranulocytosis, although this has yet to be confirmed.

A number of other costs were not included in the model or are subject to significant uncertainty. In our costing analysis we chose to take an incremental approach and thus only included the costs that differed between treatment arms. Hence, the total costs borne by the health-care system are likely to have been underestimated for all three agents, although the incremental costs are thought to be accurate within the scope of the analysis.

Furthermore, in our model we chose to take a NHS perspective and therefore only included the costs borne by the health-care system. If a societal perspective were taken, other costs such as patient time and lost earnings would be included. Given the seriousness of the condition these costs are likely to be considerable.

In terms of health benefits our model assumed that, in the short term, benefits would be restricted to quality of life gains. This assumption is based on the findings of our clinical analysis, which was unable to determine a definitive difference between the three iron chelators. However, this does not mean that such a difference does not exist. There is increasing evidence that deferiprone may offer an advantage over DFO in terms of cardiac iron loading. This is especially important for thalassaemia patients as cardiac disease is the leading cause of death in this patient group. However, the crucial factor is to what degree these surrogate outcomes such as liver, serum and cardiac iron translate into long-term outcomes such as morbidity and mortality. Until this is clarified, any small differences between iron chelators in terms of LIC, serum ferritin or cardiac iron cannot be guaranteed to translate into survival benefits. Considering the chronicity of the condition this must be the primary focus for future research.

In conclusion, deferasirox appears to be cost-effective in the short term compared with infusional DFO. However, the model indicates that deferiprone may be more cost-effective than deferasirox, largely because of the high costs of deferasirox in comparison with deferiprone.

However, it cannot be stressed enough that this analysis is exploratory in nature. The appropriateness of deferiprone as a comparator is still controversial because of its side-effect profile, something that was not explored in this analysis. Furthermore, there was a dearth of data, which necessitated a short-term analysis and a number of assumptions. To be able to form more robust conclusions, further research is required regarding:

• the long-term benefits of the three chelators in each patient population

• the costs of the three chelators in the long term

• the adverse event and adherence profiles of the three chelators in the long term.

Eligible patient populations

As described in Chapter 2 there are approximately 624 iron-overloaded patients living with beta-TM and 625 iron-overloaded patients suffering from SCD in the UK. Each year there will be an additional 15 cases of iron overload diagnosed in beta-TM patients and 16 cases of iron overload diagnosed in SCD patients. To turn these estimates into useable figures for assessing the budget impact of deferasirox, a number of assumptions must be made.

First, with regards to prevalence estimates, an age distribution needs to be applied to determine the proportion of patients at each age and their associated costs. To determine the age distribution for each disease, data were taken from clinical trials; for beta-TM the Capellini et al. trial73 was used, whereas for SCD the Vichinsky et al. trial81 was employed. A log-normal distribution was fitted to each data set to determine the number of patients at each age group (see Appendix 8). As our model categorises adults as aged 18 years plus, we needed to estimate the proportion of patients in this group. To do this it was simply a case of summing the proportions from 18 to 64 years. For both diseases adults account for approximately half of the total patient population in the RCTs (beta-TM = 49.5%; SCD = 49.6%). Here we have to make the assumption that the RCTs are reflective of clinical practice, which may not be true.

For incidence estimates the age at which patients are diagnosed and treated for iron overload had to be estimated. For the 15 cases of iron-overloaded beta-TM patients it was assumed that they would present at age 2 years. For the 16 cases of iron-overloaded SCD patients it was assumed that they would present at age 4 years. These estimates are taken from analysis of the trial data and also concur with expert opinion.

Costs

The costs used in our model were also used for estimating the budget impact. These costs include the costs of chelation therapy, monitoring and administration. As there is very little difference between male and female patients in terms of costs, the costs for male patients are used. See Table 20 for an example of the cost estimates for beta-TM and SCD male patients.

Budget impact results

For each condition we present below a range of budget impact assessments. These analyses are exploratory and aim to give an indication of the likely budget impacts rather than precise estimates.

Contributions of authors (alphabetically)

Professor Adrian Bagust developed the economic model and had input into all aspects of the economic review. Angela Boland, Research Fellow, was responsible for the economic review and writing and editing of drafts of the report. Patrick Chu, Consultant Haematologist, had input into the clinical component of the review. Rumona Dickson, Director, LriG, had input into all aspects of the clinical component of the review. Yenal Dundar, Research Fellow, was responsible for development of search strategies a and study selection and had input into aspects of the clinical component of the review. Nigel Fleeman, Research Fellow, was responsible for data management and had input into all aspects of the clinical review. Janette Greenhalgh, Research Fellow, had into all aspects of the clinical review. Jamie Kirkham, Research Associate, was responsible for statistical advice and meta-analysis and had input into all aspects of the clinical review. Claire McLeod, Research Fellow, was responsible for co-ordination of the review and the background, economic review and development of the economic model. Bernadette Modell, Emeritus Professor of Community Genetics, Ade Olujohungbe, Consultant Haematologist, and Paul Telfer, Senior Lecturer in Haematology, all had input into the clinical component of the review, and Bernadette Modell and Paul Telfer wrote the background. Professor Tom Walley was responsible for data assessment and interpretation of clinical data. All contributors took part in the editing and production of the final report.

Disclaimers

The views expressed in this publication are those of the authors and not necessarily those of the HTA Programme or the Department of Health.